COMPIT 2005 in Hamburg - TUHH
COMPIT 2005 in Hamburg - TUHH
COMPIT 2005 in Hamburg - TUHH
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
!!"<br />
<br />
<br />
#$% &'()<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
*<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
!!!" #$ %<br />
&'&%()$*!<br />
&#+#$&+%&$(<br />
<br />
<br />
*%+,!!!-,
.<br />
<br />
% <br />
<br />
<br />
-./01$* 2<br />
!"# <br />
<br />
3(*&# , <br />
$% & %'" %<br />
<br />
*+#$/$%$ 42<br />
( ) <br />
<br />
/5+067#0)&8 9<br />
%) <br />
<br />
:0; 0:?6$$08 4<br />
-,./.0*' $ <br />
<br />
<br />
@%+0+7AB.0-(& ,4<br />
)1 <br />
<br />
= (7 4<<br />
'$ $!2 ) <br />
<br />
*((@0)%$#0''30'& "<br />
-(<br />
<br />
@$@0& 9<br />
%) . )$<br />
<br />
&#@08#/$0##8#
-0'%/%+0)%$#0--## ,<br />
*! ! <br />
<br />
-('0'#: ,,<br />
"*, !$$<br />
<br />
$-0/+*& ,4<br />
$! ! <br />
<br />
'$A0&# ,42<br />
& %*,) ( 1"<br />
<br />
+@#& , 4<br />
$!* <br />
<br />
#0*#:+ ,<br />
2 # #6)/)0( <br />
<br />
!= 0@3.= $ ,"<br />
"*(' <br />
<br />
)0'%@05B ,2,<br />
78989*:7; $ ) ($ ) ( <br />
(<<br />
<br />
6E$E&@0F7+0-!$&0'+ ,"4<br />
( 5 " 3 <br />
<br />
/(@#&@+0((= 06(0-:( ,9 <br />
)$! 6)( ($<br />
<br />
B:%0$#B*B+&B 42<br />
$ * ! -$<br />
<br />
G%+B0-0-#B+3 4 <br />
) - <br />
<br />
:%#0)%$#0-$= + 4,<br />
!/ 0$!')<br />
<br />
<br />
#&#$ 44<<br />
% ) %%( <br />
<br />
&#@ 4 9<br />
2 !$<br />
<br />
-3#$0&507##0#(5 4< <br />
( ! <br />
<br />
;%$@A0&*&# 422<br />
6.$5'5' 1 <br />
'<br />
<br />
'%@0&G 4"<br />
& %( $"'=<br />
4
*# 49<br />
) ) #<br />
<br />
@(5 499<br />
' # % $<br />
<br />
F7+0#$@0F06E$E&@0-!$&0'+ 9<br />
6 2) <br />
<br />
+ ,<<br />
) & %*, <br />
$% 3<br />
<br />
:$0�'(*0@'(8 <br />
.) )% 3<br />
<br />
/+'30@$ <br />
$ 2 )<br />
<br />
/!AAB0-@H#0);0= # #!$2*!=<br />
( " 3<br />
<br />
G%+B0-#B)B <br />
( ))<br />
<br />
7B53 <br />
63 %+1 %"6 ) <br />
<br />
/#+# ,,<br />
*'$ ' %'3( 4! <br />
<br />
:#53$+0#*0+ 44<br />
<br />
<br />
807@ "<br />
),1' ' ) <br />
<br />
$?# "<br />
<br />
Impacts of Next Generation Ship Navigation and Communication Systems<br />
Abstract<br />
Martha Grabowski, Le Moyne College, Rensselaer Polytechnic Institute, Syracuse/USA,<br />
grabowsk@lemoyne.edu<br />
Over the past two decades, there have been enormous advances <strong>in</strong> comput<strong>in</strong>g, communications,<br />
network, display and control systems. These advances have resulted <strong>in</strong> <strong>in</strong>creas<strong>in</strong>gly capable, complex,<br />
and <strong>in</strong>telligent shipboard and shoreside communications, control and navigation systems. The<br />
<strong>in</strong>terface between <strong>in</strong>telligent ships and AIS-equipped shore stations is an <strong>in</strong>terest<strong>in</strong>g example of the<br />
challenges faced by ship and technology designers and developers <strong>in</strong>terested <strong>in</strong> enhanc<strong>in</strong>g ship and<br />
human performance as well as the safety of navigation. As <strong>in</strong>creas<strong>in</strong>gly capable ship and shoreside<br />
systems have proliferated, questions have surfaced about the effectiveness, usefulness and usability of<br />
such systems. This paper discusses the promises and pitfalls of current ship and shore-based<br />
advanced technology systems, describes recent empirical work to evaluate the impact of these<br />
technologies on the St. Lawrence Seaway, and provides a vision of future systems with <strong>in</strong>terest<strong>in</strong>g<br />
implications for future mar<strong>in</strong>ers.<br />
1. Introduction<br />
Over the past two decades, there have been enormous advances <strong>in</strong> comput<strong>in</strong>g, communications,<br />
network, display and control systems. These advances have resulted <strong>in</strong> <strong>in</strong>creas<strong>in</strong>gly capable, complex,<br />
and <strong>in</strong>telligent shipboard and shoreside communications, control and navigation systems. The<br />
<strong>in</strong>terface between <strong>in</strong>creas<strong>in</strong>gly capable ships and AIS-equipped shore stations is an <strong>in</strong>terest<strong>in</strong>g<br />
example of the challenges faced by ship and technology designers and developers <strong>in</strong>terested <strong>in</strong><br />
enhanc<strong>in</strong>g ship and human performance as well as the safety of navigation.<br />
As <strong>in</strong>creas<strong>in</strong>gly capable ship and shoreside systems have proliferated, questions have surfaced about<br />
the effectiveness, usefulness and usability of such systems. This paper discusses the promises and<br />
pitfalls of current ship and shore-based advanced technology systems, describes one recent study to<br />
evaluate the impact of a new technology on the St. Lawrence Seaway.<br />
2. The Pitfalls of Technology and Automation<br />
New technology is often <strong>in</strong>troduced <strong>in</strong> systems as a way of improv<strong>in</strong>g the system, and human<br />
performance <strong>in</strong> the system. In mar<strong>in</strong>e transportation, a particular accident or <strong>in</strong>cident can be the<br />
catalyst, as the accident may identify circumstances <strong>in</strong> which human error was seen to be a major<br />
contribut<strong>in</strong>g factor. Technology can be designed <strong>in</strong> an attempt to remove the source of error and<br />
improve system performance, often by automat<strong>in</strong>g functions carried out by a human operator. Often<br />
the assumption is made that technology will simplify the operator’s job and reduce errors and costs,<br />
Wickens et al. (1997), p.265.<br />
The available evidence suggests that this assumption is often supported: technology shr<strong>in</strong>ks costs and<br />
reduces or even elim<strong>in</strong>ates certa<strong>in</strong> types of human error. However, technology can also <strong>in</strong>troduce new<br />
human error forms, Wiener (1988), and “automation surprises” can puzzle the operator, Sarter and<br />
Woods (1995). These effects can reduce system efficiency or compromise safety, negat<strong>in</strong>g the other<br />
benefits that technology provides. These costs and benefits have been noted especially <strong>in</strong> the case of<br />
cockpit automation, Wiener (1988), but they also occur <strong>in</strong> mar<strong>in</strong>e transportation as well.<br />
One of the ironies of technology is that automation designed to reduce operator workload sometimes<br />
<strong>in</strong>creases it, Ba<strong>in</strong>bridge (1983, Rochl<strong>in</strong> (1997). In addition, the <strong>in</strong>troduction of technology can lead to<br />
manual skill deterioration, alteration of workload patterns, poor monitor<strong>in</strong>g, <strong>in</strong>appropriate responses<br />
to alarms, and reductions <strong>in</strong> job satisfaction, Wiener and Curry (1980). Problems can arise not only<br />
5
ecause of the technology per se but also because of the way the technology is implemented <strong>in</strong><br />
practice. Problems of false alarms from automated alert<strong>in</strong>g systems, automated systems that provide<br />
<strong>in</strong>adequate feedback to the human operator, Norman (1990), and automation that fails “silently”<br />
without salient <strong>in</strong>dications, fall <strong>in</strong>to this category. Sometimes these problems can be alleviated by<br />
more effective operator tra<strong>in</strong><strong>in</strong>g, but not always.<br />
Problems can also arise from unanticipated <strong>in</strong>teractions between technology, human operators, and<br />
other systems <strong>in</strong> the environment. These can be problems <strong>in</strong>herent to the technology, as well as to the<br />
technology’s behavior <strong>in</strong> a larger, more complex system, Tenner (1996). This can be seen <strong>in</strong><br />
<strong>in</strong>tegrated ship’s bridge systems, where system performance can be determ<strong>in</strong>ed by the <strong>in</strong>teraction of<br />
multiple sensors, systems, people, and technologies, Grabowski and Roberts (1996), Grabowski and<br />
Sanborn (2001).<br />
The benefits of new technology <strong>in</strong> mar<strong>in</strong>e transportation can <strong>in</strong>clude cost sav<strong>in</strong>gs, more precise<br />
navigation and vessel control, fuel efficiency, all-weather operations, elim<strong>in</strong>ation of some error types,<br />
and reduced operator workload dur<strong>in</strong>g certa<strong>in</strong> phases of the voyage. For <strong>in</strong>stance, some of the benefits<br />
of today’s bridge and navigation systems <strong>in</strong>clude improved awareness of hazardous conditions, the<br />
ability to detect risk of collision, elim<strong>in</strong>ation of some rout<strong>in</strong>e actions that allow the operator to<br />
concentrate on other tasks, and, on some waterways, a reduction <strong>in</strong> unnecessary verbal<br />
communications on congested voice traffic frequencies.<br />
However, the benefits of new automation and technology <strong>in</strong> mar<strong>in</strong>e transportation are not guaranteed.<br />
In some cases, the economic arguments that <strong>in</strong>itially stimulate <strong>in</strong>vestment are clearly re<strong>in</strong>forced by the<br />
f<strong>in</strong>ancial return on that <strong>in</strong>vestment. On the other hand, technology benefits can also be overcome by<br />
disadvantages or cost. For <strong>in</strong>stance, bridge automation can reduce operator workload dur<strong>in</strong>g some<br />
voyage phases, but can <strong>in</strong>crease workload <strong>in</strong> other phases of the voyage, perhaps dur<strong>in</strong>g arrival or<br />
departure, particularly if the operator is unfamiliar with the technology, Grabowski and Sanborn<br />
(2001). As a result, the anticipated workload reduction benefit of the technology may not be realized.<br />
Thus, there are promises and pitfalls to the <strong>in</strong>troduction of new technology. We can expect that some<br />
of these phenomena will be evident <strong>in</strong> new technology systems, and we can expect that new promises<br />
and pitfalls will arise as new systems are <strong>in</strong>troduced. We now explore the impact of the <strong>in</strong>troduction<br />
of one such technology, automated identification systems (AIS).<br />
New Technology Introduction: Automated Identification Systems<br />
Automated Identification Systems (AIS), a communications protocol be<strong>in</strong>g developed under the aegis<br />
of the International Maritime Organization (IMO), was <strong>in</strong>itially implemented aboard vessels around<br />
the world <strong>in</strong> 2003. This technology is designed to automatically provide vessel position and other data<br />
to other vessels and shore stations and to facilitate the communication of vessel traffic management<br />
and navigational safety data from designated shore stations to vessels. The onboard "AIS unit"<br />
(which consists of a VHF-FM transceiver, an assembly unit, and a communications transceiver) cont<strong>in</strong>uously<br />
and automatically broadcasts identification, location, and other vessel voyage data, and receives<br />
messages from other ships and shore stations, National Research Council (2003).<br />
Many benefits of AIS have been projected. For ship owners, AIS is projected to reduce transit times,<br />
with accompany<strong>in</strong>g lower fuel consumption, improv<strong>in</strong>g fleet management. S<strong>in</strong>ce arrival times will be<br />
accurately known, better schedul<strong>in</strong>g of passages through locks, m<strong>in</strong>imiz<strong>in</strong>g delays, is possible, as pilots,<br />
<strong>in</strong>spectors, and other shore-based personnel can be dispatched when needed. AIS is also projected<br />
to enhance safety by transmitt<strong>in</strong>g precise environmental <strong>in</strong>formation to vessels and by transmitt<strong>in</strong>g<br />
real-time ship-to-ship communications of course and location of each equipped vessel. AIS is<br />
also projected to enhance traffic management by cont<strong>in</strong>uously monitor<strong>in</strong>g vessel location and speed <strong>in</strong><br />
all weather conditions, permitt<strong>in</strong>g timely pilot dispatch<strong>in</strong>g, timely ship <strong>in</strong>spections, better speed control,<br />
better schedul<strong>in</strong>g of lockages and vessel tie-ups, and faster response times to accidents and/or <strong>in</strong>cidents,<br />
particularly when hazardous cargoes are <strong>in</strong>volved.<br />
6
However, there is much discussion <strong>in</strong> the maritime community about the costs and benefits of AIS,<br />
Creech and Ryan (2003), and the effectiveness of AIS displays, National Research Council (2003),<br />
particularly <strong>in</strong> the wake of collisions that might have been avoided with use of AIS, Darce (2004). As<br />
a result, <strong>in</strong>vestigat<strong>in</strong>g the impact of AIS is a topic of current research <strong>in</strong>terest. This paper reports on<br />
the impacts of the <strong>in</strong>troduction of AIS <strong>in</strong> a maritime system that was among the first <strong>in</strong> the world to<br />
require AIS carriage for commercial vessels transit<strong>in</strong>g its waterways – the St. Lawrence Seaway. The<br />
follow<strong>in</strong>g sections discuss the hypotheses under <strong>in</strong>vestigation, the motivation for the hypotheses, and<br />
the measures, variables and data collection methods used. Results and conclusions from the study are<br />
conta<strong>in</strong>ed <strong>in</strong> the f<strong>in</strong>al section of this paper.<br />
3. Hypotheses and Measures<br />
Determ<strong>in</strong><strong>in</strong>g the impacts of AIS technology <strong>in</strong>troduction on operator performance and perceptions<br />
was the focus of this study. The hypotheses and measures utilized thus assess how operators perceived<br />
that the AIS system supported them, how adequate the system’s decision support facilities were, how<br />
the technology impacted operator performance, and how operators perceived the utility and/or benefits<br />
of the AIS technology.<br />
Both <strong>in</strong>put-output and process measures are used <strong>in</strong> this study to assess the impacts of AIS technology<br />
on operator performance. Operators are def<strong>in</strong>ed as <strong>in</strong>dividuals us<strong>in</strong>g the AIS technology on vessels<br />
and <strong>in</strong> the St. Lawrence Seaway traffic management system; thus, subjects are vessel masters, mates<br />
on watch, and Seaway Traffic Management System (TMS) operators. Hypotheses 1-2 <strong>in</strong> Table I<br />
consider <strong>in</strong>put-output measures: the performance of the AIS users, and whether users consider more<br />
alternatives <strong>in</strong> their decision-mak<strong>in</strong>g. Hypotheses 3-7 consider process measures of and human<br />
reactions to the technology: whether users report lower workloads and stress, greater confidence and<br />
satisfaction, <strong>in</strong>creased effort us<strong>in</strong>g the system, greater vigilance and lower fatigue when utiliz<strong>in</strong>g the<br />
technology. The operationalizations, dependent variables and data collection methods for each<br />
hypothesis are summarized <strong>in</strong> Table II.<br />
Table I: Hypotheses<br />
1 AIS-supported operators will show better decision performance than operators not us<strong>in</strong>g AIS.<br />
2 AIS-supported operators will consider more alternatives than operators not us<strong>in</strong>g AIS.<br />
3a. AIS-supported operators will report lower workload than operators not us<strong>in</strong>g AIS.<br />
3b. AIS-supported operators will report lower stress than operators not us<strong>in</strong>g AIS.<br />
4a. AIS-supported operators will report greater confidence <strong>in</strong> the system than operators not us<strong>in</strong>g AIS.<br />
4b. AIS-supported operators will report greater satisfaction <strong>in</strong> the system than operators not us<strong>in</strong>g<br />
AIS.<br />
5 AIS-supported operators will show decreased effort <strong>in</strong> us<strong>in</strong>g the system, compared to operators<br />
not us<strong>in</strong>g AIS.<br />
6a. AIS-supported operators will report greater monitor<strong>in</strong>g vigilance than operators not us<strong>in</strong>g AIS.<br />
6b. AIS-supported operators will report less fatigue than operators not us<strong>in</strong>g AIS.<br />
7a. AIS-supported operators will perceive greater decision support capabilities provided by the technology<br />
than operators not us<strong>in</strong>g AIS.<br />
7b. AIS-supported operators will perceive greater system benefits than operators not us<strong>in</strong>g AIS.<br />
7c. AIS-supported operators will report lower degrees of distraction provided by technology than<br />
operators not us<strong>in</strong>g AIS.<br />
7
Table II: Operationalizations and Measurement of Dependent Variables by Hypothesis<br />
Operator Performance Input-Output Assessment Measures<br />
Hypothesis<br />
Dependent<br />
Variable<br />
Variable Operationalization<br />
Hypothesis 1 Operator decision performance<br />
Operator Decision Performance<br />
<strong>in</strong>:<br />
• Situation Awareness<br />
• Threat Avoidance<br />
• Situation Monitor<strong>in</strong>g<br />
• Voyage Plan Monitor<strong>in</strong>g<br />
Data Collection Method<br />
• Post-transit questionnaire<br />
Hypothesis 2<br />
Hypothesis 3a<br />
Hypothesis 3b<br />
Number of decision alternatives<br />
considered by operators<br />
Operator workload<br />
Operator stress<br />
• Ability to evaluate alternative<br />
avoidance maneuvers<br />
• Vessel and TMS operators’<br />
perceived workload<br />
• Self reported operator stress<br />
• Post-transit questionnaire<br />
• NASA Task Load<br />
Index (TLX) assessment<br />
of subjective<br />
workload<br />
• Post-transit questionnaire<br />
Hypothesis 4a<br />
Operator decision confidence<br />
• Self reported operator confidence<br />
• Post-transit<br />
questionnaire<br />
Hypothesis 4b<br />
Operator satisfaction<br />
• Self reported operator satisfaction<br />
Hypothesis 5 Operator effort • Self reported physical effort<br />
• Self reported mental effort<br />
Hypothesis 6a<br />
Operator vigilance<br />
• Self reported operator vigilance<br />
• Post-transit questionnaire<br />
• Post-transit questionnaire<br />
• NASA TLX<br />
• Post-transit questionnaire<br />
Hypothesis 6b<br />
Operator fatigue<br />
• Self reported operator fatigue<br />
• Post-transit questionnaire<br />
Hypothesis 7a<br />
Hypothesis 7b<br />
Hypothesis 7c<br />
Operator perception of<br />
decision support capabilities<br />
Operator perception of<br />
system benefit<br />
Operator perception of<br />
degree of distraction<br />
• Operator perception of support<br />
provided by technology<br />
for situation awareness<br />
• Operator perception of system<br />
benefit<br />
• Operator perception of degree<br />
of distraction provided<br />
by technology<br />
• Post-transit questionnaire<br />
• Post-transit questionnaire<br />
• Post-transit questionnaire<br />
4. Procedure<br />
Data for the AIS pre-implementation and post-implementation phases were gathered. Basel<strong>in</strong>e data<br />
for the pre-AIS condition were gathered prior to AIS carriage requirement <strong>in</strong>itiation, from May 2002<br />
to December 2002. Data were collected <strong>in</strong> the form of a survey that was adm<strong>in</strong>istered to the subjects.<br />
The survey consisted of a questionnaire and a workload <strong>in</strong>dex survey. The questionnaire solicited operator<br />
op<strong>in</strong>ions about the effects of the technology on the operators’ decision performance, satisfaction,<br />
and confidence. For each question, the answers had a range along a seven-po<strong>in</strong>t scale, from<br />
“Strongly Agree” to “Strongly Disagree.” In the analysis, “Strongly Disagree” was <strong>in</strong>terpreted to cor-<br />
8
espond to a score greater than or equal to 6 on the 7-po<strong>in</strong>t scale. “Strongly Agree” was <strong>in</strong>terpreted to<br />
correspond to a score of less than 3 on the 7-po<strong>in</strong>t scale, and “Partially Agree” was <strong>in</strong>terpreted to correspond<br />
to a score of 3 or greater, but less than 6, on the 7-po<strong>in</strong>t scale.<br />
The workload <strong>in</strong>dex referred to as the NASA (U.S. National Aeronautics and Space Adm<strong>in</strong>istration)<br />
Task Load Index (TLX) solicited op<strong>in</strong>ions about six factors that contribute to workload when us<strong>in</strong>g a<br />
technology, Hart and Staveland (1988). In the NASA TLX, operators were asked to compare which<br />
of two workload factors was dom<strong>in</strong>ant when an operator was us<strong>in</strong>g either the pre-AIS (exist<strong>in</strong>g bridge<br />
technology) or AIS technology. The operator was asked to circle one factor <strong>in</strong> each pair that was<br />
dom<strong>in</strong>ant when the operator was us<strong>in</strong>g the technology. Operators were also asked to rate on a scale<br />
from zero to 100 the six factors that contribute to workload <strong>in</strong> their daily tasks — mental demand,<br />
physical demand, temporal demand, effort, performance, or frustration. These terms were def<strong>in</strong>ed for<br />
the operator with<strong>in</strong> the body of the survey.<br />
Data for the post-AIS condition were gathered immediately follow<strong>in</strong>g AIS carriage requirement <strong>in</strong>itiation,<br />
from May 2003 to August 2003. Thus, subjects were exposed to two technology treatments – one<br />
pre-AIS, and one post-AIS technology – and were queried about their perceptions of the technology.<br />
Data for both conditions were collected aboard participat<strong>in</strong>g vessels mak<strong>in</strong>g vessel transits along the<br />
St. Lawrence Seaway between Montreal and Duluth, M<strong>in</strong>nesota and from St. Lawrence Seaway Traffic<br />
Management System (TMS) controllers operat<strong>in</strong>g the St. Lawrence Seaway Traffic Management<br />
System center <strong>in</strong> Massena, New York.<br />
4.1. Subjects<br />
Human subjects were bridge watch team members – capta<strong>in</strong>s and mates on watch aboard vessels mak<strong>in</strong>g<br />
transits of the St. Lawrence Seaway dur<strong>in</strong>g the data collection periods – as well as Seaway Traffic<br />
Management System (TMS) operators at Eisenhower Lock <strong>in</strong> Massena, New York. Thus, there were<br />
three types of subjects <strong>in</strong> this study: vessel capta<strong>in</strong>s, mates on watch, and TMS operators. Demographic<br />
<strong>in</strong>formation such as position, educational qualifications, experience at sea or at the Traffic<br />
Management Center, and familiarity with the technology under study was sought from the subjects. A<br />
total of 48 subjects participated <strong>in</strong> the study: 21 subjects for the pre-AIS condition, and 27 for the<br />
post-AIS condition. Table III provides the breakdown <strong>in</strong> types of subjects; Table IV provides <strong>in</strong>formation<br />
on the educational background of the respondents. Table V provides other demographic <strong>in</strong>formation<br />
such as age and familiarity with the system.<br />
Subject Type<br />
Number of<br />
subjects<br />
Table III: Participation by Subject type<br />
% Participation <strong>in</strong><br />
Pre-AIS survey<br />
Number of<br />
subjects<br />
% Participation <strong>in</strong><br />
Post-AIS survey<br />
Masters 6 33.32% 18 66.67%<br />
Mates 11 47.63% 5 18.52%<br />
TMS Operators 4 19.05% 4 14.81%<br />
Total 21 27<br />
In the pre-AIS condition, respondents were mostly mates, with 48% of the respondents <strong>in</strong> the pre-AIS<br />
condition mates, 33% ship capta<strong>in</strong>s (masters) and the rema<strong>in</strong><strong>in</strong>g 19% TMS operators. In the post-AIS<br />
condition, most (67%) of the respondents were ship’s capta<strong>in</strong>s; with 19% mates on watch and the rema<strong>in</strong><strong>in</strong>g<br />
respondents TMS operators.<br />
Table IV: Subjects’ Educational backgrounds<br />
9
Highest degree<br />
received<br />
Number of<br />
subjects<br />
% Participation <strong>in</strong><br />
Pre-AIS survey<br />
Number of<br />
subjects<br />
% Participation <strong>in</strong><br />
Post-AIS survey<br />
Undergraduate 5 23.81% 7 25.92%<br />
Graduate 4 19.04% 0 0%<br />
Professional<br />
Tra<strong>in</strong><strong>in</strong>g<br />
7 33.34% 16 59.26%<br />
Other 5 23.81% 4 14.82%<br />
Total 21 27<br />
In terms of educational background, <strong>in</strong> the pre-AIS condition, respondents were fairly evenly split,<br />
with 33% of the respondents report<strong>in</strong>g hav<strong>in</strong>g received professional tra<strong>in</strong><strong>in</strong>g, 23% of respondents report<strong>in</strong>g<br />
hav<strong>in</strong>g received an undergraduate degree, and 19% of respondents report<strong>in</strong>g hav<strong>in</strong>g received a<br />
graduate degree. This educational background breakdown was echoed <strong>in</strong> the post-AIS condition, despite<br />
the overwhelm<strong>in</strong>g number of ship’s capta<strong>in</strong>s <strong>in</strong> the post-AIS condition response pool. In the<br />
post-AIS condition, almost 60% of the respondents reported receiv<strong>in</strong>g professional tra<strong>in</strong><strong>in</strong>g, such as a<br />
maritime school or union tra<strong>in</strong><strong>in</strong>g course; about 25% reported receiv<strong>in</strong>g an undergraduate degree. All<br />
subjects <strong>in</strong> both treatments were highly experienced, with an average of 23 years of sail<strong>in</strong>g experience<br />
<strong>in</strong> the pre-AIS condition, and an average of 26 years of sail<strong>in</strong>g experience <strong>in</strong> the post-AIS condition.<br />
The average age of participat<strong>in</strong>g subjects <strong>in</strong> the pre-AIS condition was about 45 years while that of<br />
those participat<strong>in</strong>g <strong>in</strong> post-AIS was 47 years. Note that the mates were a younger group of subjects,<br />
particularly <strong>in</strong> the post-AIS group. All the subjects reported a familiarity level of 2.3 to 2.5 out of 7<br />
with the system <strong>in</strong> use. The familiarity level was measured on a scale of 1 to 7 with 1 be<strong>in</strong>g ‘Very<br />
Familiar’ and 7 be<strong>in</strong>g ‘Very Unfamiliar’. Table V shows that the TMS operators <strong>in</strong>dicated the highest<br />
level of familiarity with the AIS technology.<br />
4.2. Experimental Design<br />
The experiment was a 3 x 2 design: three types of subjects – Masters (capta<strong>in</strong>s), Mates and Traffic<br />
Management System Operators – were exposed to two technology treatments: the pre-AIS condition,<br />
and the post-AIS. AIS transits varied accord<strong>in</strong>g to the participat<strong>in</strong>g vessels’ schedules throughout the<br />
data collection period. Data were captured <strong>in</strong> form of the questionnaires and workload <strong>in</strong>dex surveys<br />
for both technology treatments. Not all transits <strong>in</strong>volved data collection from subjects, and because<br />
the number of respondents was small, the replications <strong>in</strong> Table VI are also small.<br />
Table V: Participat<strong>in</strong>g Subjects’ Demographic Information<br />
Subject Type Pre-AIS Post-AIS<br />
Average Age Familiarity to system Average Age Familiarity to system<br />
Overall 45.71 2.3 47 2.52<br />
Masters 50.33 2.33 50.13 2.88<br />
Mates 41.63 2.36 32 2<br />
TMS<br />
Operators<br />
50 2 49.25 1.33<br />
Table VI: Subjects Exposed to Technology Treatments<br />
10
Type of Subjects Pre- AIS Post-AIS Total<br />
Masters 6 18 24<br />
Mates 11 5 16<br />
TMS Operators 4 4 8<br />
Total 21 27 48<br />
4.3. Method<br />
The hypotheses <strong>in</strong> Table I consider the performance of AIS users, and whether users considered more<br />
alternatives, reported lower workloads, reported greater confidence and satisfaction, reported requir<strong>in</strong>g<br />
greater effort when us<strong>in</strong>g the system, and reported lower fatigue and greater vigilance when us<strong>in</strong>g<br />
the AIS. All hypotheses used a post-transit questionnaire, and H3a and H5 utilized the NASA Task<br />
Load Index (TLX) <strong>in</strong> addition to the post-transit questionnaire.<br />
Hypothesis 1 focuses on operators’ perception of decision performance pre-AIS and post-AIS. Hypothesis<br />
2 focuses on the number of alternatives considered by operators pre- and post-AIS, while<br />
Hypothesis 3 focuses on subject workload and stress associated with the use of the technology. Hypothesis<br />
4 focuses on operators’ decision confidence and satisfaction, and hypothesis 5 assesses operator<br />
effort when us<strong>in</strong>g the technology. Hypothesis 6 focuses on operator fatigue and vigilance, and<br />
hypothesis 7 focuses on operator perceptions of the technology, particularly any perceived system<br />
benefits, perceptions of decision support capabilities provided by the technology, and operator perceptions<br />
of the degree of distraction <strong>in</strong>troduced by the technology.<br />
Operator decision performance is an important variable to consider when us<strong>in</strong>g technology <strong>in</strong> a safety<br />
critical system. This dependent variable <strong>in</strong>cludes factors such as operator and decision performance<br />
with respect to situation awareness and monitor<strong>in</strong>g, threat avoidance, and voyage plan monitor<strong>in</strong>g.<br />
Each of these factors has been identified as an <strong>in</strong>put towards the operators’ decision performance,<br />
Grabowski and Sanborn (2001). Hence, the post-transit questionnaire set <strong>in</strong>cluded questions that <strong>in</strong>dividually<br />
addressed each of the factors identified above. Operators were asked their op<strong>in</strong>ion about<br />
whether the technology improved their decision performance us<strong>in</strong>g the above-identified dependent<br />
variables.<br />
The data were analyzed us<strong>in</strong>g Chi-Square tests, which provide the basis for judg<strong>in</strong>g whether two<br />
population proportions can be considered equal, and whether the hypothesis of <strong>in</strong>dependence between<br />
variables is tenable. In this study, the hypothesis of <strong>in</strong>dependence between pre-AIS and post-AIS was<br />
tested, which contradicts H o , and the results of the data analysis will be phrased <strong>in</strong> terms of accept<strong>in</strong>g<br />
or reject<strong>in</strong>g the hypothesis of <strong>in</strong>dependence; further analysis is then made <strong>in</strong> terms of support<strong>in</strong>g or<br />
reject<strong>in</strong>g H o . Data for overall subjects, and then subjects under each category <strong>in</strong>dividually, have been<br />
presented for each variable.<br />
For the Chi-Square test, the p-value is compared to the level of significance to check whether the hypothesis<br />
is accepted or rejected. P-values lower than or closer to the significance level <strong>in</strong>dicate strong<br />
or partial support to the hypothesis. The p-values reported <strong>in</strong> Table VII are compared with the level of<br />
significance to check the acceptance or rejection of the hypothesis for each of the variables measured<br />
such as situation awareness, threat avoidance, situation monitor<strong>in</strong>g, and voyage plan monitor<strong>in</strong>g.<br />
Operator workload was measured through the NASA TLX metrics scores, Hart and Staveland (1988),<br />
for the pre-AIS and post-AIS periods. The NASA TLX is a subjective workload assessment tool used<br />
to perform subjective workload assessments on operator(s) work<strong>in</strong>g with various human-mach<strong>in</strong>e systems.<br />
It is a multi dimensional rat<strong>in</strong>g procedure that derives an overall workload score based on a<br />
weighted average of rat<strong>in</strong>gs on six subscales: Mental Demand, Physical Demand, Temporal Demand,<br />
Performance, Effort and Frustration.<br />
11
The NASA comparison cards and rat<strong>in</strong>g scale were performed by each subject and cumulative workload<br />
scores for each were obta<strong>in</strong>ed by add<strong>in</strong>g up the results for each of the NASA TLX metrics. The<br />
mean for workload scores was then calculated for the three different types of subjects, as well for subjects<br />
overall.<br />
The Kruskal-Wallis test was used to analyze the workload <strong>in</strong>dex. The Kruskal-Wallis test provides a<br />
statistical procedure for test<strong>in</strong>g whether the two populations are identical. In this research, the hypothesis<br />
of identical populations between pre-AIS and post-AIS was tested, which contradicts H?.<br />
The results of the data analysis are thus phrased <strong>in</strong> terms of accept<strong>in</strong>g or reject<strong>in</strong>g the hypothesis of<br />
identical population. Further analysis is then be made <strong>in</strong> terms of support<strong>in</strong>g or reject<strong>in</strong>g H o . Data for<br />
overall subjects, and then subjects under each category <strong>in</strong>dividually, have been presented for each<br />
variable.<br />
Operator stress was measured by self-reported stress levels provided by the operator <strong>in</strong> their responses<br />
to the post-transit questionnaire. The question asked the operators whether us<strong>in</strong>g a particular technology<br />
was stressful. Operator effort was measured by operators’ self-reported effort <strong>in</strong> the post-transit<br />
questionnaire, and the effort score <strong>in</strong> NASA TLX. The first question asked the operator whether the<br />
technology helps reduce physical effort and the second question asked whether the technology help<br />
reduce mental effort <strong>in</strong> complet<strong>in</strong>g their tasks <strong>in</strong> the questionnaire. The metrics of ‘Effort’ <strong>in</strong> the<br />
NASA TLX reflect the operators’ perception of the mental and physical effort required to use the system,<br />
Hart and Staveland (1988).<br />
5. Results<br />
Table VII summarizes the f<strong>in</strong>d<strong>in</strong>gs of the data analysis. The results show a mixed reaction of subjects<br />
toward AIS. On the plus side, subjects reported overall improved voyage plan monitor<strong>in</strong>g (p= 0.038),<br />
greater monitor<strong>in</strong>g vigilance (p=0.049), and greater decision support capabilities when us<strong>in</strong>g AIS (p=<br />
0.038). At the same time, overall, more physical effort was required by subjects us<strong>in</strong>g AIS (p=0.014).<br />
Thus, overall, subjects found the contribution of AIS to be <strong>in</strong> voyage plan monitor<strong>in</strong>g, vigilance<br />
monitor<strong>in</strong>g and decision support capabilities, rather than <strong>in</strong> threat avoidance or situation monitor<strong>in</strong>g.<br />
However, these overall benefits arrive with the cost of <strong>in</strong>creased physical effort. These are important<br />
results for operators, fleets and technology managers deploy<strong>in</strong>g AIS.<br />
Interest<strong>in</strong>gly, different subject types reported significantly different results. For <strong>in</strong>stance, Masters<br />
reported that greater physical effort was required when us<strong>in</strong>g the AIS (survey p= 0.023, TLX<br />
p=0.208), as well as greater fatigue when us<strong>in</strong>g AIS (p= 0.058). Mates, <strong>in</strong> contrast, did not report the<br />
same results. They reported significantly improved voyage plan monitor<strong>in</strong>g (p=0.039), significantly<br />
reduced workload (p=0.180), and significantly less mental effort required (survey p=0.052, TLX p=<br />
0.371). F<strong>in</strong>ally, TMS Operators reported significantly improved threat avoidance (p=0.028),<br />
significantly reduced workload when us<strong>in</strong>g AIS (p=0.564), and significantly improved monitor<strong>in</strong>g<br />
vigilance (p= 0.047).<br />
6. Discussion<br />
This research provides <strong>in</strong>terest<strong>in</strong>g results with respect to early adoption impacts of new AIS<br />
technology <strong>in</strong>troduction. The research analyzes operator feedback of two technology treatments: pre-<br />
AIS and post-AIS. Seven hypotheses were def<strong>in</strong>ed to test various dependent variables about operator<br />
technology impacts. Data were collected <strong>in</strong> the form of a post-transit questionnaire and the NASA<br />
Task Load Index, a workload measurement <strong>in</strong>strument.<br />
The purpose of the study was to understand and identify the impact of AIS on operator performance <strong>in</strong><br />
the early stages of technology adoption—dur<strong>in</strong>g the first year of AIS <strong>in</strong>troduction. The results show<br />
that overall, operators perceived that AIS contributed by significantly improv<strong>in</strong>g voyage plan<br />
monitor<strong>in</strong>g, a critical facet of master, mate and TMS operators’ decision performance.<br />
12
Table VII: Results Summary<br />
H# Hypothesis F<strong>in</strong>d<strong>in</strong>gs Supported? P-value<br />
1 AIS-supported operators<br />
will show better decision<br />
performance than<br />
operators not us<strong>in</strong>g AIS<br />
Results <strong>in</strong>dicate a partial<br />
support to H1.<br />
Partially<br />
Situation Awareness<br />
Threat Avoidance<br />
Situation Monitor<strong>in</strong>g<br />
Overall, subjects report<br />
that AIS does not improve<br />
situation awareness<br />
Overall, subjects report<br />
that AIS does not improve<br />
threat avoidance; however,<br />
TMS operators <strong>in</strong>dicate<br />
that AIS improves threat<br />
avoidance<br />
AIS does not improve<br />
situation monitor<strong>in</strong>g<br />
No<br />
Partially<br />
No<br />
Overall<br />
p= 0.784<br />
Overall p=0.940<br />
TMS Operator<br />
p= 0.028<br />
Overall p=0.414<br />
Voyage Plan Monitor<strong>in</strong>g<br />
2 AIS-supported operators<br />
will consider more<br />
alternatives than operators<br />
not us<strong>in</strong>g AIS<br />
3a<br />
3b<br />
4a<br />
4b<br />
AIS-supported operators<br />
will report lower workload<br />
than operators not us<strong>in</strong>g<br />
AIS<br />
AIS-supported operators<br />
will report lower stress<br />
than operators not us<strong>in</strong>g<br />
AIS.<br />
AIS-supported operators<br />
will report greater<br />
confidence <strong>in</strong> the system<br />
than operators not us<strong>in</strong>g<br />
AIS<br />
AIS-supported operators<br />
will report greater<br />
satisfaction <strong>in</strong> the system<br />
than operators not us<strong>in</strong>g<br />
AIS<br />
Overall results <strong>in</strong>dicate<br />
that AIS improves Voyage<br />
Plan Monitor<strong>in</strong>g; Mates<br />
report significant<br />
improvement <strong>in</strong> voyage<br />
plan monitor<strong>in</strong>g<br />
Overall, the subjects’<br />
believe that use of AIS<br />
does not encourage<br />
consider<strong>in</strong>g more<br />
alternatives.<br />
Overall, subjects report no<br />
significant change <strong>in</strong> the<br />
workload. However, Mates<br />
and TMS operators report<br />
a significant reduction <strong>in</strong><br />
workload by us<strong>in</strong>g AIS.<br />
Overall, the results<br />
<strong>in</strong>dicate that AIS does not<br />
lower stress levels<br />
Overall, results <strong>in</strong>dicate<br />
that us<strong>in</strong>g AIS does not<br />
<strong>in</strong>duce greater confidence<br />
<strong>in</strong> system.<br />
Overall, results <strong>in</strong>dicate<br />
that us<strong>in</strong>g AIS does not<br />
enhance operator<br />
satisfaction <strong>in</strong> the system<br />
Yes<br />
No<br />
Partially<br />
No<br />
No<br />
No<br />
Overall p=0.038<br />
Mates<br />
p= 0.039<br />
Overall<br />
p=0.324<br />
Overall<br />
p= 0.04<br />
Mates<br />
p= 0.180<br />
TMS Operators<br />
p= 0.564<br />
Overall<br />
p=0.375<br />
Overall<br />
p=0.446<br />
Overall<br />
p=0.824<br />
13
H# Hypothesis F<strong>in</strong>d<strong>in</strong>gs Supported? P-value<br />
5 AIS-supported operators<br />
Partially<br />
will show decreased effort<br />
<strong>in</strong> us<strong>in</strong>g the system,<br />
compared to operators not<br />
us<strong>in</strong>g AIS<br />
6a<br />
6b<br />
7a<br />
7b<br />
7c<br />
AIS-supported operators<br />
will report greater<br />
monitor<strong>in</strong>g vigilance than<br />
operators not us<strong>in</strong>g AIS<br />
AIS-supported operators<br />
will report less fatigue<br />
than operators not us<strong>in</strong>g<br />
AIS<br />
AIS-supported operators<br />
will perceive greater<br />
decision support<br />
capabilities provided by<br />
the technology than<br />
operators not us<strong>in</strong>g AIS<br />
AIS-supported operators<br />
will perceive greater<br />
system benefits than<br />
operators not us<strong>in</strong>g AIS<br />
AIS-supported operators<br />
will report lower degrees<br />
of distraction provided by<br />
the technology than<br />
operators not us<strong>in</strong>g AIS<br />
Survey analysis shows that<br />
AIS significantly <strong>in</strong>creases<br />
physical effort <strong>in</strong> all<br />
subjects.<br />
- Masters report<br />
greater<br />
physical effort<br />
required.<br />
- Mates report<br />
lower mental<br />
effort<br />
required.<br />
TLX workload analysis<br />
shows significantly<br />
reduced workload for each<br />
subject type.<br />
- Overall results <strong>in</strong>dicate<br />
that us<strong>in</strong>g AIS improves<br />
monitor<strong>in</strong>g vigilance.<br />
- TMS Operators report<br />
significantly improved<br />
monitor<strong>in</strong>g vigilance.<br />
- Overall<br />
results<br />
<strong>in</strong>dicate that<br />
us<strong>in</strong>g AIS<br />
does not<br />
reduce fatigue<br />
level.<br />
- Masters report<br />
greater fatigue<br />
us<strong>in</strong>g AIS.<br />
Overall results<br />
significantly <strong>in</strong>dicate that<br />
subjects perceive AIS to<br />
provide greater decision<br />
support capabilities.<br />
Overall results <strong>in</strong>dicate<br />
that subjects do not<br />
perceive any significant<br />
system benefit by us<strong>in</strong>g<br />
the AIS.<br />
Overall results <strong>in</strong>dicate<br />
that AIS does not provide<br />
lower degree of distraction<br />
than the traditional bridge<br />
equipment.<br />
Yes<br />
No<br />
Yes<br />
No<br />
No<br />
Survey<br />
Overall p=0.014<br />
Masters<br />
p= 0.023<br />
Mates<br />
p= 0.052<br />
TLX<br />
Masters<br />
p= 0.208<br />
Mates<br />
p= 0.371<br />
TMS Operators<br />
p= 0.773<br />
Overall<br />
p=0.049<br />
TMS Operators<br />
p= 0.047<br />
Overall<br />
p= 0.340<br />
Masters<br />
p=0.058<br />
Overall<br />
p=0.038<br />
Overall<br />
p=0.712<br />
Overall<br />
p=0.169<br />
14
AIS was also found to significantly improve monitor<strong>in</strong>g vigilance for all operators, and for traffic<br />
management system (TMS) operators <strong>in</strong> particular. F<strong>in</strong>ally, AIS was perceived by operators to offer<br />
significant decision support capabilities, particularly <strong>in</strong> terms of better and more efficient decision<br />
mak<strong>in</strong>g, <strong>in</strong> plann<strong>in</strong>g the voyage, <strong>in</strong> situation monitor<strong>in</strong>g, and <strong>in</strong> reduced mental workload. Each of<br />
these benefits can contribute to improved operator performance.<br />
AIS impacts varied significantly by subject type, suggest<strong>in</strong>g that AIS technology impacted operators<br />
<strong>in</strong> different ways. Understand<strong>in</strong>g these differences is important for regulators, managers, operators,<br />
and those <strong>in</strong>terested <strong>in</strong> understand<strong>in</strong>g technology impacts <strong>in</strong> safety-critical systems. For <strong>in</strong>stance,<br />
TMS operators found that AIS significantly improved their threat avoidance ability, significantly<br />
reduced their workload, and significantly improved their monitor<strong>in</strong>g vigilance. TMS operators are<br />
responsible for monitor<strong>in</strong>g traffic <strong>in</strong> the channel as well as monitor<strong>in</strong>g the status of each vessel <strong>in</strong> the<br />
vic<strong>in</strong>ity. Improvement <strong>in</strong> TMS threat avoidance, monitor<strong>in</strong>g and workload capabilities are thus<br />
important technology impacts.<br />
Mates also reported significantly improved voyage plan monitor<strong>in</strong>g, and significantly reduced<br />
workload and mental effort when us<strong>in</strong>g the AIS. Thus, for operators with similar cognitive tasks,<br />
Grabowski and Sanborn (2001), AIS users reported similar benefits.<br />
The same could not be said for operators with more strategic cognitive task responsibilities; i.e.,<br />
vessel masters. Masters us<strong>in</strong>g AIS reported significantly more physical effort required, as well as<br />
greater fatigue. Thus, the overall AIS benefits of improved voyage plan monitor<strong>in</strong>g, improved<br />
monitor<strong>in</strong>g vigilance, and improved decision support came at the expense of <strong>in</strong>creased workload and<br />
effort for masters, but decreased workload for mates and TMS operators. These f<strong>in</strong>d<strong>in</strong>gs are<br />
consistent with earlier f<strong>in</strong>d<strong>in</strong>gs that identified differential technology impact on vessel operators with<br />
different cognitive task requirements, Grabowski and Sanborn (2001). The f<strong>in</strong>d<strong>in</strong>gs provide important<br />
<strong>in</strong>sights with respect to technology impacts <strong>in</strong> safety-critical systems.<br />
However, caution is advised with the use of these f<strong>in</strong>d<strong>in</strong>gs. First, the overall number of subjects (48)<br />
is low, which limits the statistical strength of these f<strong>in</strong>d<strong>in</strong>gs. Follow-on work is required to <strong>in</strong>crease<br />
the number of subjects <strong>in</strong> the study <strong>in</strong> order to strengthen the statistical validity of the results. In<br />
addition, the number of masters, mates and TMS operators is low, suggest<strong>in</strong>g that more of these types<br />
of subjects are required <strong>in</strong> follow-on work <strong>in</strong> order to strengthen the validity of the by-subject<br />
analyses. F<strong>in</strong>ally, no ship’s pilots were <strong>in</strong>volved <strong>in</strong> this first phase of AIS impact research. Follow-on<br />
studies to study pilot impressions of AIS technology impacts are an important next step.<br />
It is worth not<strong>in</strong>g that this evaluation is an early post-implementation study, conducted immediately<br />
after the <strong>in</strong>troduction of the AIS technology on the Seaway. It is anticipated that results from a later<br />
implementation study of the AIS technology might differ significantly from these early postimplementation<br />
results. This research provides f<strong>in</strong>d<strong>in</strong>gs from first movers <strong>in</strong> technology adoption <strong>in</strong> a<br />
large-scale system. It rema<strong>in</strong>s for future work to <strong>in</strong>vestigate the generalizability of first mover<br />
f<strong>in</strong>d<strong>in</strong>gs over the long term. Follow-on surveys for later post-AIS technology impact evaluation are<br />
required <strong>in</strong> order to develop steady state, mature technology <strong>in</strong>sertion f<strong>in</strong>d<strong>in</strong>gs.<br />
F<strong>in</strong>ally, these results reflect a relatively homogenous subject pool of well-educated, highly<br />
experienced, middle-aged English-speak<strong>in</strong>g males who were very familiar with the technology under<br />
study. Results with a less homogenous subject pool—with less well educated, less experienced, more<br />
<strong>in</strong>ternational, less technology-familiar and of vary<strong>in</strong>g ages—may provide <strong>in</strong>terest<strong>in</strong>g results and<br />
comparisons. The generalizability of the current f<strong>in</strong>d<strong>in</strong>gs across demographic, social, cultural and<br />
technology attributes can be established with such cross-comparisons, the topic of future work.<br />
The results of this prelim<strong>in</strong>ary assessment suggest that technology contributions and costs can differ<br />
widely among operators <strong>in</strong> the same safety-critical system. Managers, researchers, regulators and<br />
operators should therefore consider overall costs and benefits, as well as technology impacts on<br />
different types of operators, when deploy<strong>in</strong>g new technology <strong>in</strong> a safety-critical system.<br />
15
References<br />
BAINBRIDGE, L. (1983), Ironies of Automation, Automatica 19, pp.775-779<br />
CREECH, J.A.; RYAN, J.F. (2003), AIS: The Cornerstone of National Security?, J. Navigation (U.K.),<br />
56, pp.31-44.<br />
DARCE, K. (2004), Track<strong>in</strong>g River Traffic, The New Orleans Times-Picayune, 13 March 2004<br />
GRABOWSKI, M.; ROBERTS, K.H. (1996), Human and organizational error <strong>in</strong> large-scale<br />
systems, IEEE Transactions on Systems, Man, and Cybernetics 26, pp.2-16<br />
GRABOWSKI, M.; SANBORN, S.D. (2001), Evaluation of Embedded Intelligent Real-Time<br />
Systems, Decision Sciences, 32:1, pp.95-123.<br />
HART, S.; STAVELAND, L. (1998), Development of NASA-TLX (Task Load Index): Results of<br />
Empirical and Theoretical Research, <strong>in</strong> Hancock, P., Meshkati, N., Human Mental Workload. Elsevier<br />
Science Publishers B.V.<br />
NATIONAL RESEARCH COUNCIL (2003), Shipboard Automatic Identification System Displays:<br />
Meet<strong>in</strong>g the Needs of Mar<strong>in</strong>ers, National Academy Press<br />
http://books.nap.edu/html/SR273/SR273.pdf<br />
NORMAN, D.A. (1990), The “Problem” with Automation: Inappropriate Feedback and Interaction,<br />
not “Over-Automation”, Proc. Royal Society B237<br />
ROCHLIN, G. (1997), Trapped <strong>in</strong> the Net: The Unanticipated Consequences of Computerization,<br />
Pr<strong>in</strong>ceton University Press<br />
SARTER, N.; WOODS, D.D. (1995), How <strong>in</strong> the World Did We Ever Get Into That Mode?, Human<br />
Factors 36<br />
TENNER, E. (1996), Why Th<strong>in</strong>gs Bite Back: Technology and the Revenge of Un<strong>in</strong>tended<br />
Consequences, New York: Alfred A. Knopf<br />
WICKENS, C.D.; MAVOR, A.S.; McGEE, J.P. (editors) (1997), Flight to the Future: Human<br />
Factors <strong>in</strong> Air Traffic Control. Wash<strong>in</strong>gton, D.C.: National Academy Press<br />
WIENER, E. L. (1988), Cockpit Automation, In Human Factors <strong>in</strong> Aviation, E.L. Wiener & D.C.<br />
Nagel (editors). San Diego: Academic Press<br />
WIENER, E.L.; CURRY, R.E. (1980), Flight Deck Automation: Promises and Problems, Ergonomics<br />
23, pp.995-1011<br />
16
Robust Pareto – Optimal Rout<strong>in</strong>g of Ships<br />
utiliz<strong>in</strong>g Ensemble Weather Forecasts<br />
Jörn H<strong>in</strong>nenthal, Technical University Berl<strong>in</strong>, Berl<strong>in</strong>/Germany, h<strong>in</strong>nenthal@ism.tu-berl<strong>in</strong>.de<br />
Øyv<strong>in</strong>d Saetra, The Norwegian Meteorological Institute, Bl<strong>in</strong>dern/Norway, oyv<strong>in</strong>ds@met.no<br />
Abstract<br />
Sophisticated rout<strong>in</strong>g of ships establishes a multi-objective, non-l<strong>in</strong>ear and constra<strong>in</strong>ed optimization<br />
problem. The paper proposes an optimization approach to ship rout<strong>in</strong>g on the basis of hydrodynamic<br />
simulation and advanced weather forecast. Response amplitude operators are employed to describe a<br />
ship <strong>in</strong> waves. The probabilistic ensemble forecast is applied to account for the stochastic behavior of<br />
weather. An elaborated example is given for a conta<strong>in</strong>er service between Europe (Le Havre) and<br />
North America (New York). A severe weather situation for the North Atlantic is taken <strong>in</strong>to account.<br />
The robustness of the optimization result aga<strong>in</strong>st weather changes is assessed. A perturbation<br />
approach with respect to a parent route and a B-spl<strong>in</strong>e technique to describe the route <strong>in</strong>crease<br />
computational efficiency. The implementations uses a multi-objective genetic algorithm with<strong>in</strong> the<br />
commercial software modeFRONTIER.<br />
1. Introduction<br />
Ship monitor<strong>in</strong>g, rout<strong>in</strong>g assistance and decision support became important topics for save and<br />
reliable sea transport. Types of Route optimization regard<strong>in</strong>g weather and sea state can become an<br />
important tool to improve (i) the reliability of sea transport with<strong>in</strong> a transport cha<strong>in</strong>, (ii) comfort and<br />
safety of crew and passengers. In order to provide a mean<strong>in</strong>gful database for a route decision support,<br />
a route optimization tool has to take different oppos<strong>in</strong>g objectives <strong>in</strong>to account (e.g. estimated time of<br />
arrival and fuel consumption). Probable weather changes can affect optimized routes, abolish their<br />
advantages or even make them <strong>in</strong>feasible to sail. The approach presented here applies a stochastic,<br />
multi-objective optimization strategy on ship rout<strong>in</strong>g with regard to a determ<strong>in</strong>istic wave forecast.<br />
This serves to provide a database of advantageous Pareto optimum routes. Subsequently the<br />
robustness of the optimization result regard<strong>in</strong>g assumable weather changes is assessed. Therefore, an<br />
ensemble wave forecast generated by the ensemble prediction system of the European Centre for<br />
Medium-Range Weather Forecasts (ECMWF) is utilized. The optimization presented here belongs to<br />
a westbound Atlantic cross<strong>in</strong>g of the Hapag Lloyd panmax conta<strong>in</strong>er ship ‘CMS Hannover Express’.<br />
2. Determ<strong>in</strong>istic and ensemble wave forecast<br />
Determ<strong>in</strong>istic and ensemble forecast were provided by the ECMWF. They are wave forecasts of a<br />
severe sea condition <strong>in</strong> the North Atlantic from 20 th to 30 th of January 2002. Data for the significant<br />
wave height H 1/3 , the peak period T P and the wave angle β are given <strong>in</strong> time steps of 12 hours on a<br />
grid with 1.5° mesh size. While estimated time of arrival (ETA) and fuel consumption (FUEL)<br />
depend on all environmental conditions, accelerations and slamm<strong>in</strong>g are primarily <strong>in</strong>fluenced by<br />
swell. Swell - i.e., long-crested waves - be<strong>in</strong>g the crucial factor, it was decided to focus on this<br />
<strong>in</strong>fluence, leav<strong>in</strong>g out w<strong>in</strong>d-sea, current, w<strong>in</strong>d, drift-ice, shallow-water etc. for later and more ref<strong>in</strong>ed<br />
work.<br />
The determ<strong>in</strong>istic forecast represents one likely development of the sea state but no <strong>in</strong>formation about<br />
the probability of its occurrence. In contrast, the operational ensemble prediction system (EPS)<br />
generates a set of 50 forecasts. These ensemble members have an almost equal probability of<br />
occurrence. They are generated by superpos<strong>in</strong>g small perturbations to the operational analysis before<br />
launch<strong>in</strong>g the forecast calculation. The ensemble spread measures the “differences” between the<br />
ensemble members. It is related to the forecast uncerta<strong>in</strong>ty. Small spread <strong>in</strong>dicates low forecast<br />
uncerta<strong>in</strong>ty, and vice versa. Generally the uncerta<strong>in</strong>ty <strong>in</strong>creases while go<strong>in</strong>g further <strong>in</strong>to the forecast,<br />
ECMWF (2002). This characteristic of the ensemble forecast is used to assess the robustness of ship<br />
17
outes aga<strong>in</strong>st weather changes.<br />
For the seakeep<strong>in</strong>g calculation a cont<strong>in</strong>uous wave energy spectrum as a function of the wave<br />
frequency is needed. Therefore the data were associated with the description of a Bretschneider wave<br />
spectrum, Lewis (1998):<br />
2<br />
172.8 ⋅ H 1<br />
⎪⎧<br />
⎪⎫<br />
−5<br />
− 691.2<br />
3<br />
−4<br />
S<br />
ζ<br />
( ω)<br />
=<br />
⋅ω ⋅ exp ⎨ ⋅ ω<br />
4<br />
4 ⎬<br />
(1)<br />
T1<br />
⎪⎩ T1<br />
⎪⎭<br />
T1 = 0. 772 ⋅T p<br />
, transformation of peak to mean period, Journée (2001).<br />
3. Ship responses <strong>in</strong> waves<br />
Response amplitude operators (RAOs) and response functions for the added resistance due to waves<br />
(RFs) are employed to evaluate the ship performance <strong>in</strong> waves. Both have been calculated prior to the<br />
optimization by means of the strip theory code SEAWAY, Journée (2001), based on potential theory.<br />
Besides RAOs for the center of gravity motions for all six degrees of freedom and RAOs for the<br />
motions of selected po<strong>in</strong>ts also relative to the free surface, two different methods are available to<br />
determ<strong>in</strong>e the added resistance due to waves: The radiated energy method of Gerritsma and<br />
Beukelman and the <strong>in</strong>tegrated pressure method of Boese. Fig.1 shows RAOs of the motion on the<br />
bridge, RAOs of the relative motion at the bow and RFs of the added resistance due to waves as<br />
computed with SEAWAY. The plots serve to illustrate the <strong>in</strong>put to the response simulations with<strong>in</strong><br />
the route optimization.<br />
The ship responses and numerical results considered with<strong>in</strong> the optimization are:<br />
• RAOs of the translatory motion on the bridge to calculate the load<strong>in</strong>g on the crew by means of<br />
accelerations.<br />
• RAOs of the motion relative to the free surface (for a po<strong>in</strong>t on the keel l<strong>in</strong>e 10% beh<strong>in</strong>d the forward<br />
perpendicular) to assess the slamm<strong>in</strong>g probability.<br />
• RFs for the added resistance due to waves (preferably accord<strong>in</strong>g to the <strong>in</strong>tegrated pressure method<br />
which displays a smoother distribution over the wave frequency ω than its radiated energy<br />
counterpart).<br />
Fig.1: RAOs of motion on the bridge, RAOs of the relative motion at the bow and RFs of the added<br />
resistance due to waves<br />
The determ<strong>in</strong>ation of the calm water resistance is based on a regression analysis of model tests and<br />
full-scale data, follow<strong>in</strong>g Holtrop and Mennen (1984). The operat<strong>in</strong>g po<strong>in</strong>t of the propeller is<br />
calculated accord<strong>in</strong>g to the ITTC (1978) power prediction method, us<strong>in</strong>g the characteristic of a<br />
propeller with the same diameter, number of blades, similar pitch and blade area ratio as given <strong>in</strong><br />
18
Yasaki (1962). In a f<strong>in</strong>al step the fuel consumption and the compliance with the permitted operat<strong>in</strong>g<br />
condition is determ<strong>in</strong>ed <strong>in</strong> agreement with the project guide available from the producer of the ship's<br />
ma<strong>in</strong> eng<strong>in</strong>e, MAN B&W (2000).<br />
4. Sett<strong>in</strong>g up of the optimization task<br />
Automated optimization is the formal process of f<strong>in</strong>d<strong>in</strong>g a good (the best) solution from a set of<br />
feasible alternatives. It requires a complete mathematical problem formulation <strong>in</strong> terms of objective<br />
functions (what is to be improved), free variables (what shall be consciously changed) and constra<strong>in</strong>ts<br />
(what restricts the feasibility). The computation of the ship behavior for a given route and wave<br />
forecast is accomplished <strong>in</strong> MATLAB modules. For the optimization, these modules are embedded<br />
<strong>in</strong>to the generic optimization environment modeFRONTIER, www.esteco.it. A multi-objective genetic<br />
algorithm (MOGA) was applied to search optimum routes with<strong>in</strong> the sea condition described by the<br />
determ<strong>in</strong>istic forecast.<br />
4.1. Free Variables<br />
To provide free variables for both the spatial and the temporal description of the route, the course and<br />
the velocity profile are expressed as B-spl<strong>in</strong>es. An <strong>in</strong>itial course and velocity profile are given (parent<br />
route). Perturbations of the parent route <strong>in</strong> time and space are realized by superpos<strong>in</strong>g Greville-spaced<br />
shift spl<strong>in</strong>es to the parent spl<strong>in</strong>es. The vertices of the shift spl<strong>in</strong>es are controlled by shift parameters<br />
that are taken as the free variables of the optimization task. The velocity perturbation is performed as<br />
a reduction of the design speed of 23 kn. The spatial perturbation provides a shift of maximum 10% of<br />
the arc length of the orig<strong>in</strong>al parent route to either starboard or portside. Fig.2 presents example routes<br />
as realized with five parameters for the spatial shift. The optimization focuses on the open water part<br />
of the journey. Pilot time and estuary travel<strong>in</strong>g are deliberately left out of the <strong>in</strong>vestigation. The<br />
geographical feasibility is checked dur<strong>in</strong>g the optimization process. A restriction of shift parameters<br />
to avoid land collision, e.g. at Newfoundland, is possible, but not applied here. Investigations showed<br />
that at least 9 free variables to describe the velocity profile and 7 variables to describe the course are<br />
necessary for a sufficient temporal and spatial resolution of the route that correlates to the wave<br />
pattern as well as to the geographical conditions <strong>in</strong> the North Atlantic.<br />
4.2. Constra<strong>in</strong>ts<br />
Fig.2: Parent route, perturbation and <strong>in</strong>vestigated area<br />
Constra<strong>in</strong>ts are imposed by def<strong>in</strong><strong>in</strong>g limit<strong>in</strong>g values to the slamm<strong>in</strong>g probability (3%) and to lateral<br />
and vertical accelerations (0.2g). Furthermore, the operation po<strong>in</strong>t of the ma<strong>in</strong> eng<strong>in</strong>e has to comply<br />
with the eng<strong>in</strong>e characteristics.<br />
19
4.3. Objectives<br />
Two major objectives, the estimated time of arrival (ETA) and the fuel consumption (FUEL), are<br />
taken <strong>in</strong>to account, both of which have to be m<strong>in</strong>imized. Especially <strong>in</strong> rough weather conditions the<br />
velocity of <strong>in</strong>itial route variants has to be quiet low to avoid <strong>in</strong>feasibility due to active constra<strong>in</strong>ts. The<br />
follow<strong>in</strong>g objective functions are suited to accelerate the propagation of the <strong>in</strong>itial routes towards<br />
faster route variants with<strong>in</strong> the applied multi-objective optimization method:<br />
Objective<br />
ETA<br />
( ETA − set value ETA) 2<br />
= , (2)<br />
⎛<br />
ETA ⎞<br />
Objective FUEL = weight ⋅ ⎜ FUEL − FUEL ⋅<br />
⎟ . (3)<br />
⎝<br />
set value ETA ⎠<br />
The set value ETA is derived from optimum results at calm water conditions (ETA = 118 hours after<br />
departure, the value matches the estimated duration of a journey as stated <strong>in</strong> the sail<strong>in</strong>g lists of the<br />
ship operator) The weight assigned to the FUEL objective is 0.05, so as to produce objectives of an<br />
equal magnitude at the biggest deviants of the set value.<br />
4.4. Optimization with modeFRONTIER<br />
For the setup of the optimization task the generic optimization software modeFRONTIER uses a<br />
process flow chart, Fig.3. Initially free variables, used as <strong>in</strong>put parameter of the optimization loop, are<br />
collected <strong>in</strong> an <strong>in</strong>put file (parameter.txt). Afterwards, an application is started to build the new route<br />
by means of a perturbation of the parent route accord<strong>in</strong>g to the actual <strong>in</strong>put parameter (makeroute).<br />
Next, the route is evaluated (virtualShip) and results are given back to modeFRONTIER. F<strong>in</strong>ally, the<br />
optimization rout<strong>in</strong>e observes the constra<strong>in</strong>ts, calculates the objectives and, if necessary, launches a<br />
new <strong>in</strong>vestigation.<br />
2<br />
5. Optimization results<br />
Fig.3: modeFRONTIER, process flow chart<br />
Fig.4 shows a result of the MOGA. Each dot represents a feasible route that can be taken <strong>in</strong>to account<br />
for a route decision consider<strong>in</strong>g ETA and FUEL.<br />
20
Fig.4: Optimization results<br />
The optimization results build a database that yields <strong>in</strong>formation about weather conditions and ship<br />
behavior for many route variants. These can be filtered accord<strong>in</strong>g <strong>in</strong>dividual preferences.<br />
Consequently a more conscious decision can be made about which route to take. Sometimes this will<br />
yield fuel sav<strong>in</strong>gs for arriv<strong>in</strong>g at the desired time. Sometimes this will show which route is still the<br />
best (or least worse) for severe conditions. Typically, two l<strong>in</strong>es (1 + 2) border a solution space. L<strong>in</strong>e<br />
(1) is a Pareto frontier – the set of all solutions for which a s<strong>in</strong>gle objective cannot be further<br />
improved without deteriorat<strong>in</strong>g any other objective. All routes lie above and to the right of the Pareto<br />
frontier. Those variants that are closest to the frontier display low passage time for reasonably low<br />
fuel consumption. Apparently, it is impossible to decrease ETA below certa<strong>in</strong> limits without<br />
impair<strong>in</strong>g FUEL. L<strong>in</strong>e (2) is caused by the fact that longer routes require more fuel. This l<strong>in</strong>e cuts off<br />
the Pareto frontier. The ability to push this l<strong>in</strong>e to the left, towards faster route variants, strongly<br />
depends on the setup of the optimization and the applied optimization strategy (if not restricted by the<br />
capability of the ship). Area (3) marks the set of <strong>in</strong>itial routes, the start population of the genetic<br />
algorithm. They were randomly generated and have to be at a quiet low speed level to decrease the<br />
number of <strong>in</strong>feasible routes due to active constra<strong>in</strong>ts. Po<strong>in</strong>t (4) marks the arrival at calm sea<br />
conditions. Obviously, an arrival on schedule seems impossible with<strong>in</strong> this severe weather situation.<br />
In this case, the optimization result provides useful <strong>in</strong>formation about feasible routes and their<br />
properties. The probability of weather changes is neglected so far.<br />
6. Robust assessment<br />
After the optimization, the robustness of routes to weather changes was assessed consider<strong>in</strong>g 2000<br />
routes from the optimization result, with ETA between 130h and 150h, near the Pareto frontier. The<br />
characteristics of these routes were evaluated when launch<strong>in</strong>g them to the ensemble forecast<br />
members. The different sea conditions cause a spread <strong>in</strong> fuel consumption or make routes <strong>in</strong>feasible<br />
due to active constra<strong>in</strong>ts. Thus different approaches to assess the robustness of a route are th<strong>in</strong>kable:<br />
• Relate robustness to the number of ensembles where a route rema<strong>in</strong>s feasible<br />
• Relate robustness to a m<strong>in</strong>imum spread <strong>in</strong> the fuel consumption<br />
• Def<strong>in</strong>itions that take both aspects <strong>in</strong>to account<br />
• Include aspects that account for the gravity of constra<strong>in</strong>t violations….<br />
So far, the first aspect is taken <strong>in</strong>to account. Therefore we do not claim to establish parameters for the<br />
decision support, but to evaluate if the ensemble forecast can be used for a robust assessment with<strong>in</strong><br />
this optimization approach.<br />
21
Fig.5: Robust assessment (global view (left) and stretched zoom (right))<br />
Fig.5 shows results of the evaluation. The evaluation was conducted <strong>in</strong> steps of 1 day. For the 5 th day<br />
(120h) the evaluation of the whole routes were taken <strong>in</strong>to account, even if they take longer. The<br />
figures start at the end of day 1. Till then, all routes rema<strong>in</strong> feasible with<strong>in</strong> more than 90% of the<br />
ensemble members. Dur<strong>in</strong>g the second day the feasible route variants decrease drastically. E.g., only<br />
about 20 routes rema<strong>in</strong> feasible <strong>in</strong> more than 50% of the ensemble forecast members. The next days<br />
show a successive decrease so that f<strong>in</strong>ally only 10 routes are left that rema<strong>in</strong> feasible <strong>in</strong> 50 to 60% of<br />
the forecast members.<br />
Fig.6, optimized route, ETA = 143,5h, fuel consumption 608t<br />
Fig.6 shows the fastest variant of these routes (ETA 143.5h, FUEL 608t). Furthermore, the wave data<br />
of the determ<strong>in</strong>istic forecast are depicted. All wave fields <strong>in</strong> the figure propagate from west to east.<br />
24 hours after the departure, the ship reaches a strong wave field (upper left picture). Approach<strong>in</strong>g to<br />
this wave field dur<strong>in</strong>g the second day, this route became <strong>in</strong>feasible at about 32% of the ensemble<br />
members. Faster route variants, that took course closer to this wave field, mostly went <strong>in</strong>feasible. A<br />
second wave field, passed northerly, has only m<strong>in</strong>or affects to the route (upper right picture). Here,<br />
the reduced velocity is not necessary with<strong>in</strong> the determ<strong>in</strong>istic forecast, but enables the route to stay<br />
22
feasible with<strong>in</strong> other ensemble members. Shortly before reach<strong>in</strong>g Newfoundland, the velocity has to<br />
be reduced to pass a third wave field (lower left picture). From day 3 to day 5 additional 9% of the<br />
ensemble forecasts rank this route as <strong>in</strong>feasible. F<strong>in</strong>ally, the route has been assessed as feasible by<br />
59% of the forecast members. Still, the uncerta<strong>in</strong>ty of the weather forecast, expressed by the ensemble<br />
spread, is not explicitly considered. In other words, it is not specified if the ensemble caused<br />
<strong>in</strong>feasibility due to probable weather changes or if the uncerta<strong>in</strong>ty of the forecast was the reason for<br />
that. Nevertheless, the characteristics of the graphs <strong>in</strong> Fig.5 appear plausible. Therefore, the <strong>in</strong>clusion<br />
of the ensemble prediction weather forecast is a promis<strong>in</strong>g approach to assess the robustness of ship<br />
routes aga<strong>in</strong>st weather changes with<strong>in</strong> a route optimization.<br />
7. Summary and conclusions<br />
Initially the paper provides a brief <strong>in</strong>troduction to the applied wave forecast and the mathematical<br />
model of a ship <strong>in</strong> waves used with<strong>in</strong> the presented route optimization. A method for the spatial and<br />
temporal generation of route variants based on B-spl<strong>in</strong>es was presented. The major advantage of<br />
parametrically modeled route- and velocity profiles lies <strong>in</strong> the reduced number of free variables. This<br />
enables the application of a multi objective, stochastic algorithm with a reasonable time exposure.<br />
Here, a multi objective genetic algorithm was applied to optimize a ship route with regard to a<br />
determ<strong>in</strong>istic wave forecast (Other strategies, e.g. evolution strategies, might also be considered but<br />
were beyond the scope of this study). The decision mak<strong>in</strong>g of the navigat<strong>in</strong>g officer is supported by<br />
build<strong>in</strong>g up a significant amount of <strong>in</strong>formation from which the best compromise (Pareto optimum)<br />
can be found <strong>in</strong> accordance with <strong>in</strong>dividual preferences. Subsequent to the optimization, an ensemble<br />
forecast was used to assess the robustness of selected routes aga<strong>in</strong>st weather changes. For the time<br />
be<strong>in</strong>g, the number of ensemble forecasts at which a route rema<strong>in</strong>ed feasible measured robustness. For<br />
the future, robustness of a route should be <strong>in</strong>cluded as a further objective to the optimization, to<br />
<strong>in</strong>crease the percentage of feasible route variants. Strategies have to be employed how to handle the<br />
uncerta<strong>in</strong>ties of a weather forecast, e.g. by optimiz<strong>in</strong>g once a day regard<strong>in</strong>g the course from the<br />
current waypo<strong>in</strong>t to the port of dest<strong>in</strong>ation. A ratable relation between robustness and benefit is to be<br />
established to provide mean<strong>in</strong>gful results for a decision support.<br />
References<br />
ECMWF (2002), User guide to ECMWF products, http://www.ecmwf.<strong>in</strong>t/products/forecasts<br />
HOFFSCHILDT, M.; BIDLOT, J-R.; HANSEN, B.; JANSSEN, P.A.E.M. (1999), Potential benefit of<br />
ensemble forecast for ship rout<strong>in</strong>g, ECMWF Techn. Memo. 287, Internal Report, ECMWF,<br />
Read<strong>in</strong>g/UK<br />
HOLTROP, J.; MENNEN, G.G.J. (1984), An approximate power prediction method, International<br />
Shipbuild<strong>in</strong>g Progress, Vol. 31<br />
ITTC (1978), Performance prediction method for s<strong>in</strong>gle screw ships, 15 th Int. Tow<strong>in</strong>g Tank Conf.,<br />
The Hague<br />
JOURNEÉ, J.M.J. (2001), User and Theoretical Manual of SEAWAY, Rel. 4.19, TU Delft<br />
MAN B&W (2000), K90MC MK6 Project Guide Two-stroke Eng<strong>in</strong>es, 5 th Edition, www.manbw.dk<br />
LEWIS, E.V. (Ed.) (1998), Pr<strong>in</strong>ciples of Naval Architecture - Motions <strong>in</strong> Waves and Controllability,<br />
SNAME, Jersey City<br />
YASAKI, A. (1962), Design diagrams for modern four, five, six and seven-bladed propellers<br />
developed <strong>in</strong> Japan, 4 th Naval Hydrodynamics Symp.<br />
23
! " <br />
# <br />
<br />
<br />
<br />
! <br />
<br />
!! !! "<br />
# $ #% &! # <br />
$ #%%& ' !! $ ! ! <br />
! $$ $& $ $$ <br />
$& $ <br />
$$ ! $!!$!! #! <br />
<br />
! $! $ $ !<br />
()*+ ! $ $ !<br />
$, &! ! - - $<br />
! # $ - ! $ ! $ <br />
!! $ $ ! $ <br />
! ($ +<br />
<br />
# .-# & $ )* $ ## ## $ <br />
#! ##<br />
' ( #%%) $ / $ <br />
) * # 0 # ()*0 + ' $ <br />
# ' $ ! <br />
! 1 ! $ 2 3 $<br />
# ! $!<br />
4 !! $ <br />
5 ( # <br />
$$ +<br />
<br />
, $!#- - <br />
# $#-$ !! !!!<br />
$! $# $ !$ <br />
$!! !- <br />
# $#-<br />
<br />
$!! 67+ ! $ !<br />
( # ! -# ##+ + !! $ <br />
!3+ $ !! !
# ! ! $ # $ <br />
!! <br />
<br />
$!! $#8 # # <br />
! ! &!# 0 <br />
! 8 3 $ ! $ !! <br />
8 # ! ! $ $ !<br />
8 1! $!! !<br />
, 8 9 ! $#-<br />
<br />
"#$ <br />
<br />
8 $#-67+$ !! <br />
+! !3+ <br />
<br />
" %<br />
<br />
!! !! "<br />
# $ #% &! # <br />
$ #%%&8! $ <br />
# !) $ <br />
! # $$ !!<br />
<br />
# # ! !!: ! (6+<br />
!! !! $ # # !! <br />
!! # ! !!<br />
# $$ !! !! <br />
$ $ !!<br />
*#++)6 $ ' $<br />
!(6' + !2 # !! <br />
!! #! <br />
<br />
$! !!6! $ #%%&<br />
0 ! !!!! <br />
# !! ! # #<br />
!!! ! ! $# ! <br />
- & $ & ! , ! - <br />
! ; !!$!!!<br />
<br />
$ !! ! <br />
$ # $! $. $$<br />
! !! $ ! 8 ! ! <br />
$$ ! ' !<br />
!! <br />
<br />
0 $ ! 6 <br />
! $ #%%& ! $ $ ! $ <br />
! (! + # / 0 <br />
$ $ # <br />
<br />
! # ! $ ! !!<br />
# !! ! $ $ <br />
!5! ! !! !< <br />
$ ! , <br />
$$ #$& $!5 ! $ 6! <br />
- ;#- !! <br />
! $ ! - <br />
<br />
"" <br />
<br />
0 $ ! !! <br />
$$ # # !! ( ! < ! <br />
+ ! ' 8 $ $<br />
$ $$! (! + $$ <br />
" # #! ! <br />
! 0 # !!( +$ !(+<br />
<br />
, $ ! !!# !!!! $ <br />
#%%& !! $ ! !#<br />
7+ $ # - +# ! - <br />
!! - ! ! <br />
# - ! ! $ # - !$ <br />
!#! ## & $ )*0 <br />
( &! #+# # ##$! ##=! <br />
## ! ! $ ! , &! /. # - 1 <br />
##! ! $= ., ! <br />
! !! $ ! %- % $ <br />
# - ! ! ! - > ! <br />
! <br />
$$!!! ! $ $<br />
<br />
<br />
? ! $ $ ) ,<br />
#%%- ! $ !< $ <br />
(?.=*+# !! ( !!! <br />
*8+$ !< $$$ <br />
$$ <br />
<br />
, $ !!# ! ! # <br />
!$) ( +.#%%&.#%%) <br />
# # ! !$ $ <br />
( +$ ! ) <br />
!!& 1 $ # /1 $ <br />
! (!!& /1+#(/1+ <br />
$ !# " ! !<br />
$ !! ! <br />
# ! # ! $<br />
! !$ ! $ <br />
$ ! ! $! ! #<br />
! $$ $ # . !<br />
! ! ! $ ! ! <br />
$ $$ ! #<br />
! - ! ! $ & <br />
! $!! ! . <br />
<br />
* ##!! - ! <br />
9
6 ! #- # !!<br />
$ # ! ! #- # ! #<br />
(& ' ! ! + ! $<br />
! ! $ # $#-# <br />
$!! !$ $ <br />
! 6 <br />
$( +$ $ ! 2 <br />
$ # ! ' ! 2 $ <br />
! ! - ! $& <br />
! 2$ #- !<br />
<br />
<br />
"& <br />
<br />
!. # $ $<br />
# - ' ( #%%) $<br />
$$ # # <br />
= !! ! <br />
$ *.> *.?! $ ! !& ' <br />
(#%%)= !! !!<br />
$#!. &!<br />
<br />
, ! !! <br />
' ( #%%) # $ $ )<br />
* # 0 # ()*0 + / <br />
&! # ! ! $ )*0 <br />
<br />
$$ 6 7+ +<br />
3+ # #<br />
$6+! @- !-$# &!+<br />
! &! @ $ <br />
&! 4 ! ! $ <br />
$ $ #- !! $ <br />
$ <br />
$ (6! $+$ !! $ - <br />
$!#$ <br />
! ! $ <br />
! !8 $ <br />
<br />
<br />
# $ ! , ' $ <br />
# ' $ ! <br />
># ' ! !& <br />
! 8 1! $ 23$# <br />
! $! 5 <br />
# <br />
( +, ! $# $& (<br />
+ ! $ ! ( !+ <br />
! $$ &! $ !<br />
! ! ! ' ( ! ! +# $<br />
<br />
&<br />
<br />
8 7+ !$ !(! ++<br />
$$! $ !()*+<br />
/
& '(<br />
<br />
; !# &!$!<br />
<br />
? !! 1 ( ! +<br />
!$$' <br />
- ! !! !$ <br />
' ' 7 # ' <br />
3 ' <br />
0 #$ !! <br />
! ' # <br />
' ! 7 !! <br />
' <br />
<br />
&! !$ $ $ # ! - ! ! " $ # <br />
$ ! $# -<br />
<br />
B 8 $ ! ! $ <br />
!! 8 1<br />
B 0 ! ! $ !<br />
$! $ ! <br />
! ># # $ $ ! !!<br />
# !# ## &! <br />
B # ! $ $ !2 <br />
' ! 7 0 $! ! !$<br />
B<br />
!! 5$ #$ &!!$ ' !# <br />
, $ $ <br />
? ' # !$!( # <br />
$!+ ! ! # <br />
. $ <br />
<br />
&")<br />
<br />
0 $ !!! &!$ <br />
! $ 8 ! $ #- <br />
)*()*+ & ! !<br />
)*#! -# ##()**C0 +<br />
<br />
&" *+<br />
<br />
, )**C0 $ ! ! / <br />
$ ( 6!+ ( 6' +0 $ / <br />
( ! + # $ #<br />
! # ( ! + ! &<br />
<br />
&"",-<br />
<br />
; (D(0+ !$ 0$E7 $# <br />
! $ $# 1 FE7! <br />
!4 !$ !! $<br />
<br />
7 ,## ' !# ! - !> ! !<br />
$ ' ! ' ! $ ' * <br />
$ <br />
A
!-!$ !0 $& $ <br />
<br />
6;$)*(*C+0 <br />
Variable Type Denotes<br />
1. 0 set set of nodes (customers)<br />
2. set set of arcs (roads)<br />
3. 2 set set of vehicles<br />
4. $ <strong>in</strong>teger total available capacity<br />
5. <strong>in</strong>teger service time at node <br />
6. 1 <strong>in</strong>teger travel time from node to node 1<br />
7. 1 <strong>in</strong>teger cost of us<strong>in</strong>g #1<br />
8. <strong>in</strong>teger demand of node <br />
9. time arrival time of vehicle <br />
10. time departure time of vehicle <br />
11. " tr<strong>in</strong>ary availability function of node <br />
12. time starttime servic<strong>in</strong>g of node by vehicle <br />
13. 1 b<strong>in</strong>ary #1is used by vehicle <br />
<br />
<br />
### )*! ! 7@<br />
A7@73 $ $$ )*!6<br />
<br />
<br />
<br />
1<br />
1<br />
(7+<br />
∈ 2 ( <br />
1+∈<br />
<br />
&"& <br />
<br />
)* & # ##5!-## ! $<br />
## ! $ (! ! +<br />
! $ (! $ +!<br />
# $ $$ ##$ <br />
! -## ##$ $-## <br />
! )*0 !<br />
<br />
0 ##$ ! $ " <br />
6<br />
→ I−7F7<br />
H<br />
# ( " <br />
( + = 7+ ( " <br />
( + = −7+ $ -#<br />
# " <br />
( + = F 0 $-#$$<br />
∀ ∈ 6 "<br />
(<br />
+ ≠ F 0 $ $ $ ##<br />
#! !!<br />
<br />
; ## $ ! # ## $ <br />
$ )* # ## ()*0 + $ #%% , - # <br />
## # F E7 J F F K - DJ E7 E7 K - DJ4 - ; - K # 4 - ; - <br />
! !$ !( !+ !<br />
! !( !+! <br />
<br />
&&)<br />
<br />
; 340 L IF E7H " ! $ $ )*0 $ <br />
#%%# $ !$# 3 0 . #-<br />
<br />
># # $ ## $ $ ##<br />
$$$# ##$ ( $#%%+<br />
3 4'(+@(9+(G+@(77+ - $$#%%0 $$ <br />
<br />
G
$# #! -# ##! $ $<br />
1 #! $(7+<br />
<br />
<br />
<br />
?(+ $ & 2(3+ <br />
!2(+ <br />
2(1+ - !2(9+<br />
$ ' - 2<br />
(/+ # ( <br />
( " <br />
( + = 7+ -#( " <br />
( + = F ++2 '(A+ <br />
$ ( # ! $ !+2<br />
' (G+ ! 2 $ (7F+ (77+ <br />
<br />
<br />
&.<br />
<br />
& $ #- $# ! $ <br />
$ ! $ ! !!& 8<br />
! ! (!<br />
! + & ! ! !! ( <br />
& + <br />
<br />
5 !! $ # ! ! <br />
! 5#%% &! # $#!$$ <br />
# $ $!!! $<br />
$ !<br />
$ 2# # <br />
<br />
5 #%% # ! $ <br />
!
$! $ ! ! ! <br />
- $ <br />
<br />
5!?5* # -$! $$<br />
# $ !$ # $ $!<br />
$# !!( +!$ <br />
<br />
. '+<br />
<br />
, ! $ !! ($ )**C0 +# !! <br />
5 . #%%) 8 <br />
! $ # $ ! $ ! 0 ! # <br />
! $ ! 8 $ $<br />
! !$ ! $ <br />
<br />
. '+ <br />
<br />
! 8 79$##-<br />
$)*0 !0 &! $ <br />
<br />
<br />
'#%%%! $)*0 #.! <br />
!!6 7+ (M+ . <br />
$2+ ! !!9$$<br />
$ ! ! <br />
# # $ <br />
! - $ # <br />
! ( ' $ + # <br />
!# $ $ )*0 !<br />
<br />
7#%% $ #)* <br />
(M+ M $ <br />
M & $M $ !<br />
! $ $ 2$ &! $#<br />
J3/_A91__G7F_K# ! ( +- . <br />
M#- #!! # $ $ <br />
# # <br />
# $ $$ !! $ , <br />
# & #-. (-D7+7 #++8 <br />
M!2$ <br />
! !<br />
<br />
9 ( #+++ ! # (48+ $ )*0 "<br />
$ $$<br />
# # ! $ ! "<br />
! $ ! )*0 # ! < $ <br />
$ $ <br />
$ ! &! <br />
48! $ ' # $)*0 ># # 48<br />
$ ($$$ ! - +<br />
<br />
<br />
# $ <br />
$! <br />
37
."<br />
<br />
!$$& ' $ # <br />
!$& ! $ ( ! # -+$ $$ ! .<br />
#++- ! - $#- <br />
. #++- ! # $ <br />
' 0 $ (# # # <br />
' # ! $+<br />
<br />
,!# & $!! <br />
& 1 & . ! <br />
# $N=;$ $$<br />
# ! >=;<br />
$ !<br />
<br />
.&!<br />
<br />
0 ! &! ! <br />
<br />
.& <br />
<br />
!$7/ $/!! #3! (7+<br />
/ $ ! $? $#$ -$ <br />
!$) 7A@ !!& 7F! <br />
!! / ! !#- <br />
<br />
<br />
.&" <br />
<br />
5 ! ' 0 ! $ <br />
! ( + $ $<br />
! <br />
<br />
-4! $ $ # ! !<br />
! $ <br />
' 7+!:+' :3+ $<br />
$& <br />
<br />
/0 ! # ! 6 0 .<br />
! # #!$# #!( OD7+=<br />
( PDFF+ (PDFF7+<br />
<br />
! !!$1FF# <br />
<br />
#!! $ 71FFF <br />
<br />
)0# ! 8 3 $ $<br />
$ $ ! ( + # <br />
! 8 33<br />
<br />
% # !! # !! $ <br />
<br />
1 !6 -Q-$7339&<br />
<br />
3
$$& $ #! # <br />
# $ $ # .<br />
#!!<br />
<br />
.&&'(<br />
<br />
0 # &! &! # / ! (# 3<br />
+/ (! + &! 7# GA<br />
2 &! # 7/ " &! # <br />
=:FF*?#7M"$)=: M;&, &! <br />
?* $G1R $FR<br />
<br />
<br />
.&.- <br />
<br />
,764&! !5 =;.$ +<br />
&!$ !# $N=;$ <br />
! # # ' $ ! ( <br />
+ &! $(!+ <br />
<br />
&! ! !$ &! <br />
- # ! $,$ # <br />
! $$4&! 7 !<br />
$ 3 2 &! $ $ G 0 ! <br />
33
! &! ! !<br />
$# $8 # <br />
# 0 <br />
, !# <br />
# ($! + <br />
! $.# ># !( <br />
&+ ! $$ <br />
<br />
<br />
1 <br />
<br />
!!! ! .<br />
#%%& . #%%) ? , # &! $ <br />
! 0 ! ! # ! 6 (+ <br />
$! (+- ' !& !( $ +<br />
! &!6 !+# #! 2 !3+<br />
(1+#$ # !' ! 2 !+#<br />
#! 8 !7+ !! #! <br />
2 !9+ ! !<br />
<br />
<br />
,6: #$ !<br />
$#! 6!! (85+' <br />
! (S5+!! (8+<br />
<br />
0 !$ &! $ # !<br />
!,, !! ( !+ ! # ! <br />
$ ! ! &! 8 <br />
!$ &!, ' ! ( !+ ' ! <br />
' #! !<br />
$$ ! , # ' ! ! $<br />
!# $<br />
<br />
!!! !! !( !719,+<br />
! ! ! ( + # N=; $ <br />
! ! $ ! <br />
! ! >=; ( ,7+ ' $ <br />
$$ !<br />
<br />
3
2<br />
<br />
!! 67+ !( #! -# <br />
##++!! $!3+ <br />
!! ! !#$!<br />
! ! # 7+ ## $ ! <br />
! -# + ! ! $ ! <br />
!! # ## !! !!<br />
!<br />
<br />
## !.!! ! $ # <br />
& - ! # # $#-,$ <br />
# .! (! ! # & + . ! ( !.<br />
! ! # & + , $ #- # ! $ <br />
$$ $!!<br />
# !# !!# $ ! <br />
&( $! +<br />
<br />
$ <br />
<br />
#- -M)884=2,M4)>5;C2)54:(FF+.. <br />
!8 3A(7+!!7.7A<br />
<br />
?5):4T,2:48;4)8M2:48)584)8T285;5=5=285=8,(FF7+<br />
0*: *:( + !8<br />
!! = !!71/.7G3<br />
<br />
4"4 (FFF+ 5 6 <br />
;C ") ( + ! <br />
$4?!8! !!73.1A<br />
<br />
4"428=>T(FF3+; 5$8! <br />
<br />
,M4)>5;C(FF1+*.<br />
!6 ### - $) <br />
<br />
>5="4)M4) T2 M4>)M > (7GGG+ <br />
,5)3/!!G/.37A<br />
<br />
=?>:5*2)48T2,*4)4),2?584(FF+06<br />
0 ; ( + M 4 ?! ?$<br />
(M4??5.FF+=C$!!9GF.9G9<br />
<br />
31
=?>4;;=(7GGA+; ( =*<br />
<br />
)54:(7GG3+. 4! T5! ) /7(3+!!31.<br />
333<br />
<br />
8?>4), (7GG1+ ) ! ?8.)G19/ ? <br />
0 - $ <br />
<br />
8?> =2 C4)5* =2 ;44)8 =2 =4;8 =2 =;;4) (FF+ <br />
6 ); =;8( +79 <br />
4! ?$$ (4?FF+58*!!/11./1G<br />
<br />
8?>U;4")M =2 =;;4) 2 =4;8 =2 :4 )C "2 C4)5* =2<br />
;44)8 =2 )4) T2 8?> = (FF3+ 5 !<br />
? -: $<br />
39
!"#$ <br />
% % <br />
& '<br />
" '<br />
(<br />
% "# %<br />
( ! <br />
" # ! <br />
& )* * + "<br />
'"<br />
' % & %<br />
% "<br />
<br />
<br />
<br />
! "# $" % # #<br />
&' ##&()<br />
* # # + <br />
+ *, ! # <br />
#+ ! -# + &'<br />
&( . ! #<br />
<br />
<br />
<br />
/0,&! #&'&( )<br />
<br />
##&'&( <br />
! !1 <br />
+ # ! <br />
+ &'&(##*<br />
+&'&(#+<br />
# . "# +)#+<br />
#!
*+ <br />
# ## # ## <br />
# ) 3(4 * 4'5 <br />
! ! 6 # <br />
<br />
<br />
7 # ( $7(% # !<br />
1 # # <br />
!&'&( "#<br />
! ! #!#64'5<br />
<br />
! + <br />
) ## #<br />
# ! + <br />
<br />
<br />
<br />
! ) #&'<br />
&( # # ! ' #<br />
&'&() # <br />
! #* # <br />
# <br />
<br />
<br />
/8,&'&( <br />
<br />
<br />
# # <br />
• #+ # <br />
&'&(<br />
• + <br />
# ! &'&( ! <br />
<br />
<br />
<br />
&&'&( ! #!#<br />
# 7 + 9&# #<br />
## + #<br />
) # # <br />
#+) ! <br />
$ % +)<br />
# ! <br />
2
,&( !#+9&<br />
# <br />
<br />
! # ! 6 <br />
# . <br />
!#' # !<br />
/ # ! !+ # +<br />
# # <br />
#<br />
<br />
&. # ! # <br />
) ##! &'&(<br />
! #<br />
#" * #<br />
! +,<br />
• ! <br />
• ! ##+.! <br />
#+)<br />
• ! ! <br />
• # ) $ + %<br />
<br />
!<br />
" # <br />
<br />
#&'&( # #<br />
7 # ( $7(% 7( # #<br />
! # ! <br />
"! #6<br />
# ! # * # + !<br />
75&'&(<br />
<br />
; *####<br />
• "-0
@,() ##!<br />
<br />
<br />
+ )/!+###<br />
.! !++ * *<br />
7 # ( $7(% #+ ! <br />
!+)##!#&'&(<br />
<br />
7( #"-0
&'C&( * D# <br />
D ! #!1<br />
<br />
<br />
<br />
/>,& #&'&(<br />
<br />
6 4'5 "# ! ! &' &(<br />
! ! ##" + <br />
# # * 6<br />
#<br />
<br />
"# # ! # * # ##<br />
###) # !<br />
! ## !<br />
<br />
## . # &'E&(,<br />
• ##&' <br />
• ! #&'&(#<br />
• #9&# ##<br />
• ## * #+<br />
• #<br />
<br />
F # * # #. # <br />
&'E&(,<br />
• <br />
##1$ ++<br />
#+# ! % )#1<br />
!# . $ <br />
#. + #%<br />
• "$ <br />
! # #1 ! <br />
#.+!<br />
• % &' <br />
= #+## GH<br />
G+)#+H+)#+ # +<br />
# * <br />
+ # * <br />
" # + # # <br />
##+++ # !<br />
@0
* , #+ <br />
## +# # F<br />
<br />
#+ #<br />
! # C #!+C+! <br />
&' &( # + <br />
7 #( #1<br />
<br />
!! <br />
<br />
&' &( GH /A<br />
& G H +<br />
# # + +<br />
)## <br />
<br />
"# &' 5 F <br />
# +) +# # <br />
5"I#<br />
<br />
<br />
<br />
'<br />
5<br />
<br />
'<br />
5<br />
9 <br />
& "<br />
<br />
'<br />
5<br />
& <br />
<br />
A<br />
D<br />
P<br />
T<br />
7<br />
&'<br />
<br />
'<br />
5<br />
<br />
'<br />
5<br />
75<br />
3<br />
<br />
/A,& &'&(<br />
<br />
"# &( <br />
* ## 7(<br />
# ###<br />
<br />
&' # <br />
"#&'# &'<br />
&' # & <br />
6 + # !<br />
# &' # +) <br />
6#&'& " ?&<br />
<br />
G9! = "4 H/6 ! #5"I#<br />
&'<br />
<br />
@8
,&'C"#5 &<br />
<br />
&( # # <br />
5 # ! <br />
! ! ##<br />
&( #! &(## <br />
&(#6#&(<br />
<br />
) * <br />
<br />
#, #+ ) ## # <br />
F) ! &' <br />
#, ! # 6 <br />
#, ) # <br />
+ 9&/ <br />
# +)#+ + <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
5 <br />
-'F<br />
54'F<br />
D<br />
J' )<br />
D<br />
"#<br />
! <br />
'F<br />
…<br />
….<br />
( <br />
<br />
D<br />
5<br />
5<br />
.0<br />
/2,9 &<br />
9 <br />
<br />
8<br />
<br />
@
#, ! +&' # <br />
#+,<br />
• <br />
• I# + $*<br />
# ! % $ 6 + <br />
! % # # #+<br />
<br />
• +)<br />
• <br />
• )<br />
• <br />
• # #+<br />
• !# #<br />
<br />
6 7(#!# <br />
# ! # &' &(<br />
#&'&( <br />
+ <br />
<br />
) +<br />
<br />
+ # # <br />
5 * . # # <br />
,<br />
• <br />
• <br />
• # !#<br />
• )#<br />
<br />
# # # ## * # # <br />
# ' + <br />
)# * <br />
# <br />
#"*## <br />
<br />
) * , -<br />
<br />
5# + # )<br />
# # -# )! <br />
! ) <br />
G H ###<br />
56 + #+ ,<br />
• ) <br />
• .<br />
• 6<br />
• ### <br />
• # <br />
• <br />
• $ %<br />
• ) <br />
• <br />
@@
) .<br />
<br />
#&(#<br />
<br />
+ # ! <br />
! <br />
<br />
+ ) # <br />
# <br />
) # # GH # # ) <br />
+ <br />
3(4## ## #) <br />
<br />
# "# <br />
## +5 <br />
# 3(4 # 7 <br />
# )<br />
<br />
9&# ) 6 <br />
# ## <br />
<br />
/ !& <br />
<br />
& # &'#&(<br />
# ) "! <br />
#*+ &'&(#+ # <br />
. " #+ ! +<br />
<br />
<br />
"# # # & + #<br />
+ I ## + + <br />
##<br />
! ),<br />
• 5 ) + # # <br />
• " # + 1 # <br />
• ' # + . <br />
• "! # # + . + ) <br />
• &! #<br />
• ) # # <br />
<br />
• +)#+ !+ <br />
• "#" <br />
/ + 0 ( &<br />
<br />
#,## + #1 <br />
+) #+) -' ##<br />
+) ! #+).<br />
# + ! # J<br />
= )5' #)<br />
+ ##.<br />
<br />
5# # 6 +)#+##,<br />
<br />
@>
• &# <br />
• = <br />
• F <br />
• #<br />
<br />
6#! # ! + <br />
5 ! # ! # # )<br />
= ! ! * #) ! <br />
! !+ ) <br />
<br />
/ !& .1 ! 2&<br />
<br />
5 . * ! # ) <br />
# # # ,<br />
• = . $! %<br />
• &# <br />
• . <br />
<br />
) ! + * ) ; <br />
4 # + # <br />
) # <br />
*<br />
<br />
/ # 5+ <br />
) # ) ; 4 <br />
• # <br />
• <br />
• <br />
• # <br />
<br />
) + #) <br />
+ #, % <br />
! " ! # $<br />
)%+ <br />
<br />
3 + 0 <br />
<br />
#, ! *#,"<br />
# "# ! #<br />
• & #+<br />
• 5 # <br />
• 5 # <br />
• 7 #<br />
<br />
3 " '& <br />
<br />
/ # !<br />
+)# + #<br />
## #+ + <br />
#+ # # #,<br />
"# ! ' ! # #<br />
6 # ! # ! <br />
* + + ! <br />
@A
# # 6 ## #* <br />
# )++<br />
<br />
<br />
<br />
( <br />
<br />
-'F<br />
+ 0 <br />
( &<br />
+ 0<br />
<br />
% <br />
<br />
# 5<br />
6<br />
<br />
<br />
54'F<br />
<br />
<br />
57'F 5<br />
5 <br />
/<br />
'B<br />
+ 0<br />
<br />
$<br />
+ 0<br />
<br />
<br />
K.<br />
6<br />
4<br />
<br />
<br />
<br />
9 <br />
<br />
.<br />
<br />
0<br />
.nc2<br />
/:, D5<br />
3 <br />
<br />
' ! ## # (# 51<br />
( '## +)#+! ! <br />
! # &<br />
#! # #,<br />
# # *!<br />
. # !# <br />
+)#+ # * # * # <br />
!<br />
3 #<br />
<br />
##,## *<br />
#! 3(45"$&;% #*<br />
## + #### <br />
# # #<br />
<br />
<br />
7 <br />
<br />
! # # <br />
! 6 # <br />
! # ! <br />
<br />
! # ,<br />
• 9 +? *+<br />
# # <br />
&'&( ++<br />
#*<br />
<br />
@
• + #+ # ##<br />
+ # ! <br />
#<br />
• &' &' # <br />
#! ! #<br />
# &( + <br />
! &(+ ! # <br />
• + + ##<br />
++ #+<br />
• + ! 1 ##<br />
<br />
• "# &( <br />
! #!+# <br />
#<br />
<br />
#<br />
F477B("447FB7L-= 5BL-& $8
¨!#"$&%(')*¥*+£¦¨,¥!#%-.§¨¤/"<br />
¢¡¤£¦¥§¥§¨¤¥©¨¤¥©¨¨<br />
02143657198;:77@#ACBED7FC5HGID7JK17@MLN=6BO3QPOR)S93T=T=6DVUXWZY¦[]\&^`_a7bVc¤dOef\&^OcagbVcdVW<br />
U B‚D7P`U }~K€<br />
Ĥ‰v‘ŽT’“†ŒŽ”ˆ •–‰—ˆ,‡X‡gŠ˜’$ˆXo„›šo’OŠœŽ6„4‰ †ŒŽ”’Oˆ$•r‰N‡gžŸÔZZ…6–• ’š›¤’`7Ž”Ô…6+‰vŠ¡†,„?…|‡g†v¢<br />
ƒC„?…Q†‡4ˆ,‡V‰oŠ‹‡gŠ‰Œ†o‰o7Ž6†<br />
Î<br />
h$ì j d`k$lmkoǹ W+lm^`bZk`WVprqmlslm^OtvuEw*tbZxZxzy|{Vk<br />
ŠŒ…6¤§ˆ`Š,…6ž¦¥E§…6)Ž6„4‰¨Š¤†vŽ©‡gžŸÔZ7…Q–• †vŽf‰ª‡«Ž6„4‰¡¬¤‰ogO…6–•—¤®’O¤¯‰v7Ž6†¡Ôg›†¤„4‰¤ÔŠ°šo’OŠ˜,‰v†;šv’OŠ±ˆ9•X…Q²`‰v Ž”’`Ž”Ôž<br />
£<br />
‰°†ŒŹ¬µo‰¤$ŽzŽ”’Œ‰vŠvŽ”Ô…Q·¤’`V†ŒŽTŠ˜Ô…67Ž6†v¢¯ƒC„–…s†©…Q†©X’`+‰&M…6Ž6„§Ž6„4‰Ç¹ £ Ôžº•?’OŠŒ…6Ž6„–¤<br />
¤’`7Ž”Ô…6+‰vŠ¦žŸ’$ˆX§ˆ`Š˜‰¤…Q7…Q¤…m³<br />
¸‰¼o´7‰o7ŽT…Tˆ`ž&¹½ŹˆXOŠ˜ˆ`ŽT… £ Š’•XŠˆ`¤¾¤…6–•v¿œ’OŠŽ6„4‰¯À¾‰vžŸ?‰vŠŒÁ”›‰ˆ? Ôž •–’OŠŒ…6Ž6„?¤¾§;Ä?…v„œŠ‰Œ†Œ´4žÄŽ6†…Qň±Æ‰o…Ç•`„–Ž<br />
»<br />
’O²`‰vŠŽ6„4‰†¤„?…ȇ¢É¸+‰¤¤’O+4ž¦¥E§#…69Ž6„4‰†o‰¤¤’O+ɇgžŸÔZ7…Q–•¯†vŽf‰ª‡œŽ6„4‰žŸ’$ˆX`…Q–•·’˜šÉŽ6„V‰°,’O7Ž”Ô…6+‰vŠ,†…s†<br />
O…Q†ŒŽTŠŒ…¬o´VŽT…T’O<br />
ÊE…Qr•®…67Ž”’¯ˆ?¤¤’`ŕgŽ#Ž6„4‰’,‡+ŽT…6¤§ˆ`žzƉv…Ÿ•O„–ŽÆO…s†vŽTŠ,…”¬v´VŽT…’O2Ä–…To„Ëʆ<br />
Š‰ÔžÄ…Ÿ³‚‰·¬v¥¾„4‰o´4Š,…Q†ŒŽT…†vŽ”’Oˆ$•r‰ŠŒ´4žŸ‰v†ÉŽ”<br />
,’Og,‰oŠ,7…Q–•<br />
¤ˆ`žÇ$ŕžŸˆ`Žf‰¤2…6Ž6„4‰¨#Š¤†vŽÃ‡gžŸÔZ7…Q–•2†vŽf‰ª‡¢—ƒC„?…s†Ë‰oV†,´4Š‰Œ†®Ž6„rˆ`Ž*Ž6„4‰§†¤„?…ȇ“…s†®’,‡+ŽT…6¤®ÔžQžÄ¥±žÇ’$ˆ??‰<br />
¬’XŽ6„2Ž6„4‰°†¤„?…ȇ½Ì†;žÇ’`–•`…TŽTŹO…Q+Ôžz†vŽTŠ˜‰o–•XŽ6„‡4ˆ`ŠÔ¤¯‰$Žf‰vŠ,†§Ô+ËŽ6„V‰žÇ’•X…s†vŽT…T¯Ê†”‡V‰¤$Ž6†o¢Í‰Œ†”‡+…6Žf‰¾Ž6„4‰žÄ…6¤¾…TŽ”ˆ`ŽT…’OV†<br />
?‰vž¦§Æ…6Z…TŽT…TÔžCŠ˜‰v†,´4žÄŽ6†ÔŠ˜‰¦‡+Š’O¤…Q†,…6–•?¢<br />
’š°Ž6„V‰¤§’<br />
Ï ¥É+% ;Ð®Ñ ¨&¥<br />
jXÓ t^`q Ó koc¤ÔIxZ^ i k*t j ak&l j dXqmÔ˜tqmÕo^`lmlmn¾ÖIk$lml+Ô˜t j ÖIk${ jXÓ ^¾Õ jXÓ t^`q Ó koc©Ôx7qŸ×zWX×7^Oc¤tqmÕobZlm^Oc¤lmn¾t j ^ ì j qs{‹ck$Ô˜t j Ö&Ôoy<br />
Ò<br />
j ck`Ẁ txZkIl jXÓ dXqŸtbZ{Zq Ó ^`l`Ô˜t˜ck Ó d`tx¾×7^Oc¤^`_®kot˜koc¤ÔÃxZ^ i k©t j akÆt^Op`k Ó q Ó t j Õ jXÓ Ôqm{Vkoc¤^Otq jXÓ yÙck i q j bZÔ<br />
ØZbVc¤txVkoc¤_<br />
j ^`Õ¤xVk$ÔIt j Ô j l i ktx7qmÔ#×Zc j a7lŸk$_ÚÕ j _®×7lŸkot˜k$lŸnarn®Õ j _§×7bVt˜koc¤ÔI{Zqm{ ÓVj tÆÛTbZlŸÜglmlrtxVk Ó bZ_§koc j bZÔ#×Zc,^`ÕvtqmÕo^`l<br />
^O×Z×Zc<br />
Ò j _§×gbVt˜kocÔbV×7× j c¤t½^`lmck$^`{VnkvÝ4qmÔtÔ©Û j c&txVkd`c¤^O×7x7qmÕ{ZqmÔ˜×glm^$ǹ WVtxVk;Õ,xVk$Õp4bVס^ Ó {·txVkt˜c,^ Ó ÔqŸt j Û<br />
qmÔÔ¤bVk$Ôoy<br />
Ó7Ó q Ó d¦c¤k$ÔbZlŸtÔoyÞ&Õvtb7^`lmlŸǹ Ẁ txVk½Ô˜t j Ö½^Od`k¦×7ls^ Ó qmtÔ˜k$lŸÛZxZ^`Ôt j akd`k Ó koc,^Ot˜k${§a–n^&×koc¤Ô jXÓ‹ß ^ÉÔ˜t j ÖI^Od`k<br />
txVk½×7lm^<br />
Ó7Ó kocŒà,yMY¦xVkÉÔqŸáok j ÛKÕ jXÓ t^`q Ó koc¦Ôx7qŸ×7Ô#kvÝ4×Zck$Ô¤Ô˜k${‹q ÓËÓZj _®q Ó ^`l+Ôl j t½Õo^O×7^`ÕoqŸtªn ß Y¦âÆ[ÉàÆq Ó Õvck$^`Ô˜k${‹q Ó txVk<br />
×7lm^<br />
Ó { x7^`Ô;c¤qmÔk Ó {Vc¤^`_¯^OtqmÕo^`lmlŸn¡q Ó txZkËlm^`Ô˜t°ÛQkoÖ(ǹ k$^Oc¤Ôoy2Ùck$Ô˜k Ó tlŸǹ WÃ× j Ô˜te”×7^ Ó _®^EÝre<br />
×7^`Ô˜ttxVckok{Vk$Õo^`{Vk$Ô¾^<br />
i kÔl j t*Õo^O×7^`ÕoqŸtqŸk$Ô&bV×2t j‹ãOä`ä`ä Yâ[WgÔl j t&Õo^O×7^`ÕoqŸtqŸk$Ô*bV×Nt j¡å$æOä`ä`ä Yâ[ç^Ock°{7qmÔÕobZÔÔk${ yIØ j c<br />
ÔxZqm×7Ô©xZ^<br />
jXÓ Ô°txVk¯d`k Ó koc,^Otq jXÓœj Û©^NÔ˜t j ÖI^Od`k‹×7ls^ Ó a–n ^NxrbZ_®^ Ó ak$q Ó dNÔ˜kok$_®Ô;t j ak$Õ j _§k¯_ j ckË^ Ó {<br />
txVk$Ô˜k¯ck$^`Ô<br />
j ck{ZqŸè¯ÕobZlmtoy _<br />
k°×7ck$Ô˜k Ó txVkock°^ Ó koÖê^O×Z×7c j ^`Õ¤x2t j _ j {Vk$l txVkÔ˜t j Ö½^Od`k×glm^ ÓZÓ q Ó d j ÛM^®Õ jXÓ t^`q Ó kocÉÔxZqm×NÕ j _§×7lŸkot˜k$lŸn<br />
é<br />
j _®×7bVt˜koct^Oprq Ó d›q Ó t j Õ jXÓ Ôqm{Zkoc¤^Otq jXÓ a j txÅtxVk2ÔxZqm×zë|Ô®l jXÓ dXqmtbZ{Zq Ó ^`l½Ô˜t˜c¤k Ó d`tx“×7^Oc,^`_§kot˜koc¤Ô^ Ó {<br />
arnÕ<br />
l j dXqmÔtqmÕo^`l ^`Ô˜×k$ÕvtÔoy<br />
ì í ;Ч!¡,¨¤¥©<br />
kËÕ jXÓ Ôqm{Vkoc jXÓ k¯Õ jXÓ t^`q Ó kocÔx7qŸ×œ^ Ó { jXÓ kËckodXq jXÓ WKk`yºdVyâÆbVc j ×k j c°Þ*Ôqm^ j ctxVk¯[î4Þ(ÖqŸtx Ô˜k i koc,^`l<br />
é<br />
j ctÔ¦Ö&xVkock°txZk¾Ôx7qŸ×2Õo^`lmlmÔ$y©ï,yºk`yq ÓNj bZc*_ j {Vk$lWgtxVkÔt j Ö½^Od`k®qmÔ¦×7ls^ ÓZÓ k${·Û j c*^Ô˜×k$ÕoqŸÜ7Õ°ÔxZqŸ×2^ Ó {2Û j c<br />
×<br />
kckodXq jXÓ ð Y¦xVk;Ôx7qŸ×·Õ j _§k$Ô©Û6c j _ñckodXq jXÓ2ò ^ Ó {·^Occ¤q i k$Ô©q Ó txVkÜZc¤ÔtI× j ct j ÛÃckodXq jXÓ·ó WZÕo^`lmlmÔ©^`lsltxVk<br />
jXÓ<br />
j ctÔ j Û+ckodXq jXÓ§ó ^ Ó {§txVk Ó lŸk$^ i k$Ô#txZqsÔÃckodXq jXÓ ^ Ó {§Ô^`qslmÔÃt j ckodXq jXÓ®ô y#õÉbVc#_ j {Vk$l4^`lml j ÖÔÃt j d`k Ó koc,^Ot˜k<br />
×<br />
j ÖI^Od`k×7lm^ Ó k`yºdVyZÛ j c^`lml+txVk× j ctÔ j ÛCc¤kodXq jXÓNó t^Op4q Ó d®q Ó t j ^`ÕoÕ j b Ó t¦txVk;Õo^Ocd j ^`lŸck$^`{Vn jXÓ e”a j ^Oc¤{<br />
^§Ô˜t<br />
Ó Õo^`lmlsq Ó d^OtIÜ7c¤Ô˜t½× j ct j ÛckodXq jXÓ‹ó yY¦xVkÉÔ˜t j ÖI^Od`kÛ j c©ckodXq jXÓó qmÔI×7lm^ Ó7Ó k${Ëq Ó tªÖ j Ôt˜ko×7Ô ð ÞöÜ7c¤Ô˜t<br />
ÖxVk<br />
j ×Ztqs_®qŸáok$ÔItxVk;l jXÓ dXqmtbZ{Zq Ó ^`lgÔ˜t˜ck Ó d`tx‹×g^Oc¤^`_§kot˜koc¤Ô j ÛKtxVk;ÔxZqŸ×CyÞêÔ˜k$Õ jXÓ {·Ô˜t˜ko×N^`ÔÔ¤qŸd Ó Ô½d`c j bV× j Û<br />
Ô˜t˜ko×<br />
jXÓ t^`q Ó koc¤ÔÆt j Õo^Oc¤d j x j lm{ZÔoW?Õ jXÓ Ôqm{Vkoc¤q Ó d;× j ctÔ j ÛK{Vk$Ô˜tq Ó ^Otq jXÓ ^ Ó {ËÔ˜t˜ck Ó d`txlmqm_®qmtÔoyÞ÷_¾bZlŸtqm×7lmqmÕoqŸtªn j Û<br />
Õ<br />
i ^ Ó Õvk ß k`yºdVy;txZktxZqmc¤{2{Zqs_§k Ó Ôq jXÓ2j Û#txVk§ÔxZqm× ß xVk$qŸdXx–t,à,W txVkÕvc¤^ Ó kvefÔ˜×glmqŸtoW<br />
^`Ô˜×k$ÕvtÔÉÖqŸtx2×Zc¤^`ÕvtqmÕo^`lck$lmk<br />
Ó {{Z^ Ó d`koc j bZÔÆd jrj {ZÔ$Wrø ä`ù efÕ jXÓ t^`q Ó koc,ÔoWZÔ˜×k$Õoqm^`l7Õ jXÓ t^`q Ó koc,ÔoWVxZ^OtÕ¤x·Õ j‚i koc¤ÔoWVtxVk i qsÔqŸa7qmlsqŸtfn¾lmq Ó k‚à<br />
ckokoÛ6koc¤Ô½^<br />
xZ^ i k ÓVj t©ǹ kot&a+kok Ó t^Op`k Ó q Ó t j ^`ÕoÕ j b Ó toy½î j ÛT^Oc$W jXÓ lmnÔ˜t^ Ó {Z^Oc¤{ æOä`ù efÕ jXÓ t^`q Ó koc¤Ô*^Ock°Õ jXÓ Ô¤qm{Vkock${ y<br />
ø ã
¡ ¨¤% "Z ¢¡¤£¦¥§¥§¨¤¥© !/<br />
t Ó k$Õvk$ÔÔ^Oc¤qmlmn°krbZqm{ZqmÔt^ Ó t,àÔ˜k$Õvtq jXÓ Ô:¤ ¦ ÖqŸtx/¨ å;¢ ^`ÕoÕ j c¤{7q Ó dt j txVk&l jXÓ dXqŸtbZ{7q Ó ^`l–× j Ô¤qŸtq jXÓ Ô<br />
ßTÓVj<br />
¦ÆÖqŸtx?¨ ä@;¢ W
M<br />
j Ö*à,WZÖIk°xZ^ i k ô N£¦¥¨§© ä y<br />
a<br />
j ck`W7txVkÛ j lsl j Öq Ó d§Õ jXÓ Ô˜t^ Ó tÔ^Ock°{ZkoÜ Ó k${‹Û j c¦k$^`Õ¤x±Ô˜×k$ÕoqŸÜ7Õ j ×Ztqm_¯qŸá$^Otq jXÓ ×Zc j a7lŸk$_·y<br />
ØZbVc¤txVkoc¤_<br />
D N¤©£:¦ Åæ Y x j ls{ZÔoy<br />
ä<br />
YxVk;_®^EÝVqm_b7_ñl j ^`{Zq Ó d§ÖIk$qŸdXx–t j ÛÃk$^`Õ¤x±Ôl j tN ¤©£:¦ q Ó t jXÓ ÔEY¦â[WZÖxVkoc¤k<br />
M<br />
Y¦xVk;Ô¤xZqŸ×zë|Ô½l jXÓ dXqmtbZ{Zq Ó ^`l+Õvk Ó t˜koc j Ûd`c¤^ i qŸtªn
N ¦J 4<br />
N<br />
y"!¾RK1±=6:7D$#)3 € PX:&%('Ã=61?U‚1r=Q>)#Ã3 € U BE3 ~ JzU‚1*#¢:gF¢U‚RK1 € R3+' ð<br />
å<br />
4<br />
<br />
<br />
<br />
¦<br />
<br />
y769: € 1rP`U‚36:+F(3 € :98+1–BO=6:7D,#1*#;:¡Y¦xVk‹ak Ó {Zq Ó d±_ j _§k Ó t§^ Ó {HÔxVk$^Oc§Û j c¤ÕvkNq Ó k$^`Õ¤xÅÔ˜k$Õvtq jXÓ ¤¦ W<br />
5<br />
å;¢ _¾bZÔ˜t ÓZj t¦kvÝ4Õvkok${2×koc¤_®qmÔÔ¤qŸa7lŸk*lmqm_®qŸtÔ ð<br />
¨<br />
€ R3+')3 € 3TF¢1*=zJ3T=T3 ~ BO36J4%>:½YxVk;ÔxZqŸ×Cë|Ô½ÖÆk$qŸdXx–t^ Ó {·qmtÔ©a7b j nX^ Ó Õvn·xZ^ i kt j _®^OtÕ¤x ð<br />
øVy"!¾RK1<br />
S ß TÉà N<br />
qŸtx·tx7qmÔ©Õ jXÓ {ZqŸtq jXÓ ^ Ó {£¨ ¢Åß ÔxZqŸ×zë|ÔIa j ÖÉàIÛ j lsl j ÖÔ½Û j c¦txVkÔxVk$^Oc¦Û j c¤Õvk°q Óœßå à 𠤢¡ ä y<br />
é<br />
y"!¾RK1 € RK3+' € D73T= € :gF)1?8+1rF 5g1–14=@:&Y¦xZk°ÔxZqŸ×Cë|Ô©l jXÓ dXqŸtbZ{7q Ó ^`l Õvk Ó t˜koc j ÛMd`c¤^ i qmtfn2^ Ó {2qŸtÔl jXÓ dXqŸtb4e<br />
Y<br />
å;¢ ^OckÉ_®q Ó qm_®qŸáok${ÔbVaKJk$ÕvtÆt j ^`lmlZtxVk&Õ jXÓ Ôt˜c¤^`q Ó tÔdXq i k Ó ^Oa j i k`yØ j cItx7qmÔ#_®q Ó qm_®qŸá$^Otq jXÓ Ẁ txVk<br />
¨<br />
j c¤qŸtxZ_ ß îrk–bZk Ó tqm^`l ÉbZ^`{Vc¤^OtqsÕ*Ùc j d`c¤^`_®_¯q Ó d–à j c¦txVk k$lm{Zkoc˜eNk$^`{2^`lŸd j c¤qŸtxZ_ ÖÆkockbZÔ˜k${ y<br />
î©ÉÙÅ^`lŸd<br />
Y æ<br />
X ¦ j Û<br />
B # .%$'&)(¦$ .51$¡ )*¢ ¤£¦¥¨§(+.0© )0&@( .3*¦* (+* *:$"* I2 (>*)5<br />
j i qm{Vk${ËtxVk*Ô¤xZqŸ×‹Ô^`qmlmÔ jXÓ k i k Ó p`kok$l ß Ö&qŸtx j bVtÆt˜c¤qm_Ëà#txVkÉÔxVk$^Oc½Û j c,Õvk ¤¡ ¦ ^Ot½txVk*× j q Ó t < = ¦ qmÔdXq i k Ó<br />
Ù#c<br />
arn<br />
¤¢¡I¦) 4<br />
N K 4<br />
S ±h?å ¨˜u <br />
ßå à<br />
Ó {‹txVkak Ó {7q Ó d_ j _§k Ó t ô NP¦Ã^Ot¦txVk× j q Ó t
j c,qŸtxZ_ñ^`lŸÖ½^$n4Ô¦Û j b Ó {·^®Ô j lmbVtq jXÓ y<br />
&k$lm{Vkoc˜e2k$^`{2^`lŸd<br />
j ×Ztqm_¯qŸá$^Otq jXÓ ×Zc j a7lmk$_®ÔÛ j ca j tx±txVktfÖ j ^`lŸd j c¤qŸtxZ_¯Ô*^Oc¤k¾Û j c¤_¾bZlm^Ot˜k${±ak$l j Ö°y¥£ j b Ó {Z^Ocn9Õ jXÓ e<br />
Y¦xVk<br />
jXÓ å k Ó Ô¤bVck$Ô*txZ^OttxVk®^`lŸd j c,qŸtxZ_ { j k$Ô ÓVj tÉ×Zc j {ZbZÕvk Ó kodX^Otq i k®ÖIk$qŸdXx–tÔoy¦£ j b Ó {Z^Ocn2Õ jXÓ {7qŸtq jXÓ¨§<br />
{ZqŸtq<br />
Ó ÔbZck$Ô*txZ^OtÉtxVk j ×7tqm_®qŸáok${±ÖIk$qŸdXx?tÉÛ j cÉtxVk§Ô˜k$Õvtq jXÓ ¤¦ W§Ü å¢ qsÔ ÓVj t_ j c¤k¾tx7^ Ó¡æOä© ^Oa j‚i k<br />
k<br />
N X ¦ W+¨ å¢ Û j c&txZk°ÖIk$qŸdXx?t j ÛMtxZqmÔ¦Ôk$Õvtq jXÓ y¦Y¦xZqsÔ¦Õ jXÓ {ZqŸtq jXÓ qmÔck$lŸkve<br />
^ Ó {" *) G<br />
^Ock^Oca7qŸt˜c¤^Oc,qmlŸn®Õ¤x j Ô˜k Ó ×7^Oc,^`_§kot˜koc¤ÔoW<br />
ÖxVkoc¤k¥<br />
J ~,+ 1rP`U®U : : €<br />
æ y<br />
6<br />
X ¦3 ä<br />
<br />
0$<br />
4 2#"<br />
<br />
jXÓ Û j ctxVkËb7Ô^Od`k j Û©txZk$Ô˜kËtªÖ j {Zq¡ kock Ó t°^`lŸd j c¤qmtxZ_®Ô°qsÔ;kvÝ4×7lm^`q Ó k${›q Ó txVkËÛ j lsl j Öq Ó dVy·Y¦xVk<br />
Y¦xVk¯ck$^`Ô<br />
j c¤qŸtxZ_ Ó kok${7Ô^{7q¡ +koc¤k Ó tqm^OaglŸk j aKJk$Õvtq i k§ÛTb Ó Õvtq jXÓ ^ Ó { {Zq¡ kock Ó tqs^Oa7lŸk¾a j b Ó {Z^OcnNÕ jXÓ {7qŸtq jXÓ<br />
î©ÉÙ÷^`lmd<br />
Ó Õvtq jXÓ Ô$yY¦xVk°^OagÔ j lmbVt˜k i ^`lmbVk°Û6b Ó Õvtq jXÓ qmÔ ÓVj t*{Zq¡ kock Ó tqm^Oa7lŸk`yYxVk;a j b Ó {Z^Oc¤nËÕ jXÓ {Zqmtq jXÓ ë|Y¦xVk;Ô¤xZqŸ×<br />
ÛTb<br />
jXÓ k i k Ó p`kok$l”ëZxZ^`{t j a+kq Ó t˜kod`c¤^Ot˜k${·q Ó txVk j a J˜k$Õvtq i kÛTb Ó Õvtq jXÓ q Ó txVk &k$lm{Vkoc˜e2k$^`{·^`lŸd j c¤qmtxZ_·W<br />
Ô^`qmlsÔ<br />
j b Ó {§txZ^OtÆqmt#ÖI^`ÔIqm_§× j ÔÔ¤qŸa7lŸk½t j p`koko׋qŸtI^`Ô#a j b Ó {Z^Oc¤n¾Õ jXÓ {7qŸtq jXÓ yÃØ j cItxVk$Ô˜k*ck$^`Ô jXÓ ÔtxVk<br />
ak$Õo^`bZÔ˜k&ÖÆk&Û<br />
×Ztqm_¯qŸá$^Otq jXÓ ×7c j a7lŸk$_®ÔCl j–j p{Zq¢ +kock Ó toyÃY¦xVk^`{ i ^ Ó t^Od`k j Û4txVk©î©ÉÙ±^`lŸd j c¤qŸtx7_ j i kocÃtxVk k$ls{Vkoc˜e2k$^`{<br />
j<br />
j c¤qmtxZ_ qmÔ¦txVk°× j ÔÔqma7qmlmqŸtªn j ÛMÛ j c¤_b7lm^Otq Ó d§a j b Ó {7^OcnÕ jXÓ {ZqŸtq jXÓ Ô$yÆï Ó txVkEk$ls{Vkoc˜e2k$^`{2^`lŸd j c¤qmtxZ_<br />
^`lŸd<br />
ÓVj t©× j ÔÔqŸa7lmk`yY¦xZka j b Ó {Z^Ocn¯Õ jXÓ {ZqŸtq jXÓ Ô½xZ^ i k°t j akÛ j c¤_b7lm^Ot˜k${·q Ó {Zqmck$ÕvtlŸǹ W4txZ^Ot¦_§k$^ Ó Ô¦n j b<br />
txZqmÔ©qsÔ<br />
i kt j Ö¦c¤qŸt˜kÉ^ ÓVj txVkoc¦×7c j d`c¤^`_-dXq i q Ó dÕvkoct^`q Ó ×7^Oc,^`_§kot˜koc¤Ô½t j txVk &k$lm{Vkoc˜e2k$^`{·^`lŸd j c¤qŸtx7_ñÔ j txZ^Ot<br />
xZ^<br />
j _§k j Û©txVka j b Ó {Z^Ocn Õ jXÓ {7qŸtq jXÓ Ô;^Oc¤k_§kotoyœî j _§k j txVkoc¤ÔÕo^ ÓZÓVj t¾ak¯t^Op`k Ó q Ó t j ^`ÕoÕ j b Ó t§q Ó txZqmÔ<br />
Ô<br />
yY¦xZkockoÛ j c¤k¦ÖÆkÜZc¤Ô˜tlmk$^ i ktxZk$Ô˜k&Õ jXÓ {7qŸtq jXÓ Ô j bVtoW4Ô˜t^OctÆtxVk&k$lm{Vkoc˜e2k$^`{¯^`lmd j c¤qŸtxZ_ ^ Ó {§ÖqŸtx¯qŸtÔ<br />
Ö½^$ǹ<br />
ÓVj txVkoc;×7c j d`c¤^`_ Öx7qmÕ¤x _ j {ZqŸÜ7k$Ô&txZk®Ô j lmbVtq jXÓ t j _§kokot°^`lmlÃa j b Ó {7^Ocn2Õ jXÓ {Zqmtq jXÓ Ôoy<br />
ck$Ôb7lŸtÉÖÆk®Ô˜t^Oc¤t°^<br />
i ^ Ó t^Od`k j Û¦txVk2î©Ùç^`lŸd j c¤qŸtxZ_qmÔ¾txZ^Ot§qŸt{ j k$Ô ÓVj t§^`lŸÖ½^$n4Ô§Õ jXÓ?i kocd`k`y¤£Ik$Õo^`bZÔ˜k j Û&txVk<br />
Y¦xVk‹{ZqmÔ¤^`{<br />
j i k_§k Ó tq jXÓ k${2^`{ i ^ Ó t^Od`k j ÛtxZkî©ÉÙ)^`lŸd j c¤qmtxZ_ j i kocÉtxVkEk$ls{Vkoc˜e2k$^`{9^`lŸd j c,qŸtxZ_ ÖIk°Ü7c¤Ô˜t¦t˜cn<br />
^Oa<br />
j Ü Ó {œ^NÔ j lsbVtq jXÓ Öqmtx txVk‹î+ÉÙ«^`lŸd j c,qŸtxZ_·yËïªÛ½txVkockqmÔ ÓVj Õ jXÓ?i kocd`k${Ô j lsbVtq jXÓ ^OÛ6t˜koc¾^2Ô˜×k$ÕoqŸÜZk${<br />
t<br />
j ×Ztqm_®qmá$^Otq jXÓ ÖqŸtx±txVkk$ls{Vkoc˜e2k$^`{œ^`lŸd j c¤qŸtx7_·yØ j c j bVc;ÔÕvk Ó ^Oc¤q j W txVk<br />
tqm_§k§lmqs_®qŸtÖIk§ck$Ô˜t^Oc¤ttxVk<br />
txVkck$Õ j _®_§k Ó {Zk${ i ^`lmbVk .<br />
^ Ó t&Û j c*txVk°Ô˜k$Õ jXÓ {2×7ls^ ÓZÓ q Ó d§Ô˜t˜ko× ð é qŸtx j bVt&txZqmÔ¦Õ jXÓ {ZqŸtq jXÓ txZk j ×Ztqm_¯qŸá$^Otq jXÓ _¯qŸdXx?tdXq i k°ÖIk$qŸdXx–tÔ<br />
i<br />
Ó Ô j _§kÔ˜k$Õvtq jXÓ ÔÉÕol j Ôk¾t j txVk_®^EÝVqm_¾bZ_ ×koc¤_¯qmÔÔqŸaglŸkÉÖÆk$qŸdXx–tÔÉÛ j c&txVk$ÔkÔk$Õvtq jXÓ ÔoyÉï ÛMtxVkock§^Ock ÓVj t<br />
q<br />
ÓVj bVdXx xZk$^ i n±Õ jXÓ t^`q Ó koc¤Ô;t j Ô˜t j Ö;WKqŸtqmÔqm_§× j ÔÔqma7lŸk;t j ck$^`lsqŸáoktxVk j ×Ztqm_®qmáok${9ÖIk$qŸdXx–t;{ZqmÔ˜t˜c,qŸa7bVtq jXÓ y<br />
k<br />
j b Ó {Z^Ocn§Õ jXÓ {7qŸtq jXÓ k Ó Ô¤bVck$ÔÆtxZ^OtI^ Ó kvÝVÕvkok${Z^ Ó Õvk j Û æOä© j ÛztxZk&ck$Õ j _¯_§k Ó {Vk${ i ^`lsbVk&Û j cItxVk<br />
Y¦xVk&a<br />
j ÛM^§Ô˜k$Õvtq jXÓ ¤¦ W6¨ å¢ { j k$Ô ÓZj t¦lŸk$^`{·t j ^ Ó kvÝVÕvkok${Z^ Ó Õvk j ÛKtxVk_®^EÝ4qs_bZ_(×koc¤_®qsÔÔqŸa7lmk<br />
ÖIk$qŸdXx?t<br />
j ÛKtxZqsÔ¦Ô˜k$Õvtq jXÓ y<br />
ÖIk$qŸdXx?t<br />
j ×7tqm_®qŸá$^Otq jXÓ Õ jXÓ Ôqm{Vkoc,ÔtxVk¯lm^`Ôt× j c¤t j Û½^2ckodXq jXÓ WKak$Õo^`bZÔ˜k§txZqmÔ°qmÔtxVk¯_ j Ô˜t°qm_§× j ct^ Ó t× j ct<br />
Y¦xVk<br />
Ó Õvk*txZk Ó txZk;ÔxZqŸ×‹lŸk$^ i k$Ô¦txZqmÔIckodXq jXÓ ^ Ó {·qmtÔIl jXÓ dXqŸtbZ{Zq Ó ^`l7Ô˜t˜c¤k Ó d`tx‹×7^Oc¤^`_§kot˜koc,Ô¦xZ^ i kt j _®kokot©txVk<br />
Ôq<br />
Ó dtxVkÉl jXÓ d7J j bVc Ó kon jXÓ txVk j Õvk$^ Ó yMï Ó qmtqm^`lmlŸǹ Wr^{Zq¢ +kock Ó tÆ^`lmt˜koc Ó ^Otq i kÉÖ½^`ÔÆt˜k$Ôt˜k${ ð ï Ó k$^`Õ¤x<br />
lmqm_¯qŸtÔ{ZbZc¤q<br />
ÛCtxZk&× j ctÔ ó ¦ W ¨ å;¢ txVk j ×Ztqm_¾bZ_(ÖÆk$qmdXx?t©{ZqmÔt˜c¤qŸa7bVtq jXÓ§j ÛztxVkÕ jXÓ t^`q Ó koc¤Ô½txZ^Ot½ÖIkockÉt j l j ^`{<br />
j<br />
Ó txVkÔ˜k$Õvtq jXÓ Ô j ÛtxZk°ÔxZqŸ×‹Ö½^`Ô&Õo^`lsÕobZlm^Ot˜k${ yYxVkt j t^`lCÖIk$qŸdXx–t j ÛÃtxVk°Õ jXÓ t^`q Ó koc¤Ô&^`lŸck$^`{Vn‹Ô˜t j ÖÆk${±q Ó<br />
q<br />
jXÓ ¤¦ W©¨ å¢ ÖxVk Ó Õo^`lslmq Ó dtxZqmÔ½× j ct©Ö½^`Ô¦t^Op`k Ó q Ó t j ^`ÕoÕ j b Ó toy£½bVt©ÖIk;ck$^`lmqŸáok${NtxZ^Ot<br />
txVk°Ôk$Õvtq<br />
j ×Ztqm_¾bZ_ ÖÆk$qmdXx?tÉ{ZqmÔ˜t˜c,qŸa7bVtq jXÓ Ô¦Õo^`lmÕobZlm^Ot˜k${¡q Ó txZqmÔ&Ö½^$n2ÖIkockq Ó Û6k$^`ÔqŸa7lmk;ÖqŸtx9txVkÕ jXÓ t^`q Ó koc¤ÔÉt j<br />
txVk<br />
j ^`{Vk${ y ak;l<br />
<br />
D7=6
æ y<br />
6<br />
X ¦3 ä<br />
© °!M&¥Ð ¢¡¤£¦¥§¥§¨¤¥© !/<br />
Y ø<br />
<br />
0 <br />
4 2"<br />
<br />
2§¦ 0 <
j ct j ÛztxVkÕ jXÓ Ôqs{Vkock${ËckodXq jXÓ ð¡ ;ÓZj Öq Ó d°txVk Ó bZ_°akoc½^ Ó {txVkÉÖÆk$qmdXx?t j ÛztxVk$ÔkÉÕ jXÓ t^`q Ó koc,Ô©^ Ó {<br />
ÜZc¤ÔtÆ×<br />
jXÓ q Ó ÖxZqmÕ,x¡txVkon9^Ock®Ôt j ÖIk${ WCÖIk§p ÓZj ÖçtxVk§ck$_¯^`q Ó q Ó d‹Ôl j tÕo^O×7^`Õoqmtfn ^ Ó {¡txVkck$_®^`q Ó q Ó d<br />
txVk®Ô˜k$Õvtq<br />
j l j ^`{ j ÛMk$^`Õ¤x±Ô˜k$Õvtq jXÓ y¦Y¦xVk°Ô˜t j ÖI^Od`kc¤b7lŸk$Ô¦^OckÕ jXÓ Ô˜t˜c¤b7Õvt˜k${2q Ó txZ^OtÖ½^$nNtx7^Otck$Ô˜t j ÖÔ*^Ock<br />
ÖIk$qŸdXx?t&t<br />
ì j qs{Vk${ yÉõbVc¦×Zc j d`c,^`_ ×Zc j {Zb7Õvk$Ô^Ô˜kot j Û^`{7_®qmÔÔqma7lŸkÔ˜t j Ö½^Od`k§×7lm^ Ó ÔoWZÖxZqsÕ¤x2^Oc¤k°c¤^ Ó p`k${±^`ÕoÕ j c¤{Zq Ó d<br />
^<br />
j txVk¯Û j lml j Öq Ó dNÕvc¤qmt˜koc¤q jXÓ ð Y¦xVk$Ô˜kËck$ÔbZlmtÔ°^OckÕ¤x j Ô˜k Ó ÖxZkock®a j txœtxVkak Ó {Zq Ó d‹_ j _§k Ó t^ Ó {›txVk<br />
t<br />
j c,ÕvkËq Ó k$^`Õ,x Ôk$Õvtq jXÓ›j Û½txVkÔxZqŸ×œq Ó ^`lmlMtxVk¯× j c¤tÔ ó C ó¡j Û½txVkÕobVcck Ó t;ckodXq jXÓ ó ^Ock<br />
ÔxVk$^OcÛ<br />
ÖqŸtx7q Ó txVk×koc¤_®qmÔÔ¤qŸa7lŸk*lmqm_®qŸtÔ ß ^`{ZxZkock Ó Õvk j Ûa j tx ô N£¥§© ¦Ã^ Ó { ¤¢¡¤£¦¥¨§© ¦”à,y<br />
"! ¥I5 £¢ . J(¥¤( $"(+8<br />
k¦akodXq Ó ÖqmtxtxVk©ÜZc¤Ô˜tÃ× j ct ó C j Û j bVcÆÕobVcck Ó tMckodXq jXÓ§ó yMÙ#c¤qm_¯^Oc¤qmlŸn°^`lslrÕ jXÓ t^`q Ó koc,Ô#txZ^Ot#xZ^ i k ó C#^`Ô<br />
é<br />
j ct j ÛÃ{Vk$Ô˜tq Ó ^Otq jXÓ¡ß Ù½õ§¦;àI^Ock;b Ó l j ^`{Vk${ yMÞÛ6t˜kocÖ½^Oc¤{ZÔ©^`lml Õ jXÓ t^`q Ó koc,Ô jXÓ txVkÉnX^Oc¤{·ÖqmtxÔ¤^`_§k<br />
txVk$qŸc©×<br />
j _a7q Ó k${2q Ó·jXÓ k;d`c j bV×zyÆYxVk$Ô˜k;d`c j bZ×7Ô j ÛMÕ jXÓ t^`q Ó koc¤Ô&^OckÔ˜t j ÖÆk${±q Ó txZk;×7^Oct˜a7^ nrÔ j ÛÃtxVk<br />
ÙIõ¨¦ç^OckÕ<br />
ÓZÓ q Ó d¾Ö&qŸtxNtx j ÔkxZ^ i q Ó d¯txVkÛ6^OctxZk$Ô˜t&Ù½õ§¦ ß Õ j b Ó t˜k${±ÛQc j _ × j ct ó C,à,yY¦xVk;×7c j Õvk${7bVck j Û<br />
ÔxZqm×NakodXq<br />
j–j Ôq Ó d§txVk×7^Oct˜ag^$n·qmÔ½kvÝ4×7lm^`q Ó k${ak$l j Ö;y#Y j ^`Õ¤xZqŸk i k°txVk j ×Ztqm_®^`lÖIk$qŸdXx–t{ZqmÔt˜c¤qŸa7bVtq jXÓ q Ó txVklm^`Ô˜t<br />
Õ¤x<br />
j ct ó Õo^`lmÕobZls^Ot˜k${œq Ó txVk§ÜZc,Ô˜tÉ×7lm^ ÓZÓ q Ó dÔt˜ko×zWztxVk¯Õ jXÓ t^`q Ó koc,Ô;^Ock®t˜ck$^Ot˜k${ {Zq¡ kock Ó tlŸǹ y ß Þ&ÔtxZkock<br />
×<br />
^Ock*txVckok&×7^Oct˜a7^ nrÔ½q Ó k$^`Õ¤x‹Ô˜k$Õvtq jXÓ W4txVk j ×Ztqm_®^`l7ÖÆk$qmdXx?tIÛ j c jXÓ k*×7^Oct˜a7^ n®q Ó Ôk$Õvtq jXÓ ¤ ¦ WR¨ <br />
å ¢<br />
qmÔ§Ô˜kot§t j<br />
C<br />
©<br />
j Û*txVk j ×Ztqs_®^`lIÖÆk$qŸdXx–t j Û*txVkNÔ˜k$Õvtq jXÓ ¤§¦ªW:¨/ å¢ y¦à Ò j ^`Ô˜t^`l¦Õ jXÓ t^`q Ó koc¤ÔoWIÖxZqsÕ¤x<br />
jXÓ t^`q Ó koc¤Ô¾txZ^Ot^Ockl j ^`{Vk${ ^ Ó { b Ó l j ^`{Vk${›q Ó ckodXq jXÓ ó W#^Ock‹l j ^`{Vk${—q Ó t j txVk‹Õ¤x j Ô˜k Ó ×7^Oc¤t˜a7^$n<br />
^Ock‹Õ<br />
j qŸtÔ_®^EÝ4qs_bZ_×koc¤_®qmÔ¤ÔqŸa7lŸkÖIk$qŸdXx–toyœY¦xVk‹ÔxZqŸ×zë|ÔÕ j ^`Ô˜t^`l4J j bVc Ó konœq Ó ckodXq jXÓ—ó qmÔ¾txVkÕo^`lslmq Ó d<br />
bV×—t<br />
ÛÉ^`lmlItxVk·× j ctÔ ó C ;ó ^ Ó {txVkN_®^EÝVqm_¾bZ_×koc¤_®qmÔÔ¤qŸa7lŸk®ÖIk$qŸdXx–t§Û j c jXÓ k·×7^Oct˜a7^ n—q Ó Ô˜k$Õvtq jXÓ<br />
j<br />
W)¨¥ å¢ qmÔ;Ô˜kot°t j<br />
C<br />
¤¦ j Û½txVk¯_®^EÝVqm_¾bZ_ ×koc¤_®qsÔÔqŸa7lmk;ÖÆk$qmdXx?t j ÛÆtxVkËÔ˜k$Õvtq jXÓ ¤§¦ªW^`Ô;txVkockË^Ock<br />
©<br />
Ó k$^`Õ¤x¡Ô˜k$Õvtq jXÓ y*â#Ýr× j ct&Õ jXÓ t^`q Ó koc¤ÔoWÖxZqmÕ,x2^OckÕ jXÓ t^`q Ó koc¤Ô*ÖqŸtx9ÙIõ§¦Úq Ó txZk Ó kvÝ4t<br />
txVckok¾×7^Oct˜ag^$n4Ô&q<br />
jXÓ WZ^Ock°l j ^`{Vk${·q Ó t j txVkÕ¤x j Ô˜k Ó ×7^Oct˜a7^ n‹bZ×·t j ^§Õvkoct^`q Ó t j lmkoc¤^ Ó Õvk i ^`lmbVk ß Ô˜kok°a+k$l j Ö*àI^Oa j i kqŸtÔ<br />
ckodXq<br />
Õo^`lmÕobZls^Ot˜k${ j ×Ztqm_®^`lzÖÆk$qŸdXx–toyY¦xZk×Zc j Õvk${ZbVck j ÛÃÕ¤x jrj Ôq Ó dË^×7^Oct˜a7^ n‹qsÔ ð<br />
y¦é kÔ˜k$^Oc¤Õ,xœÛ j c;tx j Ô˜k®×g^Oct˜a7^$n4Ô;xZ^ i q Ó dN^`lŸck$^`{Vn¡Õ jXÓ t^`q Ó koc¤Ô°Û j c;txZk¯Ô^`_§k®Ù½õ§¦ñ^`ÔtxVk®d`c j bV×<br />
å<br />
ÛÉÕ jXÓ t^`q Ó koc¤Ô§t j l j ^`{ ÓVj Ö°yHïªÛ*_ j ck·txZ^ ÓHjXÓ kNÔbZÕ¤x×g^Oct˜a7^$nHqmÔ¾Û j b Ó {txVk·×g^Oct˜a7^$n j Û*txVk<br />
j<br />
jXÓ Õol j Ô˜k$Ôt©t j txVka j ÖöqmÔ©Õ,x j Ô˜k Ó WZ^`Ô¦ÖÆkÛ j b Ó {‹txZqsÔÆak Ó koÜgÕoqm^`l+Û j c¦txZk*ak Ó {7q Ó d_ j _§k Ó tÔoy<br />
Ôk$Õvtq<br />
ÛCtxZkd`c j bV× j ÛÕ jXÓ t^`q Ó koc,Ô½t j l j ^`{ ÓVj Ö«kvÝ4Õvkok${ZÔ¦ÖÆk$qŸdXx–t j c¦Ôl j t©Õo^O×7^`Õoqmtfn·lmqm_®qmtÔ j ÛKtxVk;Õ¤x j Ôk Ó<br />
ï<br />
× j ÔÔ¤qŸa7lŸk Ó bZ_akoc j ÛÕ jXÓ t^`q Ó koc¤Ô©qmÔ½Ô˜t j ÖÆk${Nq Ó t j txZqmÔÆ×7^Oct˜ag^$n‹^ Ó {txVk Ó<br />
×g^Oct˜a7^$nËtxVk_®^EÝVqm_b7_<br />
j lŸkÉ×Zc j Õvk${ZbZck;^`Ô¦q Ó å à½Ôt^OctÔ^OdX^`q Ó Û j ctxVkck$_®^`q Ó q Ó d§Õ jXÓ t^`q Ó koc¤Ôoy<br />
txZk;Öx<br />
y¦ï Û ÓZj ×7^Oct˜a7^ n—{Vk$Ô¤Õvc¤qŸak${ q Ó÷å à¾qmÔÛ j b Ó {^ Ó { txVk·d`c j bV× j ÛÉÕ jXÓ t^`q Ó koc¤Ô§tx7^Ot®qmÔ¾t j l j ^`{ JbZÔ˜t<br />
æ<br />
Ö jXÓ lŸn®Õ jXÓ Ô¤qmÔ˜tÔ j ÛzkvÝr× j ct½Õ jXÓ t^`q Ó koc¤ÔoW4k$_§×7tfn®×7^Oc¤t˜a7^$n4Ô½^Ock*Ôk$^Oc¤Õ¤xVk${‹Û j c$yÃï ÛC_ j ck*txZ^ ÓËjXÓ k<br />
ÓZj<br />
n°qmÔ^ i ^`qmlm^Oa7lŸk jXÓ k j ÛZtxVkIÛ j lml j Öq Ó dtxVckok½^`lŸt˜koc Ó ^Otq i k$ÔÃqmÔCc,^ Ó { j _®lŸnÕ,x j Ô˜k Ó yÃé qŸtx<br />
k$_®×Ztfn×7^Oct˜a7^<br />
kock Ó t°^`lŸt˜koc Ó ^Otq i k$Ô i ^Ocn4q Ó d2Ôt j Ö½^Od`k×glm^ Ó Ô;^OckËd`k Ó koc¤^Ot˜k${—ÖxVk Ó c¤b ÓZÓ q Ó d‹txVk<br />
txZk$Ô˜k¯txVc¤kokË{Zq¡<br />
j Ö½^Od`k×7ls^ ÓZÓ q Ó dÔ˜k i koc¤^`lCtqm_§k$Ô©Ö&qŸtx‹txVk°Ô^`_®k;q Ó ×7bVt©{Z^Ot^4yÆÞ&_ jXÓ d®tx j Ô˜k°{7q¡ +koc¤k Ó t¦Ô˜t j ÖI^Od`k<br />
Ôt<br />
Ó Ô½txVk;×7ckoÛQkoc¤ck${ jXÓ k°Õo^ Ó ak;Õ,x j Ô˜k Ó _¯^Oprq Ó d®bZÔ˜k j ÛÃ^®{Vk$ÕoqmÔq jXÓ Õvc¤qŸt˜koc¤q jXÓ {Zk$ÔÕvc¤qŸak${ak$l j Ö;y<br />
×glm^<br />
^?à‹YxVkd`c j bV× j ÛÆÕ jXÓ t^`q Ó koc¤Ô*t j l j ^`{ ÓVj ÖqsÔ*Ôt j ÖIk${ q Ó t j ^×7^Oct˜ag^$nNÖqmtx9k ÓZj bVdXx¡Ôl j tÉ^ Ó {<br />
ß<br />
j ctxZqmÔ©d`c j bZ× j ÛMÕ jXÓ t^`q Ó koc¤Ôoy#ï Û_ j ck;txZ^ Ó·jXÓ k°ÔbZÕ,x·×7^Oct˜a7^ n·qmÔ©Û j b Ó {<br />
ÖIk$qŸdXx–tÕo^O×7^`ÕoqŸtªn·Û<br />
jXÓ k*ÖqŸtxËlŸk$^`Ô˜t½Ôl j t½Õo^O×7^`ÕoqŸtªnËqmÔÆÕ¤x j Ô˜k Ó t j p`koko׋×7^Oct˜a7^ nrÔIÖqmtx¯xZqŸdXxZkocIÔl j tÆÕo^O×7^`ÕoqŸtqmk$Ô<br />
txZk<br />
j cÛTbVtbVckd`c j bV×7Ô j ÛÆÕ jXÓ t^`q Ó koc,ÔoyïªÛ ÓVj Ô¤bZÕ¤x±×7^Oct˜ag^$n9qsÔ&Û j b Ó {9txVkd`c j bV× j ÛÕ jXÓ t^`q Ó koc,Ô<br />
Û<br />
i q Ó d‹txVk®xZqmdXxVk$Ô˜tÔl j tÕo^O×7^`ÕoqŸtªn qmÔÕ j _§×7lŸkot˜k$lmn9l j ^`{Vk${ ^ Ó {¡txVk<br />
qsÔÔ˜×7lmqŸtoy¾Y¦xVk×7^Oct˜ag^$n¡xZ^<br />
Ó q Ó d¯Õ jXÓ t^`q Ó koc¤Ô*^Ock¾Ôt j ÖIk${ {ZbVk;t j·æ àÖxVkock°^OdX^`q Ó9jXÓ k j Û#txVk¾txVckok¾^`lŸt˜koc Ó ^Otq i k$Ô<br />
c¤k$_®^`q<br />
^?àfe ß Õ àqmÔ©c,^ Ó { j _®lŸnËÕ¤x j Ô˜k Ó y<br />
ß<br />
a+àYxVkËÖIk$qŸdXx–t j Û&^`lmlÆtxVk·Õ jXÓ t^`q Ó koc¤Ô j ÛtxZkËd`c j bZ×qsÔ¾{7q i qs{Vk${œarnœtxVk‹d`c j bV×Cë|Ô Ó bZ_akoc j Û<br />
ß<br />
jXÓ t^`q Ó koc,ÔoyMØ j c#k$^`Õ¤x®×7^Oct˜a7^ n°txVk rb j tqŸk Ó t j ÛgqŸtÔ j ×Ztqm_¯qŸáok${¾ÖÆk$qŸdXx–t#^ Ó {qŸtÔÃÔl j tMÕo^O×7^`Õoqmtfn<br />
Õ<br />
j ÖÚtxVk®d`c j bV× j Û©Õ jXÓ t^`q Ó koc¤Ô;qmÔ;l j ^`{Vk${ q Ó t j txZ^Ot;×7^Oc¤t˜a7^$n±ÖxVkock§txZk$Ô˜k<br />
qsÔ;Õo^`lmÕobZlm^Ot˜k${zy<br />
tªÖ j2i ^`lmbZk$Ô*^O×7×Zc j ÝVqm_®^Ot˜k$lmn2_®^OtÕ¤xzy°ï ÛtxVkd`c j bV× j ÛIÕ jXÓ t^`q Ó koc¤Ô{ j k$Ô ÓVj tÕ j _§×7lŸkot˜k$lŸn9ÜZt<br />
Y7Y
Ó ^Otq i k$Ô ß ^?àfe ß Õ àqmÔ½c¤^ Ó { j _®lŸnÕ¤x j Ô˜k Ó y<br />
txZk;txVckok;^`lŸt˜koc<br />
Õ à‹YxVkÆd`c j bV× j Û+Õ jXÓ t^`q Ó koc¤ÔÃqmÔÔ˜t j ÖIk${§q Ó t j ^&×7^Oct˜a7^ n°Öqmtx°k ÓVj bZdXxÔl j t^ Ó {¾ÖIk$qŸdXx–tÕo^O×7^`Õoqmtfn<br />
ß<br />
yYxZqmÔc¤bZlŸk©qmÔÃ^O×Z×glmqŸk${qŸÛ ÓZj ×g^Oct˜a7^$nqmÔÃÛ j b Ó {¾arn¾bZÔ¤q Ó d*txVk©c¤bZlŸk$Ô å àÃ^ Ó { æ à,yÃÞ¢×7^Oct˜ag^$nqmÔMÕ¤x j Ôk Ó<br />
5<br />
j ^`{Vk${9arn2Õ jXÓ t^`q Ó koc¤Ô*ÖqŸtx±ÙIõ§¦ÚÛ6^Oc¤txVkoc^$Ö½^$n±txZ^ Ó txZkÙ½õ§¦ j ÛtxVkd`c j bV×<br />
tx7^OtÉqsÔ&^`lŸc¤k$^`{Vn2l<br />
Y¦xVk Ó txZk°Ô^`_§k j ×koc¤^Otq jXÓ Ô©^`Ôq Ó × j ct ó C ^Ock°{ jXÓ k;ÔbZÕoÕvk$ÔÔ¤q i k$lŸn‹q Ó × j ctÔ ó<br />
<br />
×koc¤^Otq jXÓ Ô^OdX^`q Ó akodXq ÓZÓ q Ó dÖ&qŸtx‹txVkÜZc¤Ôt©× j c¤toy<br />
j<br />
j Û+txVk¦^`lŸt˜koc Ó ^Otq i k$ÔMq Ó®æ àKÖIk¦^Ock¦^Oa7lŸk½t j d`k Ó koc,^Ot˜kÔ˜k i koc¤^`lZ{7q¡ +koc¤k Ó tMÔ˜t j ÖI^Od`k&×7lm^ Ó ÔÛ j cMtxVk<br />
Þ*Ô#^Éck$ÔbZlmt<br />
Ó t j txZqmÔ×7^Oct˜ag^$n°{ZbZkÆt j txVk½×7^Oct˜a7^ ngë|ÔÃÔl j tÕo^O×7^`Õoqmtfǹ W?txZkÆd`c j bZ×qsÔKÔ˜×glmqŸtK^ Ó {°txZqsÔC×7^Oc¤t˜a7^$n<br />
q<br />
Ó tqŸck$lmn¯l j ^`{Vk${ yMY¦xZk&ck$_¯^`q Ó q Ó d¾Õ jXÓ t^`q Ó koc¤Ô©^OckÔ˜t j ÖÆk${·{ZbZk&t j®æ àÆÖxVkoc¤k*^OdX^`q ÓjXÓ k j Û<br />
qsÔÆk<br />
j c©txZqmÔId`c j bZ× j ÛKÕ jXÓ t^`q Ó koc¤ÔoyMï ÛK_ j ckÉtxZ^ Ó‹jXÓ k j ÛKÔb7Õ¤x‹×7^Oct˜ag^$n4Ô©qmÔIÛ j b Ó {ËtxZk jXÓ kÖqŸtx<br />
Û<br />
j tÆÕo^O×7^`ÕoqŸtfnqmÔÕ,x j Ô˜k Ó yMï Û ÓVj Ô¤bZÕ¤xË×7^Oct˜a7^ n®qmÔ#Û j b Ó {zW–^`lŸt˜koc Ó ^Otq i ka+à#qmÔ^O×7×7lmqŸk${<br />
txZk&lŸk$^`Ô˜t½Ôl<br />
t j Ü Ó {·^§×7^Oc¤t˜a7^$ǹ y<br />
ÛKÕ jXÓ t^`q Ó koc¤ÔIt j l j ^`{ ÓVj Ö;yMï Ó txZqmÔÆÕo^`Ô˜kÉtxVkÉÕ jXÓ t^`q Ó koc¤Ô½^Ock&×koc¤x7^O×7ÔM×7bVt jXÓ t j × j ÛzÕ jXÓ t^`q Ó koc,Ô<br />
j<br />
ÓVj txVkoc°Ù½õ§¦¾yJk i koctxZk$lŸk$ÔÔoWzc¤k$Ô˜t j ÖÔ°^Ock®^ ì j qm{Zk${ ak$Õo^`bZÔ˜k§txVk§Õ jXÓ t^`q Ó koc¤Ô×gbVt jXÓ t j ×<br />
Ö&qŸtx¡^<br />
ÛKtxVk j txVkoc¤Ô©^OckÉk$^Oc¤lsqŸkocIb Ó l j ^`{Vk${ yMY¦xZk*Õ,x j qmÕvk j ÛCtxVkÉ×7^Oc¤t˜a7^$nqmÔ½_®^`{VkÉ^`Ô½Û j lml j ÖÔoyMÙÃ^Oct˜a7^ nrÔ<br />
j<br />
ÖqŸtx Õ jXÓ t^`q Ó koc,Ô;xZ^ i q Ó d·^·Ù½õ§¦ jXÓ k§× j ctÉÛ6^OctxZkoc°^$Ö½^$n txZ^ Ó txZk®ÙIõ§¦ j ÛItxVk<br />
^Oc¤k¯Ô˜k$^Oc¤Õ,xVk${<br />
j bV× j ÛÆÕ jXÓ t^`q Ó koc¤Ô*t j l j ^`{ ÓVj Ö;y*ïªÛ ÓVj ÔbZÕ,x2×g^Oct˜a7^$n4ÔÉ^Ock¾Û j b Ó {2×7^Oc¤t˜a7^$n4Ô*ÖqŸtx2Õ jXÓ t^`q Ó koc,Ô<br />
d`c<br />
j × j c¤tÔ½Û6^OctxZkoc^$Ö½^$nN^Ock;Ôk$^Oc¤Õ¤xVk${zWZkotÕOyÆÙc j Õvkok${7q Ó d§q Ó txZqmÔ½Ö½^$ǹ WgtxVk×7^Oct˜ag^$n4Ô<br />
Ö&qŸtx·ÙIõ§¦]tªÖ<br />
j ^`{Vk${a–nÕ jXÓ t^`q Ó koc¤Ô©xZ^ i q Ó d jXÓ k j ÛKtxVklm^`Ô˜t½× j ctÔ j ÛCtxZkÉÕobZcck Ó tIc¤kodXq jXÓ ^`Ô<br />
tx7^Ot¦^Ock^`lŸck$^`{Vnl<br />
Ó akÉb7Ô˜k${Û j c¦l j ^`{7q Ó d^`lslgtxZk$Ô˜k;Õ j ^`Ô˜t^`lzÕ jXÓ t^`q Ó koc¤Ô©x7^ i q Ó d¾× j c¤tÔ j ÛK{Vk$Ô˜tq Ó ^Otq jXÓ<br />
txZk$qŸc©ÙIõ§¦]Õo^<br />
yMïªÛgtxVk¦d`c j bZ× j Û+Õ jXÓ t^`q Ó koc,Ô{ j k$Ô ÓVj tÕ j _®×7lŸkot˜k$lŸn¾ÜZt#q Ó t j txVkÕ,x j Ô˜k Ó ×g^Oct˜a7^$n{ZbVk©t j qŸtÔ<br />
^`xZk$^`{<br />
j tÕo^O×7^`ÕoqŸtªǹ WgtxZk;d`c j bV×2qsÔ¦Ô˜×7lmqmtoyY¦xVk;Õ,x j Ô˜k Ó ×g^Oct˜a7^$nNqsÔ©k Ó tqmck$lŸn‹l j ^`{Vk${N^ Ó {NtxVk;ck$_®^`q Ó q Ó d<br />
Ô¤l<br />
Õ jXÓ t^`q Ó koc,Ô^Ock°Ôt j ÖIk${2{ZbVkt j 5 à,y<br />
Û ÓVj ×g^Oct˜a7^$nœqsÔÛ j b Ó {œbZÔq Ó d·c¤bZlŸk$Ô å àÉt j 5 àqŸt¾qmÔ Ó k$Õvk$ÔÔ¤^Ocn t j Ô˜t j Öñq Ó t j ×7^Oct˜a7^ nrÔ¾^`lŸck$^`{Vn<br />
øVy¦ï<br />
j ^`{Vk${¯arn®Õ jXÓ t^`q Ó koc¤Ô½xZ^ i q Ó d^¾ÙIõ§¦ö^`xZk$^`{ j Û txZk&d`c j bZ×zë|Ô j ÛzÕ jXÓ t^`q Ó koc,ÔÆ× j ct j Ûz{Vk$Ô˜tq Ó ^Otq jXÓ y<br />
l<br />
j ^`{Zq Ó d®q Ó t j txVk$Ôk;×7^Oct˜a7^ nrÔ&Õo^ Ó Õo^`bZÔ˜k°ck$Ô˜t j Ö&Ô¦ÖxZqsÕ¤x·lŸk$^`{·t j xZqŸdXxVkoc¦Õ j Ô˜tÔ$y<br />
ó y<br />
qŸtxZq Ó txVkË×g^`Õp4q Ó d±c¤bZlŸk$ÔÖÆkt^Op`k·Õo^Ock j Û¦txVk‹_®^EÝVqm_b7_ ×koc¤_®qsÔÔqŸa7lmk¾ak Ó {7q Ó d2_ j _®k Ó t j Û¦k$^`Õ¤x<br />
é<br />
jXÓ ¤¦ªW ¨ å;¢ Ô¤q Ó Õvk‹txZqmÔ§xZ^`{Å^`lŸÖI^ nrÔ®a+kok Ó txZk2Õvc¤qŸtqmÕo^`l rbZ^ Ó tqŸtfnq Ój bVc®Ô¤Õvk Ó ^Oc¤q j y“Y j<br />
Ô˜k$Õvtq<br />
j bVc°×Zc j d`c¤^`_ ×Zc j {Zb7Õvk$Ô^Ot°lŸk$^`Ô˜t jXÓ kËÔ˜t j Ö½^Od`k¯×7ls^ Ó Û j c;k$^`Õ,x›qm_¯^OdXq Ó ^Oa7lŸkÔÕvk Ó ^Oc,q j W<br />
_®^Op`k¯Ô¤bVcktxZ^Ot<br />
j Ö txVk9ak Ó {7q Ó d›_ j _®k Ó tÔ‹q Ó Ô j _§k¡Ô˜k$Õvtq jXÓ Ôt j ak9xZqmdXxVkocËtx7^ Ó txVk _®^EÝVqm_¾bZ_<br />
ÖIk9×Zc,qm_®^Oc¤qmlmn—^`lml<br />
Ó {7q Ó d_ j _§k Ó tÔ j ÛCtxVk$Ô˜k;Ôk$Õvtq jXÓ Ôoy#Y¦xZkockoÛ j c¤k`W4^ Ó ^Oc¤a7qŸt˜c¤^Oc¤qslŸn®Õ¤x j Ôk Ó ×7^Oc¤^`_§kot˜kocÔkotÔ<br />
×koc¤_®qmÔ¤ÔqŸa7lŸk©ak<br />
jXÓ ¤¦ W)¨ å;¢ bZ× t j ÖxZqsÕ¤x¡txVk§ak Ó {Zq Ó d_ j _§k Ó tÔxZ^ i kËt j ak§lŸk$ÔÔ j c;krbZ^`lMt j txVk<br />
txVk¯Ôk$Õvtq<br />
Ó {Zq Ó d*_ j _§k Ó tÔ#Û j cMk$^`Õ,x¯Ô˜k$Õvtq jXÓ yÃÞ)Ô˜k$Õ jXÓ {¯^Oca7qŸt˜c¤^Oc,qmlŸn;Õ¤x j Ô˜k Ó ×7^Oc¤^`_®kot˜koc<br />
_®^EÝVqm_¾bZ_×koc¤_®qmÔÔ¤qŸa7lŸkMak<br />
Ó t^Od`k§txZ^Ot^`lml j Ö&Ô&txZkak Ó {Zq Ó d¯_ j _§k Ó tÔq Ó txVk§Ô˜k$Õvtq jXÓ Ô ¤¦HGIC ; ¤ ›ß qyºk`y Õol j Ô˜kt j<br />
Ô˜kotÔÉtxVk¾×koc¤Õvk<br />
j Ö*à;t j akËxZqmdXxVkoc°txZ^ Ó txVk_®^EÝ4qs_bZ_ ×koc¤_®qmÔÔ¤qŸa7lŸkak Ó {Zq Ó d‹_ j _§k Ó tÔq Ó txVk$Ô˜k‹Ô˜k$Õvtq jXÓ Ô$yNïªÛ<br />
txVkËa<br />
Ó d;txVk*×7^`Õ¤prq Ó d j ×koc¤^Otq jXÓ q Ó txVkÉÔ˜k$Õ jXÓ {‹Ô˜t˜ko×·q Ó ^ Ó nË× j cttxVkÉak Ó {Zq Ó d;_ j _§k Ó t©q Ó ^ Ó n¯Ô˜k$Õvtq jXÓ<br />
{ZbVc,q<br />
Ó txVk i ^`lsbVk°ck$Ô¤bZlŸtq Ó d§Û6c j _ txVkÔ˜k$Õ jXÓ {2×7^Oc,^`_§kot˜koc*{ZkoÜ Ó k${2^Oa j‚i k`W j cÉ^ Ó nNÕ jXÓ t^`q Ó koc,Ô<br />
qmÔd`ck$^Ot˜kocÉtxZ^<br />
ÓZÓVj tak2l j ^`{Vk${ j bVc×Zc j d`c¤^`_ Ô˜t j ×7ÔËqm_®_§k${Zqs^Ot˜k$lŸnHq Ó txZqmÔ¯× j ct¯^ Ó {)Ô˜t^OctÔËtxVk2Ö&x j lŸkN×7^`Õ¤prq Ó d<br />
Õo^<br />
Ó ^Oc¤q j ^`_ jXÓ dtx j Ô˜k^;×7lm^ ÓZÓ kocMÕo^ Ó Õ¤x jrj Ô˜k¦txVk jXÓ kxVk©×ZckoÛ6koc¤ÔoyÃY j _®^Op`ktxZqsÔM{Vk$ÕoqmÔq jXÓ k$^`Ô¤qŸkoc<br />
Ô^`_§k&ÔÕvk<br />
j c;xZqm_ ÖÆk§Ö½^ Ó tt jNj +koc;x7qm_ Ô j _§k®Õvc¤qŸt˜koc,qm^t j ×7qmÕp jXÓ k j ÛtxZk§×7lm^ Ó ÔoyÞ½t;txVk®_ j _§k Ó t jXÓ lŸn jXÓ k<br />
Û<br />
jXÓ WXtxVk©ak Ó {7q Ó dÉ_ j _§k Ó tÕvc¤qŸt˜koc¤q jXÓ WXtxZ^Ot#×gqmÕp4ÔMtxVkÔt j Ö½^Od`k&×7ls^ Ó ÖxVkock©txVka+k Ó {Zq Ó dÉ_ j _§k Ó t<br />
Õvc¤qŸt˜koc,q<br />
Ó k$^`Õ¤x§Ô˜k$Õvtq jXÓ q Ó ^`lsl?txVkI× j ctÔ ó C ;ó qsÔKlŸk$Ô¤ÔKtx7^ Ó txZkI_®^EÝVqm_¾bZ_]×koc¤_®qmÔÔ¤qŸa7lŸkÃak Ó {Zq Ó d_ j _§k Ó t<br />
q<br />
Ó txVkÉ×7^OctqsÕobZlm^Oc½Ô˜k$Õvtq jXÓ WZqmÔ½qm_®×7lŸk$_§k Ó t˜k${ yÃï Ój bVc¦bZ×Ve”t j e ÓVj ÖêÔÕvk Ó ^Oc¤q j ^Ô˜t j ÖI^Od`k×7lm^ Ó _®^Oprq Ó d§bZÔ˜k<br />
q<br />
Û©txVkËa+k Ó {Zq Ó d·_ j _§k Ó tÕvc,qŸt˜koc¤q jXÓ ÖI^`Ô¾^`lŸÖ½^$n4Ô°Û j b Ó { y·Y¦xZk¯{Vk i k$l j ×7_®k Ó t¾^ Ó {›qm_®×7lŸk$_§k Ó t^Otq jXÓ j Û<br />
j<br />
i ^ Ó Õvk·lmqŸp`ktxVkÕvc,^ Ó kvefÔ˜×7lmqmtqmÔ¾×7lm^ Ó7Ó k${ y±ïªÛ ÓZj Ô˜t j ÖI^Od`k2×7lm^ Ó qmÔ<br />
ÛTbVctxVkocÕvc¤qŸt˜koc¤qs^2Öqmtx›×Zc,^`ÕvtqmÕo^`lIck$lŸk<br />
i k Ó ^`Ô&^¯ck$ÔbZlmt©ÖxVk Ó b7Ôq Ó d§txVkak Ó {Zq Ó d§_ j _®k Ó tÉÕvc¤qŸt˜koc¤q jXÓ W j bVc*{Vk i k$l j ×k${Na7^`lslm^`Ô˜te”Ö½^Ot˜koc˜ef_ j {Vk$l<br />
dXq<br />
{Vk$ÔÕvc,qŸak${q Ó txVk Ó kvÝ4t¦Ô˜k$Õvtq jXÓ qmÔ©^O×Z×7lmqmk${ y<br />
Y
t^ Ó p4ÔxZ^ i k®{7q¡ +koc¤k Ó tÉ_®^EÝVqm_b7_‘Üglmlmq Ó d –bZ^ Ó tqŸtqŸk$Ô$W3N§¦ ¥©¨¨¥ W © ¦ W3Ü<br />
å¢ y°Y¦xVkl jXÓ dXqmtbZ{Zq Ó ^`l<br />
Ó t˜koc j ÛId`c¤^ i qŸtfn j ÛItxVk§a7^`lmlm^`Ôte”ÖI^Ot˜koc$W < W ¦ ¥©¨¨¥© © ¦ WKq Ó txVk§t^ Ó p j ÛÆtxVk®Ôk$Õvtq jXÓ ¤¦ WJ¨F å¢ qmÔ<br />
Õvk<br />
j Ôk${¯t j ak*p ÓVj Ö Ó yMï ÛztxVkÉak Ó {Zq Ó d_ j _§k Ó t©q Ó ^Ot¦lŸk$^`Ô˜t jXÓ kÔ˜k$Õvtq jXÓ‹j ÛKtxVkÔ˜t j Ö½^Od`k°×glm^ Ó dXq i k Ó<br />
ÔbV×7×<br />
kok${Vk${zWVÖxVkock;Ô j _®k j ÛKtxZk°Õo^`lmÕobZlm^Ot˜k${N×7^Oc¤^`_§kot˜koc¤Ô&Õ¤xZ^ Ó d`k ð<br />
Ó<br />
j t^`lÖIk$qŸdXx–t/NP¦ j Û¦^±Ô˜k$Õvtq jXÓ ¤¦qmÔ;dXq i k Ó arnœ^ Ó ^`{7{ZqŸtq jXÓ j Û¦qmtÔlmqmdXx?tÔx7qŸ× ÖÆk$qmdXx?t/NPO@Q7¦ WÃtxVk<br />
Y¦xVkËt<br />
X ¦ txZ^Ot¯^Ock2Ôt j ÖIk${¢q Ó txZk2Ô˜k$Õvtq jXÓ ¤ ¦ W½^ Ó {HtxZkNt j t^`l¦ÖIk$qŸdXx–t j Û*txVk<br />
a7^`lmls^`Ô˜te”ÖI^Ot˜koc N¦ ¥¨¨¥© © ¦ ×7bZ_§×k${‹q Ó txVk;t^ Ó p j ÛÃtxZqmÔ©Ôk$Õvtq jXÓ ð<br />
N%¦) N O@Q7¦32 N X ¦32 N§¦ ¥¨¨¥ © ¦<br />
N O@Q7¦ 0 < O@Q7¦”à 2 ß N X ¦ 0
å ¥©¨¨¥ © ¦ 3 ä<br />
æ<br />
yN§¦<br />
¥©¨¨¥ © ¦ N¦ ¥©¨¨¥© W © ¦<br />
yN§¦<br />
¡ !"+ѧ¡˜+"<br />
N<br />
X ¦ Û j c<br />
ÖxVkoc¤k¥ "—^ Ó {¨¦ ) G<br />
^Ock°^Oca7qmt˜c¤^Oc¤qmlŸnËÕ¤x j Ô˜k Ó ×7^Oc¤^`_§kot˜koc,ÔoW<br />
€ J ~,+ 1rP`U®U : :<br />
5 y < ¤¢¡)¦ < ¤¢¡£¥§© ¦<br />
Y ã<br />
Ôx j ÖÔ½txVk;{ZqsÔ˜t˜c¤qŸa7bZtq jXÓ§j ÛtxVkÕ jXÓ t^`q Ó koc¤Ô j i koctxVktxVckokÉ×7^Oct˜ag^$n4Ô j ÛCtxVk;Ô¤xZqŸ×Û j c©txVk;{Zko×7^Oc˜e<br />
ØÃqŸdVy<br />
Ó × j ct*Ô˜k i k Ó W+ÖxVkock;txZkÕ jXÓ t^`q Ó koc¤ÔÉ^Ock¾Ô j ct˜k${±^`ÕoÕ j c¤{Zq Ó dËt j txZk$qŸc× j ctÔ j Û#{Vk$Ô˜tq Ó ^Otq jXÓ y½é k<br />
tbVck¾q<br />
×7qmÕ¤p j bVtÉ× j ct Y<br />
Û j ctxVk× j c¤tÔ å t j WK^`ÔkvÝ4^`_®×7lŸk`yÞ*Ô_§k Ó tq jXÓ k${ q Ó txVk§Ü7c¤Ô˜t*×7lm^ ÓZÓ q Ó dËÔ˜t˜ko×CW txVk<br />
jXÓNj‚i koc;txVk§Ô˜k$Õvtq jXÓ Ô j ÛÆtxVk§ÔxZqŸ×¡qmÔ jXÓ lŸn j ×Ztqm_®qŸáok${¡q Ó txVk§lm^`Ôt*× j ct j ÛÆ^‹c¤kodXq jXÓ y<br />
ÖIk$qŸdXx?t{ZqsÔ˜t˜c¤qŸa7bZtq<br />
§ e åoä cko×Zc¤k$Ô˜k Ó t&txVk^`Õvtb7^`lKÔ˜t^Ot˜k j Û#txVk¾ÔxZqŸ×9^OttxVk¾{Vko×7^OctbVc¤kq Ó × j ct Y ß Ù½õ Y à,yï Ó × j c¤t Y<br />
ØÃqŸdXÔoy<br />
ÓVj Õ jXÓ t^`q Ó koc¤Ô&^Ock°b Ó l j ^`{Vk${‹^ Ó {·ø 5 Õ jXÓ t^`q Ó koc¤Ô¦Öqmtx·^t j t^`lCÖIk$qŸdXx–t j Û 555 å y 5 t jXÓ Ô¦^Oc¤k°l j ^`{Vk${ y<br />
Assigned Patterns of POD<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
¢¡¢<br />
£¡£<br />
¤¡¤<br />
¥¡¥¡¥¡¥<br />
¦¡¦¡¦¡¦ §¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
§¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
¦¡¦¡¦¡¦ ¥¡¥¡¥¡¥ ¤¡¤ £¡£<br />
£¡£<br />
¤¡¤<br />
¥¡¥¡¥¡¥<br />
£¡£<br />
¦¡¦¡¦¡¦ §¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
§¡§¡§ ¨¡¨¡¨ ¥¡¥¡¥¡¥ ©¡©¡© ¡¡<br />
¨¡¨¡¨ ©¡©¡© ¡¡ §¡§¡§ ¦¡¦¡¦¡¦<br />
¤¡¤ ¥¡¥¡¥¡¥<br />
£¡£<br />
£¡£<br />
¤¡¤<br />
£¡£<br />
¥¡¥¡¥¡¥<br />
¦¡¦¡¦¡¦ §¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
§¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡ ¥¡¥¡¥¡¥ ¤¡¤<br />
£¡£<br />
£¡£<br />
¤¡¤<br />
¥¡¥¡¥¡¥<br />
£¡£<br />
¦¡¦¡¦¡¦ §¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
§¡§¡§ ¨¡¨¡¨ ¥¡¥¡¥¡¥ ©¡©¡© ¡¡<br />
¨¡¨¡¨ ©¡©¡© ¡¡ §¡§¡§ ¦¡¦¡¦¡¦<br />
¤¡¤ ¥¡¥¡¥¡¥<br />
£¡£<br />
¡¡¡ ¡¡ ¡¡ ¡¡<br />
¡¡ ¡¡ ¡¡¡ ¡¡<br />
¡¡<br />
¡<br />
¡ ¡<br />
¡ ¡<br />
¡ ¡ ¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡<br />
¡ ¡¡<br />
¡¡¡ ¡¡ ¡¡ ¡¡<br />
¡¡ ¡¡ ¡¡¡ ¡¡<br />
¡¡<br />
¡<br />
¡ ¡<br />
¡ ¡<br />
¡ ¡ ¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡<br />
¡ ¡¡<br />
¡¡¡ ¡¡ ¡¡ ¡¡<br />
¡¡ ¡¡ ¡¡¡ ¡¡<br />
¡¡<br />
¡<br />
¡ ¡<br />
¡ ¡<br />
¡ ¡ ¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡<br />
¡¡ ¡¡ ¡¡<br />
¡¡<br />
¡<br />
¡ ¡ ¡¡¡<br />
¡¡ ¡¡ <br />
¡¡ ¡¡<br />
¡¡ ¡¡ <br />
¡¡ ¡¡ <br />
<br />
¡¡ ¡¡<br />
¡¡ ¡¡ <br />
¡¡ ¡¡ <br />
<br />
¡¡ ¡¡<br />
¡¡ ¡¡ <br />
¡¡ ¡¡ <br />
1 2 3 4 5 6 7 8 9 10 11 12<br />
POD<br />
¢¡¢<br />
¤¡¤ ¥¡¥¡¥¡¥ ¦¡¦¡¦¡¦ §¡§¡§ ¨¡¨¡¨ ©¡©¡© ¡¡<br />
£¡£<br />
¡¡ ¡¡ ¡¡ ¡¡ ¡<br />
¡ ¡ ¡¡¡<br />
¡¡ ¡¡ <br />
¢¡¢<br />
¤¡¤ ¥¡¥¡¥¡¥ ¦¡¦¡¦¡¦ £¡£<br />
¡ ¡¡<br />
ØÃqŸdVy ærð Ckod`k Ó {<br />
10<br />
MC ges<br />
=41264.7<br />
9<br />
8<br />
7<br />
MC(i) [1000 t]<br />
6<br />
5<br />
4<br />
3<br />
2<br />
1<br />
0<br />
0 2 4 6 8 10 12 14 16 18 20<br />
Section i<br />
ØÃqŸdVy 5 ð é k$qmdXx?t*{ZqmÔ˜t˜c¤qma7bVtq jXÓ¯j i koc*txVk°ÔxZqm×zë|Ô½Ô˜k$Õvtq jXÓ Ô&q Ó × j ct §<br />
x V<br />
(T)=137.1523, x M<br />
=136.1293<br />
3.5 x 105 Section i<br />
1 x 104 Section i<br />
BM(i)<br />
3<br />
2.5<br />
2<br />
1.5<br />
1<br />
0.5<br />
0<br />
0 2 4 6 8 10 12 14 16 18 20<br />
SF(i)<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
0<br />
−0.2<br />
−0.4<br />
−0.6<br />
−0.8<br />
−1<br />
0 2 4 6 8 10 12 14 16 18 20<br />
ØÃqŸdVyºø ð £Ik Ó {Zq Ó d®_ j _§k Ó tÔq Ó × j ct §<br />
ØÃqŸdVy Y ð î4xVk$^Oc¦Û j c¤Õvk$Ôq Ó × j c¤t §
BM(i)<br />
Weight <strong>in</strong> t<br />
Weight <strong>in</strong> t<br />
Weight <strong>in</strong> t<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
¡ ¢¡¢<br />
£¡£¡£<br />
¤¡¤<br />
¡ ¢¡¢<br />
£¡£¡£<br />
¤¡¤<br />
¥¡¥¡¥<br />
¦¡¦<br />
¡<br />
=¡=<br />
>¡><br />
?¡?<br />
@¡@<br />
=¡=<br />
ö o ö ö o<br />
p¨p<br />
u¡u<br />
v¡v ö<br />
ö o ö ö o ö ö o ö ö o ö ö o<br />
p¨p<br />
s¡s<br />
t¡t<br />
u¡u<br />
v¡v<br />
ö<br />
q¡q<br />
r¡rs¡s<br />
t¡t<br />
£¡£¡£<br />
¤¡¤¡¤<br />
£¡£¡£<br />
§¨§¨§¨§<br />
©¨©¨©<br />
¡<br />
¡<br />
¡<br />
¡<br />
¡¡<br />
¡¡<br />
¡<br />
¡<br />
¡<br />
¡<br />
¡¡<br />
¡<br />
¡<br />
¡<br />
<br />
<br />
<br />
<br />
Ÿ¨Ÿ¨Ÿ¨Ÿ¨ ¨ ¢¡¢¡¢¡¡<br />
£¡£¡£<br />
¤¡¤¡¤<br />
§¨§¨§¨§<br />
§¨§¨§¨§<br />
©¨©¨©<br />
§¨§¨§¨§<br />
¡<br />
¡<br />
¡¡<br />
¡¡<br />
¡<br />
¡<br />
¡¡<br />
¡¡<br />
¡¡<br />
¡¡<br />
¡<br />
¡<br />
¡¡<br />
¡<br />
¡<br />
¡ <br />
¡¡<br />
¡<br />
"!"<br />
"!"<br />
#<br />
$#<br />
$<br />
%!%& '¨'¨'<br />
(¨(<br />
+¨+¨+<br />
,¨,<br />
/¡/¡/<br />
0¡0<br />
43<br />
7¡7<br />
8¡8<br />
9¡9¡9<br />
:¡:¡:<br />
;!;<br />
;!;<br />
;!;<br />
Weight <strong>in</strong> t<br />
Weight <strong>in</strong> t<br />
Weight <strong>in</strong> t<br />
1400<br />
1200<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
1400<br />
1200<br />
1000<br />
800<br />
600<br />
400<br />
200<br />
¡¡<br />
¡<br />
¡¡<br />
¡<br />
¡<br />
¡¡<br />
¡<br />
¡¡<br />
¡<br />
¡¡<br />
¡<br />
¡<br />
s¡s<br />
x¡x¡x<br />
y¡y¡y r¡r<br />
3¡3¡3<br />
9¡9 8¡8 2¡2¡2<br />
D¡D<br />
GFE¡E<br />
¨¨<br />
¨<br />
0 0 5 10 15 20<br />
0<br />
0<br />
¡<br />
¢¡¢¡¢ £¡£¡£<br />
¥¡¥¡¥ ¤¡¤¡¤<br />
§¨§¨§<br />
§¨§¨§<br />
©¨©<br />
©¨©<br />
§¨§¨§ ©¨©<br />
¡¡ ¡<br />
¡¡<br />
¡ ¡¡ ¡¡<br />
¡<br />
!¡!¡! "¡" #¡#<br />
$¡$<br />
%¡%<br />
©¨©<br />
&¡&¡&<br />
¦¡¦<br />
'¡'¡'<br />
,¡,¡, -¡-¡- .¡.¡. /¡/ 0¡0¡0 1¡1¡1 2¡2¡2 3¡3¡3<br />
(¡(¡(<br />
)¡)<br />
*¡*¡*<br />
+¡+<br />
6¡6<br />
D¡D<br />
7¡7<br />
:¡:<br />
E¡E<br />
;¡;<br />
4¡4<br />
5¡5<br />
>><br />
@¡@<br />
A¡A<br />
CCCCBBBBA¡A<br />
H¡H<br />
I¡I<br />
J¨J¨J<br />
K¨K<br />
L¡L<br />
M¡M<br />
N¡N<br />
O¡O<br />
P<br />
QP<br />
Q<br />
SR T<br />
T<br />
T<br />
U<br />
U<br />
VU<br />
WV<br />
W<br />
Ẍ Ẍ Ẍ X Ẍ Ẍ Ẍ X Ẍ Ẍ Ẍ X Ẍ Ẍ Ẍ X Ẍ Ẍ Ẍ X<br />
Ÿ Y Ÿ Y Ÿ Y Ÿ Y Ÿ Y<br />
Z¡Z¡Z<br />
[¡[¡[<br />
\¡\<br />
]¡]<br />
^¡^<br />
_¡_<br />
`¡`¡`<br />
a¡a b¡b¡b c¡c d¡d¡d e¡e f¡f¡f g¡g<br />
h¡h¡h<br />
i¡i<br />
j¡j<br />
k¡k<br />
l<br />
ml<br />
m<br />
'¡'¡' &¡&¡& ¡<br />
¡<br />
§¨§¨§ ©¨©<br />
¦¡¦ ¤¡¤¡¤ ¥¡¥¡¥<br />
£¡£¡£<br />
¤¡¤¡¤<br />
£¡£¡£ ¢¡¢¡¢ ¡ ¡<br />
¡<br />
¡<br />
p¡p q¡q r¡r<br />
s¡s<br />
¡<br />
v¡v¡v ¨<br />
w¡w¡w<br />
x¡x¡x<br />
y¡y¡y<br />
x¡x¡x<br />
¨¨¨<br />
€¡€<br />
‚¡‚<br />
„¨„¨„<br />
…¨…<br />
ƒ¡ƒ<br />
†¡†¡†<br />
‡¡‡¡‡<br />
ˆ¨ˆ¨ˆ¨ˆ<br />
‰¨‰¨‰<br />
Š¨Š¨Š¨Š<br />
‹¨‹¨‹<br />
Œ<br />
Œ<br />
Œ<br />
ŒŽ¡Ž<br />
¡<br />
‘¡‘<br />
’¡’<br />
“¡“ ”¡” •¡•<br />
–¨–¨–¨–<br />
—¨—¨—<br />
¡<br />
˜¡˜<br />
¡<br />
œ¡œ¡œ¡ ž¡ž<br />
Ÿ¡Ÿ<br />
š¡š¡š<br />
›¡›<br />
¨ ¨<br />
¡¨¡<br />
¢¡¢¡¢<br />
£¡£¡£<br />
¤¡¤¡¤<br />
¥¡¥¡¥¦¡¦ §¡§<br />
¬¡¬¡¬ ¡ ®¡®<br />
¯¡¯<br />
°¡°<br />
±¡±<br />
²¡²<br />
³¡³ ´¡´ µ¡µ<br />
¨¨¨<br />
·¨·<br />
ª¡ª¡ª<br />
«¡«<br />
¸¡¸¡¸<br />
¹¡¹<br />
˜¡˜ †¡†¡† ˆ¨ˆ¨ˆ¨ˆ ~¡~ ¡ {¨{ w¡w¡wz¨z¨z<br />
}¡}<br />
|¡|<br />
v¡v¡v ¨<br />
¨¨¨<br />
u¡u¡u ¡ t¡t¡t<br />
¡<br />
n¡n<br />
o¡o<br />
q¡q t¡t¡t u¡u¡u p¡p o¡o<br />
n¡n<br />
º¨º¨º<br />
»¨» ¼¡¼<br />
½¡½<br />
¾¡¾¡¾<br />
¿¡¿¡¿ º¨º¨º<br />
¿¡¿¡¿ À¡À Á¡Á<br />
¡Â<br />
áà ¾¡¾¡¾<br />
{¨{ †¡†¡† ‡¡‡¡‡ ˆ¨ˆ¨ˆ¨ˆ ‰¨‰¨‰ Œ z¨z¨z<br />
Ä¡Ä<br />
Å¡Å<br />
Æ¡Æ<br />
Ç¡Ç<br />
”¡”<br />
•¡•<br />
”¡”<br />
˜¡˜<br />
¡<br />
LEFT<br />
CENTER<br />
¢¡¢¡¢ £¡£¡£<br />
–¨–¨–¨– —¨—¨— ˜¡˜ ¡ š¡š¡š ›¡› ¨ ¨ ¡¨¡ ¢¡¢¡¢ £¡£¡£ ¤¡¤¡¤ ¥¡¥¡¥ ‘¡‘<br />
RIGHT<br />
Ø¡Ø¡Ø Ù¡Ù<br />
Ú¡Ú¡Ú<br />
Û¡Û<br />
Þ¡Þ¡Þ ß¡ß¡ßà¡à<br />
á¡áâ¡â<br />
ã¡ã<br />
ä¡ä å¡å<br />
ס×<br />
Ö¡Ö¡Ö<br />
¦¡¦<br />
§¡§<br />
5 10 15 20<br />
ë¨ë<br />
ì¡ì¡ì<br />
í¡í¡íî¡î<br />
ï¡ï ê¨ê¨ê<br />
¨ ¨¡¨¡¨ ©¡© ¨¨¨<br />
Ä¡Ä<br />
Æ¡Æ Ç¡Ç Å¡Å<br />
È¡È Å¡Å<br />
É¡É<br />
Ä¡Ä<br />
Ê¡Ê<br />
Ë¡Ë<br />
Ì¡Ì<br />
Í¡Í Î¡Î Ï¡Ï<br />
СÐ<br />
Ñ¡Ñ<br />
Ò¡Ò<br />
Ó¡Ó<br />
Ö¡Ö¡Ö ×¡×<br />
Ô¡Ô¡Ô<br />
Õ¡Õ<br />
Ü¡Ü¡Ü<br />
Ý¡Ý¡Ý Þ¡Þ¡Þ ß¡ß¡ß<br />
à¡à<br />
á¡á<br />
ä¡ä<br />
å¡å<br />
æ¡æ<br />
ç¡ç<br />
è¡è¡è<br />
é¡é¡é<br />
ê¨ê¨ê<br />
ë¨ë<br />
ì¡ì¡ì<br />
í¡í¡í<br />
î¡î<br />
ï¡ï<br />
ð¡ð<br />
ñ¡ñ ò¡ò<br />
ó¡ó ô¡ô õ¡õ ö¡ö ÷¡÷<br />
ø¡ø¡ø<br />
ù¡ù¡ù<br />
Ê¡Ê É¡É<br />
Å¡ÅÈ¡È<br />
Ä¡Ä<br />
0 0 5 10<br />
15<br />
È¡È É¡ÉÊ¡Ê<br />
Ë¡ËÌ¡Ì<br />
Í¡ÍСÐ<br />
Ñ¡Ñ Ô¡Ô¡Ô Õ¡Õ Ü¡Ü¡Ü Ý¡Ý¡Ýà¡à<br />
á¡áæ¡æ<br />
ç¡ç<br />
è¡è¡è<br />
é¡é¡éê¨ê¨ê<br />
ë¨ë<br />
ì¡ì¡ì<br />
í¡í¡í<br />
î¡î<br />
ï¡ïð¡ð<br />
ñ¡ñ ø¡ø¡ø ù¡ù¡ùú¡ú<br />
û¡û<br />
Å¡Å<br />
20<br />
Section i<br />
Section i<br />
Section i<br />
ØÃqŸdVy åoä4ð ¦ÉqmÔ˜t˜c¤qŸagbVtq jXÓËj ÛtxZk;Õ jXÓ t^`q Ó koc¤Ô j‚i koc&txVk×g^Oct˜a7^$n4Ôq Ó × j c¤t Y<br />
' &¥#¡,Ñ"+¨¤&¥ £¦¥Ð ¡ ÑÑ% !þý % ÿ<br />
ü<br />
Ó koÖ Õ jXÓ Õvko×Zt·Û j c·txVkœÔ˜t j ÖI^Od`k ×7lm^ ÓZÓ q Ó d›Û j cNÕ jXÓ t^`q Ó koc2ÔxZqm×7Ô¯t^Op4q Ó dHq Ó t j Õ jXÓ Ôqs{Vkoc¤^Otq jXÓ a j tx<br />
Þ<br />
jXÓ dXqŸtb7{Zq Ó ^`lVÔ˜t˜c¤k Ó d`tx¯×g^Oc¤^`_§kot˜koc¤Ô½^ Ó {l j dXqmÔ˜tqsÕ¦^`Ô˜×k$ÕvtÔIÖI^`Ô½q Ó t˜c j {ZbZÕvk${ yÃØZbVctxZkoc¤_ j ck`W4^<br />
txVkÉÔxZqŸ×Cë|Ô#l<br />
j {Vk$l4Û j c#txVk j ×Ztqm_¯^`lV{ZqmÔ˜t˜c¤qma7bVtq jXÓ°j Ûga7^`lslm^`Ô˜te”Ö½^Ot˜kocÖ½^`ÔM×Zck$Ô˜k Ó t˜k${ yîrk i koc¤^`l7^`Ô˜×k$ÕvtÔMÖqŸtx×Zc,^`ÕvtqmÕo^`l<br />
_<br />
i ^ Ó Õvk ÓVj t¾ǹ kot¾t^Op`k Ó q Ó t j ^`ÕoÕ j b Ó t ß tx7qŸc¤{›{7qm_§k Ó Ôq jXÓœj Û¦txVk‹ÔxZqm× ß xVk$qŸdXx–t,à,W#Õvc¤^ Ó kvefÔ×7lmqŸtoWMxZ^OtÕ,x<br />
ck$lŸk<br />
j i koc,ÔoWZ^ Ó { i qmÔqma7qmlmqŸtªnlmq Ó k‚àÃÔxZ^`lslVakck$^`lmqmáok${¯lm^Ot˜kocIa–n®^ Ó kvÝ4t˜k Ó {Vk${‹_ j {Vk$lyï Ó ^`{Z{ZqŸtq jXÓ W j bVcÆ×Zck$Ô˜k Ó t<br />
Õ<br />
j {Vk$lÔ¤xZ^`lml©ak2kvÝ4t˜k Ó {Vk${¢t j q Ó ÕolmbZ{Zk·ckokoÛQkoc,ÔË^ Ó {¢{Z^ Ó d`koc j b7Ô¯d jrj {7ÔoW©ø ä`ù efÕ jXÓ t^`q Ó koc¤Ô·^ Ó {)Ô˜×k$Õoqm^`l<br />
_<br />
jXÓ t^`q Ó koc¤Ôoy Õ<br />
j _§kœ_ j ck›Õvc¤qŸt˜koc,qm^Û j c2txVk›^`ÔÔ˜k$Ô¤Ô_§k Ó t j Û¾^ÅÔ˜t j Ö½^Od`k ×glm^ Ó xZ^ i k t j akœ{Vk i k$l j ×k${ y Y¦xZk Ó txVk<br />
î<br />
Ó7Ó kocMÕo^ Ó Õ¤x jrj Ô˜k^`_ jXÓ d;Ô˜k i koc¤^`lZÕvc,qŸt˜koc¤qm^ ß k`yºdVyMak Ó {Zq Ó d*_ j _®k Ó tÔoW–Õvc¤^ Ó kvefÔ˜×7lsqŸtoẀ kotÕOy¦à¾txVk¦Õvc¤qmt˜koc¤q jXÓ<br />
×7lm^<br />
Õvc¤qmt˜koc¤q jXÓ j bVc×7c j d`c¤^`_<br />
xVk×ZckoÛ6koc¤Ôoy±é)qmtx›txZqsÔ°×Zc¤koÛQkocc¤k${<br />
j {ZbZÕvk${‹arn j bVc_ j {Vk$ly<br />
×Zc<br />
ÖqmlmlÆÕ¤x jrj Ô˜k‹txVkak$Ô˜t×glm^ Ó ^`_ jXÓ d±tx j Ô˜k<br />
å
A Multi-Swarm Algorithm for Multi-Objective<br />
Ship Design Problems 1<br />
Antonio P<strong>in</strong>to, Emilio F. Campana, INSEAN - Italian Ship Model Bas<strong>in</strong>, Rome/Italy<br />
A.P<strong>in</strong>to@<strong>in</strong>sean.it, E.Campana@<strong>in</strong>sean.it<br />
Abstract<br />
A Multi-Objective Determ<strong>in</strong>istic Particle Swarm Optimization algorithm (MODPSO) is modified<br />
to avoid premature collapse of the swarm toward the center of the feasible design space, and<br />
to improve the search along the decision space bounds. MODPSO search is based on concept<br />
of Pareto dom<strong>in</strong>ance. In order to build an accurate Pareto front, after an <strong>in</strong>itial search, the<br />
orig<strong>in</strong>al swarm is decomposed <strong>in</strong> more small sub-swarms, select<strong>in</strong>g the particles on the basis<br />
of their distance from the <strong>in</strong>itial Pareto front. Each new sub-swarm has its own leader and<br />
all the subs-warms cooperate to f<strong>in</strong>d out the f<strong>in</strong>al Pareto front of the multiobjective problem.<br />
Effectiveness and efficiency of this approach are <strong>in</strong>vestigated by solv<strong>in</strong>g a set of test functions,<br />
which summarize most of the difficulties that could be encountered <strong>in</strong> multiobjective optimization,<br />
and by compar<strong>in</strong>g the results with that of two well-known Evolutionary Algorithms, namely<br />
the Nondom<strong>in</strong>ated Sort<strong>in</strong>g Genetic Algorithm and the Strength Pareto Evolutionary Algorithm.<br />
F<strong>in</strong>ally, MODPSO is used to solve a ship design problem, reduc<strong>in</strong>g the heave and pitch motion<br />
peaks of the Response Amplitude Operator of a conta<strong>in</strong>ership advanc<strong>in</strong>g at fixed speed <strong>in</strong> head<br />
seas.<br />
1 Introduction<br />
In many <strong>in</strong>dustrial design applications (naval, aeronautical, automotive) specific strategies have<br />
to be used <strong>in</strong> look<strong>in</strong>g for trade-off solution, rang<strong>in</strong>g from aggregated approaches to those <strong>in</strong>volv<strong>in</strong>g<br />
the concept of Pareto optimal solution of the multiobjective (MO) problem. Among the<br />
latter, the <strong>in</strong>terest of optimization community has been shift<strong>in</strong>g toward Evolutionary Algorithms<br />
(EAs) due to their eas<strong>in</strong>ess <strong>in</strong> deal<strong>in</strong>g simultaneously with a set of possible solutions, which <strong>in</strong><br />
turn allows to f<strong>in</strong>d several optimal solutions <strong>in</strong> a s<strong>in</strong>gle run <strong>in</strong>stead of runn<strong>in</strong>g a series of separate<br />
simulations. In comparison with other techniques, EAs are less <strong>in</strong>fluenced by the Pareto front<br />
shape, easy to implement and to parallelize. A recent survey of the exist<strong>in</strong>g EAs variants is<br />
provided by Coello (1999,2003).<br />
The development of efficient and fast MO codes is fundamental, especially when the design optimization<br />
requires the solution of partial differential equations (PDE), which are very expensive<br />
from the standpo<strong>in</strong>t of computational time. Also, the more the geometries are complex, the<br />
more the cost of simulation process <strong>in</strong>creases and if the optimizer is not efficient only a small<br />
part of the design space can be explored.<br />
Particle Swarm Optimization (PSO) is an EA technique <strong>in</strong>troduced by Kennedy and Eberhart<br />
(1995). S<strong>in</strong>ce it was proposed, it has received great attention from the optimization community,<br />
Parsopoulos and Vrahatis (2002a). PSO has been also successfully used to solve eng<strong>in</strong>eer<strong>in</strong>g<br />
problems, Shi and Eberhart (2001), Boer<strong>in</strong>ger and Werner (2004), demonstrat<strong>in</strong>g its ability<br />
<strong>in</strong> deal<strong>in</strong>g with non-convex feasible spaces. Application may be found on naval hydrodynamics<br />
(P<strong>in</strong>to et al. (2004,<strong>2005</strong>)), structural (Fourie and Groenwold (2000),2001), Fukuyama and<br />
Yoshida (2001), Schutte and Groenwold (2003)), aerodynamics (Ng et al. (2003)) and multidiscipl<strong>in</strong>ary<br />
optimization problems (aerodynamic and structural, Venter and Sobiezczanski-Sobieski<br />
1 This work has been supported by the M<strong>in</strong>istero delle Infrastrutture e dei Trasporti <strong>in</strong> the framework of the<br />
research plan “Programma di Ricerca sulla Sicurezza”, Decreto 17/04/2003 G.U. n. 123 del 29/05/2003.<br />
62
(2004)). Parallel implementations of the PSO algorithm, which allow to solve complex large-scale<br />
eng<strong>in</strong>eer<strong>in</strong>g optimization problems, are described <strong>in</strong> Baker et al. (2000), Schutte et al. (2004).<br />
Historically, the basic PSO algorithm was orig<strong>in</strong>ally def<strong>in</strong>ed for s<strong>in</strong>gle-objective unconstra<strong>in</strong>ed<br />
optimization problems but soon modified versions able to deal both with constra<strong>in</strong>ts, Parsopoulos<br />
and Vrahatis (2002b), and MO problems, Coello and Lechuga (2002), have appeared.<br />
In the present work, a s<strong>in</strong>gle objective PSO algorithm (the DPSO) is <strong>in</strong>itially illustrated as well<br />
as its MO counterpart, namely the MO Determ<strong>in</strong>istic Particle Swarm Optimization algorithm<br />
(MODPSO), Campana and P<strong>in</strong>to (<strong>2005</strong>)). A new version of the MO algorithm is then presented,<br />
modified to avoid the premature collapse of the swarm toward the center of the feasible space<br />
and to improve the search along the decision space bounds. After an <strong>in</strong>itial search the orig<strong>in</strong>al<br />
swarm of particles is decomposed <strong>in</strong> more small sub-swarms, depend<strong>in</strong>g on the distance of each<br />
particle from the current Pareto solutions. In this way, each sub-swarm has its own guide and<br />
all collaborate to build the Pareto front. The f<strong>in</strong>al aim is to reduce the number of objective<br />
function evaluations needed to build the Pareto front improv<strong>in</strong>g the accuracy of the Pareto front<br />
reconstruction at the same time.<br />
A systematic analysis of the proposed MODPSO formulation on a set of algebraic test problems<br />
is <strong>in</strong>itially carried out. Tan et al. (2003) and Boer<strong>in</strong>ger and Werner (2004) tested swarm<br />
algorithms by compar<strong>in</strong>g it with different EAs <strong>in</strong> specific applications. Here the validation<br />
is performed by us<strong>in</strong>g algebraic test functions that summarize most of the difficulties which<br />
could be encountered by EAs, Deb (1999). Then MODPSO is compared with other well-known<br />
EAs, follow<strong>in</strong>g a comparison test proposed by Zitzler et al. (2000). The Nondom<strong>in</strong>ated Sort<strong>in</strong>g<br />
Genetic Algorithm (NSGA), Sriv<strong>in</strong>as and Deb (1995), and the Strength Pareto Evolutionary<br />
Algorithm (SPEA), Zitzler et al. (1999), are used for the success demonstrated <strong>in</strong> their category<br />
on the same set of test functions, Zitzler et al. (2000).<br />
F<strong>in</strong>ally, an optimum ship design application with functional and geometrical constra<strong>in</strong>ts is given,<br />
namely the m<strong>in</strong>imization of the heave and pitch motion peaks of the Response Amplitude Operator<br />
(RAO) of a ship (conta<strong>in</strong>ership S175) advanc<strong>in</strong>g at a fixed speed <strong>in</strong> head seas.<br />
2 Set-up of a MO problem<br />
Let us consider a system whose behavior is described by a set of equations, assume that its<br />
status depends by a N-dimensional design vector ¯x = [x 1 , x 2 , ..., x N ] T and let’s <strong>in</strong>troduce the<br />
follow<strong>in</strong>g bounds constra<strong>in</strong>ts:<br />
b m<strong>in</strong><br />
i<br />
≤ x i ≤ b max<br />
i , i = 1, ..., N (1)<br />
The application of bounds <strong>in</strong> equation (1) def<strong>in</strong>es a parallelepiped Π <strong>in</strong> the N-dimensional design<br />
variables space. In a similar way, functional constra<strong>in</strong>ts are also imposed, assum<strong>in</strong>g that<br />
c m<strong>in</strong><br />
j<br />
≤ g j (¯x) ≤ c max<br />
j , j = 1, ..., l (2)<br />
where the g j (¯x) is a generic function of the design variables vector ¯x. The constra<strong>in</strong>ts <strong>in</strong> equation<br />
(2), applied to Π, def<strong>in</strong>e the shape of the feasible solution set X <strong>in</strong> the N-dimensional design<br />
space, i.e. X ⊂ Π is the set of design solutions that satisfy all the conditions. The vector function<br />
¯f(¯x) = [f 1 (¯x), f 2 (¯x), ..., f M (¯x)] T <strong>in</strong>cludes all the objective functions of the problem, which we<br />
want to optimize simultaneously. The MO problem may be f<strong>in</strong>ally formulated as follows 2 :<br />
m<strong>in</strong>imize<br />
for<br />
¯f(¯x)<br />
¯x ∈ X<br />
2 The case of the m<strong>in</strong>imization of all the objective functions is here depicted, the maximization of the generic<br />
¯f(¯x) be<strong>in</strong>g equivalent to the m<strong>in</strong>imization of − ¯f(¯x).<br />
(3)<br />
63
e<strong>in</strong>g f k : R N → R.<br />
Consider now the above mentioned MO problem with M objectives, f k (¯x), and N decision<br />
variables, x i . The solution of the MO problem is given by the vector ¯x ∗ = [x ∗ 1 , x∗ 2 , ..., x∗ N ]T , which<br />
satisfy the constra<strong>in</strong>ts (1) and (2) and yields the optima values of all the objective functions.<br />
Unfortunately, this result can not be achieved <strong>in</strong> typical MO problems, s<strong>in</strong>ce usually does not<br />
exist a s<strong>in</strong>gle solution that is optimum with respect to every objective function, Miett<strong>in</strong>en (1999).<br />
In fact, objective functions may be easily <strong>in</strong> conflict with<strong>in</strong> each other. Let therefore <strong>in</strong>troduce<br />
two vectors ū = [u 1 , u 2 , ..., u L ] T and ¯v = [v 1 , v 2 , ..., v L ] T . We can say that a vector ū dom<strong>in</strong>ates<br />
a vector ¯v (also written as ū ≽ ¯v) if and only if u i ≤ v i , i = 1, 2, ..., L and there exists at least an<br />
<strong>in</strong>dex with u i < v i . Thus, a solution ¯x ∗ of the MO problem is said to be Pareto optimal if and<br />
only if does not exist another solution ¯x ′ so that ¯f(¯x ∗ ) is dom<strong>in</strong>ated by ¯f(¯x ′ ). The decision vector<br />
¯x ∗ , correspond<strong>in</strong>g to a solution <strong>in</strong>cluded <strong>in</strong> the Pareto optimal set P ∗ , is called non-dom<strong>in</strong>ated.<br />
The Pareto front is def<strong>in</strong>ed as follow:<br />
PF ∗ = {(f 1 (¯x), f 2 (¯x), ..., f M (¯x)) | ¯x ∈ P ∗ } (4)<br />
3 Determ<strong>in</strong>istic Particle Swarm Optimization - DPSO<br />
S<strong>in</strong>ce the pioneer<strong>in</strong>g work of Kennedy and Eberhart (1995), PSO simulates the social behavior of<br />
a group of <strong>in</strong>dividuals by shar<strong>in</strong>g <strong>in</strong>formation among them while they are explor<strong>in</strong>g the design<br />
space. Each particle of the swarm has its own (<strong>in</strong>dividual) memory to remember the places<br />
visited dur<strong>in</strong>g the exploration, whereas the swarm has its own (collective) memory, to memorize<br />
the best locations ever visited by any of the particles. The particles have an adaptable velocity<br />
and <strong>in</strong>vestigate the design space analyz<strong>in</strong>g their own fly<strong>in</strong>g experience, and the one of all the<br />
particles of the swarm. Each particle is a potential solution of the optimization problem under<br />
consideration. DPSO, Campana and P<strong>in</strong>to (<strong>2005</strong>), is a determ<strong>in</strong>istic version of the PSO for<br />
constra<strong>in</strong>ed s<strong>in</strong>gle objective problems, recalled here for sake of completeness.<br />
Step 0. (Initialize) Distribute a limited number of particles on the decision variables space<br />
bounds.<br />
Step 1. (Compute velocity) For each particle i calculate a velocity vector v i us<strong>in</strong>g the particle’s<br />
memory and the knowledge ga<strong>in</strong>ed by the swarm.<br />
– If constra<strong>in</strong>ts are satisfied then:<br />
– else if constra<strong>in</strong>ts are violated:<br />
vi n+1 = χ [w n vi n + c 1 (p n i − xn i ) + c 2 (p n b − xn i )] (5)<br />
v n+1<br />
i<br />
= χ [ c ′ 2 (p n b − xn i ) ] (6)<br />
where χ is a constriction factor on the particles speed, w is called <strong>in</strong>ertia weight, c 1 , c 2<br />
and c ′ 2 are positive constants, pn i is the best position ever found by the particle i and p n b is<br />
the best position ever discovered by the swarm at step n.<br />
Step 2. (Update position) Update the position of each particle x i us<strong>in</strong>g the velocity vector and<br />
the previous position:<br />
x n+1<br />
i<br />
= x n i + vn+1 i<br />
(7)<br />
64
Step 3. (Check convergence) Go to Step 1 and repeat until some convergence criterion is<br />
matched.<br />
The <strong>in</strong>ertia parameter w regulates the trade-off between global (wide-rang<strong>in</strong>g) and local (nearby)<br />
exploration abilities of the swarm. A large <strong>in</strong>ertia weight facilitates global exploration (search<strong>in</strong>g<br />
new areas), while a small one tends to facilitate local exploration, i.e. f<strong>in</strong>e-tun<strong>in</strong>g of the current<br />
search area. In order to have good accuracy on both global an local search, an <strong>in</strong>itial value for<br />
w is <strong>in</strong>itially set, and then it is decreased accord<strong>in</strong>g to a given decrease rate (8). At the step n:<br />
w n = w n−1 K w (8)<br />
where w n−1 is the previous value for the <strong>in</strong>ertia weight and K w is the decreas<strong>in</strong>g rate (K w < 1).<br />
F<strong>in</strong>e tun<strong>in</strong>g of parameters <strong>in</strong> Eq.(5) is crucial for the optimization process, Shi and Eberhart<br />
(1998), Parsopoulos and Vrahatis (2002a). Indeed, f<strong>in</strong>al solution and comput<strong>in</strong>g time are strictly<br />
l<strong>in</strong>ked to these parameters sett<strong>in</strong>g. A f<strong>in</strong>e calibration of these parameters was suggested by<br />
Carlisle and Dozier (2001), Venter and Sobieszczanski-Sobieski (2003), <strong>in</strong> order to speed up the<br />
overall process and to reduce the risks of be<strong>in</strong>g trapped by local m<strong>in</strong>ima. Clerc and Kennedy<br />
(2002) describe the l<strong>in</strong>k between the dynamic of the swarm and its parameters.<br />
3.1 Enhancement to the DPSO Algorithm<br />
This section focuses on the development of an enhanced DPSO version. The proposed algorithm<br />
<strong>in</strong>corporates a modification to avoid the premature collapse of the swarm toward the center of<br />
the feasible space, and to improve the search along the decision space bounds, when an <strong>in</strong>itial<br />
face distribution is selected, Campana et al. (<strong>2005</strong>).<br />
In the attempt of keep<strong>in</strong>g low the number of objective function evaluations necessary to sample<br />
the design space, the size of the swarm is reduced as much as possible. Hence, <strong>in</strong> the attempt of<br />
not exclud<strong>in</strong>g a priori any region of the feasible design space, particles are <strong>in</strong>itially distributed<br />
on along the bounds, Fig.1. However, accord<strong>in</strong>g to Eqs.(5) and (7) each particle is attracted by<br />
the others. As a consequence, they can not escape from the dashed region <strong>in</strong>dicated <strong>in</strong> Fig.1,<br />
i.e. the corners of the design space cannot be reached by any particle of the swarm. A possible<br />
strategy could have been to <strong>in</strong>crease the <strong>in</strong>ertia, allow<strong>in</strong>g therefore the particles to move with<br />
more freedom. This solution is however <strong>in</strong>efficient, lead<strong>in</strong>g to frequent violation of the constra<strong>in</strong>ts<br />
and to a non exhaustive prob<strong>in</strong>g of the boundary regions.<br />
In order to force the <strong>in</strong>itial design exploration along the boundaries of the feasible space we then<br />
<strong>in</strong>troduce a dynamically chang<strong>in</strong>g speed limiter on that component of the particle’s velocity<br />
directed toward the center of the feasible space, <strong>in</strong> the form:<br />
|v n+1<br />
i<br />
| ≤ |bmax i<br />
− b m<strong>in</strong><br />
i |<br />
(9)<br />
g k<br />
This threshold value is hence a function of the distance between the bounds of the whole design<br />
space and of parameter g k . By chang<strong>in</strong>g g k , the <strong>in</strong>ner speed limiter (9) may be relaxed after<br />
a given number of iterations. Numerical tests have proved that by impos<strong>in</strong>g condition (9) it is<br />
possible to explore the corners of the design space avoid<strong>in</strong>g the premature collapse of the swarm<br />
toward the center of the feasible space, and f<strong>in</strong>d the correct global m<strong>in</strong>imum, Fig.1. The path<br />
of the four swarm particles for a n = 2 test function (Griewank) is depicted with and without<br />
condition (9). The dashed (<strong>in</strong>ner) region is obta<strong>in</strong>ed by jo<strong>in</strong><strong>in</strong>g together the <strong>in</strong>itial position<br />
of the particles. Without the speed limiter (a), particles are attracted each other and rema<strong>in</strong><br />
65
conf<strong>in</strong>ed <strong>in</strong>side the <strong>in</strong>ner region, while us<strong>in</strong>g it the swarm f<strong>in</strong>ds the right optimum which lay<br />
outside the dashed region.<br />
Global<br />
M<strong>in</strong>imum<br />
a<br />
b<br />
Fig.1: The path of four swarm particles for a n = 2 test function (Griewank). Without the<br />
speed limiter (a), particles are attracted each other and tend to be conf<strong>in</strong>ed <strong>in</strong>side the dashed<br />
region. The use of the normal speed limiter (b) allows the particles to explore the design space<br />
outside the <strong>in</strong>ner region and to f<strong>in</strong>d the global optimum.<br />
4 A Sub-Swarm Strategy for Multiobjective Optimization<br />
In Campana and P<strong>in</strong>to (<strong>2005</strong>), the MODPSO formulation was presented and numerical tests<br />
demonstrated its ability <strong>in</strong> solv<strong>in</strong>g MO problems, localiz<strong>in</strong>g the real Pareto front and produc<strong>in</strong>g<br />
a large number of solutions on the frontier. For sake of completeness the MODPSO algorithm<br />
is here briefly recalled.<br />
The algorithm adopts a decomposition approach of the <strong>in</strong>itial swarm S choos<strong>in</strong>g different guides<br />
(the Pareto solutions) for the particles of the small sub-swarms. Hence, for the i th particle of the<br />
swarm the value of p b <strong>in</strong> the collective term (Step 1) is assumed now to be the nearest po<strong>in</strong>t on<br />
the Pareto front. The <strong>in</strong>itial swarm S is hence decomposed <strong>in</strong>to smaller sub-swarms, each one<br />
follow<strong>in</strong>g a different guides. As a consequence, the i th particle moves toward an optimum po<strong>in</strong>t<br />
which could be different from the j th particle’s optimum. The <strong>in</strong>dividual velocity term <strong>in</strong> Step<br />
1 (p n i − xn i ) is obta<strong>in</strong>ed by choos<strong>in</strong>g as p i the m<strong>in</strong>imum of function g def<strong>in</strong>ed <strong>in</strong> Eq.(10) as <strong>in</strong> the<br />
Conventional Weighted Aggregation approach (CWA), Parsopoulos and Vrahatis (2002a).<br />
g =<br />
Here W j is a given distribution of weights for the l<strong>in</strong>ear comb<strong>in</strong>ation.<br />
M∑<br />
W j f j (¯x) (10)<br />
j=1<br />
For all the n sw particles of the <strong>in</strong>itial swarm S (<strong>in</strong>clud<strong>in</strong>g those currently on the Pareto front),<br />
the fundamental steps of the algorithm can be summarized as follows:<br />
Step I. (Distance evaluation) the i th particle evaluates its distance, <strong>in</strong> the design space, from<br />
all the K Pareto po<strong>in</strong>ts (¯x ∈ P ∗ ) known at the current iteration;<br />
Step II. (Guide selection) the i th particle selects the closest Pareto optimal solution as guide,<br />
p k,b ;<br />
66
By apply<strong>in</strong>g this procedure at each iteration the orig<strong>in</strong>al swarm S is divided <strong>in</strong>to k new smaller<br />
sub-swarms S k , where k ≤ K, be<strong>in</strong>g K the current number of the Pareto optimal solutions.<br />
Therefore, each sub-swarm S k has a guide which is different from all the other k − 1 sub-swarms<br />
guides, Fig.2. Once selected all the guides p k,b , each sub-swarm follows its own guide by us<strong>in</strong>g<br />
Eq.(5).<br />
f 2<br />
f 1<br />
i=1<br />
S 1<br />
S<br />
i=2<br />
S 2<br />
p 1,b<br />
i=3<br />
p 3,b<br />
i=4<br />
S 3<br />
p 2,b<br />
i=5<br />
p b<br />
Fig.2: S is the orig<strong>in</strong>al swarm and S k are the new sub-swarms; black dots <strong>in</strong>dicate the Pareto<br />
optimal solutions and white dots are the current positions of the particles; p k,b is the best particle<br />
of the sub-swarm S k ; <strong>in</strong> this case k = 3 and K = 4.<br />
5 Multiobjective Test Functions<br />
The key po<strong>in</strong>t <strong>in</strong> this paper is to compare MODPSO with exist<strong>in</strong>g, well assessed algorithms.<br />
In particular numerical comparisons are presented among MODPSO and two well-known MO<br />
evolutionary algorithms: the Nondom<strong>in</strong>ated Sort<strong>in</strong>g Genetic Algorithm (NSGA), Sr<strong>in</strong>ivas and<br />
Deb (1995), and the Strength Pareto Evolutionary Algorithm (SPEA), Zitzler et al. (1999). The<br />
choice of these two algorithms among many available is mostly based on the results presented<br />
<strong>in</strong> Zitzler et al. (2000), where a systematic comparison was carried out and NSGA resulted to<br />
be the best Genetic Algorithm (GA), while SPEA demonstrated to be the best EA algorithm<br />
among those tested. All the numerical data of this comparison are available on the web 3 . We<br />
anticipate here that MODPSO compares surpris<strong>in</strong>gly well aga<strong>in</strong>st those two referred algorithms,<br />
<strong>in</strong> particular giv<strong>in</strong>g a more <strong>in</strong>terest<strong>in</strong>g Pareto front with respect to NSGA and about the same<br />
front of the one f<strong>in</strong>d out by SPEA but with about one order of magnitude less objective function<br />
evaluations.<br />
In order to evaluate effectiveness and efficiency of MODPSO, simulations are carried out on<br />
a number of different algebraic problems commonly used to test multiobjective optimization<br />
algorithms, e.g. Zitzler et al. (2000), J<strong>in</strong> et al. (2001), Parsopoulos and Vrahatis (2002a), which<br />
provide a wide spectrum of different MO tasks, Deb (1999).<br />
A set of three MO test problems is <strong>in</strong>itially def<strong>in</strong>ed, all hav<strong>in</strong>g two objective functions, f 1 (¯x)<br />
and f 2 (¯x). Each problem is then solved for two different set of design variables (N = 2 and<br />
N = 30) to look for scale effects. Test functions T 1 , T 3 and T 5 are solved with N = 2 while<br />
N = 30 is used for the test functions T 2 , T 4 and T 6 . The def<strong>in</strong>ition of the algebraic functions are<br />
the follow<strong>in</strong>g, Zitzler et al. (2000):<br />
3 http://www.tik.ee.ethz.ch/∼zitzler/testdata.html<br />
67
• Test functions T 1 and T 2 have a convex Pareto optimal front:<br />
f 1 (¯x) = x 1 (11)<br />
(<br />
f 2 (¯x) = H 1 − √ )<br />
f 1 (¯x)/H<br />
(12)<br />
• Test functions T 3 and T 4 have a nonconvex Pareto optimal front :<br />
f 1 (¯x) = x 1 (13)<br />
f 2 (¯x) = H ( 1 − (f 1 (¯x)/H) 2) (14)<br />
• Test functions T 5 and T 6 have a Pareto optimal front consist<strong>in</strong>g of several noncontiguous<br />
convex parts:<br />
f 1 (¯x) = x 1 (15)<br />
(<br />
f 2 (¯x) = H 1 − √ )<br />
f 1 (¯x)/H − (f 1 (¯x)/H)s<strong>in</strong>(10πf 1 (¯x))<br />
(16)<br />
It is evident that test functions T 3 and T 4 represent the nonconvex counterpart to T 1 and T 2 ,<br />
while T 5 and T 6 have discrete features. In all test problems, function H is def<strong>in</strong>ed as follow:<br />
H = 1 + 9<br />
N − 1<br />
N∑<br />
x i (17)<br />
and all the exact Pareto optimal fronts are given by H = 1. Be<strong>in</strong>g focused on MO problems with<br />
expensive objective functions, we decided to l<strong>in</strong>k the maximum number of functions evaluations<br />
to the scale of the problem by impos<strong>in</strong>g a limit of 100N. Bounds on the design variables are<br />
(0 ≤ x i ≤ 1). Try<strong>in</strong>g to keep low the number of particles <strong>in</strong> the swarm (and hence the number<br />
of function evaluations) we selected an <strong>in</strong>itial distribution composed by only 4N + 1 particles,<br />
with just two particles per each hyper-face plus one particle <strong>in</strong> the middle of the design variables<br />
space. Their <strong>in</strong>itial velocity was set equal to zero. The set of MODPSO parameters needed <strong>in</strong><br />
(5) and (8) used to solve the six test functions are reported <strong>in</strong> Table I.<br />
i=2<br />
Table I: MODPSO parameters<br />
Constriction coefficient (χ) : 1.0<br />
Initial <strong>in</strong>ertia weight value (w 0 ) : 1.4<br />
Decreas<strong>in</strong>g coefficient for the <strong>in</strong>ertia weight (K w ) : 0.975<br />
Cognitive parameter (c 1 ) : 1.0<br />
Social parameter (c 2 ) : 0.8<br />
Constra<strong>in</strong>ed social parameter (c ′ 2 ) : 0.75<br />
Swarm size (n sw ) : 4N + 1<br />
Max number of objective functions evaluations : 100N<br />
The comparison is carried out for all N = 30 cases, i.e. T 2 , T 4 and T 6 .<br />
In Zitzler et al. (2000), both NSGA and SPEA were executed 30 times on each test problem, and<br />
each simulation was carried out us<strong>in</strong>g the parameters reported <strong>in</strong> Table II. Details concern<strong>in</strong>g<br />
the values of the parameters and implementations schemes are described <strong>in</strong> Zitzler et al. (2000).<br />
Table II: NSGA and SPEA parameters, Zitzler et al. (2000)<br />
Number of generations : 250<br />
Population size : 100<br />
Crossover rate : 0.8<br />
Mutation rate : 0.01<br />
Nich<strong>in</strong>g parameter σ share : 0.48862<br />
Dom<strong>in</strong>ation pressure t dom : 10<br />
68
T 2<br />
P * = 120<br />
54<br />
4<br />
100N<br />
P * = 167<br />
5<br />
4<br />
T1 2<br />
f 2<br />
3<br />
3<br />
2<br />
2<br />
T 1<br />
f 1<br />
f 2<br />
3<br />
2<br />
1<br />
1<br />
0<br />
1<br />
0.25 0.5 0.75 0 0.2 0.4 f 0.6 0.8 1<br />
1<br />
0<br />
0 0.2 0.4 0.6 0.8 1<br />
5<br />
4<br />
3<br />
T 3<br />
P * = 171<br />
5<br />
4<br />
3<br />
T 4<br />
P * = 141<br />
f 2<br />
2<br />
f 2<br />
2<br />
1<br />
0<br />
0 0.2 0.4 0.6 0.8 1<br />
f 1<br />
1<br />
0<br />
0 0.2 0.4 0.6 0.8 1<br />
f 1<br />
5<br />
4<br />
T 5<br />
P * = 47<br />
5<br />
4<br />
T 6<br />
P * = 116<br />
f 2<br />
3<br />
2<br />
f 2<br />
3<br />
2<br />
1<br />
0<br />
-1<br />
0 0.2 0.4 0.6 0.8 1<br />
f 1<br />
1<br />
0<br />
-1<br />
0 0.2 0.4 0.6 0.8 1<br />
f 1<br />
Fig.3: Test functions T 1 (N = 2) and T 2 (N = 30) have a convex Pareto front; T 3 (N = 2) and<br />
T 4 (N = 30) have a non-convex Pareto front; T 5 (N = 2) and T 6 (N = 30) have a discrete Pareto<br />
front; P ∗ is the number of Pareto optimal solutions found by MODPSO after 100N objective<br />
functions evaluations.<br />
69
5.1 Test Functions Results<br />
Fig.3 shows numerical results for the solution of the six multi-objective tests. The solutions<br />
belong<strong>in</strong>g to the Pareto optimal set (empty circles) are compared with the real Pareto fronts<br />
(black l<strong>in</strong>es) formed with H = 1. For each test the number of Pareto optimal solutions P ∗ is<br />
also reported.<br />
MODPSO technique reproduces with good accuracy the real Pareto front for all the tests with<br />
N = 2 design variables, i.e. T 1 , T 3 and T 5 . In particular, it is able to pick up all the branches<br />
of the discont<strong>in</strong>uous Pareto front <strong>in</strong> test problem T 5 . The number of Pareto optimal solutions,<br />
except <strong>in</strong> test function T 5 , is always greater than 100, giv<strong>in</strong>g a good resolution of the obta<strong>in</strong>ed<br />
Pareto fronts. By <strong>in</strong>creas<strong>in</strong>g the complexity of the problem, i.e. for the higher number of design<br />
variables ( N = 30) the results are still very satisfy<strong>in</strong>g. A reduction <strong>in</strong> the number of Pareto<br />
po<strong>in</strong>ts on the fronts is observable for the case T 2 and T 4 but the resolution of the calculated<br />
Pareto front is still good. In test problem T 6 the solution is even better than that of the case of<br />
two design variables (T 5 ).<br />
1.54<br />
1.54<br />
T 4<br />
13<br />
T 2<br />
T1 1<br />
MODPSO = 120<br />
100N<br />
NSGA = 45<br />
SPEA = 144<br />
13<br />
100N<br />
T1 2<br />
f 2<br />
T1 2<br />
f 2<br />
0.52<br />
01<br />
0.25 0.5 0.75 0 0.2 0.4 0.6 0.8 1<br />
T1 1<br />
f 1<br />
0.52<br />
MODPSO = 141<br />
NSGA = 30<br />
SPEA = 76<br />
01<br />
0.25 0.5 0.75 0 0.2 0.4 f 0.6 0.8 1<br />
1<br />
T1 2<br />
f 2<br />
1.54<br />
1<br />
3<br />
0.5<br />
T 6<br />
MODPSO = 116<br />
NSGA 100N = 63<br />
SPEA = 115<br />
0<br />
2<br />
-0.5<br />
1<br />
0.25 0.5 0.75 1<br />
0 0.4<br />
T1f 0.6 0.8<br />
1<br />
Fig.4: Comparative results on test function T 2 , T 4 and T 6 (N = 30)<br />
NSGA and SPEA were compared for the case N = 30. Accord<strong>in</strong>g to Zitzler et al. (2000) per<br />
each test function the outcomes of the first five runs are stored together and then only the<br />
non-dom<strong>in</strong>ated solutions are plotted and compared with the Pareto fronts found by MODPSO,<br />
Fig.4. In test problems T 2 and T 4 , MODPSO was the only method capable to localize the<br />
global m<strong>in</strong>imum of functions f 2 . Compar<strong>in</strong>g both the Pareto solutions as well as their number<br />
MODPSO and SPEA solutions are better than those obta<strong>in</strong>ed with NSGA <strong>in</strong> all the cases<br />
analyzed.<br />
70
¿From the standpo<strong>in</strong>t of the method’s efficiency however, it must be underl<strong>in</strong>ed that the number<br />
of objective function evaluation used by MODPSO <strong>in</strong> f<strong>in</strong>d<strong>in</strong>g the Pareto front was only 3000,<br />
and this has to be compared to the 25000 function evaluations required by NSGA and SPEA<br />
methods <strong>in</strong> each one of the three test problems. We can therefore conclude that the efficiency<br />
of MODPSO <strong>in</strong> the solution of these MO tests is about 24 and 10 times higher than those of<br />
NSGA and SPEA, respectively.<br />
6 MODPSO test for seakeep<strong>in</strong>g improvement<br />
In order to show the capabilities of the MODPSO procedure <strong>in</strong> solv<strong>in</strong>g MO design problems,<br />
conta<strong>in</strong>ership S175 has been selected as test case. The goal is the m<strong>in</strong>imization of the heave<br />
and pitch motion peaks of the RAO, while the ship is advanc<strong>in</strong>g <strong>in</strong> head seas at the speed<br />
of 16 knots (Fr = 0.198). The optima solutions are searched for non-dimensional frequencies<br />
ω 0 / √ L pp /g > 0.4. For the evaluation of the RAO’s heave and pitch motion, a standard striptheory<br />
approach has been applied. The method is widely accepted <strong>in</strong> prelim<strong>in</strong>ary design; 25<br />
sections, each one with 10 segments, have been used to describe the hull, accord<strong>in</strong>g to the<br />
limitations of the adopted code, Meyers et al. (1981). However, <strong>in</strong> the attempt of <strong>in</strong>creas<strong>in</strong>g<br />
the freedom of the optimizer, sections are not equally spaced along the hull, be<strong>in</strong>g <strong>in</strong>stead<br />
stretched <strong>in</strong> the control region R c of the optimization problem, Meyers and Baitis (1985). As a<br />
consequence, dur<strong>in</strong>g the optimization cycle, up to seven sections are used <strong>in</strong> the bow description<br />
and are allowed to move from the orig<strong>in</strong>al position.<br />
To modify the control region R c , a perturbation approach, previously developed, Peri et al.<br />
(2001), was adopted here: a surface patch – a Bezier polynomial surface – is superimposed to<br />
the orig<strong>in</strong>al hull shape. The ship’s geometry is modified only <strong>in</strong> the transversal direction by<br />
us<strong>in</strong>g a Bezier patch B y applied to the hull. As a consequence, the keel l<strong>in</strong>e is left unchanged<br />
dur<strong>in</strong>g the optimization process.<br />
The f<strong>in</strong>al design should satisfy some predef<strong>in</strong>ed constra<strong>in</strong>ts, geometrical and functional. In this<br />
test case, three different constra<strong>in</strong>ts have been adopted. For the displacement ∆ and the beam<br />
B, a range has been def<strong>in</strong>ed:<br />
2400t ≤ ∆ ≤ 2460t (18)<br />
25m ≤ B ≤ 26m (19)<br />
These two constra<strong>in</strong>ts are nonl<strong>in</strong>ear and <strong>in</strong> the present work they are treated as a black-box.<br />
Also, bounds are applied directly on the design variables x i . With reference to the def<strong>in</strong>ition<br />
(1), m<strong>in</strong>imum and maximum values have been imposed:<br />
b m i = −20.0, b M i = 20.0, i = 1, ..., 6 (20)<br />
To deal with the constra<strong>in</strong>ts <strong>in</strong>feasible solutions were rejected by assigned a high-function value<br />
and no further objective function evaluation are required until the particles come back <strong>in</strong>to the<br />
feasible design, see Campana and Peri (<strong>2005</strong>) for details. For this design problem the maximum<br />
number of objective functions evaluations has been limited to 600 (100N). The <strong>in</strong>itial particles<br />
distribution has been selected perform<strong>in</strong>g a LPτ-net, Statnikov and Matusov (1995), distribution<br />
of 65000 trial po<strong>in</strong>ts, but only 25 (4N + 1) feasible po<strong>in</strong>ts have been chosen from the above<br />
mentioned distribution as start<strong>in</strong>g po<strong>in</strong>ts. As a consequence the swarm size is equal to 25. The<br />
parameters utilized to solve the conta<strong>in</strong>ership S175 design problem are summarized <strong>in</strong> Table I.<br />
71
7 Numerical Results<br />
MODPSO has been used to perform the seakeep<strong>in</strong>g optimization of conta<strong>in</strong>ership S175. The<br />
detected Pareto front is shown <strong>in</strong> Fig.5. RAOs heave and pitch motion peaks are respectively<br />
reported on x and y nondimensionalised by the values of the orig<strong>in</strong>al S175 hull shape. The<br />
number of Pareto optimal solutions, is 25. Three solutions from the Pareto front correspond<strong>in</strong>g<br />
to RAOs heave and pitch motion peaks lowest values, O #1 and O #2 respectively, and the po<strong>in</strong>t<br />
with the best trade-off value between these two optimal po<strong>in</strong>ts, O #3 are selected to compare<br />
their seakeep<strong>in</strong>g performances, Fig.6, and the optimized ship hull geometries, Fig.7.<br />
PITCH *<br />
1<br />
0.95<br />
0.9<br />
O #1 MODPSO (P * = 25)<br />
0.85 O #3<br />
O #2<br />
0.8<br />
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1<br />
HEAVE *<br />
Fig.5: Pareto optimal solutions achieved from the multi-objective optimization of S175<br />
Fig.6 compares among the RAOs of the orig<strong>in</strong>al hull shape and the three optimized ones. On the<br />
horizontal axis the non-dimension frequency ω 0 / √ L pp /g is reported, while on the vertical axis<br />
the amplitude of the RAOs heave and pitch motions is plotted. As to the RAO’s heave motion<br />
curves, it is possible to see that <strong>in</strong> the optimized configurations O #1 the peak is absent. In case<br />
of O #3 , the peak is quite dumped if compared with the orig<strong>in</strong>al one, while it is still present <strong>in</strong><br />
the configuration O #2 . On the contrary, as to the RAO’s pitch motion curves, O #2 reduce most<br />
the peak. Table III summarizes the seakeep<strong>in</strong>g characteristics for the selected po<strong>in</strong>ts, <strong>in</strong> term of<br />
RAOs heave and pitch motion peak values, and improvement achieved by the three optimized<br />
hulls.<br />
A h<br />
RAOs (heave motion)<br />
0.7<br />
RAOs (pitch motion)<br />
A p<br />
ω 0<br />
/(L pp<br />
/g) 0.5<br />
1.2 Orig<strong>in</strong>al<br />
1<br />
0.8<br />
O #1<br />
O #2<br />
0.6<br />
0.5<br />
0.4<br />
Orig<strong>in</strong>al<br />
O #1<br />
O #2<br />
O #3<br />
O #3 (L pp<br />
/g) 0.5<br />
0.6<br />
0.4<br />
0.2<br />
0.3<br />
0.2<br />
0.1<br />
0<br />
0.2 0.4 0.6 0.8 1<br />
ω 0<br />
/<br />
0<br />
0.2 0.4 0.6 0.8 1<br />
Fig.6: RAOs for heave (left) and pitch (right) of the S175 orig<strong>in</strong>al and modified hull shapes<br />
72
Table III: Orig<strong>in</strong>al and optimized seakeep<strong>in</strong>g performances of conta<strong>in</strong>ership S175<br />
Configuration Heave Pitch Heave Improv. Pitch Improv. Mean Improv.<br />
Orig<strong>in</strong>al 1.320 0.640 - - -<br />
O #1 0.878 0.628 33.46 % 1.91 % 17.69 %<br />
O #2 1.220 0.554 7.58 % 13.48 % 10.53 %<br />
O #3 0.912 0.592 30.86 % 7.50 % 19.18 %<br />
- - - S175: __<br />
O<br />
S175 ; O #1<br />
10<br />
5<br />
-10 -5 0 5 10<br />
- - - S175: __<br />
O<br />
S175 ; O #2<br />
0<br />
10<br />
5<br />
Y<br />
- - - S175: __<br />
O<br />
S175 ; O #3<br />
-10 -5 0 5 10<br />
0<br />
10<br />
5<br />
-10 -5 Y0 5 10<br />
Fig.7: Orig<strong>in</strong>al S175 (dashed l<strong>in</strong>es) and optimized hulls (cont<strong>in</strong>uous l<strong>in</strong>es)<br />
0<br />
73
Fig.7 compares section l<strong>in</strong>es of the orig<strong>in</strong>al and the modified hull shapes. Configuration O #1<br />
shows that a ship with a f<strong>in</strong>e bow and a large stern offers great reduction <strong>in</strong> term of RAO’s<br />
heave motion peak. On the other side, hull O #2 shows better results regard<strong>in</strong>g the RAO’s pitch<br />
peak value. Configuration O #3 is a trade-off between the previous ones, even if it is more close<br />
to O #1 . In fact, the heave improvement is larger than the pitch improvement.<br />
8 Conclusions<br />
In this paper, a modified version of the MODPSO algorithm has been presented. The method<br />
is based on concept of Pareto dom<strong>in</strong>ance, and the orig<strong>in</strong>al swarm is decomposed <strong>in</strong> more small<br />
sub-swarms depend<strong>in</strong>g on the distance of each <strong>in</strong>dividual from the current Pareto solutions. As a<br />
result, we are able to def<strong>in</strong>e a different guide for each sub-swarm. Effectiveness and efficiency of<br />
the proposed approach were shown by solv<strong>in</strong>g a set of multiobjective test functions. MODPSO<br />
simulations were compared with those of other EAs, namely NSGA and SPEA, Zitzler et al.<br />
(2000). Numerical results showed that MODPSO accuracy <strong>in</strong> f<strong>in</strong>d<strong>in</strong>g Pareto solutions is comparable<br />
to that of SPEA, and higher than that of NSGA. Furthermore, MODPSO proved to be<br />
computationally one order of magnitude more efficient than the other tested algorithms. F<strong>in</strong>ally,<br />
MODPSO has been used to reduce RAOs heave and pitch motion peaks of the conta<strong>in</strong>ership<br />
S175. Many Pareto optimal solutions are offered to the design eng<strong>in</strong>eers, among which the preferred<br />
f<strong>in</strong>al design can be chosen.<br />
References<br />
BAKER, C.A.; WATSON, L.T.; GROSSMAN, B.; HAFTKA, R.T.; MASON, W.H. (2000),<br />
Parallel global aircraft configuration design space exploration, 8th AIAA/NASA/ISSMO Symp.<br />
Multidiscipl<strong>in</strong>ary Analysis and Optimization, Paper No. 2000–4763–CP, Long Beach<br />
BOERINGER, D.W.; WERNER, D.H. (2004), Particle swarm optimization versus genetic algorithms<br />
for phased array syntesis, IEEE Trans. Antennas and Propagation, 52(3)<br />
CAMPANA, E.F.; PINTO A. (<strong>2005</strong>), A multiobjective particle swarm optimization algorithm<br />
based on sub-swarms search, submitted to Evolutionary Computation<br />
CAMPANA, E.F.; LIUZZI, G.; LUCIDI, S.; PERI, D.; PICCIALLI, V.; PINTO A. (<strong>2005</strong>), New<br />
global optimization methods for ship design problems, Techn. Rep. INSEAN, <strong>2005</strong> – 1, submitted<br />
to Optimization and Eng<strong>in</strong>eer<strong>in</strong>g<br />
CARLISLE, A.; DOZIER, G. (2001), An off-the-shelf PSO, Workshop on Particle Swarm Optimization,<br />
Indianapolis<br />
CLERC, M.; KENNEDY, J. (2002), The particle swarm-explosion, stability, and convergence <strong>in</strong><br />
multidimensional complex space, IEEE Trans. Evolutionary Computation, 6(1)<br />
COELLO, C.A.C. (1999), A comprehensive survey of evolutionary-based multiobjective optimization,<br />
Knowledge and Information System, 1(3), pp.269-308<br />
COELLO, C.A.C.; VAN VELDHUIZEN, D.A.; LAMONT, G.B. (2002), Evolutionary algorithms<br />
for solv<strong>in</strong>g multi-objective problems, Kluwer Academic Publishers, New York, ISBN: 0306467623.<br />
COELLO, C.A.C.; LECHUGA, M.S. (2002), MOPSO: A proposal for multiple objective particle<br />
swarm optimization, IEEE World Congress on Computational Intelligence (CEC00), pp.1051-<br />
1056<br />
COELLO, C.A.C. (2003), Evolutionary multiobjective optimization: Current and future challenges,<br />
Advances <strong>in</strong> Soft Comput<strong>in</strong>g-Eng., Design and Manufactur<strong>in</strong>g, pp.243-256, Spr<strong>in</strong>ger,<br />
74
ISBN 1-85233-755-9<br />
DEB, K. (1999), Multi-objective genetic algorithms: Problem difficulties and construction of test<br />
problems, Evolutionary Computation, 7(3), pp.205-230<br />
FOURIE, P.C.; GROENWOLD, A.A. (2000), Particle swarms <strong>in</strong> size and shape optimization,<br />
Int. Workshop on Multidiscipl<strong>in</strong>ary Design Optimization, Pretoria, pp.97-106<br />
FOURIE, P.C.; GROENWOLD, A.A. (2001), The particle swarm algorithm <strong>in</strong> topology optimization,<br />
4th World Congress of Structural and Multidiscipl<strong>in</strong>ary Optimization, Dalian<br />
FUKUYAMA, Y.; YOSHIDA, H. (2001), A particle swarm optimization for reactive power and<br />
voltage control <strong>in</strong> electric power systems, IEEE Congress on Evolutionary Computation (CEC<br />
2001), Seoul<br />
JIN, Y.; OLHOFER, M.; SENDHOFF, B. (2001), Dynamic weighted aggregation for evolutionary<br />
multi-objective optimization: Why does it work and how?, Genetic and Evolutionary<br />
Computation Conf., pp.1042-1049<br />
KENNEDY, J.; EBERHART, R. (1995), Particle swarm optimization, IEEE Int. Conf. Neural<br />
Networks, Perth, IV:pp.1942-1948<br />
MEYERS, W.G.; APPLEBEE, T.R.; BAITIS, A.E. (1981), User’s manual for the standard ship<br />
motion program, SMP, David Taylor report DTNSRDC/SPD-0936-01<br />
MEYERS, W.G.; BAITIS, A.E. (1985), SMP84: improvements to capability and prediction<br />
accuracy of the standard ship motion program SMP84, David Taylor report DTNSRDC/SPD-<br />
0936-04<br />
MIETTINEN, K. M. (1999), Nonl<strong>in</strong>ear multiobjective optimization, Kluwer Academic Publisher<br />
NG, K.Y.; TAN, C.M.; RAY, T.; TSAI, H.M. (2003), S<strong>in</strong>gle and multiobjective w<strong>in</strong>g planform<br />
and airfoil shape optimization us<strong>in</strong>g a swarm algorithm, 41st Aerospace Sciences Meet<strong>in</strong>g and<br />
Exhibit, Reno, AIAA 2003-45<br />
PARSOPOULOS, K.E.; VRAHATIS, M.N. (2002a), Recent approaches to global optimization<br />
problems through particle swarm optimization, Natural Comput<strong>in</strong>g 1: pp.235-306<br />
PARSOPOULOS, K.E.; VRAHATIS, M.N. (2002b), Particle swarm optimization method for<br />
constra<strong>in</strong>ed optimization problems, Euro-Int. Symp. Computational Intelligence 2002.<br />
PERI, D.; ROSSETTI, M.; CAMPANA, E.F. (2001), Design optimization of ship hulls via CFD<br />
techniques, J. Ship Research 45/2, pp.140-149<br />
PERI, D.; CAMPANA, E.F. (2003), Multidiscipl<strong>in</strong>ary design optimization of a naval surface<br />
combatant, J. Ship Research 47/1 pp.1-1.<br />
PERI, D.; CAMPANA, E.F. (<strong>2005</strong>), High fidelity models and multiobjective global optimization<br />
algorithms <strong>in</strong> simulation based design, J. Ship Research, to appear<br />
PINTO, A.; PERI, D.; CAMPANA, E.F. (2004), Global optimization algorithms <strong>in</strong> naval hydrodynamics,<br />
Ship Technology Research, 51/3, pp.123-133<br />
PINTO, A.; PERI, D.; CAMPANA, E.F. (<strong>2005</strong>), Multiobjective optimization of a conta<strong>in</strong>ership<br />
us<strong>in</strong>g determ<strong>in</strong>istic particle swarm optimization, submitted to J. Ship Research<br />
SCHUTTE, J.F.; GROENWOLD, A.A. (2003), Siz<strong>in</strong>g design of truss structures us<strong>in</strong>g particle<br />
swarms, Structural and Multidiscipl<strong>in</strong>ary Optimization, 25/4, pp.261-269<br />
SCHUTTE, J.F.; REINBOLT, J.A.; FREGLY, B.J.; HAFTKA, R.T.; GEORGE, A.D. (2004),<br />
Parallel global optimization with the particle swarm algorithm, J. Num. Meth. Eng., 61(13),<br />
pp.2296-2315<br />
75
SHI, Y.H.; EBERHART, R.C. (1998), Parameter selection <strong>in</strong> particle swarm optimization, 7th<br />
Annual Conf. Evolutionary Programm<strong>in</strong>g, San Diego<br />
SHI, Y.H.; EBERHART, R.C. (2001), Particle swarm optimization: developments, applications<br />
and resources, IEEE Congress on Evolutionary Computation, pp.27-30<br />
STATNIKOV, R.B., MATUSOV, J.B. (1995), Multicriteria optimization and eng<strong>in</strong>eer<strong>in</strong>g, Chapman<br />
& Hall, USA<br />
SRINIVAS, N.; DEB, K. (1995), Multiobjective function optimization us<strong>in</strong>g nondom<strong>in</strong>ated sort<strong>in</strong>g<br />
<strong>in</strong> genetic algorithms, Evolutionary Computation, 2(3), pp.221-248<br />
TAN, C.M.; RAY, T.; TSAI, H.M. (2003), A comparative study of evolutionary algorithm and<br />
swarm algorithm for airfoil shape optimization problems, 41st Aerospace Sciences Meet<strong>in</strong>g and<br />
Exhibit, Reno, AIAA 2003-102<br />
VENTER, G.; SOBIESZCZANSKI-SOBIESKI, J. (2003), Particle swarm optimization, AIAA<br />
J. 41/8, pp.1583-1589<br />
VENTER, G.; SOBIESZCZANSKI-SOBIESKI, J. (2004), Multidiscipl<strong>in</strong>ary optimization of a<br />
transport aircraft w<strong>in</strong>g us<strong>in</strong>g particle swarm optimization, Structural and Multidiscipl<strong>in</strong>ary Optimization<br />
26/1-2, pp.121-131<br />
ZITZLER, E.; DEB, K.; THIELE, L. (1999), Multiobjective evolutionary algorithms: A comparative<br />
case study and the strength Pareto approach, IEEE Trans. Evolutionary Computation,<br />
3(4), pp.257-271<br />
ZITZLER, E.; DEB, K.; THIELE, L. (2000), Comparison of multiobjective evolutionary algorithms:<br />
Empirical results, Evolutionary Computation 8(2), pp.173-195.<br />
76
!! !!! " # <br />
! <br />
<br />
$ ! !%&<br />
<br />
8(<br />
*# <br />
5<br />
*# 6<br />
7<br />
(<br />
5 6<br />
5<br />
* <br />
<br />
%&'$ (<br />
<br />
) * &++, !<br />
# ! (- *.(- */<br />
$ ! 0!<br />
(- * <br />
<br />
!" <br />
<br />
( # <br />
)1) # !<br />
! (!<br />
$ *! )1)# &++, <br />
2 # ! # 34 !!
)1) :! (- **<br />
!;! !;<br />
!(- *<br />
<br />
(' 5!(- <br />
# <br />
& !<br />
<br />
<br />
<br />
<br />
, <br />
<br />
<br />
<br />
<br />
B" <br />
<br />
<br />
(('B"-A CB" A<br />
#<br />
<br />
- <br />
)<br />
- <br />
&&% <br />
!<br />
&,-# <br />
<br />
&
(,32(! B" <br />
(,=2 ! <br />
(,92-5( <br />
(
)0<br />
) 5<br />
5 <br />
5 5<br />
5C$ <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
5<br />
- 5<br />
(<br />
)<br />
5!<br />
<br />
%$$ <br />
%3'(- *<br />
<br />
(- * !# !' <br />
<br />
! !(- *.)0) <br />
55 5- 5/# <br />
# !(!$ *.)5<br />
! / ( # ! !# # '<br />
!!!!!<br />
<br />
(- * ! # # !<br />
!! ! )1) !# -!D*)D5B !<br />
-()*@%# ! (- * !)1)<br />
!# '<br />
2 !! <br />
2 ! ! # ! # !<br />
<br />
2 # !<br />
<br />
2 2! !'# " !<br />
# # <br />
<br />
(- * ! A2A ! #<br />
! # # ! <br />
B( ; ! . ! / <br />
! (- * <br />
D D #? (D-@? *D# -$*' **G *G<br />
<br />
$ % <br />
<br />
@ ! ! <br />
H!(! !#<br />
H'<br />
9&
2 !! # 0 ! 0 !<br />
! !# <br />
! <br />
2 5 # 0 <br />
<br />
2 ! ! !<br />
( !! H <br />
!! 7 <br />
!!! <br />
$ # <br />
<br />
(! $ * # )1) <br />
! ! !! ! !! "<br />
( !! ! <br />
!#'<br />
2 !I<br />
2 !! I<br />
2 !!H<br />
<br />
!! !!H <br />
./) !'<br />
2 !!2# " I<br />
2 !I<br />
2 ! !! <br />
) !!H +234J <br />
! !! ! <br />
<br />
@ ! !!<br />
# !! ! <br />
!! #<br />
" !! # <br />
7 <br />
!! <br />
2! ! 0 !! <br />
!: <br />
!! ! <br />
2!(!<br />
,44, ! ! !#$%# # <br />
; !## ' !<br />
!!# <br />
# ( ! <br />
# <br />
!! "! 50 ! !!./ 0<br />
! '<br />
<br />
<br />
(<br />
<br />
<br />
<br />
<br />
<br />
=<br />
& =<br />
&<br />
= =<br />
<br />
<br />
=<br />
&<br />
'<br />
&<br />
<br />
<br />
<br />
=<br />
&<br />
'<br />
&<br />
<br />
<br />
<br />
# (Σ' C<br />
' )! <br />
& C! <br />
C ! <br />
.&/<br />
9,
7 ! . / 0'<br />
<br />
*& + <br />
.,/<br />
# C "! <br />
<br />
%!#! 0'<br />
=<br />
<br />
<br />
=<br />
&<br />
<br />
<br />
=<br />
<br />
<br />
=<br />
&<br />
<br />
<br />
('<br />
<br />
<br />
<br />
=<br />
&<br />
.&<br />
<br />
⋅ + <br />
/ = &<br />
<br />
+ = ⋅&<br />
<br />
+ = ('<br />
<br />
+ = (, + .
=<br />
<br />
('<br />
<br />
=<br />
&<br />
<br />
=<br />
<br />
(,<br />
=<br />
1<br />
α<br />
⋅ %,<br />
<br />
(,<br />
<br />
( ! 0<br />
<br />
1 ⋅ %, 1 ⋅<br />
<br />
α<br />
<br />
α<br />
%,<br />
<br />
&<br />
<br />
<br />
= &<br />
<br />
+ <br />
= &<br />
<br />
+ &<br />
<br />
= &<br />
<br />
+<br />
<br />
.&&/<br />
(,<br />
(, <br />
<br />
*!# # !#! !<br />
'<br />
<br />
1 ⋅ %, <br />
<br />
α<br />
&<br />
<br />
<br />
= &<br />
<br />
+<br />
<br />
.&,/<br />
(, <br />
# (,C<br />
<br />
#!! %= !!!<br />
!# ! ! <br />
$ !<br />
!! 0!!!# !#<br />
!!!!<br />
<br />
6<br />
.&4/<br />
5<br />
pi/Oi<br />
4<br />
3<br />
2<br />
1<br />
0<br />
0.0505<br />
0.0446<br />
0.0401<br />
0.0359<br />
0.0322<br />
0.0287<br />
0.0256<br />
0.0228<br />
0.0179<br />
0.0139<br />
0.0107<br />
0.0082<br />
0.0062<br />
0.0047<br />
Bankruptcy probability R<br />
0.0035<br />
0.0026<br />
0.0019<br />
0.0014<br />
3.2000E-05<br />
%='!!! <br />
<br />
# <br />
!<br />
#<br />
!!# ! <br />
! ## ! <br />
<br />
<br />
93
& <br />
<br />
&'" <br />
<br />
? (- * <br />
!! !! 6<br />
!250 (- *!"<br />
! ! !# # <br />
! ##!<br />
# <br />
! <br />
<br />
!# ! <br />
"( !!#<br />
! # <br />
!( # <br />
! * <br />
<br />
<br />
( H!!# <br />
1!! !#0 '<br />
2 !! !!H!I<br />
2 !HI<br />
2 <br />
I<br />
2 !! ! !I<br />
2 !! ! ( I<br />
2 ! # H 2 H <br />
( I<br />
2 " !! # # # H<br />
I<br />
2 " !! # <br />
:!I<br />
2<br />
2<br />
<br />
! :! "I<br />
!! " <br />
<br />
!#<br />
!#<br />
<br />
<br />
<br />
<br />
<br />
-<br />
<br />
<br />
# <br />
<br />
<br />
<br />
<br />
%>'(- *<br />
<br />
9=
5!! !# !1!!<br />
# !! ( !!<br />
! !! <br />
! ! !! <br />
# !0 #! !<br />
# # ## ! #<br />
# # # # <br />
# ! 5@) L,BB DB<br />
) 5@)#!<br />
DB ! L,BB # ! <br />
# !! (! ! ! (- * <br />
# %> ((( # # ! <br />
! <br />
<br />
((('-!! <br />
<br />
# %<br />
! ! <br />
B # 6 ;! <br />
<br />
- <br />
- <br />
<br />
&! %% <br />
<br />
? (- *( # <br />
AA ! ! !)! <br />
! " $! *G * $*G<br />
*G @ " <br />
( # # 5@)<br />
!!!!5@)!!<br />
! " @ <br />
-## M? # !<br />
! 5@) ! !<br />
DB L,BB ! <br />
(1<br />
<br />
(1' ! <br />
*! L,BB DB<br />
'<br />
5!? B ! " !<br />
! # <br />
<br />
5 ! .<br />
/<br />
L*<br />
L<br />
'<br />
L* <br />
L* <br />
L <br />
'<br />
1*<br />
L*<br />
'<br />
)* DB<br />
5@$N) M<br />
# L-5 )-@DB<br />
<br />
() <br />
<br />
!! ! ! !<br />
" # ! *!;!<br />
!;!(- <br />
*<br />
9>
!#! ( '<br />
2 !2 # # ! <br />
<br />
2 !(- *#!!!<br />
<br />
2 ! ! <br />
<br />
* .* / (* # !<br />
.% 2) / A 2B" A ) !<br />
!# !!<br />
!# !! ! !# <br />
!(- * !!<br />
!! !!! <br />
! # <br />
<br />
! <br />
# !! # <br />
!<br />
<br />
# (- * ( # <br />
;! ! #! <br />
! # ( !! # # <br />
! # ( <br />
! # <br />
# ! " # <br />
<br />
? ! !<br />
! : ; " % " 5@) !! 0 <br />
! <br />
$! DB ! <br />
# DB !? # <br />
<br />
* <br />
<br />
D@1(*EO I 7)D@1) $I 1(E@@1) BI )F@P(D 1 .,44
!"!#$ <br />
<br />
<br />
<br />
!"""#"$"<br />
"" """%&#<br />
$ """"""<br />
"""$"<br />
"""""' """<br />
" " " <br />
"" " ""<br />
" ("" "$ " " <br />
" !""#"""""<br />
" " "" " " " "<br />
")"""""""<br />
<br />
*"""" $""""""<br />
" " ' " +,*#,* " "<br />
""""" ""(""<br />
" ")"""-<br />
!""""+,*#,* <br />
".,//'0 ' " "<br />
" "1""""$<br />
$"""""""""<br />
("" $ " " 1 " <br />
""$ ""-""<br />
"$ ""<br />
<br />
" " " 2" ' % <br />
"""*""<br />
" ' $" "" "<br />
"-"" $" "<br />
'"")<br />
" ""<br />
" 2 ' "$ " " " <br />
""""" 2"<br />
" " 23 " (" <br />
"""" )<br />
<br />
<br />
<br />
"" " " "" " " "<br />
""1""""""
"#"""<br />
")-" <br />
"" ""<br />
."""""$""."<br />
" ""'""$"<br />
"("<br />
<br />
2" " " " "" " <br />
$"(" " " " """ <br />
""")"""" <br />
" 5 " " " " "<br />
) "" "$"" "<br />
"0""<br />
<br />
! "#<br />
<br />
."" """(" <br />
""6$" "$"$"<br />
"$"" 25<br />
<br />
<br />
<br />
<br />
<br />
⋅ <br />
<br />
( − ) 8 = 7<br />
$<br />
<br />
<br />
≠ <br />
= γ ⋅<br />
<br />
⋅<br />
6 <br />
=<br />
<br />
= <br />
<br />
− <br />
=<br />
%<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
9%:<br />
<br />
; ""$""$"""5<br />
<br />
<br />
<br />
<br />
( ) <<br />
<br />
( )<br />
<br />
( + ∆) = <br />
( ) + <br />
( )<br />
⋅ ∆<br />
+ ⋅ ∆<br />
<br />
( + ∆) = <br />
( )<br />
+ ⋅ ∆<br />
9
$<br />
<br />
'"" "".,//'0==5<br />
<br />
Integer, parameter MAX_DATA=100<br />
Integer N<br />
Real, dimension(MAX_DATA) x0, x1, y0, y1, z0, z1,<br />
+ vx0, vx1, vy0, vy1, vz0, vz1, m<br />
common /planets/ x0, x1, y0, y1, z0, z1,<br />
+ vx0, vx1, vy0, vy1, vz0, vz1, m, N<br />
<br />
' $" ")" "<br />
$"7%$" " → + ∆<br />
5<br />
[ ]<br />
[ ]<br />
[ ]<br />
[ ]<br />
[ ]<br />
[ ]<br />
7<br />
<br />
%<br />
7<br />
<br />
%<br />
<br />
<br />
<br />
( )<br />
= 7<br />
8<br />
<br />
> <br />
= <br />
( + ∆)<br />
= %<br />
8<br />
<br />
( )<br />
= 7<br />
8<br />
<br />
> <br />
( + ∆)<br />
= %<br />
<br />
<br />
<br />
<br />
& 7 <br />
&%<br />
& 7 <br />
& %<br />
<br />
) """"""" ""<br />
)""""5<br />
<br />
<br />
<br />
[ ]<br />
[ ]<br />
[ ]<br />
5= % <br />
5= % <br />
" ≤ $ " ≤ $ <br />
<br />
= 7 <br />
<br />
<br />
<br />
<br />
= > <br />
<br />
5=% <br />
= > <br />
[ ]<br />
<br />
[ ]<br />
[ ] <br />
<br />
<br />
= 7 <br />
= <br />
" ≤ $<br />
<br />
<br />
⋅<br />
<br />
− <br />
.<br />
<br />
<br />
<br />
= γ ⋅ ⋅( <br />
− <br />
)<br />
6<br />
<br />
<br />
<br />
<br />
<br />
<br />
= +% <br />
<br />
<br />
<br />
= + <br />
<br />
<br />
> <br />
= +% <br />
<br />
<br />
<br />
<<br />
= <br />
+ <br />
⋅ ∆<br />
+ ∆<br />
<br />
< ⋅ <br />
<br />
<br />
<br />
<br />
<br />
> <br />
= <br />
+ ∆<br />
<br />
<br />
= +% <br />
<br />
<br />
<br />
<br />
" " " "" ? $ <br />
""<br />
47
$<br />
<br />
2""""5<br />
<br />
<br />
<br />
##<br />
! %<br />
<br />
<br />
9%50:<br />
#' 5@"A<br />
<br />
# <br />
<br />
<br />
5B6C<br />
<br />
# <br />
<br />
#9# 5:<br />
<br />
5B6C<br />
<br />
#5<br />
#"95:<br />
<br />
#5D*<br />
<br />
<br />
<br />
#.9# 5:<br />
<br />
#0)" ( )<br />
<br />
<br />
#"$0)" ( )<br />
<br />
<br />
-" """""<br />
' ( " " <br />
" 5' 2) '<br />
2 ' ( 2 ) . "<br />
" " " " " <br />
) **")+* ,- "" <br />
"*""") <br />
"<br />
<br />
"""""" "" <br />
"""") ""<br />
"5<br />
<br />
<br />
B"5D*55"95:<br />
<br />
{ <br />
<br />
<br />
→' <br />
<br />
{ <br />
<br />
<br />
→0)"9:<br />
<br />
<br />
}<br />
<br />
<br />
→' <br />
<br />
<br />
{ <br />
<br />
<br />
→"$0)"9:<br />
<br />
<br />
}<br />
<br />
<br />
<br />
<br />
"' (" "5<br />
<br />
4%
B"5# 550)"95:{ <br />
<br />
5 = 7 <br />
<br />
→→' { <br />
<br />
<br />
<br />
= <br />
<br />
+ → 9 :<br />
<br />
}<br />
<br />
<br />
<br />
5 <br />
→ <br />
<br />
< <br />
<br />
<br />
= → + → ⋅ +<br />
⋅<br />
⋅ →<br />
<br />
<br />
<br />
<br />
→ 5=<br />
→ + ⋅ <br />
<br />
→ <br />
<br />
} <br />
<br />
B6C5# 55.9# :{ <br />
<br />
<br />
( → =<br />
<br />
→ ) <br />
<br />
<br />
<br />
{ 5 = 7}<br />
<br />
<br />
<br />
<br />
→ ⋅ → <br />
( )<br />
<br />
5= γ ⋅ ⋅ → − → <br />
<br />
→ − → <br />
<br />
<br />
<br />
<br />
<br />
} <br />
<br />
<br />
'$$ $ "5<br />
<br />
<br />
<br />
<br />
B"5# 550)"95:{ <br />
<br />
<br />
{ 5 = 7}<br />
<br />
<br />
→→' { <br />
<br />
<br />
<br />
= <br />
<br />
+ → 9 :<br />
<br />
<br />
}<br />
<br />
<br />
<br />
<br />
<br />
5 <br />
→ <br />
< <br />
<br />
<br />
= → + → ⋅ +<br />
⋅<br />
⋅ →<br />
<br />
<br />
<br />
<br />
<br />
→ 5=<br />
→ + ⋅ <br />
→ <br />
<br />
} <br />
<br />
<br />
<<br />
<<br />
<br />
<br />
4
&$"$$"<br />
<br />
$ 2" " " " ' <br />
""#$""<br />
" "" " " )"" ""<br />
D""""""' ( "".)<br />
""" " 5<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! %<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! <br />
<br />
#'D #D <br />
#<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
' $ 2" " " " <br />
$.+/001-1"$""<br />
"""1""""<br />
2 $"$1<br />
2 "" " ' <br />
"$ " ) " "" ' "<br />
""""" "$"<br />
<br />
<br />
&$'" "$"<br />
<br />
" " " "$ # <br />
5<br />
<br />
% - " 2" " $ <br />
2""+EEF$<br />
< ' """ " " .,//'0 " "" " <br />
"$<br />
<br />
2"""" ""<br />
2 2"$"-""<br />
2'"""""<br />
""""""-"$" <br />
29":" " $"<br />
"$" ""$" "<br />
." 2"""""""<br />
"" """, 2"<br />
" " " " " " "" $" <br />
" #$ " " $" " <br />
"+"".,//'0"""'<br />
" 2" " """ <br />
<br />
<br />
<br />
46
2 " " . ) .,//'0== " <br />
-"""" 2"".,//'03+<br />
$"""<br />
$ """ ' ""<br />
""$*""" """<br />
<br />
" "C.,//'0 <br />
+EEF$ " " "<br />
C" " $ " """ " <br />
$'"$"$1<br />
"""""1""<br />
" " 3 " " <br />
" " ) " +< 0 <br />
"9.,//'0:"<br />
2"9+EEF$:" 2"'<br />
" $" $"<br />
; """3-".,//'0$<br />
+EEF$3.""" <br />
$" """$")<br />
$*" "$""<br />
""" <br />
$-" 2"$"")"1")"""<br />
$")<br />
) $ " +( " " <br />
""""".)"<br />
"$"""+D1"+D<br />
" 2 " " ) "" <br />
$ " 2"<br />
,"$ """"$""<br />
" " " " ) " <br />
" " " " " ' "<br />
""" "<br />
'"+EEF$9 2":""<br />
" " "" 0 +EE 1 $ " <br />
""""""+EE""<br />
" +D - F$ " " <br />
" " """""<br />
""""""$'""<br />
-"'$""<br />
) " " $" ' " "<br />
"""$"" """)""""<br />
1"" "<br />
<br />
-"""C"<br />
"""""-"""""<br />
" " " 2" 2" " <br />
""<br />
<br />
&$ ( )*)+%<br />
<br />
; " 2" " " .,//'0 $" <br />
" $" 3 " " " .,//'0 <br />
# " $" " 2" $ <br />
".,//'0- """ 2<br />
""-" 2" <br />
4!
" " " " 2 2" )""" "<br />
""1"2C$<br />
" 0 *<br />
)""$ " 2"".,//'0" <br />
<br />
&$( )*)+,,<br />
<br />
-""""""" <br />
5<br />
<br />
Subrout<strong>in</strong>e xyz_new()<br />
Real : a,b,c<br />
Save<br />
A=0.0<br />
B=0.0<br />
C=0.0<br />
return<br />
Entry xyz_delete()<br />
…<br />
return<br />
entry xyz_calc(u,v,w)<br />
…<br />
return<br />
end subrout<strong>in</strong>e<br />
<br />
$ " " ) " 2 " "<br />
9 :""9)G>)G>)G>:<br />
1 $" " " $ " <br />
""" $"""$" 2C"<br />
2 " " " 1 " <br />
"$$" "<br />
<br />
""""$)"$" 2""$<br />
""""""""1"<br />
"""""<br />
<br />
&&$$"! $<br />
<br />
""" 2("<br />
"$","""<br />
"$" .)' "<br />
-.,//'047""""$5<br />
<br />
type MyClass_type<br />
<strong>in</strong>teger :: i<br />
real :: a<br />
end type MyClass_type<br />
<br />
0"" "5<br />
<br />
Type (MyClass_type), dimension(MAX_OBJECTS) :: MyClass_data<br />
<br />
"" 2'"""<br />
"<br />
4&
Logical, dimension(MAX_OBJECTS) :: MyClass_isUsed<br />
<br />
' """"".'I* I$"<br />
"""")"" 2<br />
' ""5<br />
<br />
Function MyClass_new()<br />
Integer :: MyClass_new<br />
Integer :: I<br />
Logical :: ok<br />
Ok = .false.<br />
Do I=1,MAX_OBJECTS,1<br />
If (MyClass_isUsed .eq. .false) then<br />
MyClass_isUsed[I] = .true.<br />
MyClass_data[I]%i = 0<br />
MyClass_data[I]%a = 1.0<br />
MyClass_new = I<br />
Return<br />
End if<br />
End do<br />
MyClass_new = -1<br />
Return<br />
End function MyClass_new<br />
<br />
' "") <br />
"$" - 2"5<br />
<br />
Procedure MyClass_delete(id)<br />
Integer :: id<br />
MyClass_isUsed[id] = .false.<br />
End procedure MyClass_delete<br />
<br />
-' "" <br />
""5<br />
<br />
MyClass_dat[id]% = …<br />
<br />
- " " $" "" " "" "<br />
"" 2""<br />
<br />
&-"# %$<br />
<br />
- 2""""""","<br />
""""$" "1"<br />
"""""""""<br />
2""" 2-""" "<br />
2("$'"")"5<br />
<br />
type ChildClass_type<br />
<strong>in</strong>teger :: parent_object<br />
real :: c,d<br />
end type ChildClass_type<br />
<br />
4H
"" 5<br />
ChildClass_data[id]%parent_object = ParentClass_new()<br />
<br />
'5<br />
<br />
Call ParentClass_delete(ChildClass_data[id]%parent_object)<br />
<br />
." "5<br />
<br />
Procedure CildClass_(id, …)<br />
Integer :: id<br />
Call ParentClass_(ChildClass_data[id]%parent, …)<br />
End procedure ChildClass_<br />
<br />
'" """ " 2"D<br />
$" 2 $" 2<br />
- 2" ""<br />
- $)""""'""<br />
-"")" 2 2<br />
<br />
$" "")"" 2"""<br />
"""5<br />
<br />
function ChildClass_castToParent(id)<br />
<strong>in</strong>teger :: ChildClass_castToParent<br />
ChildClass_castToParent = ChildClass_data[id]%parent<br />
Return<br />
End function ChildClass_castToParent<br />
<br />
""""" 2"""<br />
"" " 2 "1<br />
2"" """" <br />
""5<br />
<br />
function ParentClass_castToChild(id)<br />
<strong>in</strong>teger :: ParentClass_castToChild<br />
ParentClass_castToChild = ParentClass_data[id]%child<br />
Return<br />
End function ChildClass_castToParent<br />
Function ParentClass_newFromChild(child_id)<br />
Integer :: ParentClass_newFromChild<br />
Integer :: child_id<br />
Integer :: id<br />
Id = ParentClass_new()<br />
ParentClass_data[id]%child = child_id<br />
ParentClass_newFromChild = id<br />
Return<br />
End function ParentClass_newFromChild<br />
<br />
0"<br />
" "" " " " "" <br />
)"""1"" ""<br />
4=
&.$"/! "<br />
<br />
'" 2""""""$""'<br />
$""""" ""$ <br />
"" $",""<br />
" "" " 2 " <br />
D""-<br />
"""" $"""1<br />
"""<br />
$"""$"""$""<br />
"$""""""<br />
"5<br />
<br />
D = GeneralShip_getDisplacement(ship_id)<br />
<br />
-" .)"2 <br />
".)" 2"<br />
".) "" )-<br />
2"""""" "<br />
$"'"$<br />
" " <br />
"" 2"""" <br />
" """-" "(<br />
" "5<br />
<br />
Integer,parameter :: GeneralShip_classNo = 1<br />
Integer,parameter :: FerryShip_classNo = 2<br />
Integer,parameter :: TankShip_classNo = 3<br />
…<br />
<br />
" " " <br />
0)"""""""<br />
""5<br />
<br />
type GeneralShip_type<br />
<strong>in</strong>teger :: child_object<br />
<strong>in</strong>teger :: child_classNo<br />
…<br />
end type GeneralShip_type<br />
<br />
0"".),2 "5<br />
<br />
Function GeneralShip_getDisplacement(id)<br />
Real :: GeneralShip_getDisplacement<br />
Integer :: child<br />
Real :: d<br />
Child = GeneralShip_data[id]%child_object<br />
<br />
<br />
Select case(GeneralShip_data[id]%child_classNo))<br />
Case(FerryShip_classNo)<br />
d = FerryShip_getDisplacement(child)<br />
Case(TankShip_classNo)<br />
d = TankShip_getDisplacement(child)<br />
Case default<br />
d = -1.<br />
4
End select<br />
GeneralShip_getDisplacement = d<br />
Return<br />
End function<br />
<br />
0 " " $ " <br />
"'" """"-"<br />
" """ "<br />
<br />
-<br />
<br />
1"$"" <br />
"" " " ' "" 2 " " "" $<br />
" "" " . <br />
"""$"$""""<br />
"" , " " " " ""<br />
)" " "" "" ' " " <br />
9" "": "" " "<br />
"""""$"<br />
)" $ , 2" " $ <br />
""<br />
<br />
; " " .,//'0 2" " " " "<br />
" " " ' $ " <br />
" " " " )" $ "" #$ " "<br />
""" "" " <br />
"" 2" C" " "" <br />
""2"""'.,//'047<br />
) """"""$"<br />
""""2"")"<br />
""<br />
<br />
)<br />
<br />
'##' 9
!" <br />
# <br />
!" <br />
$ % <br />
$ &'<br />
<br />
% <br />
$<br />
<br />
<br />
! "#$%<br />
&'' () * + ,% - &. <br />
*!- # . () <br />
( )*++,- ! # # <br />
% * . *++,- # <br />
!/ # 0 % <br />
, 1# % ## ,# $ ! <br />
# 1# ## # #<br />
! # % #% % <br />
% % # <br />
##<br />
<br />
##!23$4*2 3$4-%#<br />
! % # # <br />
*# 5## 1 5 #<br />
# ! ,# - 23$4 , # 2+5 <br />
*25-## # <br />
6 7 ,#%<br />
% <br />
# 5 # 1# # <br />
# !<br />
##%% <br />
<br />
23$4%##8 ##*8-! <br />
# # <br />
23$4 # # !<br />
* , # ,# #- 5 <br />
# ! *#<br />
-#23$4<br />
0%# % #<br />
9% 23$4 # % <br />
# # # <br />
# #
23$4 %#%%<br />
# # %# <br />
# 5 % )/00*- # <br />
8 ##*8- # 0 51<br />
2)*+++-### <br />
# # * #- <br />
#0 # <br />
%##1# !<br />
5%#% % # <br />
# # <br />
#23$4 1##%# :<br />
#% 0 % # ;% <br />
## #0 # <br />
#%<br />
% !# <br />
<br />
1 # <br />
1# #23$4 < #<br />
#%<br />
!"# <br />
<br />
! 0#1# %#<br />
# 1 !)/003- %#<br />
06#+ 7 <br />
,# !##%#<br />
+ % # 0 5 4 )/003- <br />
# #='()#1>= <br />
>++' ## !<br />
,# + + 5 % #<br />
# ?@#2)*+++-<br />
<br />
55)*++/- ##0#<br />
+ ## ##<br />
+ # ## #<br />
+ ## # ! # # <br />
% +#9%%<br />
# !% 67%!<br />
# )*++*- ,# # <br />
# # %# # <br />
# # 8 <br />
# % #<br />
67 ## #% <br />
#1 # # # <br />
%% 5##+ <br />
# ! %:<br />
##0 #<br />
<br />
#%#! #11<br />
# (1# <br />
0!,# %1<br />
677)/000- 8)/000- <br />
%## ! % <br />
#1*%##9*++,-3#%%
6 )/00*-:)/003-:")*+++-:<br />
( )*++/-:")*++/-:9!)*++*-:1)*++*- <br />
.)*++*-% , ! <br />
##% ! #<br />
!#8 %<br />
$%&' %<br />
<br />
23$41 #%% ##52*<br />
5 2 - . )*++,- 23$4 <br />
#23$4 <br />
#!# ! <br />
% %#!# 1# <br />
# % <br />
<br />
<br />
23$4 %# 2 A ? !BB # <br />
C # + ## *<br />
.*++;-25<br />
%$8! #% !<br />
%' # #%D, :#:#:# <br />
0 *!- # <br />
: # # # <br />
#D 7:: <br />
#<br />
23$4 <br />
% % %#<br />
! 1 # % E #<br />
# % <br />
<br />
D23$4 # <br />
<br />
( & )<br />
7 ! %# #<br />
# ! #<br />
! # %# ,<br />
<br />
<br />
<br />
! #<br />
% <br />
2% ,<br />
,<br />
<br />
% <br />
<br />
F #<br />
%<br />
<br />
5 %<br />
<br />
<br />
( #+ <br />
#0<br />
! <br />
2% <br />
# <br />
F # <br />
G" <br />
<br />
%##<br />
#<br />
<br />
# 23$4 % % #<br />
*.*++,- (+ # %<br />
#%<br />
# #% <br />
&
##5# 23$4<br />
#%< <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
#<br />
<br />
#$ 05 <br />
<br />
<br />
#5 <br />
5 <br />
<br />
, #D8#(#8# .
5 <br />
()1'()<br />
#<br />
.<br />
<br />
<br />
()1'() <br />
(1#<br />
6 <br />
<br />
!<br />
! ! <br />
+<br />
<br />
K <br />
()1'()<br />
#<br />
7<br />
=<br />
J <br />
<br />
<br />
()1'()<br />
(1#<br />
! <br />
<br />
<br />
<br />
<br />
<br />
<br />
#% %!#< <br />
= 3 # # % # + + %<br />
3 0 !3 <br />
# ##<br />
$<br />
%&* <br />
23$4 % # # <br />
!2% )/0?0-<br />
6 #0 ## # <br />
## 7 <br />
%# 23$4 1 ! <br />
5# :%# <br />
%#* 1# #
– #F<br />
• #3*#! -D<br />
– 4 L ##<br />
– @ !% #<br />
MN!<br />
• D<br />
• 6 @#1&!<br />
<br />
<br />
<br />
<br />
<br />
<br />
,- 6<br />
.H ? '.<br />
& & N &N<br />
. =& & ==<br />
= . .M ?N<br />
' & ' .'<br />
? & & ,+<br />
H &= & ;B<br />
M &= = BC<br />
N .= &M B*<br />
=M =. N<br />
O()<br />
. &&= . +<br />
#*5 - <br />
<br />
D4!2 3 #+! <br />
<br />
Sum of Sum of<br />
Conta<strong>in</strong>er Term<strong>in</strong>al<br />
Distance Ship Turn -<br />
Management Policies Ship / Traveled by Around<br />
Bay<br />
Scenario<br />
SCs <strong>in</strong> Times <strong>in</strong><br />
Stowage meters Hours<br />
Config.<br />
Berth<strong>in</strong>g<br />
policy<br />
Schedul<strong>in</strong>g<br />
policy<br />
Stack<strong>in</strong>g<br />
policy<br />
Sum of<br />
Conta<strong>in</strong>er<br />
Ship<br />
Operational<br />
Profit <strong>in</strong><br />
Euro<br />
Scenario 1 BCSP HEF Dest. Random 2760759 44:07:00 € 226 500<br />
Scenario 2 BCSP HEF Dest. Fixed 2698905 42:03:00 € 221 755<br />
Scenario 3 BCSP HEF Ship l<strong>in</strong>e Random 3067510 45:59:00 € 223 000<br />
Scenario 4 BCSP HEF Ship l<strong>in</strong>e Fixed 3050050 44:19:00 € 217 570<br />
Scenario 5 BCSP FIFO Dest. Random 2781045 43:51:00 € 212 540<br />
Scenario 6 BCSP FIFO Dest. Fixed 2753205 42:45:00 € 200 220<br />
Scenario 7 BCSP FIFO Ship l<strong>in</strong>e Random 3090050 44:15:00 € 220 200<br />
Scenario 8 BCSP FIFO Ship l<strong>in</strong>e Fixed 3005045 43:22:00 € 206 560<br />
Scenario 9 STTP HEF Dest. Random 2763045 42:38:00 € 198 800<br />
Scenario 10 STTP HEF Dest. Fixed 2725450 41:28:00 € 185 040<br />
Scenario 11 STTP HEF Ship l<strong>in</strong>e Random 3070050 43:07:00 € 205 550<br />
Scenario 12 STTP HEF Ship l<strong>in</strong>e Fixed 2984655 42:34:00 € 200 250<br />
Scenario 13 STTP FIFO Dest. Random 2763045 42:31:00 € 196 800<br />
Scenario 14 STTP FIFO Dest. Fixed 2704050 41:10:00 € 180 750<br />
Scenario 15 STTP FIFO Ship l<strong>in</strong>e Random 3070050 42:57:00 € 205 250<br />
Scenario 16 STTP FIFO Ship l<strong>in</strong>e Fixed 2984655 42:07:00 € 198 450<br />
0#1 2 "<br />
<br />
1# 23$4%#%1#<br />
! 8 <br />
% % ! # <br />
, # 23$4 G" !<br />
! ## <br />
#% #+ %<br />
! #% 0 # <br />
#% !# <br />
##! #<br />
# , %*.*++,- <br />
<br />
#% # <br />
1# 23$4 !<br />
<br />
# <br />
#% #% <br />
23$4 <br />
<br />
#1 ! <br />
# 0 1# 1# <br />
! # <br />
##%<br />
<br />
23$4 # <br />
1 ;% 23$4 <br />
<br />
N
# ! #; #<br />
% #%<br />
0 <br />
# # * # - 23$4 <br />
% <br />
# 1 <br />
#% #1
;(J((I F : C (4J(8 < : 85A8$J 3 *&.- $#9 ( (<br />
& ! !# 5## <br />
2*!$23/.-;K <br />
<br />
5 2 : ;5I5J ( : K;$ < *NNN- <br />
Q38 UF2 .=<br />
##..+.==<br />
<br />
F(5;4)2 Q: 54F(4K F *&- 7 <br />
3!# ! !!Q<br />
<br />
F( +C : 3 J+ : F(( 8+C *&&- 9 ! " <br />
1 (52(&&3 !3 5 <br />
2( <br />
<br />
254!$J)F *&.- 3 ! 4# &. K L Q <br />
&.2 4 F <br />
<br />
2FF(4 *&&-JA5 )# <br />
<br />
J;2)45( :255 :3535824$) *&-6 <br />
(# Q$# 4.##&M&+&N&<br />
<br />
$Q5F5F *NN&- ( <br />
5 < <br />
<br />
3(4IJ< *NNN-
-3<br />
<br />
5## 15 !3! (,# 8%#<br />
<br />
! K P'<br />
2! <br />
P&'<br />
<br />
P<br />
!<br />
P?<br />
4< F P.<br />
2TF <br />
P'<br />
3, P'<br />
D254!$J)F3 ! 4#&.<br />
K LQ &.2 4 F <br />
5## 1 3
!" <br />
#$%!&! <br />
' <br />
<br />
#(<br />
<br />
<br />
<br />
<br />
!"! #!<br />
$% ! & "<br />
# ' $ $$($<br />
( ' # <br />
<br />
$ <br />
) # <br />
! " * <br />
$ + <br />
! #<br />
! ( , <br />
# !<br />
! <br />
<br />
)*!<br />
<br />
! "!!! #$% !<br />
# ! ! # ! ! ! #& <br />
! #' ' ! ! #! ! #<br />
###! '! ! &<br />
<br />
# !! !!!! !<br />
( #!# #! #! # <br />
! !' ! ! )<br />
! !! (#!! '<br />
" * % # ! ! ! # ' + <br />
# #! !!# # )! <br />
,# #!#'!<br />
# ! ! ! !' <br />
#!#!!,!,!!( <br />
#! , ! ## ! ! ! # <br />
' '! & #<br />
%#!!# # !!! #<br />
<br />
( -./0./1 ## 2 <br />
!# 3') ! ! ' # 3<br />
!! ! '%4<br />
!# # )( '# <br />
!#!./ !<br />
! ' 2 ! ! <br />
!
-+ !#<br />
<br />
#! ./ # #! # ! <br />
%'##'" ./ !<br />
!#! #!2 !'! !./<br />
! ' # #! # <br />
#!+(!! #, ! <br />
##'!! *! %#! #'<br />
' ! ./(! <br />
## #!! ## './<br />
<br />
<br />
+*!(! <br />
<br />
# !!! !6!! 2 <br />
' ! 01 #! <br />
./ ! '!0! 7#7##1<br />
! # !!!! #'! <br />
'# ./! !#2!#!! #!<br />
#!! #./ 0 8' 91!<br />
'! -: #!!<br />
! ' 4! #! , # ! '<br />
% ! ! '! ( ! #' ! # <br />
#! ,'!! <br />
! #! ! # <br />
!!! ! ) '# ' ' # <br />
./<br />
<br />
( #! ! # 0 '1 # ##<br />
-. /00/ ' ## ! <br />
4! ./! !#!# # ##<br />
5
-:+4!! ! <br />
<br />
- ' # ./ ! % ! <br />
! ! ! # ## - ! ' <br />
!! #!! # # ! <br />
#! #! !!3<br />
# ! #! ! ! !# <br />
! '0 # 1!!!!,<br />
#!' ' # (!! #<br />
'## !# ! !',!<br />
#! ' # ! <br />
! & #<br />
<br />
<br />
<br />
-+2! # ##<br />
<br />
# ! '! ! ##% <br />
!!#! <br />
###' # #!#!'<br />
!! ! ! !# ./! ! #<br />
<br />
;
# ## # ! ! 2<br />
''( ! ' = #'<br />
####! 3<br />
" ##! '+= <br />
3 ' !!<br />
!> # <br />
" ## #! 3<br />
(3 ' ! # <br />
!<br />
<br />
,*<br />
<br />
#! ! ./ ## # ##! - <br />
## # ! ! ./ <br />
#!# &# '-5+<br />
<br />
<br />
<br />
<br />
-5+>#!<br />
<br />
,*)-. !<br />
<br />
#!!!%'+4# <br />
' ! '#!<br />
'#!<br />
!! ! #! #!<br />
<br />
!!! ! ## <br />
! !! #./ ./%# #!<br />
> ! ! ! ' ' ' <br />
'3(3! #4 ! #<br />
!! '###!'!'<br />
! ! 3 % ! '! # <br />
#! 0;1!! <br />
# (.? 0' ( . ?' @ -#!<br />
* 1 # ( (.? ! ' #! :; % :; A #<br />
#"! 2 ! ! : 0 1 ! ! # < - <br />
!! : # ./ <br />
!# ! #(.? #<br />
;C# #"## ) ! <br />
#"!B;%:; A<br />
<br />
,*+. <br />
<br />
( # #! ! ! <br />
! # # # # ' + = <br />
! # # 0 ! # <br />
# #! 1# ! # #!<br />
#(#! ! 3! '<br />
! #!)1 /002>!<br />
! !#!<br />
<br />
# % #!! # !+<br />
!!<br />
<br />
!<br />
# <br />
,#! #<br />
!## ' '<br />
!# ''! !<br />
'#! ' # (.?<br />
' ' <br />
<br />
,*,-! <br />
<br />
D #!# ! +- <br />
! ! #"!'! #<br />
'' !2 ##<br />
#! ! ## ! ! <br />
!3 # # = ! ! <br />
./! + ! # ! #<br />
# ! !' ' ! ' ! !<br />
'% 012# !#&' <br />
!! 2'# ./ <br />
! ! !! ! ! ./ ! <br />
<br />
<br />
#! '3(3 2!<br />
! '# ! <br />
# ! 2 ! # <br />
% !' #4 #!! <br />
"# '?!''! #"!<br />
<br />
/* 0<br />
<br />
2 !' ! ! #!,<br />
! #!2-; ##! ! <br />
3(3 ./(#"!#<br />
+ ##! !! ##!#+ (<br />
!# ! ! # <br />
! !#<br />
<br />
B
!!# ' ' # ( !! <br />
<br />
<br />
<br />
-;+3#! ./ -
-B+.+H! %<br />
<br />
<br />
<br />
-E+./'#<br />
<br />
. ! # #3 <br />
!! # #= ''#<br />
!! #!2' ./ !#!<br />
'#-E#!! +<br />
<br />
01. #<br />
# ' !# <br />
2 # # <br />
#<br />
<br />
0:1. # <br />
<br />
G
!+H! !#<br />
<br />
01(' <br />
(!:## ./ <br />
<br />
051 ' 3(3<br />
! ' #!## #!! *<br />
!! ! !!##! '<br />
./3# !# !!#! ' <br />
!'<br />
<br />
' # #! ! !! <br />
./#!2' ( !#!<br />
' # ! ! # <br />
#+<br />
<br />
,2<br />
' 3 ! ! ! !<br />
! # '+ <br />
# ##! ! ! <br />
! ! 6/>-2 086!<br />
/)2 91!-.5/00/<br />
# # ' # ! # ! ! <br />
# # !#- ## <br />
! ! ./ ! ( # ' <br />
%! 63H 06!3H 1 <br />
#! <br />
<br />
3<br />
.# ! 3 ! <br />
##! '! #%'# # ! <br />
# "#! ! ./ ### <br />
' # ! <br />
# #! !#!<br />
' % # ' ! <br />
!6 #!#! #<br />
./ # # ##!! <br />
<br />
<br />
!###!! #!!<br />
! ! ### <br />
!! !' #3 ! <br />
!!!#! #! ./ <br />
!' <br />
./ #<br />
<br />
0(<br />
$# ' '' '<br />
#'! # #! ! '<br />
./ ! # #<br />
!!./! #<br />
#! ! ' #! <br />
!# ! ./ # <br />
#!# ! !!#<br />
<br />
:C
= '# '#!01./#!<br />
2' ! ! " 063H#63H##./1<br />
0:1 './# !-.5/006<br />
<br />
4*. !<br />
<br />
# ! ! ! !! <br />
!! !! !!<br />
!#! ( ! '<br />
!!! './<br />
!#!'(620##!# !1 <br />
# !# ./ ! '#<br />
% ! !!#' # <br />
!##!#./ !<br />
# !!<br />
<br />
<br />
-G+2!<br />
<br />
<br />
<br />
<br />
-C+D#! #!! <br />
<br />
:
( -G !! ./ ! 0 ./ 1 <br />
# ! 2! 63H<br />
(3 !! '#'!<br />
#'!!! ! <br />
! 2 ! ! !! <br />
' !! # ! # # ' <br />
! ! #!! # <br />
-C #! #!<br />
! <br />
<br />
5*!6<br />
<br />
( !' #! , # # ! ! <br />
!! ! !./!<br />
# #!! # ! ! <br />
#! # !./ ' !<br />
+#! 2 <br />
#!'#!!!<br />
! # ! # ! ./ ##! ## <br />
'!# "#!#<br />
<br />
2 #!! '%<br />
!!./ ! ' <br />
###!# #!(<br />
!./#!2' <br />
<br />
<br />
<br />
4/3?(I>2--?60:CC1 = $J<br />
<br />
-/($>-?/260:CC17 8(!<br />
<br />
$?3?**I-/2?3?= (D3(IK?/*?$0:CC51$9 (<br />
2&$($*(5B ##:;:<<br />
<br />
$?3?** I -/2?3?= (D3 (I K?/*? $ 0:CC;1 9 ( <br />
2H ::CC;4##B:B;<br />
<br />
$?3?** I -/2?3?= (D3 (I K> J4 0:CC:1 * 1 $-% :<br />
; # 2+ (D4( 0?1+<br />
*+ <br />
<br />
.>$ DK(* I *D? I (LK( I *?DD?/ HI = >D- ( 0:CC51 * $<br />
**!!50:CC515/ !##5G;G<br />
<br />
<br />
<br />
::
!!!"#$%& "<br />
!' !!' !()<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! " <br />
# <br />
<br />
$ $ <br />
%&' ( $<br />
<br />
<br />
<br />
"<br />
<br />
* ! ! + ! !! !+ %%+ % +# ! <br />
+! ! #" %!+ ! !++ ! + #<br />
! !! !#! # %, ! ! # ! <br />
! # ! # %!+!# # # )*+," - # ! <br />
#!++ #.!! ++!+ #/!+ <br />
+!#! !!"0#+#!! !%%<br />
! ! % # ++ ! ! - .//." - .! #<br />
! !++ ! # % +! # 1! ! 2" 3+! % !,!<br />
+ +!! !+#! !!! !+ 4 #<br />
% # ! ! !4 # % # +! ! ! + <br />
+!#!# "+!!!+,+"0123- <br />
.//.%!!#!++%+!+! ! !"<br />
<br />
3+! % #!+ % !# !+ % ) #!# .#!" !#<br />
!!+#! )# !"! !# !!+!#<br />
%)+!+"+)!+#%!+<br />
#%!+!!!+!!#!%)+#<br />
! . ! ! ! %# ! # + % ) +# %<br />
" ) ! + % !! 5!+ 3! $45 .//6 !<br />
!+.%)! !!)!<br />
! # !! % ! % #!+ % ! + !<br />
!!6 1! !/ % ! ! ! % !<br />
! % # , !! % ! #! % )<br />
% !2$45.//6 "7!!+!!#+#<br />
% ! # ++# !" 8 ! + <br />
!+% !!!)!# !+!<br />
!!!.++#"<br />
<br />
"" # $%<br />
<br />
- !+ % ) % ! ! %!+ + ! +! %!"<br />
5+#7 -.//8% !+ !+%)"!+
!!" ) #! % ! % # + + !# ! ! %<br />
!!+!+%).//," + %<br />
)! !%!++!-9:<br />
;: $45.//6%+, %#!"<br />
! # !+ + # !!+ % ! ! %%+ % <br />
+#!+!!+!+%+! #!%)$45.//6"<br />
:;%%+#! !! ###!+! <br />
?@A!!! !+! !!#%!!"0#!<br />
!!#!++"<br />
<br />
B%!#!+ ! !+#!++,!!<br />
! ! #" ' +!# !! ! # + #+ !!<br />
#!" C!! . !# !+ # % %<br />
# .+%+##!%+!#!#!<br />
#!# # ! # .+ !+ +# <br />
# #!!#D#!++!4 ()**,-<br />
!+ ! !!# !!+ % ! ! D%D! % !# <br />
!++ + ! % .! # ! % % ! ! ! % <br />
+!%+ #!; &)**+4 ()**,"'!+<br />
% + !+ # !# ! ! !! % ! !+ !<br />
!#! #!++!!!#%% !#!++"+!!##<br />
!!+# !+!+!++%<br />
##+##!#!!!4 (<br />
)**,"<br />
<br />
"&' !()<br />
<br />
),)%!!+! !#!+!#<br />
% ! # )D, ! +! !" ! + % ! E !<br />
+ ! ! ! " ) .!# !! ="" E< # + <br />
+!=+!
)%+!DD !!#!DD !!#%<br />
#!%%++=.//.")! ##! ! !!! "<br />
<br />
7 .//,!)## %# #B8,!#<br />
! % ! ! % ) ! # !!# #<br />
!# + +!" .//, ! ! % ) # ! ! # %<br />
7-+!"3%!!)##! +!! !+!<br />
#"8!!#! %#!7-1+!++2!!+!<br />
# !+!#"1+!2+!%#<br />
# ! # ! " !+ %<br />
+ +!!+ +!+! !!%%% !<br />
!"<br />
<br />
"*+<br />
<br />
!+%!+ +! +% !!+!!!!<br />
D!+!%+ +!"!!! "<br />
# ! ! + ! "<br />
7-+ +!+!! #!+!! !!! !<br />
! !!#!"+7-+ +!! #!+#+#%<br />
#.+!"0#!+D!+!!<br />
!#!++*#!4!%!+###%#+!7-<br />
+! #!% #!">)**
& <br />
<br />
0!1 #2+!.#!%)!O3+!!!!#<br />
!+!%## !%+!!+!!<br />
%+#&.///"@DG ! ! +# %# !! + <br />
! ! %,# +!+ 880 ! ! ! # % DD + +!<br />
+ !!" ) % !! + ! ! %+ ! ! +%+ <br />
+ # "<br />
<br />
!#!+!!!% #!+<br />
+!%%+"!#9%%+=#+!! !!+!<br />
!!! ! % !#!"8! # !%<br />
+!!%% !!%! ! + #+!!#!<br />
!)+# ! !%%+"+,+!!! !<br />
!#+! !!!BDL++"<br />
<br />
) %+ ! # ! ! # ! +%+ +# ! !% ! <br />
!!#+ # P ::"%!##!<br />
+%)!! !+!%+"+<br />
%!+!!#! !.+% +%+!+!"<br />
!:!9!%!.+!%% !!#<br />
) !" #! ! + !# + +F<br />
!!%!! #!%)F!%)D!!F%)! !+<br />
% ) F !!! ! !!! % ! )F ! % ! % <br />
)"%+!!!!!+ #<br />
!.+%!+!%+ !%!##!!#+"<br />
!# ! !+ ! %! !+ ! ! !!! %6<br />
%!#+%+%%!++!# !%1>+ !.2!!!%%<br />
!! !!"-! #!)<br />
#!+ ! " - <br />
# ! DD + +! ! ! <br />
!+ % ) !++" % ! +# ; ! ! <br />
#!! % !!#% !#+!! <br />
>+"<br />
<br />
*, -./0$<br />
<br />
! ! % !!!%%+%)7-+ +!"<br />
)D7- !!# !%.! #B8#)"G?=
) ! # !++ B8 = !# ,< 1 #% % !<br />
.+!=""+!# !!!# ##7-"0#%! %+)!<br />
1+!# !!2!!%#7-!%%+!<br />
!" ) ! %!+#! ++%# 7- + +! # ! <br />
%#:=.//).#+!#+!+!! %<br />
+ +!")+ !) !7-+ +!!+!!%<br />
)!#!!6<br />
<br />
)6 # ! +! # ! !! + ! ! ! % <br />
+ !4!IQ J"3 ! ! ! !+! #!<br />
!#7- %!!!!!#%%+"4#!!# !<br />
# !#!!%!#"=9=
*6R#4!##4#! !!!#<br />
!+"=9=N
7- +! #" % !> % ! +#! ! ! + ! 7- +! # <br />
+!!+%+"<br />
<br />
*& <br />
<br />
)# #!)!%L+ ::9,!#! %##"<br />
+! # 4!4 # % ) ! ! !++ <br />
% ! )D% 7- !" 8 4!D4 #<br />
#!+ +!++% !" #!+!+!# %+!"0<br />
)D% + +! ! !+ ! #! <br />
+! #! !" 0 + +! .+! % + % !<br />
!!)D !!"3%+! #!+7-<br />
+#%!!!% !"P+ +!+ <br />
+ %%+#+%!!)"++! ###% <br />
! !!+!%+%+=!<br />
!!
%%"3!#+% !7-!#!!!! % #/<br />
+!+% H#+/H!/+!#61%!!#<br />
!+!!!++!#.4#2=9=@
! !!% !#% !!!!++#%)"'<br />
# !#%!!!!! +<br />
! " B! !# ! ! !" ) %# + %<br />
% !M#.#+#!! !M!!!#%)D!"<br />
<br />
)#!%D !!!!% !!!!#%<br />
"B!% #%4!!#%#++!!%%!4+! #<br />
" %#+!!+#!+ %+ +!#!! #<br />
! % ! % +! +" +! ) #! # + +! M <br />
!+ !+!#++%#!!!#")# !<br />
+!!!#!>!# + #!+ #*#<br />
!"% !!#%!!#> ! ! % <br />
+ ! !!" +! ) % ! ! !!# #! <br />
!!+%+!!+ +!!+ #+#!!#!<br />
! !H+/"<br />
<br />
)!# !.#!!!M!!!!!<br />
! #!M !4! !# !+#!!<br />
=! #! ! ! % # ! !! +! # %# <br />
!!!!! #!!%!4!"8%% !<br />
!!! #!!!!++!"+#!#!<br />
"% !+ !%% !)D<br />
!!% !!#6<br />
<br />
)6E!++!+#"4!! "0!+!!!! !"*<br />
!$IU#!UJ")"IQ J0!! OL!!#<br />
"3#+!!!+!$O=9=
!!+ +!%#!!#+!#+#0123.//,0123<br />
;.//8"<br />
2' <br />
<br />
))!! #%#!%% !=!<br />
!!#+ #!*#!)% !
+! ! +!!!+!+"(<br />
!!##!!+ +!!+%!"<br />
<br />
!!%)D% !<br />
0 )D% ! ! ! !+ + "<br />
!!!! !%##'*L<br />
! ! + + # # # % ! <br />
!%!+!!!!! !"0)% !#!<br />
! > ! ,!# % ! !% )D<br />
% !"<br />
B! %,# .+ ! !# ! % )D ! +! %!# =<br />
#+ % ! #O )% ! # + % ! ) +!# <br />
# "<br />
-! #%.!!D%+!#)#!#!<br />
+>+ ! ! ! %!+ +!# ! # !<br />
!+!%%+"<br />
$4<br />
<br />
%!+!# % B+!# B! -! !%# !+ #"<br />
0 !# ! ! # !+! ! +! # !+ # !<br />
#3E%B)"0!!# !<br />
% !# # % ! +! # # % #+ % !" ! <br />
!!%!#!>+8 #3# +!!% "<br />
<br />
,<br />
<br />
))P5"7"=::9
6 ""#"++Y+#::9." #<br />
<br />
WV7X- B"7"F L'(('5 "0")" =::< 8 R 0!+6 ) ! 3 S"<br />
P!!;;=
The Design of PDE Hull Surfaces Us<strong>in</strong>g Genetic Algorithms<br />
Abstract<br />
Tim W. Lowe, Cranfield University, Shrivenham/UK, T.W.Lowe@cranfield.ac.uk<br />
In this paper, the automatic generation of a range of flat-sided hull forms which satisfy constra<strong>in</strong>ts<br />
on various primary and secondary geometric parameters is considered. The curved<br />
section of each hull surface is generated us<strong>in</strong>g the PDE method form<strong>in</strong>g surfaces as solutions<br />
to a partial differential equation with appropriate boundary conditions. The method enables a<br />
wide range of smooth surface shapes to be represented us<strong>in</strong>g relatively few design parameters.<br />
The design parameter space of the hulls is searched us<strong>in</strong>g a genetic algorithm to locate regions<br />
satisfy<strong>in</strong>g the geometric requirements. The designs found could be used as start<strong>in</strong>g po<strong>in</strong>ts of a<br />
functional design optimisation as demonstrated <strong>in</strong> an example.<br />
1 Introduction<br />
The prelim<strong>in</strong>ary stages of the design of a new vessel often <strong>in</strong>volve specify<strong>in</strong>g values, or a range<br />
of possible values, of various global geometric properties of the hull. These properties, often<br />
known as form parameters, <strong>in</strong>clude the primary dimensions of the vessel, the displacement,<br />
the positions of the centres of buoyancy and flotation and various non-dimensional quantities<br />
such as the block and water-plane coefficients. Whilst some <strong>in</strong>itial work can be performed <strong>in</strong><br />
terms of such measures, it soon becomes necessary to f<strong>in</strong>d a hull surface hav<strong>in</strong>g these desired<br />
characteristics. Ideally, this form should be as close as possible to the f<strong>in</strong>al hull design to<br />
m<strong>in</strong>imise the overheads aris<strong>in</strong>g from subsequent design modifications, hence support<strong>in</strong>g a timely<br />
and economic design cycle. The hull form should also be represented <strong>in</strong> a manner consistent<br />
with the computer-aided design tools currently used.<br />
The two possible approaches to the determ<strong>in</strong>ation of such forms are to distort an exist<strong>in</strong>g hull<br />
us<strong>in</strong>g, for example, Lackenby (1950) transforms or to generate a completely new design from<br />
first pr<strong>in</strong>ciples. The latter approach has the advantage of not restrict<strong>in</strong>g the hull found to be<strong>in</strong>g<br />
close to an exist<strong>in</strong>g form thus permitt<strong>in</strong>g the generation of novel, possibly improved, designs.<br />
The <strong>in</strong>teractive manipulation of geometric models to generate new hull forms hav<strong>in</strong>g specified<br />
form parameters, for example through the movement of surface control po<strong>in</strong>ts, is however, very<br />
time consum<strong>in</strong>g. Recently, several automatic approaches have been suggested.<br />
Peacock et al. (1997) obta<strong>in</strong> a hull hav<strong>in</strong>g specified values of certa<strong>in</strong> form parameters by first<br />
f<strong>in</strong>d<strong>in</strong>g B-spl<strong>in</strong>e representations of various traditional characteristic design curves, for example<br />
the section-area curve or the design waterl<strong>in</strong>e, which correspond to the required form whilst<br />
simultaneously maximis<strong>in</strong>g a simple measure of curve fairness. This is achieved us<strong>in</strong>g the decision<br />
support problem technique. A similar process is then used to form hull sections based on these<br />
curves.<br />
Characteristic design curves are also used by Harries and Nowacki (1999) who aga<strong>in</strong> generate<br />
them by maximis<strong>in</strong>g a measure of curve fairness subject to the satisfaction of constra<strong>in</strong>ts on<br />
the values of specified form parameters. In their work, this is achieved us<strong>in</strong>g a Newtonian<br />
optimisation scheme and the constra<strong>in</strong>ts applied through the use of Lagrange multipliers. Harries<br />
and Nowacki use these curves to obta<strong>in</strong> hull sections and hence a B-spl<strong>in</strong>e representation of the<br />
hull itself, by aga<strong>in</strong> maximis<strong>in</strong>g a fairness measure. In later papers, for example Abt et al. (2003),<br />
the geometric generation of hull forms us<strong>in</strong>g this method is comb<strong>in</strong>ed with the optimisation of<br />
the hydrodynamical properties of the vessel to f<strong>in</strong>d functionally optimal hulls.<br />
Birm<strong>in</strong>gham and Smith (1998) by-pass the use of traditional characteristic curves by directly<br />
seek<strong>in</strong>g the control po<strong>in</strong>ts of a B-spl<strong>in</strong>e representation of the hull surface that possesses the<br />
136
equired values of the form parameters and maximises a measure of fairness. This is achieved<br />
us<strong>in</strong>g a genetic algorithm (GA) to search the design space.<br />
A similar GA search is performed by Islam et al. (2001). The GA used is <strong>in</strong>itialised with<br />
random variations of a s<strong>in</strong>gle hull form match<strong>in</strong>g the required primary dimensions, <strong>in</strong>stead of<br />
the more usual general random population. This form is based on historical data and adjusted<br />
to the specific problem be<strong>in</strong>g considered us<strong>in</strong>g a neural network. The GA is used to f<strong>in</strong>d a<br />
hull match<strong>in</strong>g the required secondary parameters. Hull fairness is ma<strong>in</strong>ta<strong>in</strong>ed by fair<strong>in</strong>g each<br />
<strong>in</strong>dividual <strong>in</strong> each generation of the GA before its merit is evaluated. This seed<strong>in</strong>g of the GA<br />
improves the speed of convergence of the algorithm, but may also prevent the discovery of forms<br />
far removed from the base hull s<strong>in</strong>ce the variation <strong>in</strong> the <strong>in</strong>itial population is reduced.<br />
Kawashima and H<strong>in</strong>o (2004) take a different approach to the problem and represent the hull by<br />
the set of curves bound<strong>in</strong>g the surface together with l<strong>in</strong>es def<strong>in</strong><strong>in</strong>g the hollows and ridges with<strong>in</strong><br />
it. Sequential quadratic programm<strong>in</strong>g is used to vary these l<strong>in</strong>es to f<strong>in</strong>d a surface match<strong>in</strong>g the<br />
required form parameters. This procedure starts from either an exist<strong>in</strong>g hull form or a set of<br />
basic hull l<strong>in</strong>es drawn from a database.<br />
All these methods deliver a s<strong>in</strong>gle fair hull form satisfy<strong>in</strong>g the design requirements. However,<br />
<strong>in</strong> general there is no reason to suppose that there exists only one hull possess<strong>in</strong>g the desired<br />
properties. Indeed, the fewer restrictions placed on the hull at this early stage <strong>in</strong> the design<br />
process the larger the number of possible hull surfaces that satisfy the requirements. Lowe<br />
and Steel (2003) used a GA was used not to seek the s<strong>in</strong>gle form that possesses the required<br />
characteristics and maximises some measure of fairness, but to search the design parameter<br />
space for those regions which conta<strong>in</strong> hull forms satisfy<strong>in</strong>g the design requirements. Surface<br />
fairness was ensured by means of the method of surface representation used, namely the PDE<br />
method, Bloor and Wilson (1990a). By generat<strong>in</strong>g surfaces parametrically as the solutions to<br />
an elliptic partial differential equation, a smooth surface without superfluous humps or hollows<br />
is guaranteed.<br />
This approach to design generates a range of possible hull forms that can either be taken as<br />
the basis for subsequent geometric modifications by a skilled naval architect or used as start<strong>in</strong>g<br />
po<strong>in</strong>ts for an automated optimisation of the vessels functional properties, for example wave<br />
resistance, Lowe et al. (1994), Markov and Suzuki (2001).<br />
This previous work is extended to the design of flat-sided hull forms satisfy<strong>in</strong>g geometric requirements<br />
given <strong>in</strong> terms of either specified values or a range of permissible values of certa<strong>in</strong><br />
form parameters. The use of the designs found as start<strong>in</strong>g po<strong>in</strong>ts for a hydrodynamic design<br />
optimisation is also illustrated.<br />
2 Hull generation<br />
The geometry and ma<strong>in</strong> dimensions of the hull forms considered <strong>in</strong> this work, together with the<br />
Cartesian coord<strong>in</strong>ate system used, are shown <strong>in</strong> Fig.1. The hulls considered are symmetrical<br />
port and starboard, and hence only the starboard half is labelled. The region with<strong>in</strong> the closed<br />
curve BCDE is flat. The rema<strong>in</strong><strong>in</strong>g curved surface is generated us<strong>in</strong>g the PDE method.<br />
The length of the deck is denoted l d and its beam b d . The deck parallel section (BE) is of length<br />
l dp and its mid-po<strong>in</strong>t positioned a distance x dp aft of the bow. Similarly, the lower boundary of<br />
the flat side (CD) is a distance d f below the deck, of length l f and its midpo<strong>in</strong>t positioned x f<br />
aft of the bow. The keel-l<strong>in</strong>e (KH) is of length l k , centred x k aft of A. The draught of the hull<br />
is denoted T and the freeboard f. Note that if either of the length (l f ) or depth (d f ) of the flat<br />
region of a particular hull is less than some small tolerance, the flat region is omitted on that<br />
vessel.<br />
137
z<br />
✛<br />
✻<br />
l d ✲<br />
y ✛ l dp ✲<br />
A ✒ B<br />
E<br />
✲ x C ✻ ❄<br />
d f D<br />
F<br />
G <br />
✛ l f ✲<br />
<br />
b d<br />
f ✻<br />
✠<br />
❄<br />
✻<br />
T<br />
❄<br />
✛<br />
K J<br />
l k<br />
I H<br />
✲<br />
Fig.1: Hull geometry<br />
The PDE method generates surfaces parametrically as the solution to an elliptic partial differential<br />
equation (PDE). Information is specified around the edge of the doma<strong>in</strong> of solution (the<br />
surface boundaries) and a PDE solved to generate the surface <strong>in</strong> the <strong>in</strong>terior. Us<strong>in</strong>g an appropriate<br />
parametrisation of these boundary conditions, a wide range of hull forms can represented<br />
us<strong>in</strong>g relatively few design parameters. This property is advantageous to any automated search<br />
of the design space, s<strong>in</strong>ce it reduces the dimension of that space.<br />
In the current work, follow<strong>in</strong>g Bloor and Wilson (1990b), the hull is generated <strong>in</strong> its two symmetric<br />
halves. Consider<strong>in</strong>g the starboard half of the vessel, the curved surface is bounded by<br />
the deck and so-called flat-of-side curve (ABCDEF G), the bow and stern profile l<strong>in</strong>es (AK and<br />
GH) and the keel l<strong>in</strong>e (KJIH). If the surface is represented parametrically as<br />
X(u, v) = (x(u, v), y(u, v), z(u, v))<br />
where 0 ≤ u, v ≤ 1 are the surface parametric coord<strong>in</strong>ates, the boundaries of the parametric<br />
space can be identified with the boundaries of the curved hull surface, Fig.2. The u and v axes<br />
are oriented such that u runs from the deck to the keel and v from bow to stern.<br />
0<br />
deck profile<br />
and flat-of-side<br />
A B C D E F G<br />
1<br />
✲ v<br />
bow<br />
profile<br />
stern<br />
profile<br />
1<br />
K<br />
J<br />
keel<br />
I<br />
H<br />
❄<br />
u<br />
Fig.2: Surface parametric space<br />
The upper boundary curve (X(0, v)) is formed from:<br />
138
• A cubic parametric curve AB <strong>in</strong> the plane z = 0, hav<strong>in</strong>g tangent at A <strong>in</strong> the direction<br />
(α db , β db , 0) (and thus a deck entrance angle of 2 tan −1 β db<br />
α db<br />
) and tangent at B parallel to<br />
the x axis with magnitude t f<br />
• A quadratic parametric curve <strong>in</strong> BC <strong>in</strong> the plane y = b d<br />
2<br />
with horizontal tangent at C<br />
• The straight l<strong>in</strong>e CD<br />
• A quadratic parametric curve DE <strong>in</strong> the plane y = b d<br />
2<br />
with horizontal tangent at D<br />
• A cubic parametric curve EF with tangent (t a , 0, 0) at E and (0, t t , 0) at F<br />
• The straight transom l<strong>in</strong>e F G of length bt<br />
2<br />
The lower boundary curve (X(1, v)) is the straight, horizontal keel l<strong>in</strong>e KH. The bow profile l<strong>in</strong>e<br />
is generated from a cubic parametric curve with tangent (α db , 0, γ db ) at the deck and (α kb , 0, 0)<br />
at the keel. The stern l<strong>in</strong>e is similarly generated with tangents (α ds , 0, γ ds ) and (0, 0, γ ks ) at the<br />
deck and keel respectively.<br />
To provide more detailed control over the form of the surface, <strong>in</strong> addition to these positional<br />
conditions, the tangents to the surface ( ∂X ∂X<br />
du<br />
or<br />
dv<br />
) at the boundaries are also specified. These<br />
vary smoothly around the surface boundary between the values illustrated <strong>in</strong> Fig.3. The surface<br />
tangent along the keel l<strong>in</strong>e is constant between J and I, a length denoted l kp the midpo<strong>in</strong>t of<br />
which has coord<strong>in</strong>ates (x kp , 0, −f − T ).<br />
β db<br />
B<br />
E<br />
β ds<br />
A ✒ ✲ C<br />
D<br />
F<br />
α db ❄ ❄ G ✒ ✲ γf γf α ds<br />
❄<br />
γ ks ❄<br />
γ db ✒ β kp<br />
✒ β kp<br />
✒ β<br />
✛ kb<br />
✻ γ ds<br />
α kb<br />
K J<br />
I H<br />
Fig.3: Hull surface tangents<br />
The curved part of the hull surface is obta<strong>in</strong>ed by solv<strong>in</strong>g an elliptic partial differential equation<br />
subject to these conditions. S<strong>in</strong>ce conditions on both X and its normal derivative around the<br />
boundary of the doma<strong>in</strong> of solution are specified, a fourth order equation is necessary. As <strong>in</strong><br />
previous work us<strong>in</strong>g the PDE method, the equation used to generate the surface is<br />
(<br />
∂<br />
2<br />
) 2<br />
∂2<br />
+ a2<br />
∂u2 ∂v 2 X = 0 (1)<br />
The parameter a is known as the smooth<strong>in</strong>g parameter and affects the extent of surface over<br />
which the boundary conditions have <strong>in</strong>fluence. Vary<strong>in</strong>g a changes the “fullness” of the surface:<br />
large values of a give a surface ly<strong>in</strong>g close to the centre-plane of the vessel (whilst, of course,<br />
satisfy<strong>in</strong>g the imposed boundary conditions) and small values give surfaces that are far from it.<br />
In the present work, equation (1) is solved numerically <strong>in</strong> the u-v parameter space us<strong>in</strong>g the<br />
f<strong>in</strong>ite difference method. A solution grid hav<strong>in</strong>g m po<strong>in</strong>ts <strong>in</strong> the u direction and n <strong>in</strong> v is taken.<br />
The banded nature of the discretised system of equations permits efficient sparse matrix storage<br />
and solution methods to be used.<br />
139
The hull geometry is thus characterised by the 29 design variables l d , T , f, b d , l dp , x dp , l f , x f ,<br />
d f , b t , t a , t f , t t , l k , x k , l kp , x kp , α db , β db , γ db , γ f , α ds , β ds , γ ds , α kb , β kb , β kp , γ ks and a. Note<br />
that not all comb<strong>in</strong>ations of values of these parameters lead to valid hull forms. For example, it<br />
is necessary that the length of parallel deck side (l dp ) is not greater than the length of the deck<br />
itself (l d ), and that it does not protrude beyond the bow or stern. The full set of constra<strong>in</strong>ts the<br />
parameters must satisfy to give a valid hull are:<br />
x dp − 1 2 l dp > 0<br />
x dp + 1 2 l dp < l d<br />
x f − 1 2 l f > x dp − 1 2 l dp<br />
x f + 1 2 l f < x dp + 1 2 l dp<br />
x k − 1 2 l k > 0<br />
(2)<br />
b t < b d<br />
x k + 1 2 l k < l d<br />
x kp − 1 2 l kp > x k − 1 2 l k<br />
x kp + 1 2 l kp < x k + 1 2 l k<br />
3 Geometric design space search<br />
The geometric requirements of the hulls sought consist of both specified values and permitted<br />
ranges of various characteristics. The example requirements considered here are taken to be:<br />
f = 1 2 T (3)<br />
T < 15m (4)<br />
V = 26000m 3 (5)<br />
1 < b w<br />
T < 4 (6)<br />
0.55 < c b < 0.65 (7)<br />
0.75 < c wp < 0.90 (8)<br />
1% aft of midships < l cb < 1.5% aft of midships (9)<br />
|l cb − l cf | < 0.5% l w (10)<br />
140
0.3 < A f<br />
A s<br />
(11)<br />
where V is the displacement of the hull, b w the waterl<strong>in</strong>e beam, c b the block coefficient, c wp the<br />
water-plane coefficient, l cb the longitud<strong>in</strong>al location of the centre of buoyancy, l cf the longitud<strong>in</strong>al<br />
location of the centre of flotation, A f the area of flat side, A s the hull surface area and l w the<br />
waterl<strong>in</strong>e length. The constra<strong>in</strong>ts were chosen as be<strong>in</strong>g typical of a generic cargo vessel. The<br />
last (11) is <strong>in</strong>cluded to ensure the hull forms generated do <strong>in</strong>deed conta<strong>in</strong> the flat side sought.<br />
The <strong>in</strong>dividual form parameters are calculated by <strong>in</strong>terrogat<strong>in</strong>g the surface X obta<strong>in</strong>ed by the<br />
solution of equation (1).<br />
Note that the first constra<strong>in</strong>t (3) can be satisfied by simply remov<strong>in</strong>g f from the list of design<br />
variables, its value be<strong>in</strong>g determ<strong>in</strong>ed by T . This reduces the number of active design variables<br />
to 28. Hulls satisfy<strong>in</strong>g the rema<strong>in</strong><strong>in</strong>g constra<strong>in</strong>ts ((4)–(11)) are found by search<strong>in</strong>g the design<br />
parameter space us<strong>in</strong>g a genetic algorithm.<br />
GAs are heuristic search methods <strong>in</strong>spired by the processes of Darw<strong>in</strong>ian evolution and natural<br />
selection, Holland (1975). An <strong>in</strong>itial randomly generated population of <strong>in</strong>dividuals, <strong>in</strong> this case,<br />
a collection of hull forms generated from random values of the design variables, “evolves” over<br />
time by seek<strong>in</strong>g to optimise some merit function through the use of so-called genetic operations.<br />
GAs have been described widely <strong>in</strong> the literature and the method used here is described <strong>in</strong> Lowe<br />
and Steel (2003). For completeness, the method may be summarised as follows.<br />
The “fitness” of each <strong>in</strong>dividual <strong>in</strong> the population is first assessed us<strong>in</strong>g a measure of merit, F .<br />
Here, this is taken to be the average of the square of the amount by which each requirement is<br />
violated:<br />
F = 1 7∑<br />
c 2 i<br />
7<br />
where c i is size of the violation of constra<strong>in</strong>t i. Low values of F imply a high hull fitness.<br />
Individuals are then randomly selected to form an <strong>in</strong>termediate population, with the selection<br />
be<strong>in</strong>g biased towards the fittest <strong>in</strong>dividuals. Each of these <strong>in</strong>dividuals may undergo a genetic<br />
operation, the probability of selection for which is a parameter of the method. Here, two<br />
operations are used: simple crossovers <strong>in</strong> which the values of the design parameters subsequent<br />
to a randomly selected parameter are swapped between two <strong>in</strong>dividuals, and whole cross-overs<br />
<strong>in</strong> which two <strong>in</strong>dividuals are replaced by two new <strong>in</strong>dividuals ly<strong>in</strong>g on the l<strong>in</strong>e jo<strong>in</strong><strong>in</strong>g their<br />
correspond<strong>in</strong>g locations <strong>in</strong> parameter space. F<strong>in</strong>ally, the result<strong>in</strong>g designs are subject to random<br />
mutations <strong>in</strong> which one parameter value of a selected <strong>in</strong>dividual is randomly replaced with a<br />
value from the permitted range of that parameter. This process of form<strong>in</strong>g a new generation of<br />
<strong>in</strong>dividuals is repeated a specified number of times.<br />
In the present study, it is not a s<strong>in</strong>gle hull form that satisfies the imposed constra<strong>in</strong>ts that is<br />
sought but a range of hulls representative of the regions of parameter space <strong>in</strong> which suitable<br />
designs can be found. As the GA progress, every hull discovered that satisfies the requirements,<br />
that is, has a merit value less than some small tolerance, is thus recorded.<br />
Due to the stochastic nature of the GA, different results may be obta<strong>in</strong>ed each time the search<br />
is conducted. For this reason the whole process is repeated a number of times to sample the<br />
solution space more widely.<br />
A consequence of represent<strong>in</strong>g the <strong>in</strong>dividuals with<strong>in</strong> the population us<strong>in</strong>g values of the design<br />
parameters described <strong>in</strong> the previous section is that many of the hull forms considered will<br />
violate the geometrical constra<strong>in</strong>ts of equation (2). In the previous work, this was accounted<br />
for <strong>in</strong> the design process by assign<strong>in</strong>g an artificially low fitness to any forms that violate the<br />
conditions, prevent<strong>in</strong>g the GA select<strong>in</strong>g them for the next generation. This, however, degrades<br />
i=1<br />
141
the efficiency of the method by effectively reduc<strong>in</strong>g the variation <strong>in</strong> the population from which<br />
<strong>in</strong>dividuals are selected and modified. In the current work, search parameters are used that are<br />
related to the geometric variables of Section 2, but chosen <strong>in</strong> such a way so as to guarantee<br />
the satisfaction of the constra<strong>in</strong>ts of equation (2). For example, <strong>in</strong>stead of randomly specify<strong>in</strong>g<br />
the length of the parallel deck section, l dp , the fraction of the deck length that is parallel is<br />
given. Undesirable hull forms, such as those whose sides descend below the level of the keel<br />
before turn<strong>in</strong>g to meet this boundary are, nevertheless, prevented by assign<strong>in</strong>g an artificially<br />
high merit function value to these <strong>in</strong>dividuals.<br />
The search space is non-dimensionalised with respect to the length scale of V 1 3 , this be<strong>in</strong>g the<br />
only constant length scale of the problem. The limits of the search space are taken to be those<br />
for which the non-dimensional search parameters lie between 0 and 1, except for a and the<br />
non-dimensionalised equivalents of l d , b d and T . To permit hulls of realistic dimensions to be<br />
accessible to the method, l d /V 1 3 is allowed to take values <strong>in</strong> the range 0 to 10, and b d /V 1 3 <strong>in</strong><br />
the range 0 to 2. The range of T/V 1 3 is specified so as to ensure satisfaction of constra<strong>in</strong>t (4).<br />
The smooth<strong>in</strong>g parameter, a, is permitted to vary between 0.01 and 1. Although hulls can be<br />
generated with a greater than one, as expla<strong>in</strong>ed <strong>in</strong> Section 2 these have a surface that rema<strong>in</strong>s as<br />
close to the hull centre-plane as the boundary conditions permit. Such forms are not desirable <strong>in</strong><br />
this search and are excluded by this choice for the range of a. This aga<strong>in</strong> improves the efficiency<br />
of the method by reduc<strong>in</strong>g the size of the search space. The lower limit of 0.01 is necessary to<br />
avoid numerical difficulties that are encountered at very small values.<br />
The process described above yields a large number of hulls which possess the desired geometrical<br />
characteristics. Due to the nature of the GA, many of these hulls will be similar, or <strong>in</strong>deed<br />
identical, to other forms found. A cluster<strong>in</strong>g algorithm, Thorp and Pierson (1998) is thus used<br />
to identify group<strong>in</strong>gs of the hulls found with<strong>in</strong> the parameter space, and to locate a s<strong>in</strong>gle design<br />
representative of each of these group<strong>in</strong>gs. It is these designs which are characteristic of the<br />
various forms found that are presented to the user as candidate hull designs.<br />
4 Results<br />
The f<strong>in</strong>ite difference mesh size used <strong>in</strong> solv<strong>in</strong>g equation (1) to generate the hull surfaces was<br />
m = 20, n = 100. Prelim<strong>in</strong>ary computational experiments <strong>in</strong>dicated that such a mesh permits<br />
the geometric parameters of <strong>in</strong>terest to be calculated with<strong>in</strong> 0.3%. The computational time<br />
taken to calculate the surface and evaluate the properties of an <strong>in</strong>dividual hull form was typically<br />
0.02 secs us<strong>in</strong>g an Intel Pentium 4 L<strong>in</strong>ux workstation.<br />
The genetic algorithm used 100 <strong>in</strong>dividuals <strong>in</strong> each population and was run for 100 generations.<br />
Any hull form with a merit value of less than 10 −6 was recorded as satisfy<strong>in</strong>g the constra<strong>in</strong>ts.<br />
This process was repeated 5 times before the hulls representative of the design clusters found<br />
were identified.<br />
A total of 2656 hulls were recorded as satisfy<strong>in</strong>g the geometric requirements which may conta<strong>in</strong><br />
multiple <strong>in</strong>stances of various <strong>in</strong>dividual forms, and 6 cluster representatives were identified. The<br />
entire process took 21.0 m<strong>in</strong>utes.<br />
Figs.4-7 illustrate a selection of the hull forms found. These show the body plans for each hull<br />
together with a perspective view superimposed with hull sections and buttock l<strong>in</strong>es. Whilst<br />
the perspective view is scaled appropriately for each form, the scale used for the body plans<br />
rema<strong>in</strong>s the same <strong>in</strong> each figure. The primary dimensions of the forms presented, together with<br />
the values of the constra<strong>in</strong>ed quantities are given <strong>in</strong> Table I.<br />
A range of hull forms have been obta<strong>in</strong>ed which are representative of those regions of the<br />
142
design parameter space satisfy<strong>in</strong>g the design requirements. The small violations of a few of the<br />
constra<strong>in</strong>ts seen <strong>in</strong> Table I are a result of the GA be<strong>in</strong>g governed by a merit function F which<br />
is the average of the <strong>in</strong>dividual constra<strong>in</strong>t violations, and the fact acceptable designs have been<br />
recorded on the basis of small values of this function. An alternative method would be to accept<br />
a design only if each constra<strong>in</strong>t was <strong>in</strong>dividually satisfied. The errors <strong>in</strong>troduced due to the<br />
scheme used are, however, small.<br />
Fig.4: Hull form A<br />
Fig.5: Hull form B<br />
Fig.6: Hull form C<br />
Fig.7: Hull form D<br />
143
Table I: Geometric properties of the designs presented<br />
Hull A Hull B Hull C Hull D<br />
l d (m) 123.8 187.2 193.6 230.1<br />
l w (m) 108.9 179.1 179.6 224.1<br />
b d (m) 31.2 21.5 24.0 17.7<br />
f (m) 6.3 5.9 5.5 5.0<br />
T (m) 12.7 11.8 11.0 10.0<br />
V (m 3 ) 25968 25996 26020 25995<br />
b w /T 2.46 1.82 2.19 1.75<br />
c b 0.603 0.572 0.549 0.652<br />
c wp 0.880 0.889 0.902 0.855<br />
l cb (% aft of midships) 1.08 1.28 1.19 1.24<br />
|l cb − l cf |/l w (%) 0.40 0.48 0.31 0.23<br />
A f /A s 0.31 0.30 0.30 0.30<br />
Across the range of designs found, many of the constra<strong>in</strong>ed parameters take values throughout<br />
their acceptable range. The exceptions to this are the water-plane coefficient which tends to<br />
take a value <strong>in</strong> the upper-end of the permitted range, and the ratio of the area of flat-side to<br />
the total surface area of the hull which is close to the m<strong>in</strong>imum permitted value <strong>in</strong> each case.<br />
5 Functional design<br />
The hull forms presented above were created purely on the basis of satisfy<strong>in</strong>g the imposed<br />
geometric requirements, with no regard for the hydrodynamic performance of the vessel. This<br />
aspect of design can be addressed by us<strong>in</strong>g the forms found by the GA as start<strong>in</strong>g po<strong>in</strong>ts for a<br />
determ<strong>in</strong>istic optimisation scheme to improve hull performance. As an illustration of this, the<br />
wave resistance of hull form B has been reduced us<strong>in</strong>g a Newtonian optimisation scheme, namely<br />
the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method. The search was conducted us<strong>in</strong>g the<br />
same scaled design parameters and parameter space limits as used for the GA. Constra<strong>in</strong>t<br />
(3) on the freeboard of the vessel was aga<strong>in</strong> satisfied by remov<strong>in</strong>g f from the set of active<br />
design variables and constra<strong>in</strong>t (4) by the prescription of limits on the permitted values of the<br />
scaled design variable related to T . The equality constra<strong>in</strong>t (5) on the hull displacement was<br />
satisfied by restrict<strong>in</strong>g the search to be<strong>in</strong>g over that surface with<strong>in</strong> parameter space on which<br />
the displacement takes the specified value. This was achieved us<strong>in</strong>g gradient projection. The<br />
rema<strong>in</strong><strong>in</strong>g constra<strong>in</strong>ts ((6)–(11)) were <strong>in</strong>corporated us<strong>in</strong>g penalty methods. Further details of<br />
the methods used are described <strong>in</strong> Lowe (2001).<br />
The functional design search was guided by the need to reduce the wave resistance of the vessel<br />
at a design speed of 20 knots. In the present work, this was estimated us<strong>in</strong>g th<strong>in</strong>-ship theory,<br />
although a more sophisticated method could easily be used. Table II gives the parameter values<br />
of the design obta<strong>in</strong>ed (together with those of the <strong>in</strong>itial form taken). The wave-resistance<br />
coefficient c w was def<strong>in</strong>ed with respect to the water-l<strong>in</strong>e length of hull B:<br />
c w =<br />
R w<br />
1<br />
2 ρU 2 l w (B) 2<br />
with U be<strong>in</strong>g the vessel’s speed and ρ the fluid density. The result<strong>in</strong>g form is illustrated <strong>in</strong> Fig.8,<br />
with the body l<strong>in</strong>es of both this and the <strong>in</strong>itial hull be<strong>in</strong>g shown <strong>in</strong> Fig.9.<br />
The design optimisation reduced the wave resistance of the hull, as estimated by th<strong>in</strong>-ship theory,<br />
by 90% whilst ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g the required geometrical characteristics. This large reduction <strong>in</strong> c w<br />
144
is partly accounted for by the <strong>in</strong>appropriateness of the <strong>in</strong>itial hull form B to the specified design<br />
speed, Fig.10 (where the Froude number F n is also def<strong>in</strong>ed <strong>in</strong> terms of the waterl<strong>in</strong>e length of<br />
hull B). The functional optimisation has <strong>in</strong>creased both the length and the draught of the vessel,<br />
both properties known to reduce wave resistance. As a consequence of the constra<strong>in</strong>t on the<br />
displacement of the vessel, the beam has correspond<strong>in</strong>gly fallen.<br />
Table II: Hull parameters of the functional optimisation<br />
Design Initial F<strong>in</strong>al<br />
(Hull B)<br />
l d (m) 187.2 198.0<br />
l w (m) 179.1 186.2<br />
b d (m) 21.5 18.6<br />
f (m) 5.9 6.7<br />
T (m) 11.8 13.4<br />
V (m 3 ) 25996 26000<br />
b w /T 1.82 1.39<br />
c b 0.572 0.560<br />
c wp 0.889 0.810<br />
l cb (% aft of midships) 1.28 1.10<br />
|l cb − l cf |/l w (%) 0.48 0.34<br />
A f /A s 0.30 0.35<br />
c w (×10 −3 ) 2.156 0.212<br />
Fig.8: Hydrodynamically optimised hull form Fig.9: L<strong>in</strong>es of optimised hull (—)<br />
and <strong>in</strong>itial form B (· · ·)<br />
3 x 10−3 F n<br />
2.5<br />
c w<br />
2<br />
1.5<br />
1<br />
0.5<br />
0<br />
0.15 0.2 0.25 0.3 0.35<br />
Fig.10: c w vs. F n for optimised hull (—) and hull B (- -). The design speed is <strong>in</strong>dicated.<br />
145
6 Conclusions<br />
The use of a GA to locate regions of design parameter space conta<strong>in</strong><strong>in</strong>g hull forms hav<strong>in</strong>g<br />
specified geometrical characteristics has been demonstrated. An example set of geometric requirements<br />
was used, consist<strong>in</strong>g of both equality and <strong>in</strong>equality constra<strong>in</strong>ts on parameters of<br />
<strong>in</strong>terest. Further constra<strong>in</strong>ts are easily <strong>in</strong>corporated to remove excessive design freedoms, for<br />
example by restrict<strong>in</strong>g the length or beam of the vessel.<br />
The hull forms considered were generated us<strong>in</strong>g the PDE method. This permits a succ<strong>in</strong>ct parameterisation<br />
of the geometry hence reduc<strong>in</strong>g the dimension of the design space to be searched.<br />
The PDE method has the additional advantage of generat<strong>in</strong>g hull forms without superfluous<br />
surface ripples. Surfaces obta<strong>in</strong>ed us<strong>in</strong>g this method cannot, however, be directly imported <strong>in</strong>to<br />
current computer-aided hull design packages. This can be circumvented by represent<strong>in</strong>g PDE<br />
surfaces <strong>in</strong> terms of B-spl<strong>in</strong>es as described by Bloor and Wilson (1990c).<br />
The designs discovered may not represent every region of parameter space that satisfies the<br />
design requirements; further designs may be discovered if the GA was run more times. The<br />
forms identified are also restricted to those hulls representable by the PDE method us<strong>in</strong>g the<br />
boundary conditions described. This, nevertheless, is seen to be a broad class.<br />
Whilst the GA has effectively searched the design parameter space for those regions of<br />
<strong>in</strong>terest, the disadvantage of this method is the number of <strong>in</strong>dividual hull designs that must be<br />
considered. Each of the 100 generations of the GA required the calculation and <strong>in</strong>terrogation<br />
of 100 candidate hulls and this was repeated 5 times. S<strong>in</strong>ce the computational time required<br />
for the calculation of the hull surface and its geometric properties is small, this is not too<br />
burdensome. However, the use of such a method to optimise more computationally <strong>in</strong>tensive<br />
merit functions, for example wave resistance, would require significant computational expense.<br />
A more computationally efficient approach would be to use the GA to f<strong>in</strong>d those regions of<br />
space conta<strong>in</strong><strong>in</strong>g valid hull forms, and beg<strong>in</strong>n<strong>in</strong>g a determ<strong>in</strong>istic functional optimisation scheme<br />
from such po<strong>in</strong>ts. This approach has been demonstrated above.<br />
Acknowledgement<br />
The author would like to thank J. Steel for writ<strong>in</strong>g the genetic algorithm library used <strong>in</strong> this work.<br />
References<br />
ABT, C.; HARRIES, S.; HEIMANN, J.; WINTER, H (2003), From redesign to optimal hull<br />
l<strong>in</strong>es by means of parametric model<strong>in</strong>g, <strong>COMPIT</strong>’03, <strong>Hamburg</strong><br />
BIRMINGHAM, R.W.; SMITH, T.A.G (1998), Automatic hull form generation: a practical tool<br />
for design and research, PRADS’98, The Hague, pp.281-287<br />
BLOOR, M.I.G.; WILSON, M.J. (1990a), Us<strong>in</strong>g partial differential equations to generate freeform<br />
surfaces, Computer-Aided Design 22/4, pp.202-212<br />
BLOOR, M.I.G.; WILSON, M.J. (1990b), Geometric design of hull forms us<strong>in</strong>g partial differential<br />
equations, CFD and CAD <strong>in</strong> Ship Design, Elsevier, pp.65-73<br />
BLOOR, M.I.G.; WILSON, M.J. (1990c), Represent<strong>in</strong>g PDE surfaces <strong>in</strong> terms of B-spl<strong>in</strong>es,<br />
Computer-Aided Design 22/6, pp.324-331<br />
HARRIES, S.; NOWACKI, H. (1999), Form parameter approach to the design of fair hull shapes,<br />
ICCAS’99, Cambridge, MA.<br />
HOLLAND, J.H. (1975), Adaptation <strong>in</strong> Natural and Artificial Systems, University of Michigan<br />
146
Press, Ann Arbor, MI.<br />
ISLAM, M.M.; KHONDOKER, M.R.H.; RAHMAN, C.M. (2001), Application of artificial <strong>in</strong>telligence<br />
techniques <strong>in</strong> automatic hull form generation, Ocean Eng<strong>in</strong>eer<strong>in</strong>g 28, pp.1531-1544<br />
KAWASHIMA, H.; HINO, T. (2004), A hull form generation method on <strong>in</strong>itial design stage,<br />
PRADS’04, Lubeck-Travemunde<br />
LACKENBY, H. (1950), On the systematic geometrical variation of ship forms, Trans. Institution<br />
of Naval Architects 92, pp.289-316<br />
LOWE, T.W.; BLOOR, M.I.G.; WILSON, M.J. (1994), The automatic functional design of hull<br />
surface geometry, J. Ship Res. 38/4, pp.319-328<br />
LOWE, T.W. (2001), Variable-complexity hydrodynamic optimisation of a yacht hull, Mathematical<br />
Eng<strong>in</strong>eer<strong>in</strong>g <strong>in</strong> Industry 8/3, pp.253-274<br />
LOWE, T.W.; STEEL, J. (2003), Conceptual hull design us<strong>in</strong>g a genetic algorithm, J. Ship Res.<br />
47/3, pp.222-236<br />
MARKOV, N.E.; SUZUKI, K. (2001), Hull form optimization by shift and deformation of ship<br />
sections, J. Ship Res. 45/3, pp.197-204<br />
PEACOCK, D.; SMITH, W.F.; PAL, P.K. (1997), Hull-form generation us<strong>in</strong>g multi-objective<br />
optimisation techniques, 6th Int. Mar<strong>in</strong>e Design Conf., Newcastle, pp.405-419<br />
THORP, N.A.; PIERSON, B.L. (1998), Cluster analysis after a partial genetic algorithm search<br />
Eng<strong>in</strong>eer<strong>in</strong>g Optimization 31, pp.225-246<br />
147
!"#"$%&'(<br />
! "!"#"$$%)<br />
<br />
#<br />
<br />
<br />
<br />
<br />
! "# <br />
<br />
$ %& ' <br />
( ( ' <br />
<br />
) $ % <br />
<br />
( ) <br />
<br />
*+ <br />
" <br />
<br />
<br />
$% <br />
<br />
*+ * <br />
* * * * * * <br />
** <br />
,*<br />
<br />
*+ * * <br />
(** (*<br />
* ** <br />
* '<br />
* * * ** *-<br />
*<br />
<br />
& <br />
<br />
. ** /%01* *<br />
2* %0, ) -.//013 <br />
) -23301*** 2<br />
<br />
, %02<br />
3** * * *2<br />
, *** <br />
( *' 4 5 *-<br />
** <br />
6 ( 7 )) ) 3 * <br />
4% 85960!3 :,<br />
*((*
2 *: * * <br />
*!*' **<br />
(< -23321<br />
<br />
3*'+ */=1<br />
*' *8* *<br />
,*< <br />
*7*/=1 <br />
* * <br />
*&$ -23341<br />
, 7%02<br />
• 62<br />
) *: * <br />
* *<br />
+** +/,1** *<br />
' ( *3,( +*<br />
* * > ) ? />)?1<br />
* -23341 * -23301 * <br />
(*: <br />
*<br />
• 02<br />
3** **: * **<br />
* * <br />
** *3 * ***<br />
* * *** ***<br />
** ) -23301 3( . @60 ?"<br />
)*>)?***( **<br />
( * * *: : ( <br />
5 -23301<br />
, #%02<br />
) * * ** '<br />
* * *3 <br />
+(* <br />
* <br />
!+ >)? * <br />
* *3 * <br />
< * * <br />
* -23341<br />
<br />
'" <br />
<br />
'$"()*+,<br />
<br />
* ***<br />
!"# " $ $% > !"# " 0 $% *<br />
"*/ 1<br />
$>A,< * <br />
( #.<br />
+* * , <br />
< *( * *<br />
<br />
+ : * < <br />
* *' * + <br />
**/ *B$*1<br />
<br />
;
2*!"#$) <br />
<br />
<br />
>A2$CB# -23341<br />
<br />
'&" <br />
<br />
'&$ <br />
<br />
3* * */<br />
##, < *1 <br />
, ( / *1 <br />
* 2 * * *<br />
( * '$<br />
* ( * * 9 ,(<br />
* < *. <br />
* * * 4 *5 <br />
2* , <br />
4*5 <br />
*<br />
<br />
< *2##,<br />
*<br />
CB
! + * *+ + * <br />
* < * * * * (<br />
* < * A ! <br />
* * *<br />
<br />
'&& + <br />
<br />
. * * (*(24'59<br />
( * * * <br />
** 4* **5**<br />
*/4'5,1( D <br />
'* <br />
4'5*E ***'/<br />
1 * (<br />
*/4*54*5,1 <br />
* 4( 5 < <br />
. <br />
( > ' * <br />
* / <br />
(1) 2* <br />
/*1* *' <br />
E * <br />
* / '1 * * <br />
/ 1 * ** * 3* * <br />
** ' + * <br />
>D *<br />
<br />
Current production plann<strong>in</strong>g<br />
Mach<strong>in</strong>ed parts<br />
Mach<strong>in</strong>ed parts<br />
to assemble at<br />
shipyard<br />
Production<br />
plann<strong>in</strong>g<br />
Draw<strong>in</strong>gs<br />
BOM’s<br />
Plan and<br />
coord<strong>in</strong>ate<br />
mach<strong>in</strong><strong>in</strong>g<br />
Material<br />
request<br />
Planned &<br />
resleased<br />
job<br />
Collect<br />
material<br />
Controled<br />
parts<br />
Planned &<br />
resleased<br />
job<br />
Mach<strong>in</strong><strong>in</strong>g<br />
Mach<strong>in</strong>ed<br />
parts<br />
Assemble <strong>in</strong><br />
mach<strong>in</strong><strong>in</strong>g<br />
centre or<br />
Store parts<br />
Mach<strong>in</strong>ed<br />
parts<br />
Control of<br />
mach<strong>in</strong>ed parts<br />
& X X<br />
Mach<strong>in</strong>ed parts<br />
for third parties<br />
Outsourced<br />
mach<strong>in</strong><strong>in</strong>g jobs<br />
Change of job and/<br />
or plann<strong>in</strong>g<br />
>D20*<br />
<br />
** 4* **54'5,< <br />
("-( /<br />
1 * * /<br />
F 1 <br />
( * <br />
*! <br />
) < <br />
** ( ****.* <br />
+* * (<br />
<br />
A 3*+*<br />
D ?' *<br />
<br />
<br />
C
'&' <br />
<br />
* * *<br />
* * *<br />
2<br />
<br />
* 2<br />
, !*2<br />
• 3** *<br />
• **<br />
• <br />
, ! *, 2<br />
• +* * /1( < <br />
( * (<br />
<br />
<br />
3 * * ** ' 3<br />
* 0,7 + *(<br />
', C /9)1**<br />
<br />
- <br />
<br />
-$"(!<br />
<br />
$,+*** G <br />
* 4 **5 ,, H , <br />
* * ( 0,7 ,<br />
* (**<br />
( +* *,4 5/4'5<br />
4< 54*51 >*(#DAA<br />
*3 6& 7<br />
/ 14'5/*<br />
(,, 1. * <br />
6%7>CC<br />
<br />
<br />
.',./..12.', <br />
! *<br />
!..**/ 1<br />
/1.'/A1 <br />
.' <br />
' !', *<br />
* < (<br />
*+ -233.1<br />
C 9)29) 2+* !<br />
* + /0,71<br />
45* + /<br />
1) < (4'*5 +<br />
/ 1E ****<br />
<br />
G * !9>B!9>D <br />
5 *-23301<br />
H 3+* < ,, *>D<br />
><br />
3 / 1! **<br />
* <br />
CA
20 * <br />
<br />
. 4*5 *<br />
6%7 >C >2* <br />
' 4<br />
*5 ; !4'5< 6* 7>C 7<br />
4'5( 2*<br />
B (** 'E /1<br />
( *4* 5*' <br />
* *<br />
<br />
-&% <br />
<br />
* * ** A / <br />
*( 1 <br />
*!* ** <br />
0,7* <br />
<br />
* ** 2<br />
, 7 2 * * * *<br />
<br />
, 7/**12< ** <br />
, 7/*'12 <br />
* **'0 /*1<br />
* *' 3 <br />
< * /1<br />
<br />
<br />
; 0 *2* <br />
>C<br />
B * /*<br />
I* * I* *1<br />
<br />
$* * 2/*1(<br />
( * ( ,<br />
* (* **4* 5*<br />
*<br />
A . * * ' /*<br />
*1<br />
<br />
CD
C2J *><br />
>C2J2* (*<br />
>C2* **(/+,+2,+ *KLM1<br />
<br />
!* * *2<br />
, !** ** ' +*<br />
, !' <br />
, ?*'<br />
, 9 *<br />
, #* <br />
, % * * 4*5<br />
, * * <br />
C
-'. <br />
<br />
-'$. <br />
<br />
!* * 2<br />
, 3*****<br />
, 7 **/*<br />
*1<br />
, 7 *'/*<br />
1<br />
, * / * 1<br />
, . ***( <br />
<br />
<br />
>G2? *2'<br />
>G2?2(+*<br />
<br />
-'& <br />
<br />
* (**/0,7<br />
9+, >C : * <br />
* * *2 0) 7' > <br />
*/4'5,1' <br />
' (?3*<br />
<br />
CC
* *( **>G+<br />
*/4*5,1* ** <br />
* >G ! * * <br />
( <br />
<br />
-'' <br />
<br />
! * * ** * * <br />
* /1!(+**<br />
* <br />
$ *'<br />
, * ! <br />
*< <br />
< < * <br />
3 * <br />
(* *<br />
#< *<br />
*<br />
<br />
/ ! <br />
<br />
3( * ** <br />
* <br />
<br />
/$#!<br />
<br />
> ( <br />
* ** * * <br />
* >*' <br />
* <br />
, , ( <br />
) * (<br />
*< : <br />
>)?* -23301 787<br />
" D &$ -23341 > (*<br />
*!#, *!<br />
*!#, * (<br />
* /701/8 23391 <br />
* **<br />
<br />
<br />
/&#+ . <br />
<br />
) * : < ** <br />
!#,**. * *!"#$<br />
* **.<br />
*** .<br />
, ,<br />
* * * * <br />
*2 ( * * * / <br />
(**-*1 <br />
<br />
D 787" 7 ,7,8 " * * *<br />
< * N * * <br />
* <br />
CG
* * 0 * <br />
4* (5 * * * <br />
( * 4*<br />
5* , (<br />
* < <br />
<br />
/'0 <br />
<br />
> / 1 ( <br />
$* * * <br />
* <br />
* * ( <br />
! , <br />
+* <br />
* '<br />
"+* 4*5<br />
* *:$ <br />
4 <br />
*5 * <br />
# < *( <br />
* ( <br />
<br />
1 <br />
<br />
7 ****< * <br />
** *+ /* 1* *'<br />
<br />
* $* ( *<br />
** * * <br />
* > +* * <br />
< ( (!<br />
* ( * *< <br />
* <br />
E * *** <br />
< *<br />
0 ** * +*<br />
** <br />
<br />
"( <br />
* (* !#,** <br />
$*** ! <br />
**<br />
<br />
( * * <br />
*** ** <br />
<br />
<br />
<br />
<br />
367"9"O?P3O!9"!)OE 3?@#/ABB1# % <br />
9 *)* 8*#, 6<br />
<br />
367"9 " /ABB1 # % : ) ;<br />
* % / 1 <br />
<br />
<br />
CH
$3667/ABB1 ! *<br />
*2 (Q* <br />
<br />
?P3O!9"!)OE 3?@#/ABBC1 -1 :<br />
( #.07!ABBC" <br />
<br />
!"#".663/ABBD1 93.0 * !<br />
*2 R* CBR*<br />
<br />
&$)!/ABB1! ?$ @ <br />
<br />
<br />
7.7&. 9)O $3863#" # /ABBD1 ! %= # *<br />
*2 *% R)**<br />
<br />
)9!"398 /ABBD1 : <br />
< #.07!ABBD,H0ABBD" **ABD,AB;<br />
<br />
)9!"398O"9!9030/ABB1*( ? "3)33 <br />
ABBB**AC,AH<br />
<br />
)"! @?O 699 &&O E .. @"O &!0 E O 699 @"O &!0 )"O 738& @PO P!0 "<br />
/ABB1) % * & & * , * @<br />
)*7 ABA**H;,D<br />
<br />
8.P0!/ABBA1* $ 2. E <br />
@ ABBA**,A<br />
C
!"# ! $<br />
$ $% <br />
$ % & <br />
<br />
<br />
!<br />
<br />
! "# $ % &'<br />
! "#( ) %) %<br />
*+<br />
%%(& ),-.<br />
),-%%% <br />
/))( ))<br />
! "# % % ! '001# 2 *<br />
"32%)()) )<br />
% / () ) <br />
)))<br />
<br />
" <br />
<br />
"#$$ <br />
<br />
4))<br />
% . <br />
)) $<br />
)))()<br />
) )). <br />
%%) *<br />
%<br />
<br />
<br />
56) )<br />
<br />
'56))
89$)4<br />
++. +)<br />
.( ) .<br />
% /.<br />
.%4%). +:<br />
89$4:))<br />
<br />
""%$ & <br />
<br />
))'/<br />
) ( )%<br />
%; () ) <br />
; )<br />
"- %<br />
);%)<br />
%.8) <br />
) ) ); %<br />
4%%<br />
) /%.<br />
<br />
"'<br />
<br />
) * 1 % ) % % %. <br />
%. 6..<br />
% . ) ( ? ) <br />
. <br />
! "#<br />
<br />
)))%)@5<br />
• % 88)%%<br />
• %.).<br />
%)<br />
• % %%<br />
• AA ) <br />
) ,-<br />
• %,B-<br />
<br />
<br />
15B"C! "#<br />
<br />
D5<br />
<br />
70
*5! "#<br />
?) E0 B 6BB<br />
+ '70 7<br />
$. 7 D<br />
&1 '1<br />
6% 0D07 " '00&0.F <br />
) ''. " '1<br />
? 100 <br />
6 'D& <br />
<br />
'# $ $<br />
<br />
4 % %.%) %<br />
) ( % ) /<br />
%))A <br />
)*%)%*G7D<br />
,'000-5<br />
<br />
B6%6 5 1<br />
B 5 D<br />
G.5 D<br />
"% (4<br />
))%A<br />
%.%%>D<br />
)%)<br />
<br />
% )<br />
<br />
%) <br />
) <br />
) %. 4 )<br />
%%/% <br />
<br />
. ) ( <br />
%% )(5<br />
• >(5 A )' )1 <br />
HD HD ) H7 H 2<br />
• B(5 ' 1 <br />
<br />
%)<br />
*);%<br />
;% ) ))% <br />
7
(") <br />
<br />
+ %) )<br />
) I.*/<br />
% / % (<br />
/%$%<br />
.%(5*/ ' <br />
,-1 D ,- <br />
%<br />
<br />
<br />
5628"$.<br />
(<br />
<br />
<br />
* ! !+ !<br />
<br />
4)%)) ) .<br />
)) %<br />
4) ;<br />
)%)<br />
) %%.<br />
4(7% 4(E2A<br />
/ )I(02A <br />
('2A7 ) ) %)<br />
% )J()) <br />
& ("<br />
%).EED <br />
H(,'1- 4<br />
%,-.1%<br />
<br />
M.E. Vib 5: Cyl<strong>in</strong>der Station fore<br />
<br />
y<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
75"<br />
<br />
<br />
<br />
<br />
<br />
&5 ))<br />
),-<br />
7'
) % H.1%<br />
) <br />
%)%%7 H<br />
% ()) 1<br />
<br />
*%) I<br />
) ) 4 ) % <br />
; / <br />
*% <br />
)()* <br />
% )(<br />
*)%H) <br />
/I<br />
* <br />
)% %<br />
<br />
<br />
E5. <br />
<br />
5 * <br />
<br />
<br />
% 7 <br />
)J.%<br />
(%)))<br />
) )) >"<br />
)@ 2%)<br />
% ))%<br />
%%)<br />
%)<br />
<br />
, ! -!<br />
<br />
, -<br />
<br />
@)<br />
<br />
%.)@A/) <br />
>"%%<br />
% <br />
. %<br />
%) %<br />
<br />
71
," -!!!! <br />
<br />
) %/<br />
> <br />
)<br />
))% . ;)% 5<br />
• 4 5 ) .<<br />
2<br />
) <br />
%<br />
• F 5 2 % <br />
) (<br />
• F 5 % <br />
) ) <br />
% <br />
4)%%.<br />
;% <br />
*)%%<br />
$ ) % ; <br />
)0 %%5<br />
% %7 <br />
<br />
<br />
05> <br />
<br />
%<br />
4<br />
% . ! "# 4 % <br />
/ <br />
( ))<br />
),-)/ )0'2A<br />
))<br />
%% )'10<br />
%%'0% %<br />
' ) ) . 11<br />
)I% @<br />
7D
.4))%<br />
'07<br />
<br />
<br />
56<br />
'56 . <br />
(<br />
<br />
* 2%<br />
)) %) ) )) <br />
) 4 ) <br />
/)% )<br />
4))% <br />
4 4<br />
)%%<br />
<br />
.#!<br />
<br />
@. (4<br />
% )).<br />
) ) )<br />
) )<<br />
)% (<br />
)/>)<br />
) / % <br />
)( <br />
/$)(<br />
$ %<br />
<br />
7
! <br />
"# !"<br />
" !"<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
$%&<br />
<br />
#$ %#& <br />
$$ $$&$ # '<br />
% '$ % $ $ &$($ &% #<br />
%& ) * ( <br />
$+ '$ $$,<br />
$& #&& & <br />
( & -$ + .% $ <br />
% !"###$ !"##%$ !&'''$ * <br />
$ & & $$% & &$<br />
#& %$$ &$ $$!<br />
% # %%$$%$ &$<br />
$ %() !&'''$<br />
<br />
/$% $(&&$ *%%&"' !<br />
"-' %0<br />
o 1&<br />
o +&& &2<br />
o * &&$%$$2<br />
# ' & ( &&$<br />
&3%,$ & ,$<br />
$ ( $$ <br />
$$% #&,&$$ <br />
$ & $% *<br />
$$%,$ $%<br />
&-$+.%% <br />
<br />
-$ $($ $ $ <br />
$%$% * $ &$% <br />
% !&'''$<br />
<br />
# $% $% $( &&$ & $(<br />
& & && &&$#<br />
$ $ $&$$% & %$ !<br />
%( !&''*$( !&''*$)!"##+$3 $ %<br />
/. &$% $$ $<br />
$ &% &&$!"###$
5 $. /$6,- !<br />
, &'$ 1 &<br />
$ '$$ $$&$7$!<br />
%$( &&$%$ !<br />
6,- - &&. /$%#$ !<br />
,$ $ $% '$ !<br />
<br />
<br />
'%<br />
<br />
-%$%,&,$ $%. /$<br />
6, - $ 3 $ # & <br />
$% $$%$& $% /.$<br />
$ , # &%& <br />
7% , !&''&$ # / 1 % - $ /<br />
.$ % $ % $ % <br />
!"##%$#-$- & 0<br />
o #/1 %- $ 8/ 9<br />
$ & % <br />
o #6*&- &$% :3 !<br />
$;:-%;:# ;# /.$<br />
($'&$$! <br />
o #-$%- %& <br />
$%&&& 7$!<br />
$ $ %%%&<br />
#6,- & -$- $$ <br />
# $$ % %$ &,$0<br />
o & <br />
o 3 %& <br />
o 6&& <br />
# $ $ $ $ <br />
7 %8%$%&$%$9 "<br />
8 9 8% $%9 # $ <br />
$ ' 7%<<br />
# $ % /.$& # &<br />
$ & $ # <br />
$ $$%&& " &<br />
% &" # &,%<br />
3%$ &" $ $ % %$ <br />
$ # & 3$% * <br />
$% & $ &%" $%$ <br />
/1 %- 8/.$9<br />
-$- <br />
/8/9<br />
1 %- <br />
6*&<br />
- <br />
-$%- <br />
6<br />
+$$<br />
6,- <br />
/+$$<br />
.+$$<br />
3<br />
-$%+$$<br />
6,+<br />
7%05 && $%$$<br />
<br />
<br />
4
! % , !"##-$ , !"##+$ 5 $% $ <br />
$$&$ $# $-$3$%<br />
/3$%<br />
<br />
# $ &$$%& & 0<br />
o $<br />
o '$& <br />
o &<br />
o $$<br />
$$ & $. /$6,- 0<br />
89>$& %, !&''&$0<br />
* &&$ $$'$&* %!<br />
%&'$& & &&$'$ "<br />
#'$ &# '$&<br />
& &&'$$$ #&'$ &<br />
, #&$'$$ '$&!<br />
& $'$$&& $<br />
89+ &%, !&''&$0<br />
# & $$&$ $ <br />
$ &( $,& #&( <br />
% #$$& <br />
$ $ &&& $ $<br />
%$ #& $ <br />
& $ & & %<br />
$%?&$%$ $% <br />
893$$%, !&''*$0<br />
#& %$ &<br />
#% %& % !<br />
#& & %$ &!<br />
3$ $ $ :$; $ $ <br />
%% & $ $<br />
<br />
<br />
(%!<br />
<br />
# < &&$$%,&<br />
,$%&*%$ & !<br />
Product1<br />
Product2<br />
Assemble<br />
Product3<br />
Assemble<br />
Product4<br />
Product1<br />
Product2<br />
(1) (2)<br />
Assemble<br />
Product3<br />
Storage<br />
Assemble<br />
Product4<br />
Assembly1<br />
Assembly2<br />
Assembly1<br />
Stockyard1<br />
Assembly2<br />
Operation Model<br />
b) Operation as Relation<br />
Product<br />
Place<br />
Time<br />
Operation<br />
Product<br />
Place<br />
Time<br />
Product1<br />
Product2<br />
(3)<br />
Assemble<br />
Trans.<br />
Product3<br />
Storage<br />
Trans.<br />
Assemble<br />
Product4<br />
Target Subject Content<br />
Assembly1 Stockyard1 Assembly2<br />
a) Operation as Object Event<br />
<br />
7%
)<br />
)<br />
B<br />
B<br />
B<br />
3$$ &<br />
%0<br />
C<br />
#%-"+&<br />
<br />
C<br />
3 %<br />
C<br />
/&<br />
6 <br />
-$ <br />
<br />
<br />
<br />
<br />
B<br />
B<br />
B<br />
. <br />
+ <br />
/$%@ <br />
B<br />
B<br />
B<br />
B<br />
)<br />
>$<br />
D <br />
7<br />
1$%<br />
<br />
7%E- & $& $%<br />
& - $ $& <br />
%&& $$#& & $ $<br />
$%$ 7%E#$ & &<br />
&$$%&' $$ $<br />
$ %)!"###$0<br />
o #$!/ $$$ $<br />
o >$$$'$& <br />
o + &<br />
o 3$$<br />
o @$$$% &&$%$<br />
<br />
# & ' & % $ & &<br />
$$$&& $$. /$6,!<br />
1F 0<br />
&G<br />
3 %&/<br />
/&6<br />
%0<br />
.$<br />
1<br />
/.$<br />
*<br />
+$<br />
-$<br />
- <br />
6<br />
*<br />
6,<br />
%<br />
6<br />
+ .$<br />
/<br />
. <br />
3$$3 &<br />
6 0<br />
. <br />
+ <br />
. <br />
-3<br />
/3<br />
/$%@ <br />
>$.$<br />
@ .$<br />
7%H05 &. /$6,- $%$% <br />
<br />
<br />
A
- #$$&$ $% <br />
&&$%$ % #&!<br />
$% $% 7%H * $ <br />
$& $$% &"7, <br />
$%," %# %$ &, 0<br />
o 3$$&&$ $<br />
o 3 %% <br />
o -'% <br />
#," 0<br />
o . & %& $& %$ <br />
&"<br />
o + &$ &"<br />
o /$%@ $% &&$%$%<br />
<br />
# $ "%%$%% <br />
($ <br />
<br />
*%+ <br />
<br />
*%$%+ <br />
<br />
#$% %" & !<br />
$$$* '&$% &$$!<br />
$ G $$ +$ /<br />
.8+/.9 %&$$!&'&% $ %.<br />
!"###$ % $$ % <br />
$* $ && % <br />
$ #" $& $<br />
$* $$&(<br />
% %( & <br />
, !"###$ * && & $$% ( <br />
'$ & && $$( # (<br />
% $ $.+$-$ &% <br />
$% %,!"###$.!"###$.+$ !<br />
' $$ $%(&$% &<br />
%# ,$% 3<br />
":/$%@ ; '&&% $<br />
<br />
/$%@ <br />
1$&6/<br />
J<br />
5&&%/$<br />
6<br />
*$<br />
6<<br />
6E<br />
6H<br />
#<br />
3 $<br />
-%<br />
3 $<br />
-%<br />
3 $<br />
7%K0/$%@ .$<br />
<br />
4I
# $$ $ ( %<br />
# $ $ !<br />
& '%$$$ # & '%!<br />
$& & $$$ <br />
& & & $&!<br />
#$% & $( !<br />
% && & $ <br />
$* $ $&%%$ %&!<br />
7$&& $ % &!<br />
<br />
<br />
# $ $,$$& $ <br />
$ &$$%<br />
<br />
*%'%! <br />
<br />
# & $ #$ $!<br />
( & % $$* $<br />
$ $ & &$ $ & <br />
#$ &$ L# #!<br />
$ LM#M % (<br />
$L$$* $ /8-95&&$ $<br />
$. $#$%&$%$- /<br />
.- $ &$$/.%0 <br />
!&''&$<br />
<br />
#$ & 0<br />
<br />
<<br />
<br />
<br />
−<br />
. <br />
<br />
<br />
<br />
<br />
+<br />
−/<br />
<br />
1<br />
( / 1 <br />
−<br />
<br />
< <br />
<br />
σ<br />
<br />
<br />
/ −<br />
<br />
<br />
<br />
8'H9<br />
<br />
<br />
<br />
.<br />
8.<br />
9 = <br />
⋅ (<br />
<<br />
<br />
σ<br />
/<br />
< Π<br />
. <br />
+<br />
<br />
( /<br />
1 <br />
11$N-D%&G$%$NF<br />
(-Nσ<br />
F-F<br />
<br />
3& && !<br />
&#$&& %!<br />
(($ O %&<br />
#$$ 0 & 0<br />
<br />
−λ⋅0<br />
809<br />
= λ ⋅ <br />
8'H
$ #& $$ & $ !<br />
$ $$0<br />
<br />
89+ $& $$& $ $$<br />
'% %# $$ $ & &!<br />
#" & $$ &$$ 0<br />
&<br />
<br />
8<br />
<br />
9<br />
= 89<br />
89<br />
89<br />
<br />
8'HE9<br />
+$<br />
<br />
-# = P + ? + Q<br />
<br />
( − −<br />
+<br />
4< ( −+<br />
4<br />
(<br />
0<br />
<br />
3−56 8.<br />
≤ ( 9 = 89<br />
89<br />
<br />
89<br />
I I I<br />
<br />
(=<br />
-%<br />
<br />
<br />
P?Q $$ &$&<br />
$6/ $ $$ &$$ R'HHS0<br />
$ &$$( %<br />
& <br />
0 8'HH9<br />
8509<br />
= 03<br />
−56 ⋅ ( −<br />
0<br />
)<br />
<br />
#$$$$ $ 0<br />
8'HK9<br />
0<br />
= ∏ 0<br />
850<br />
9<br />
<br />
6/=<br />
6 <br />
<br />
# $$$ $& * <br />
$ % ( %-$% <br />
$ $ $7 !<br />
%$&($$<br />
<br />
8
$ $$%$$$ && P?<br />
Q R'HS$$ &$&7$$<br />
' R'HES $!<br />
#& &$% &&%3$ %<br />
& &(&&% &&!<br />
% $ ' $$ &<br />
<br />
<br />
8509<br />
= 8'H49<br />
0 <br />
0 3 −2 <br />
<br />
0 <br />
= ∏ 0 <br />
850<br />
9<br />
<br />
<br />
8'H=9<br />
6/=<br />
6<br />
<br />
5$ & $ $%<br />
& G$& $$"<br />
:/$%@ ; &$$ 0<br />
7 = 0 8'HA9<br />
+ 0 <br />
<br />
* " (,#&/$%@ .$ %$<br />
& , '& $% '$ 8%9 <br />
.+$-$* $'$ <br />
$% & /$%@ .$<br />
<br />
*%(& <br />
<br />
#@ .$ &" :@ ;:1 ;"#:@ ;<br />
&& &&<br />
σ$& λ&%#%<br />
&:@ ;" &7$$%&&!<br />
$ :@ ; $%$& <br />
& :@ ;" $,$ $$<br />
& & #$%&$%$ '$$<br />
$:1$;:5%%1;:%1;& <br />
$ $ G & <br />
:$1 ;:($1 ; $,":@ ; &<br />
7$$$ "<br />
:@ ;# 7%<br />
<br />
#/1 %- $%$%&$ &!<br />
$$ $$ # % <br />
%&$ $$&% %<br />
&7%4<br />
<br />
#&&/3$% $&$%" <br />
#/$ & $%E1 $$ &#$$<br />
E1 3 && %0<br />
<br />
< < <<br />
8'HI9<br />
= + + <br />
<br />
<br />
<br />
<br />
4E
3 $6<br />
5%%U%<br />
D%&$%$<br />
<<br />
1 <br />
1 7<br />
@ <br />
σ -<br />
µ -<br />
78(9 -<br />
λ 5<br />
78#9 5<br />
/81$9<br />
/85&&%9<br />
/859<br />
$1 <br />
($1 <br />
+$<br />
-+$<br />
$1 <br />
@F$<br />
σ 1$<br />
µ 1$<br />
6"<br />
-+$<br />
<br />
51 <br />
#<br />
λ<br />
/<br />
G$ / +$ <br />
<br />
<br />
6" I<br />
<br />
6"<br />
J<br />
6"<br />
<br />
&<br />
6"<br />
IJ<br />
<br />
7%01$&@ .$<br />
<br />
7%40*&$% &<br />
<br />
<br />
*%*%, <br />
<br />
#/$%@ .$ & %&.+$-$%8<br />
!&'''$/!"##9$3$% $ $$<br />
& $ µσ<br />
$ & & # % % <br />
5(!.$$ & & $ %0 <br />
!&''&$# $$ &.+$-$ !<br />
3 '# /$!<br />
%@ .$ %$ & $7<br />
$$&:/$%@ ; $7 $ &$<br />
",$+K&.+$-$ $ !<br />
EIIII $ & & $#$*<br />
$ & # $ %/$%@ .$ <br />
$$ $$&$%.+$-$<br />
<br />
3<br />
5<br />
+<br />
1<br />
<br />
#$*0+ &/$%@ .$.+$-$<br />
.<br />
81$<br />
9<br />
-26.73<br />
-8.00<br />
-14.61<br />
-72.57<br />
138.61<br />
.+$-$<br />
-%<br />
8 9<br />
12.62<br />
21.13<br />
20.48<br />
37.38<br />
11.24<br />
3$<br />
/$<br />
8$WXI9<br />
0.97<br />
0.38<br />
0.61<br />
<br />
I<br />
/ <br />
.<br />
D%8/$%<br />
@ 9<br />
@<br />
3%<br />
/$<br />
#/$%@ .$ $ % %$ $ #'$<br />
# $$$ & $ $$<br />
V/$%@ :%7,/$%& $!<br />
$/$%@ .$ <br />
3.17<br />
5.93<br />
3.49<br />
1.44<br />
4.49<br />
'<br />
H<br />
(<br />
$<br />
K<br />
+ <br />
@<br />
%<br />
/$%@ <br />
'<br />
K<br />
(<br />
$<br />
H<br />
<br />
<br />
4H
&$% &&$ &.+$<br />
-$#%& /$%@ .$ $$<br />
, $3#&'$$$ &$&<br />
& &,#/$%@ .$ &$$%<br />
:/$%@ ;& $$$$+K#.+$-$<br />
& $EIIII$$ ' 3$
$& $% %!<br />
$$ $ $ 0<br />
o # & % % <br />
$$$ <br />
o #$ &&<br />
<br />
3 $* <br />
-* <br />
3 * <br />
-
* $%% %! ( %3$$<br />
$ %" &$$%% <br />
<br />
@ K<br />
Cost/Risk<br />
@ K<br />
Makespan/Risk<br />
Cos t<br />
Cost/Risk<br />
M ak e s pan (s e c.)<br />
Makespan/Risk<br />
4 9 14 19 24<br />
@ I<br />
Plann<strong>in</strong>g Risk<br />
7%I0+ $% <br />
<br />
<br />
4 9 14 19 24<br />
@ I<br />
Plann<strong>in</strong>g Risk<br />
7%0. /$%@ <br />
<br />
3 $/$<br />
-/$<br />
3 /$<br />
-
3 $*<br />
-* <br />
3 *<br />
-<br />
-E<br />
-
$$$ & -%$ !<br />
$%)<br />
!"###$0<br />
o #$/ $$$ $<br />
o >$$$'$& <br />
o + &<br />
o 3$$<br />
o @$$$% &&$%$<br />
+$ & , ($ " <br />
8 $$9 , &$$ 0<br />
o 6,%$ %$" % <br />
o #" $$% % <br />
% <br />
o * ,%$$% $$ <br />
o *+ . , %$,%$($$<br />
",<br />
<br />
<br />
# /& # 6.6#6Z,36?3.3& &#<br />
& /.$-613- <br />
<br />
<br />
<br />
36?3.3 ZN 6.6#6 #N 3.313 ZN #3Z+* - 8AAA9 : <br />
*++3-[AAF$+%
Z..@3N*DD--38AAA9/
! <br />
"#$ %&!" #"! $%<br />
<br />
#%<br />
<br />
<br />
<br />
<br />
<br />
<br />
!"<br />
<br />
'( <br />
<br />
& '(""##'%)#<br />
' *( %'<br />
#* %" + # ' # <br />
# '(" $ # <br />
(' $# %<br />
"'# #" $% ('<br />
' # ## ' '( <br />
"',%'#-./0" <br />
$ ' ' " '' '' "<br />
''1$ % $'# $<br />
'<br />
<br />
)( *%%%<br />
<br />
)('(+%%<br />
<br />
,2"3% 42'(#$$$##<br />
" %2#$%&&%'5 #"",<br />
""'$' % <br />
#" %6#&""<br />
'<br />
<br />
6""" % $4<br />
• %"%"$' <br />
• ,' %"<br />
• , %"$"<br />
• )%''#"7#%#<br />
• )%"<br />
• ) #$"<br />
6"8' 9%'%"8" 9&<br />
" " % # " <br />
##%#, % 8' ##9<br />
$" # - ''0 ' $# $<br />
:$ % " , #" "<br />
%'%"$%#"6#'<br />
' " " $" % ; %" -0<br />
&-0-' 0
&""%7"%7" #"<br />
' &%7" - "0 $ #<br />
" # " '' # #; ' -<br />
#0-0%<br />
$ '"-'0<br />
<br />
<br />
Process function structure<br />
Functional decomposition<br />
energy<br />
<strong>in</strong>formation<br />
material<br />
Processfunction 1<br />
Task 1<br />
Task 2<br />
Processfunction n<br />
Element 1 Element 2 Element n<br />
Fulfils<br />
Fulfils<br />
Relations, form<strong>in</strong>g the structure of<br />
the (sub)system<br />
energy<br />
<strong>in</strong>formation<br />
material<br />
Task 3<br />
Fulfils<br />
Pr<strong>in</strong>ciple / technology<br />
<br />
6#46$"<br />
<br />
" #"<br />
"#%$4<br />
• $''%"$" $<br />
"*<br />
• #*<br />
• *<br />
• %$ *<br />
• $$<br />
<br />
)()(%,<br />
<br />
&" )## '" ' 8 =<br />
#"= "##"<br />
# "9&")##'<br />
)))
Customer needs<br />
Requirement analyses<br />
1. Analysis of objectives and environment<br />
2. Identify functional requirement<br />
3. Def<strong>in</strong>e performance requirements and prior design conditions<br />
Analyse loop<br />
Process management<br />
• Risk management<br />
• Interface management<br />
• Requirement management<br />
Functional analyses<br />
1. Decompose sub-functions<br />
2. Def<strong>in</strong>e performance requirements and prior conditions for sub-systems<br />
3. Def<strong>in</strong>e functional <strong>in</strong>terfaces (<strong>in</strong>tern and extern)<br />
4. Def<strong>in</strong>e an <strong>in</strong>tegrated functional architecture<br />
synthese loop<br />
Verification and<br />
Life Cycle loop<br />
System syntheses<br />
1. Transform functional architecture to a physical architecture<br />
2. Def<strong>in</strong>e alternative system concepts, configurations- and system elements<br />
3. Select preference-products and process solutions<br />
4. Def<strong>in</strong>e physical <strong>in</strong>terfaces (<strong>in</strong>tern and extern)<br />
Product design<br />
<br />
6#
%'$$'-%0"$<br />
#'$%"<br />
" '%##%7"$' "#<br />
" '$%"% '$"%<br />
$ ' ' $ '" '% $ " <br />
''%%''#<br />
'#' /012$%&&' #"4<br />
• %%#; "-%# "7<br />
''0<br />
• % % - (' $"0 ,/& ' -7+ 7<br />
#7%70<br />
• %%-0$'-#7%"<br />
###0<br />
<br />
Whole life cycle of functional object “pump with tag 130P01”<br />
Functional object :<br />
pump with tag 130P01<br />
Part of whole ”realiz<strong>in</strong>g<br />
functional object pump 130P01”<br />
Pump of manufacturer A<br />
Serie nr 34562<br />
Pump of manufacturer B<br />
Serie nr 3A_4562<br />
Pump of manufacturer A<br />
Serie nr 104562<br />
Beg<strong>in</strong> of life<br />
end of life<br />
Part of whole life cycle<br />
<br />
6#D4@ 7"'"'<br />
<br />
)(.(<br />
<br />
, " # #; ' "<br />
## ''" " % (' E $ <br />
'%"/"/-//0@-0''% "<br />
##E-';#0' + "## +<br />
$%$
6#E$''$##$';#<br />
(; %# # %+ ; -7 <br />
% 0'$# ('%#<br />
' ) $%&&' '' $# # <br />
# " $" #; <br />
$#$" $#%"$<br />
%"(<br />
<br />
)(/(0!%<br />
<br />
#'"",'' %"<br />
+"++"+$"'<br />
%#'%;%" "'#"<br />
,''$'%'"7'"-#0,<br />
' ( '' + '' 7' '<br />
-'" # ## %# # #0 ) ' <br />
'$"(%-%$'<br />
'$%$'$''#;0<br />
(-%$'#;#<br />
'0 %"( <br />
'$''''' <br />
' '( # " $ $ % #<br />
$,'%" "<br />
<br />
,' %$' $#8#9(6#G4<br />
• .4 $ 2 3 2 <br />
$'3'$<br />
• .44'$'"',<br />
+%<br />
• .'4 $" ' # $ " " # <br />
+''''%<br />
(<br />
<br />
Requirements<br />
People<br />
Projectmanagement<br />
Risks<br />
Interfaces<br />
Profit<br />
Technology<br />
Processes<br />
Functionality<br />
Product<br />
skills<br />
Methods<br />
performanc<br />
Geometry<br />
Planet<br />
<br />
6#G4('<br />
<br />
.'%'''''<br />
'.'%;%"#".<br />
' % ; %" #" % %" <br />
<br />
E
'' /012$%&&'6#'$"<br />
##''"#% <br />
'(" ' #% " ' <br />
# +" " $ (<br />
- '0 ' $ '( <br />
#"7#<br />
<br />
)(1(0!2 %<br />
<br />
.+% $"4<br />
• '" + ## ( $ ' <br />
+%#%"'''% ' "<br />
%### #%<br />
• '"+## ' +%<br />
#%"'''% ' ''<br />
"'$'"+##<br />
($'<br />
• C# + $ ' $ $<br />
##'%<br />
<br />
6#H$'%$$#' $"<br />
$' '$" "#"<br />
'%&'#'#%<br />
+ # % # %" <br />
/" % ' ' +<br />
5 ";#' $+ #"<br />
$" +('%'<br />
%% $"';+"<br />
+ %" + ( <br />
;'<br />
<br />
Viewpo<strong>in</strong>ts…<br />
Experiences…<br />
Terms ...<br />
Objectives….<br />
Solutions…<br />
client<br />
subject<br />
specification<br />
2 <br />
Implicit<br />
<strong>in</strong>formation<br />
3 4<br />
4<br />
3 4<br />
84<br />
3 4<br />
3 4<br />
7<br />
contractor<br />
5 <br />
<br />
5 <br />
%<br />
6#!<br />
+<br />
<br />
0%#!<br />
explicit <strong>in</strong>formation<br />
<br />
%%<br />
<br />
%<br />
<br />
6#H46'' (''<br />
<br />
)(9(0!<br />
<br />
%##%%"' #<br />
" <br />
<br />
G
6"$'4<br />
• , $" ' $ -# #"0 <br />
'% *<br />
• /# '*<br />
• /# $ " $ -'' <br />
''0$"" -%'"%0*<br />
• /#%$ *<br />
• /#$$*<br />
6"##$'4<br />
• +#*<br />
• ? #+*<br />
• ##+#*<br />
• , (' $ ,/& ' -$ ''0 #<br />
*<br />
6 "$'4<br />
• %" %# <br />
$"#-#+7#7%<br />
7'0<br />
6%##$'4<br />
• .%" (# # <br />
%$'$#"# <br />
6'$'4<br />
• .%"-0$'*<br />
• %"''*<br />
• #*<br />
• $''*<br />
• ##"#<br />
• ,$" +('(' <br />
%'%% %"+<br />
• ,$" '-# 0<br />
'*<br />
6$##$4<br />
• ,$"#; $#$"<br />
$#"$%"(<br />
• , (" " $ <br />
%<br />
('$"'''<br />
'%$'"# "'-6(%"<br />
' '" ' 0 & <br />
%" (%$"(#'%$', #<br />
'( & <br />
'("'%4%""%""<br />
%"' ,%./"'$%<br />
($''% $<br />
# "<br />
<br />
-(0!<br />
<br />
-('%% !<br />
<br />
,";#$'' ;
# #" $3445' $ ' 8.<br />
. . F$# /#9 -A.F/0 $ 6# A.F/ $<br />
($#"%"%"<br />
",#$ ''#$ 4<br />
% % ' -0 ''" -0 <br />
- #$'0 ''#$#<br />
<br />
<br />
Information<br />
objective<br />
function<br />
Information representation<br />
life form<br />
process<br />
activity<br />
System<br />
content<br />
geometry<br />
relation<br />
System<br />
behavour<br />
physical object<br />
System<br />
structure<br />
Collection<br />
Property<br />
<br />
6#46$'<br />
<br />
)$'%"$';";%-<br />
%"("0 '%%"$<br />
6#I A.F/%"'$'%"%"<br />
## @%"-@0 "%"%#<br />
$ $ % % % - % <br />
'%#''$$"0 #%"%<br />
';%%6#I$" <br />
1?,<br />
<br />
View model<br />
level<br />
Bus<strong>in</strong>ess object<br />
level<br />
Specialization of Bus<strong>in</strong>ess object<br />
(library object level)<br />
Specialization of library object<br />
Objective process objective a for human acceptable <strong>in</strong>side climate<br />
system objective climate control<br />
to avoid enter<strong>in</strong>g of smells <strong>in</strong>to other spaces<br />
Function process function realiz<strong>in</strong>g desired humidity<br />
realiz<strong>in</strong>g desired air quality<br />
realiz<strong>in</strong>g desired <strong>in</strong>side air temperature<br />
system function HVAC system function air distribution<br />
regulat<strong>in</strong>g air quantity.<br />
circulat<strong>in</strong>g water<br />
controll system function<br />
remote monitor<strong>in</strong>g and alarm<strong>in</strong>g<br />
Physical object system HVAC system<br />
plant<br />
proces unit<br />
proces unit section<br />
air-condition<strong>in</strong>g plant<br />
air handl<strong>in</strong>g unit<br />
chilled water unit<br />
filter section<br />
heat<strong>in</strong>g section<br />
cool<strong>in</strong>g section<br />
Property absolute property exhaust capacity<br />
return water temperature<br />
m<strong>in</strong>imum ambient air <strong>in</strong>side <strong>in</strong> w<strong>in</strong>ter<br />
pr<strong>in</strong>ciple system pr<strong>in</strong>ciple s<strong>in</strong>gle pipe<br />
direct-on-l<strong>in</strong>e<br />
6#I4& $%"- $$#1?,0
&%$%$$+%"'<br />
6#>$'' #%$%$<br />
$$""'"%-''0% <br />
$%%",''%<br />
('" &" $ $ % 6#> $ <br />
%2''"#"3-'"0 $ 23 ' <br />
'-'0' '
7>E<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
Pump 45P1 is valid s<strong>in</strong>ce 01-02-2000<br />
Pump 45P1 is referred on P&ID 45<br />
Pump T345 represents pump 45P1<br />
Pump T345 is of manufacturer xyz<br />
Pump T345 is valid s<strong>in</strong>ce 01-02-2002<br />
Pump T345 is term<strong>in</strong>ated at 29-01-<strong>2005</strong><br />
Pump R654 is successor of pump T345<br />
Pump R654 is of manufacturer abc<br />
Pump R654 is valid s<strong>in</strong>ce 01-02-<strong>2005</strong><br />
Pump R654 represents pump 45P1<br />
6#A4)('(' " <br />
<br />
-(. <br />
<br />
''% " $#%%('<br />
7$#$' '$";#'<br />
%# $6#< (' "<br />
%% ' '' ")"%<br />
% # ' $ '%# #<br />
'%#'%#@ "%% 7<br />
$'%##'' 6#D4<br />
'/( ''' <br />
-#&:%"0#%"';'<br />
I>
' $#$<br />
%+-##$0<br />
<br />
concept template chilled water unit<br />
Bus<strong>in</strong>ess generic def<strong>in</strong>ition of a<br />
chilled water unit<br />
Development<br />
Re-use<br />
class template chilled water unit<br />
Description of a specific<br />
chilled water unit<br />
by yard for build<strong>in</strong>g nr xxx<br />
Development<br />
Re-use<br />
Instantiated class template chilled water unit<br />
Description of a unique<br />
chilled water unit<br />
by supplier for build<strong>in</strong>g nr. xxx<br />
<br />
6#D4 #''%'%+<br />
<br />
''"%%('"'% -<br />
'0%%$%# $<br />
%%$%('%"% <br />
$"'("% "%%"('% $ <br />
#%#$#;%"% <br />
<br />
-(/<br />
<br />
(' $A.F/ $<br />
%% 4<br />
;"'*<br />
< %";%*<br />
A ; # *<br />
D # $ ' ' $ <br />
+% $$#<br />
-#'0<br />
"## %%"$ A.F/<br />
%"4$-0%%%"%<br />
"#';A.F/% " %"<br />
%%(#" '""%# %<br />
''%A.F/%''$ <br />
%# # $ '7' % % ' <br />
# ## # #<br />
"% '$ <br />
<br />
.(<br />
<br />
.('"%,0#<br />
<br />
A.F/ $ % ' # C % -C 60 C <br />
## & ).% " & ).% % ' 8 9 , <br />
( & ).% (' - (' + 70<br />
'# $ % - 7 #7 %0 %" # <br />
"' C ##-#'06#E( $<br />
%%%"$"'$)%$' (<br />
'<br />
<br />
<br />
I
has UID<br />
object<br />
grammar part<br />
has UID<br />
object<br />
Fact-ID<br />
Lefthand object<br />
Has relation x with<br />
Righthand object<br />
“ma<strong>in</strong> fact”<br />
has context<br />
has description<br />
<br />
<br />
<br />
<br />
:% 6 ."% <br />
6#E4 $A.F/ $*!J+ <br />
1 2 101 60 3 15 201 65 9<br />
which forms the connection between<br />
two remote locations by means of a<br />
1010829 10430 pip<strong>in</strong>g route 1146 is a specialization of 1483 route<br />
sequence of pip<strong>in</strong>g components.<br />
3/16/1998<br />
1130271 130207 pump impeller 1191 can be a part of a 130058 centrifugal pump<br />
which is a set of rotat<strong>in</strong>g vanes with<strong>in</strong><br />
an enclosure which is used to impart<br />
energy to a fluid through dynamic force<br />
<strong>in</strong> order to generate a required head or<br />
pressure 3/26/1996<br />
1130956 130207 pump impeller 1146 is a specialization of a 130144 impeller<br />
which forms part of a rotat<strong>in</strong>g assembly<br />
of a pump impart<strong>in</strong>g k<strong>in</strong>etic energy to<br />
the liquid be<strong>in</strong>g pumped.<br />
3/26/1996<br />
1910911 911231 shutdown procedure 1146 is a specialization of 910002 procedure<br />
to take a facility out of operation.<br />
3/20/1998<br />
1194089 192495 heat transfer 1146 is a specialization of 192923 unit operation<br />
transport energy and entropy from one<br />
location another. Typically the<br />
exchange of heat between two 10/19/1995<br />
1521473 520876 w<strong>in</strong>ch 1146 is a specialization of a 731016 mach<strong>in</strong>e<br />
for materials. haul<strong>in</strong>g or pull<strong>in</strong>g; especially: a<br />
powerful mach<strong>in</strong>e with one or more<br />
drums on which to coil a rope, cable, or<br />
cha<strong>in</strong> for haul<strong>in</strong>g or hoist<strong>in</strong>g.<br />
2/12/2002<br />
<br />
6#G4' C % -C 60<br />
<br />
<br />
<br />
6#H4 7% C 6<br />
I
," $#'4<br />
• &<br />
• &<br />
• .%" - "#0<br />
• #<br />
• @## 7%-7##%0<br />
<br />
) ' % $ 6#E $ ' <br />
& ).% $ C 6 6#H 0 $%&&' )# > <br />
'#$>K%K <br />
<br />
& ).%7"$6# %##"<br />
(,+*35.;,*.;
% & ).% -% >>>>0 $ A.F/ $ A.F/<br />
$ & ).% ' % %" # ';<br />
& ).%$$A.F/'%"A.F/-$%<br />
%%0(%%%& ).%%,"<br />
%"%$"6#I;''<br />
<br />
.()7#!<br />
<br />
,(' $(#' <br />
'' 1?," ""; '$<br />
'" 1?, <br />
$A.F/ $#'#%7$"#'<br />
# 6#< , ' ( $ $ #<br />
% % $ ' & ' % <br />
)((#%$',$" -"0<br />
' L/@ $" % $ A.F/ "<br />
''L/@, @%"-@0& ).%%<br />
8 9A.F/'%"-'' '0,<br />
'%"''@%(%"%-'<br />
0'' %'' <br />
<br />
Reference data library<br />
for shipbuild<strong>in</strong>g<br />
(standard template’s<br />
and objects)<br />
“chilled water unit” object <strong>in</strong>formation model (template):<br />
chilled water unit is the whole for compressor<br />
compressor is of type semi-hermetic<br />
compressor is of manufacturer Carrier<br />
chilled water unit is the whole for condenser<br />
condenser has property cool<strong>in</strong>g medium<br />
cool<strong>in</strong>g medium is qualified as LT water<br />
LT water is supplied by client<br />
chilled water unit is the whole for liquid cooler<br />
liquid cooler is the whole for circuits<br />
circuit is the whole for pump<br />
pump has function circulat<strong>in</strong>g the chilled water<br />
pump has property pump capacity<br />
pump capacity is qualified as 50% of the unit capacity<br />
liquid cooler is the whole for starter panel<br />
Exchange of product<br />
<strong>in</strong>formation, based<br />
on object <strong>in</strong>formation<br />
models (templates),<br />
exchanged by XML<br />
or excel sheets<br />
Sub-sub-system def<strong>in</strong>ition,<br />
requirements etc.<br />
Sub-system def<strong>in</strong>ition,<br />
requirements etc.<br />
system def<strong>in</strong>ition,<br />
requirements, etc.<br />
Electrical contractor HVAC contractor<br />
Yard<br />
<br />
6#4.'(#'<br />
<br />
) '" " ' '%" ' $ ##<br />
% ' @ # ' ##<br />
'" #'$##$<br />
'-$' "#''06# % <br />
4<br />
• "'" ' <br />
"*<br />
• !#; ' <br />
% *<br />
• ,$"'' 3$"<br />
$'' '" %"<br />
• # % % $ ' '% ' <br />
% " % ' # <br />
<br />
ID
(%<br />
<br />
A.F/ $ - ''% " "0 7 - <br />
''70 $ '%# <br />
%'""%#%''('<br />
$"%"'%""'%$%<br />
""'# "' %"##"<br />
'' " # # %7 + " '' <br />
'''''A.F/ $<br />
###%# (%"<br />
$#''"5 '# $<br />
''A.F/ $)('$"<br />
% % " % '' " , ' <br />
'"7%#"$ #%<br />
$ A.F/ ' % ' . " $" ''<br />
$ ' 7 A.F/<br />
<br />
<br />
<br />
%<br />
<br />
?)@M3->HA>GEH<br />
<br />
F::))C11-II0&I>>DEA07&I>EEID>EE><br />
<br />
:&)->D0&-:&)7 .7>A7>G7>
!<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
"#<br />
<br />
!""""!#"" ""!$<br />
" " " " $% " & " $ " #"<br />
!"'' ""(<br />
• &" ) * " '% +!" " $% * <br />
$ " ! +"$ " " &" " " * ""<br />
" $<br />
• , ) * " "" * !" ! &"<br />
""" #!" !!" "&"""<br />
*""# ""$*"'<br />
• )")*""!" !&"""<br />
" " " "!"&"""<br />
*"""" "!"$"! "<br />
<br />
- "$ " !""" $"'<br />
"."*""!* "#"""""!! !$<br />
"*!"$"(<br />
<br />
&""#"( −/"<br />
−0*" 1"*"$<br />
1<br />
1$"!23 4<br />
−/"""!"<br />
&""( −,"""<br />
<br />
"#"( −)" " 15<br />
1<br />
1$"!<br />
−0!"$"<br />
−0*!" "<br />
"( −,#<br />
<br />
)""#"( −)" " 16<br />
1.<br />
1$"!<br />
−)"!" "
)""( −8"'<br />
−,! <br />
<br />
)""$!$#"#!"'<br />
*"(<br />
1<br />
1<br />
1<br />
)#$"!#""""!#*" !"<br />
'. - !. " "! " # *"' , ) ,""<br />
)*"')"),)$<br />
-!" +'.""" "<br />
" " ' ". " " " +!" " <br />
!9"""* <br />
'.*"*$$"*"""$<br />
)" ) - ## " " 1" # '. "<br />
"""!:"<br />
<br />
& " $ !9 #!" " ! "' $"! <br />
#""" "$"(<br />
<br />
1<br />
1<br />
1<br />
- "" " " " " <br />
<br />
-#""!!" " "#!"<br />
-# "'"#2" $*!"4" !"<br />
#-""" " !"<br />
<br />
-"##' "*!"'$"#"!$ <br />
"""*"'""""";)! "-"<br />
*" #,6)$"!*" !""#" <br />
" !"""<br />
<br />
$#%<br />
<br />
$#"#%<br />
<br />
" "! ' !" " * ' $ " #<br />
"" " "" " "" " * "$ "' "<br />
#!" ""$"! $# *"" " <br />
"1$"!6 "'""# "" "<br />
(<br />
<br />
1<br />
1<br />
1<br />
1<br />
1<br />
#!""#""""' <br />
" " """"$<br />
#"*" "!'"!!" "'<br />
" !" !"""$"$<br />
#!!" "<br />
* * " * '" ' " !#!" ' !<br />
!"*'""*"'" !#<br />
**"!!* !"**$ ""<br />
<br />
$#$#&'( %!)*<br />
<br />
-"*;)8"'"!"$"""*"<br />
"!" ""-!*""" "'.<br />
<br />
7
*#" "*9""*" !" !"!"<br />
" - ! "" " " ! " '<br />
<br />
<br />
,"" ";)"!""","")!""<br />
!" " *.' " " " "<br />
"#*"$"!$" ""!"*"<br />
" *""""* "$'"""**".&<br />
"'$1*" ""'."*#$"!-""!<br />
!""!"" ""!"!"*#$"#<br />
**""!6"#" ;)"!!"<br />
#"!""" "###1"!""!"<br />
! 2"" " ""4 & " '$ " <br />
"* " " ! - # ! !"" '" $"! '" 1<br />
"*"""<br />
<br />
6!* $'*"!#" """ !<br />
"-#""""!"*#<br />
!"$ *$ "$ ' * $ " "" " "<br />
!" => "! ="> !*" " =&"!<br />
8">'""'""'&"!""<br />
!"*$!"#"" !$ "!<br />
"'."" "$? !"<br />
"" ";)""!"* - "<br />
"**"*$"""!"*#*$"<br />
<br />
<br />
/(;1! *1*""!""<br />
<br />
"" " * " ! " " " ! <br />
( .*.&"$"!" " <br />
*$ ! @ 21 ! " 14 A " " !<br />
"*(<br />
1 6" " !" """<br />
1 " +"$ ". " !"$ " !" " " !<br />
<br />
1 " " !" $" " !"<br />
1<br />
""! <br />
&" "" !" 2 '* ! " "4 *$ "<br />
" " *$ " "" " ' '" " ! <br />
&""*"'"!"".""'!""<br />
'"<br />
<br />
6 !"" "" " $"! " 9"" " ' "" !" <br />
" $""!" $-!#" " "'<br />
!""""*""1"""! <br />
"!" $ " " ! !" :" *$ "! & " "<br />
" " '"9" ' $!!!."" " <br />
! " " '. $ '" !"$ 6 " ! " !" <br />
!""*"""*!" !"!-!'<br />
+!" " "$ ' '" ! !"" ' *'"<br />
+!"B6"'.*# ? 6"'.<br />
<br />
$#&#!%!+<br />
+*<br />
<br />
? " "*,6),6,6$"! "" !""<br />
!9"$ "$"!#! " '"""<br />
!" !" " 9*'$6""!"!""!* <br />
'.!"$"! """% !"&<br />
""+!""*"""*""" !""$"" $<br />
&" $ " "* ' " " " ""*" !" ' "<br />
"$ "* ' # " " !" " !" " <br />
!' ""*" "!")"$"*""<br />
" "$ "* "" ,6),6,6 $"! '" !! "* '"<br />
""*#*"-! "'"<br />
'"''"*#*$"6" ""<br />
*"$""!" !"" *"'""*-<br />
"#$!""B6*""? 6''."$:"<br />
! ""- *:"1""*"" '$"<br />
'$ " " "" $ *" *:"1"<br />
"**$*" "!!""*<br />
<br />
$#,#+*&'( %<br />
*++<br />
<br />
6 " " "!$ " " !" ;) ! " *<br />
!"'"1"" """"!"A""<br />
!""* " $'"1" "<br />
":"!"6""""!"" *""<br />
!" "" ,6),6,6 $"! " " !. 9"# "<br />
""$"""A!#" " ""<br />
'" ' * / " " !
"" ' $ " " '" " ' " " !"$ !"<br />
* "!" $ " - !" " " " " ;)<br />
""!* "!"$ 6"!""#"'<br />
*!"$"$""" ""*-##"<br />
" *'" +!" ? 6 "" " '. !<br />
#"<br />
<br />
/C()*1*""!!'"" <br />
$#-#<br />
<br />
<br />
-.$" $ "'"' "#!"" "*<br />
" " "$ " # !" "" *"' " "<br />
2# '. $4 9*"$ " *! " # "'<br />
" '" " " !" !! - ! " " *<br />
# """*"'"(<br />
1 6:"!" " !" " " *"" "" !" <br />
"*!""<br />
1 )"""*"!"<br />
1 .*#<br />
1 "$<br />
<br />
-'$ !""""*"""'$ !""<br />
*"" " " " * 5 " " " $ '" <br />
""*#!"*#"*" " !"#"<br />
"""'$-""*"""<br />
" $" " " * " " $" ""*" '" "<br />
""""1"!"-"""$<br />
. " !*. "#"-# ""$<br />
<br />
CDD
!"$*"""*"8"""*"#"""*!""<br />
&"#!"""""! $.$""*" <br />
#<br />
<br />
-*.*# "!""""""<br />
*"'"6"" !""* *.*# <br />
" ' *# " *. *"" " "" "" " " *.<br />
*"" # "!" " " ' " *. *"" ! ""<br />
"" *. *"" ' 9! " " "<br />
#1" ""<br />
<br />
<br />
/;(.*"" """"*"<br />
<br />
/E(0!"<br />
<br />
<br />
CD
6!""""" """"$"-<br />
"$'"" #"$#-""*#*$<br />
"' " / " - " * #" ""<br />
#""$"<br />
<br />
)','*""!" $ !"! "<br />
"""*"".""!3 *"!" $""'-<br />
'.""'""!" "!! "'<br />
"!" $"- " $ '"""<br />
'! "#$!!<br />
<br />
<br />
/F("<br />
<br />
<br />
$#.#/<br />
<br />
6 ="" "> " " " !" 2 <br />
" " ".3 4 "" !" " "" * " !" :"<br />
=""">' *"*""*$" "<br />
"$"!"- """' "<br />
"" " "#"""<br />
<br />
&" =""">'""+"#' "" "<br />
!9" "!'=""">!!"<br />
" """"61' "" *"*#*<br />
:"A"*#* :""""":"<br />
*$""! !"*$" "$!"<br />
<br />
<br />
CDC
(8 "<br />
<br />
<br />
/7("'<br />
<br />
<br />
CD;
$#0#1!23%<br />
<br />
&""*#!"""#!""!""# <br />
*"" " " "" "! !$ +"$ :" # "<br />
! :" " " $ . ' 1 " :" "<br />
"$ # .' *! !: - * #"<br />
'..'#"'"*#6#$<br />
:" # " * * " $ " "<br />
"$""!""*"A$1' **" <br />
$!* *$"1' *#* :"-<br />
'$ **$"$1' 9!"#*".<br />
" #* "$ "" !" " "$ 1" #* <br />
"$" """$ :")' *"'<br />
'$ 6 " "" " ! +" " .<br />
"""""" "' '$#"*"$ <br />
=*.>""#!!9"<br />
<br />
1<br />
<br />
66 &A 2CDDD4 ! " <br />
,A8&-8"!<br />
<br />
6BAA/G,ABB A6G0AH6G A) &0@H62CDDC4" <br />
&,,6%CDDC<br />
<br />
A) &0@H6G0AH6BH,G0@ 6&GAB6AB2CDD;4# <br />
$ C ,A8&-5!*<br />
<br />
A) &0@H 6G I&IA G I&6,@6 6 2CDDD4 % " <br />
,A8&-8"!<br />
<br />
AB6ABG0@ 6&G @-82CDDD4 !E"&/&8<br />
? 0FC? .J'&"#,6)J&,1E@,? $8!<br />
<br />
AB6A BG 0@ 6 &G @- 82CDDC4 ! K/!<br />
J'&"#,6)"J'&"#%@,L? $24<br />
J'6!8*<br />
<br />
6BAA /G 6)@L6 ,G @- 8G 06 ,&6 BG 6I6HA &G I&6,@6 6 24<br />
& &"'();1E;<br />
<br />
)@ 6 LG 8@H6 LG 6BAA /G 06 ,&6 B 2F4 <br />
"$"* $+ ,6)68MF<br />
<br />
06 ,&6BG/ 6)HIG-A AL6L2E4 "$" $"<br />
""*'(,<br />
<br />
<br />
CDE
!"# $ # <br />
% &%' <br />
!"# $ # '' &%' <br />
!("()%'*<br />
)'&' <br />
<br />
"<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! "# $<br />
<br />
<br />
<br />
#$% <br />
<br />
+" "" ),)'-' "# )'<br />
). ' *%' ) "# #<br />
) $ )" )%' ' '# " ' " )'<br />
" ) & ")")'/'#$<br />
" )%' ). # )' " ) ) <br />
)#) )""<br />
0 *1'))#0 &)'%#<br />
% &''(" & )'' /'*' &<br />
' ' )')' #' & &% ' <br />
%')" ) " &+)#) <br />
))<br />
<br />
1 #')&' /' " )<br />
%)' )$ &'')"')<br />
/' ' '#)) "# ' <br />
)& )% '#) "<br />
) % #) &##<br />
&" ' ' % % )<br />
+ /' # ) "' ' <br />
* )++("+ '# "#')) <br />
) ')) "%%"" )<br />
' )# ) 2 1*$ *3 41$35 )<br />
& )'$#" &'<br />
)"')' " ) ' <br />
<br />
<br />
0 & 162 ) " ) ' ) ') $<br />
<br />
<br />
<br />
! !<br />
"
" ) & )%')<br />
) )' )# ) ' 4 '<br />
&%&))8 5 &) <br />
')) ') ) ))%'9<br />
&%'% " ) %')%#)#<br />
)")') $%#) <br />
) %"' 4# ) 5 ) )' )<br />
)1 &4 :)')''5,) ''<br />
% ) & &#+ ' <br />
%' :<br />
1"' ) % ).-<br />
1"' /' )" &)'-<br />
1"' ' %#) &#*<br />
4,)#)' '# -<br />
)" ') ) -<br />
1"' )") &")<br />
4%&% ' ' % ))"<br />
' & ' %" )%'-<br />
1"') ) )-<br />
)" %&))#)')<br />
$ ) & ) ) ) ' % ) %# :<br />
')' /'#;' '" &*<br />
%% ))%'):<br />
$) %# )')%"' ) & #)')<br />
& )' &'-<br />
$ ,%# ) )'& <br />
)))<br />
<br />
&$%<br />
<br />
&$#$<br />
<br />
% &''(" # ,) ' <br />
"%#) )%') #1""#" %"& #<br />
) )''3 %" 162)<br />
< )< ) +" %&<br />
"%&) %&#'%* '))"%<br />
" "'.;=<br />
<br />
>? " ' % ' $<br />
) ' ) ) <br />
# $ ' ) %& )# <br />
)* &>1? " & '<br />
/' )' * ) ' <br />
& ' %# & )' + )' <br />
), ##'%#4 ) ' '<br />
')) " ')# '5) # ),@#) )& <br />
# ) ' '# 4 ' A5 /'<br />
#') B#")' ') , <br />
%' #))%' ) #' <br />
& " % ) &" "# ) # &# <br />
.' # ')'/' )')#<br />
/'"<br />
7
Process model<br />
Process owner<br />
communiqué Process yard communiqué Sub-contractor<br />
Start<br />
project<br />
Prelim<strong>in</strong>ary<br />
study<br />
Feasibility <strong>in</strong>formation<br />
Feasibility <strong>in</strong>formation<br />
Attributions<br />
to prelim<strong>in</strong>ary<br />
study<br />
Attributions<br />
prelim<strong>in</strong>ary<br />
study<br />
Generat<strong>in</strong>g<br />
program of<br />
requirements<br />
Tender<br />
request<br />
Start<br />
project<br />
Advice prelim<strong>in</strong>ary design<br />
Shar<strong>in</strong>g<br />
expertise<br />
Evaluat<strong>in</strong>g<br />
offers<br />
Adapt<strong>in</strong>g tender & specs<br />
+ updates (x times)<br />
mak<strong>in</strong>g<br />
Prelim<strong>in</strong>ary<br />
design<br />
Tender<br />
request<br />
Start<br />
project<br />
Project phase<br />
Mak<strong>in</strong>g<br />
Shortlist and<br />
choos<strong>in</strong>g<br />
yard<br />
Contract design<br />
on ma<strong>in</strong> features<br />
specs & price<br />
Contract<br />
owner - yard<br />
Adapt<strong>in</strong>g tender & specs<br />
+ updates (x times)<br />
Tender<br />
request<br />
Mak<strong>in</strong>g<br />
contract<br />
design (x<br />
times)<br />
Start<br />
project<br />
Detail eng<strong>in</strong>eer<strong>in</strong>g system design phase<br />
evaluat<strong>in</strong>g &<br />
choos<strong>in</strong>g<br />
eng<strong>in</strong>eered<br />
systems<br />
components,<br />
layouts<br />
Evaluat<strong>in</strong>g<br />
work<strong>in</strong>g<br />
draw<strong>in</strong>gs<br />
Ma<strong>in</strong> draw<strong>in</strong>gs for<br />
approval and system<br />
schemes for approval (x<br />
times)<br />
approval<br />
Draw<strong>in</strong>gs for approval<br />
approval<br />
Mak<strong>in</strong>g<br />
technical<br />
(systems)<br />
design<br />
field<br />
specialists<br />
Mak<strong>in</strong>g<br />
work<strong>in</strong>g<br />
draw<strong>in</strong>gs<br />
Approved<br />
eng<strong>in</strong>eer<strong>in</strong>g<br />
package<br />
Adapt<strong>in</strong>g tender & specs<br />
+ updates (x times)<br />
Contract<br />
yard - subcontractor<br />
Suggestions for approval (x<br />
times)<br />
Suggestions for approval (x<br />
times)<br />
;=:2 ')#<br />
Mak<strong>in</strong>g<br />
contract<br />
design (x<br />
times)<br />
Mak<strong>in</strong>g<br />
eng<strong>in</strong>eer<strong>in</strong>g<br />
for approval<br />
Mak<strong>in</strong>g<br />
detailed<br />
eng<strong>in</strong>eer<strong>in</strong>g<br />
Approved<br />
eng<strong>in</strong>eer<strong>in</strong>g<br />
package<br />
<br />
&$& <br />
<br />
1 ) # % ) & D)' B <br />
# ) 1,) ' ' ' <br />
" ) ) ' )) & ' DB<br />
' 2% ' # ' ' 4)5<br />
& ) ) ) '<br />
'# #" /') "))&<br />
&%"' ' "<br />
<br />
, )++-"" ) '% & )' " ) )<br />
& ' '* '3 ' " )<br />
% & ) /' ' # # ' ' %<br />
'% "#&1)):E$ ' *)<br />
& ' )%%# %# ) %# "& ,) ( <br />
,) '" ) ' "')<br />
<br />
<br />
C
% ' "' " ) '' "' )'<br />
" )) 4 &') # ')"%#<br />
5"' ' ', "<br />
,) ' & ' & %" & <br />
)'%'# %'&' ')%'<br />
' ' ?<br />
<br />
+ ))&'%#%# . &'''"& )) )%*<br />
% '))% ' )<br />
3 ) )D #)B<br />
$ )' #, )<br />
"G" .# ) <br />
' ' 4 / )5 $<br />
'/''%# % ),# #<br />
<br />
+ )) ' )<br />
D@ &# )* ,' <br />
',) /' )) # ' ' ) # % <br />
&# "%# # ) B%# <br />
&''(" $ ' %# ')) & "' )) <br />
$ ')' %#'D ) B<br />
% )
'' #" %5%# ' ,)&<br />
&& ;) &<br />
)%#% , )<br />
<br />
+)%# ;%) 0 )<br />
,4) "%# '5 )'& <br />
) & 4")"'5$ ) &<br />
&/'&&"#+,)'><br />
' '% )K@?$& '<br />
"& %# ' ' % ' # $ ) <br />
'& - )0 )<br />
# ' ') '&<br />
"$ ' )43 3)= 3)5<br />
%# )%%# %' ) ! " ) <br />
&) /' 45&; ,):%#<br />
/' ' & " <br />
<br />
$) )H#)'))')<br />
) ) ' ' )%%# %' ) <br />
'% % ) ))$<br />
)') )*)
*#<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
"<br />
<br />
<br />
!<br />
<br />
<br />
+<br />
Not OK<br />
*<br />
<br />
<br />
<br />
<br />
<br />
!<br />
OK<br />
<br />
$<br />
<br />
!<br />
+<br />
ok<br />
<br />
&'<br />
<br />
<br />
()<br />
Not OK<br />
<br />
Not ok<br />
Not ok<br />
- +<br />
.<br />
/<br />
<br />
<br />
<br />
<br />
<br />
<br />
"<br />
<br />
#$<br />
<br />
%<br />
;:1,) )/'<br />
Task: SP_FO1<br />
<br />
OK<br />
<br />
,<br />
Not OK<br />
*#<br />
Not ok<br />
Effort required (hours)<br />
8<br />
7<br />
6<br />
5<br />
4<br />
3<br />
2<br />
1<br />
y = 19,719e -0,3313x<br />
0<br />
0 2 4 6 8 10<br />
sum of parameters<br />
;6:%& )<br />
<br />
&#"%" ) )%%#%'<br />
) ) )) " <br />
' )) L= ' #)''<br />
& ) LOL # ' ' &; <br />
,)& ' #''#% '6&% <br />
" &" & ' &<br />
" L;36 ' %# %' '<br />
&) % ) <br />
<br />
=
3#)"%')# %&' <br />
'#'"%' $ ) % <br />
) ) # # ' <br />
"# & ;' )/'; )')<br />
) & # &" ) # # /'"<br />
"' "% ; ,): %& # ')) "# %& %<br />
$ /'""'/'%#,) <br />
=* =*64) )5$/'% % <br />
' '? 6 ) '% <br />
)5$)/'# ' ,)%) <br />
,)"%'% ",))<br />
<br />
$ ) ") ''# % <br />
'' ' $ ) #':<br />
# 4"'% ))& ) <br />
#"#% '5#)4" )<br />
# "# % ,' 5 + "# ) <br />
)) $ &) &":<br />
* -<br />
* ) -<br />
* ) ;'3)')-<br />
* #
'$'- <br />
<br />
H % )) ' % " $ <br />
%&)') ) )%%#%'%<br />
& )# )"%# &" )#<br />
% ' )" %# ' %#<br />
) ' " & %' <br />
'"0 '& *)') ""'<br />
)" ' ' ) &%' 'H "&<br />
& ) " % "' ) & '<br />
'&)'"')0 " ) <br />
%&'"'%' '"'&' " "# <br />
#)%%#%' ) '% <br />
& 0 ) ' "' <br />
%' ' & ±P ' "' $ ' " %#<br />
'% ' # ) " '% ,) <br />
" ' % % <br />
<br />
.<br />
<br />
0 )) ' & '% ) ' %&<br />
& , '# )'%'<br />
' ) )'$')))<br />
& ' &"$' :<br />
* 1"' ')') ))O=-<br />
* 1"' ' )' 'O<br />
<br />
.$#% <br />
<br />
+) %&# " ) &" ) &<br />
)'"#')') % )'" )@#<br />
)') ,%' )) %#() <br />
0 0 &''(" + )' ) : E ) ' <br />
) & " )' #E + )' # <br />
') )' ' ) '%*# 4M )'<br />
) 5H ' ) )' )')<br />
'/' ' )' /'$' )') &)<br />
) )) ) )' 0 <br />
0 &''("* )++("<br />
<br />
<br />
<br />
;O:+)) &+*) O 4 ':&&&" & <br />
&&&' &&& *' 5<br />
<br />
& ' $!( )* +!,--<br />
.!,,!,-- /!)* 0 %'"<br />
=
; ))& ' /'#) ) )) /' <br />
$/' &)' %<br />
6+)' 162 &":<br />
* 1 2<br />
+ ) '' )%' ')) <br />
) %# # '# # ';# '<br />
' :<br />
o '% )& #-<br />
o $%&'%/' -<br />
o " )/'-<br />
o )& #<br />
<br />
+)' # # 3 24325 $)& #<br />
" %' % '% 45 ) @<br />
):<br />
$%:322';#<br />
32*+ 32*H 32* 32* 32*;<br />
945: 6 7F C7 FO J=<br />
0 45: J == = =6 =6<br />
45: O 6= 6O 6C 6C<br />
()4 5: F<br />
$ 2 & C ==C =6 =J<br />
40 5:<br />
)<br />
4 5<br />
OO FJ ==J6 =OF =OO<br />
&: F OF F C F<br />
#): ( <br />
''<br />
#<br />
9 <br />
''<br />
("#<br />
"<br />
;#<br />
)<br />
"<br />
)<br />
"# "<br />
"#<br />
<br />
<br />
H ,)"' # '%:<br />
o '=4 32
"# )') " )') )') ' )'# ' <br />
)) /') $ ) # & <br />
.
.$#$&$ <br />
<br />
H # ' % )' )<br />
)"%; )' )%<br />
'&=)$'JP "'"& *<br />
& =P "#' "%<br />
%1# &''(" "' <br />
% ' )' ) )) & ' <br />
% 'CP)) ) )$%$ ' <br />
',&) /' ) #=<br />
<br />
$%:;#= /'4,5<br />
() 32*; 32* 32*+ 32*H 32* $ 1 ;#<br />
3* ) = F= J= F=F F6 O6F<br />
2') )) = C6 FOC C7 CC O=6<br />
G'4P5 CC C C C7 C<br />
<br />
$%:;# /'4,5<br />
() 32*; 32*+ 32* 32* 32*; $ 1 ;#<br />
3* ) = J== F=7 CFF F= O66<br />
2') )) = FOF C C7 C6 OF<br />
G'4P5 7F C CJ C7 C<br />
<br />
;' % ' #"%'' <br />
,' "&$' ' # <br />
" ) ' )' ) )) <br />
# ' ' + # ) ) <br />
D )') B)%%#%<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
occurence (%)<br />
35<br />
30<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
69 - 70,6<br />
70,6 - 72,1<br />
72,1 - 73,7<br />
73,7 - 75,3<br />
75,3 - 76,8<br />
76,8 - 78,4<br />
78,4 - 80<br />
80 - 81,5<br />
81,5 - 83,1<br />
83,1 - 84,7<br />
Eng<strong>in</strong>eer<strong>in</strong>g Effort (simulated)<br />
84,7 - 86,2<br />
86,2 -87,8<br />
87,8 - 89,4<br />
89,4 - 90,9<br />
90,9 - 92,5<br />
92,5 -94,1<br />
94,1 - 95,6<br />
95,6 - 97,2<br />
normalised time categories<br />
(total effort)<br />
972 - 98,8<br />
98,8 - 100,3<br />
100,3 - 101,9<br />
;:;/'#%', )= #=<br />
101,9 - 103,5<br />
103,5 - 105,0<br />
105,5 - 106,6<br />
106,6 - 108,2<br />
108,2 - 110<br />
Ship 1, oneoff/product<br />
platform<br />
Ship 2, one-off<br />
Ship 2, product<br />
platform<br />
<br />
=
.$&<br />
<br />
+ ) )# *' & <br />
" ) &" ) ')' '4
) ' # )% ' '+)' ' <br />
' &#<br />
!4'% ''# '5:'' <br />
)' ' & ' '% ' '% <br />
'') #B ' <br />
95 4 ** )##5:<br />
% ' ' )' ' ' ) <br />
4& , ) 5#<br />
* 5 : ,) '%* &<br />
),# <br />
&%<br />
<br />
.$&$# <br />
<br />
+$%;7 &,)% "D)' ' B<br />
7PJP 4 ' "5%4 ' <br />
'5)"#; DB4 ' " %&' <br />
)' ' ' %# '% 5 ' ' <br />
,) " =P ' $ , ' ' %# ) <br />
' ,))#"%##$,)))"<br />
)#) ' %' $' ' & '<br />
#%%#) #,)4)5' % <br />
) % ) " " + )"'# ) %<br />
D B"% ' ;C%' <br />
'' &+ # '%#) #<br />
' ) ) '' )'<br />
' '%#<br />
<br />
$%:)' ' /'4,5<br />
% <br />
<br />
, ) <br />
))' ' <br />
JOP J=P JP<br />
frequency<br />
16%<br />
14%<br />
12%<br />
10%<br />
8%<br />
6%<br />
4%<br />
2%<br />
0%<br />
71-73<br />
77-79<br />
83-85<br />
89-91<br />
95-97<br />
101-103<br />
107-109<br />
113-115<br />
119-121<br />
125-127<br />
131-133<br />
ref<br />
case a<br />
case b<br />
case ade<br />
<strong>in</strong>dexed eng<strong>in</strong>eer<strong>in</strong>g efforts<br />
<br />
;7:;/'#%' ' )' ' <br />
<br />
=C
!<br />
<br />
$' &%# ') ##<br />
)) & & ) )<br />
# " ) " ' 1 ) '<br />
#% "') % &"&)))<br />
& -;' # )#" )$ <br />
%, ') ) #'% <br />
% %#( # & ) <br />
#)/' )* ''<br />
&% )*&$# '&% ' <br />
& %# )) ') % "% <br />
1!0 1!(UU1!(!4O5: <br />
6 3@2$('.))=J*6<br />
<br />
1!0 1!(UU-1!(!4O%5: <br />
OF O'%"))OO=*OO7<br />
<br />
1GHG+1 +- 1U1 20 - H3($+19*H93 0 4O5 <br />
!"# $ #<br />
<br />
<br />
=J
!" <br />
<br />
<br />
<br />
<br />
! " # #" $ # " !" # "# ! " #!% "&<br />
" '" " % " # % % # % <br />
" " % " #! " (# " # "<br />
" ! ) " # % # # " <br />
") *!! $ " ""! !! " ""!<br />
" (# %!! #+ " #" % " !<br />
" " ,! ### #" "<br />
<br />
!" <br />
<br />
# # " ##!! ")!"!<br />
## "#!" ""!" #$!#%&%!#%'(<br />
)&*++,'<br />
+ '$%)!! "- ! #" !!$ %)!" "<br />
""#%! "&%! ! )" <br />
#" " " ! <br />
# # #" $ ! )")!! <br />
+ ) " % - (# " #" " ! " <br />
+$ #(# "# # %!".+<br />
# ## <br />
+ $)!"%"-(# " !% ""%# !<br />
#) # " ! # % #) "%<br />
(# ### " " $## %%!<br />
. (# "#) )" "!")!"$<br />
+ %% !- %!# %<br />
# " " )!"$!!#" <br />
!#) # ")!" , %!<br />
+ '$%%!-# %!"%# # <br />
#) "" # % ! % # " '$ <br />
# " "! "" % # / % <br />
#) # "$" )!"!! <br />
+ '$% -!!" "" "% $0<br />
" # " 1.
+ 3&! - # " " " "%% <br />
" # #""$ ! "# %%<br />
# " )" "# (# & "4) 5<br />
"# "4## " ) # !!5##) ) <br />
#"!$%"#&<br />
<br />
# ) # #!! !" 6<br />
+ / ) ! % +! " - (# )!) ""! , <br />
")")!!) ")!"%%<br />
+ / %"""%-""%# !) ")!!")!"<br />
%% %% !# # #!")%# " . <br />
% # # !! " + 7$ $" 8 #!" <br />
! % # # " ! #1 # <br />
"# $$" %!! "% # <br />
" <br />
+ 9% -" "# +! #"%"<br />
# " " % (# %" % " ) ! " <br />
.%# )!"% 3) # " "##"%<br />
# %!"# ## !<br />
+ :"" #-/" %!#!" ! )# <br />
" %! " # !! " # <br />
##!" "<br />
+ !# %% +" !-%% ! "%% <br />
" #"))!!<br />
+ +" " ! - +" " ! ! # " .$!<br />
" "% ,"" ! " %# <br />
# ! ! ! # .$! " " " )" # % ##<br />
"# $ % "" <br />
+ !" -(## )!! # !<br />
%,%# " ## #" (# <br />
!" ## # # " ") "%% %# <br />
%##" ,<br />
<br />
# <br />
<br />
#$ <br />
<br />
; #"! % ! %"%<br />
# !! " $ "#/ "# ) ! %<br />
# #) ! ! # # ;# " <br />
!" #"!!# "!)!! " <br />
<br />
# %<br />
<br />
/ ! " # " ## 7 %% 8 # <br />
%!# ! "#7! # #% 8" %% %<br />
#!""# !7# $!#28(# ) !")" "# #! #<br />
#+# ## ## %!0<br />
,! 6<br />
<br />
;?$ 2 ,' ,,/, @ <br />
<br />
;# 6<br />
;!## #<br />
$ 2 %% " " #&" <br />
<br />
2
' ! # "!<br />
<br />
/" # " $<br />
!$ %% <br />
<br />
(# %% $ 2 !!"%#!")# %!" )! <br />
%' /" % ,#<br />
<br />
$ 2 ?;A7' ,,/, @ 8<br />
<br />
(# ! # ,## # " "# )!"!! ;7$ 2 %#<br />
" !8<br />
<br />
##&$<br />
<br />
! # "" # !## # (# !## # ")" " <br />
# #% #"+""# !!## #(# %<br />
! # !B+C " <br />
<br />
! ## ##02&#$'6<br />
<br />
<br />
<br />
Light-ship Weight<br />
<br />
<br />
<br />
Equipment<br />
Hull<br />
Mach<strong>in</strong>ery Rema<strong>in</strong>der<br />
<br />
<br />
026,! !!" # #<br />
<br />
!) !) %)!#;$$"# %! ) !")" #<br />
#0(# # # #02" ," "! ) !%" !<br />
!#""% ! " "# ! ,%# #"# ! ) !%" !%<br />
# " # " ! " " ! <br />
# %! " %% # ! % " " <br />
$ %# " 7 )! 8".%! 7 <br />
%# !8 ,! +" " # #; #<br />
<br />
Light-ship Weight<br />
Hull Structure<br />
Propulsion<br />
Electric Plant<br />
Command<br />
Auxiliaries<br />
Outfit<br />
Armament<br />
Eng<strong>in</strong>eer<strong>in</strong>g<br />
Assembly<br />
<br />
<br />
06,! )!!" # #<br />
<br />
#' () ) <br />
<br />
(# ! ) ! % %" 7 8 % # # !# 7;D8 " " <br />
!" )!""$ ("!! )" "# <br />
) % # " # # ! " " # # %%
"" # " ) % # +! # # # ! <br />
%" ! ) ! % # " " !! % ! # # !<br />
# " %% ) !") %%" ! ) !) #%" <br />
! ) ! )" .) # # )!" % # # ! # <br />
%" ! ) !".$!" # # # ! ) !! "# %# $<br />
%) +" + ## <br />
<br />
;#! # "# """ ) !!"" " %# <br />
%" %# (## # "#; # %"% ,#<br />
)!! # )! %# """ ) "!!# (#!!! <br />
)" % )!# )!"%# " #(# """ )<br />
" # !)! <br />
<br />
' <br />
<br />
'$ <br />
<br />
(# ! # %# ! %#"E<br />
## !!! # #"%# ) "! !!<br />
!%!!"! $&)*++,'<br />
<br />
' %<br />
<br />
# # ! "## #<br />
%!" %% %#!""# !0 # <br />
%%!6<br />
+ :!-:! # "# #""!" # !<br />
! " ) )" " +#" ! <br />
!" "!<br />
+ '-' # #%#" ! # #<br />
(# # !" % # #"1 ! " <br />
) 7 ) # "!"8<br />
<br />
# % # %! !" ! %% " ! # ,! <br />
+" " %"0!#/ <br />
<br />
'#&$<br />
<br />
!# "" # !"!#%"# #<br />
7# # ! ," " !" " # "<br />
"!8(# # # # # ## #" <br />
!# #"<br />
<br />
* ?$ ,;<br />
<br />
;# 6<br />
* # %# #!!<br />
$ %% " " # <br />
<br />
""! %% !" "# %!%# "%<br />
$#%!" # %!!6<br />
+ :!# # 7 " # % !!8<br />
+ %! " ,# # 7 % # #" # % %<br />
8<br />
+ /%% " 7 % !!! ,! 8<br />
<br />
C
+ " %% <br />
+ / # #%# !7 #!! #!8<br />
<br />
# ## %% !!"%#!"# <br />
%! " "!!6728!" . "78!# "<br />
) %# #!! %")!" # !$%)!! "(##!! <br />
,"# !$%)!" %% ;#! 4 5 %% !" ,# #!"<br />
%!# %!#"##"#!"" ) ! %% #<br />
"# . "!(# ! ) #<br />
""%###" %% " ) "3 ,! F <br />
#" $ ) + " C/ ! " ! # <br />
#" ) 7$*++,8<br />
<br />
'' ) ) <br />
<br />
% # " ) % # .% # <br />
%" ! ) ! % ! !# 7D8 (# ) % # <br />
% " #) ! % ! !! # " ) " !!<br />
# .(# #"" )# "<br />
) !!%" ! ) !%# ;D"D"## ! "<br />
<br />
*+)), <br />
<br />
0C " # " " ) " ! # #1<br />
#"## #" !#!" %" <br />
! ) !<br />
<br />
Weight Estimat<strong>in</strong>g<br />
Coefficients<br />
Cost Estimat<strong>in</strong>g<br />
Coefficients<br />
Ship Characteristics<br />
- Length<br />
- Beam<br />
- Shaft Power<br />
- Fuel Capacity<br />
- Others...<br />
Weight Parametric<br />
Formulas<br />
Cost Parametric<br />
Formulas<br />
Initial Design<br />
Cost Estimate and Level<br />
of Confidence<br />
Weight Confidence<br />
Levels<br />
Cost Confidence<br />
Levels<br />
<br />
<br />
0C63) ) %!" #+ " <br />
<br />
(# % # # # 0B +" "<br />
!!) !! (# "# !!% ) !6<br />
+ # "## #"-" "! <br />
" ) # % "# "!!<br />
+ %!- " #!#&" "<br />
) "# ) <br />
+ : % " " - # " ! " %% " " ! <br />
!"" " ! ) !##%% " !)!! " ) !"##<br />
%!#!"" " # .%"!! #<br />
.$" " "%! # %" <br />
+ %%-# #"%! %#<br />
# %!%" ! ) !" " #"<br />
B
-) ) <br />
<br />
") """) %# #"!#) # "% "" ) ! "<br />
"# %!! <br />
<br />
-) <br />
+ (# #"! !! " " 7# %% " %! <br />
)!""8#!")" " "! <br />
+ (# " %$# ## #! % ! !!<br />
+ (# # ! # #!$"! ) !%" !<br />
+ (# #"%% .$#$! #!!!%#<br />
%<br />
+ (# #"!"" " <br />
+ % "# # ! <br />
"##) " <br />
<br />
-) <br />
+ (# "") #"" ) ! %! %% "%" <br />
! ) ! ")!"" )!! # #"<br />
<br />
./ $<br />
<br />
(# %!! ,! ### #; #% " #"<br />
# 0!#/ ##; #"0!#/ ! <br />
%) "!" # ,! % # "!# !<br />
" (# ,! # D!!+AD!!+%%A ) ! 7D,8 " !<br />
" (# %# # "(! <br />
<br />
(! 6,! D,# <br />
<br />
! 01 2310&4 &5 +( "(3<br />
(!%" 7G8 @B<br />
"#78 C<br />
!$ %% @=<br />
/ ## / $78 / 2=@<br />
/! 7 8 /! 2BB<br />
/ ## :/ $78 / <<br />
/ " #7 8 / C
(# "#; #% #% #: %# <br />
%# % ### ! # ! #!%%0B0# % #<br />
%% !!" " "! " #!! # # ! %+#"<br />
%0B" %# "! "E# ,! !! "<br />
) !!"%# 0#) # %% $ 2 " ) " 27#<br />
# #" # 4I! 5 ! % # ! # #+#" % 0B (# <br />
!## # # !!"# %!## %# %6<br />
<br />
;?$ 2 ,' ,,/, @ <br />
<br />
(# # " !!!!# ## # #<br />
<br />
0B6'## # #; #<br />
<br />
(# !!" # !! " / (# ! <br />
! , !! "##!$# %(# <br />
%0@## D,!% (# ! %0@ " !%# <br />
##!#" 4'56 ")! ## "<br />
<br />
H
0@6,! # "" !%! #) !!<br />
<br />
(# # " # #%# D, ,! # <br />
# ,! # # # # ! ) ! # 0H / " # ! ) ! % " !<br />
" "# # # # ," """!! ) !<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
0H6,! #!" # #<br />
<br />
>
0H ! %# #"! ) !%# # ###!#"6';27. %8<br />
;# "! +!$ " # ! ," # # 0> ## "! # <br />
%!!6<br />
• %6 "/ <br />
• D !#6 )" " % # 7 8 7% ! JK #<br />
,! +""8"'7# #"% 8<br />
# !# # %!!6<br />
o ,-# !#%### )! " ) "<br />
o I! -# ."!% , # ,<br />
o L/ )- " )%# 7# """ )<br />
"%# +" % " #"! 8<br />
o / )-# !)! %# " )<br />
o D!!+" -# "%% # )! "# %# )! %# <br />
,! ! ) !# # #(# !!+" ! # # ,! <br />
! ) ! !" " !! " # ! ! ) ! !"<br />
! # ! ! ) ! )! % ) " " ! " " # <br />
%!!!"# !!! # ! ) !(## # <br />
" "$ # # +) # <br />
o 4; #5!6'M(MIM-'"!) ") ! <br />
%)!!"!# #")" "% %<br />
<br />
# # % ! ! " % ! * # % !E # <br />
%#" )# ! )%!<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
0>6; #"%%D, . "<br />
<br />
<br />
=
6<br />
<br />
(# ,!! #"% #"%# !" %#(# <br />
#" "%" %! ## %!" ""(# <br />
"%! %% "%" ! ) ! !! "# "<br />
!!%"%!#<br />
<br />
<br />
<br />
<br />
!$(<br />
$<br />
<br />
I-"$#$ $$.$
!!<br />
"! <br />
<br />
#$<br />
!!#$<br />
<br />
<br />
<br />
<br />
! "" " " <br />
# $%&!'(()$%&(*+,-<br />
. / % , $ % 0 % ),$%0%* & ! <br />
1" #1 <br />
2 0 /-)0/-* 3<br />
4 $%&( / )+,&* ! 1 56 ! "<br />
# +, 3 " <br />
"7! " <br />
5<br />
<br />
" 1 0/, ' ! # " <br />
"# !" "8<br />
" !3 3!<br />
! <br />
<br />
9 "! " 0/-! !<br />
3 " " " ! <br />
+ "! ! #<br />
! :! ! ! <br />
)!*! #" ! <br />
3! ! ! )!*<br />
<br />
;3 " !: 0/- 4 ! ! <br />
! ! 1<br />
&! "<br />
"! "<br />
""" !: &<br />
" 0/- !: 4<br />
" 4<br />
0 / "" 3 " 0/- <br />
+)+!*&"< ='<br />
<br />
<br />
' 4""0/->! ! %)0/,*
='?+&"< !0 /%"&'''<br />
<br />
" # <br />
<br />
4!!:! ! " 0/- " <br />
"! ! 4 <br />
!!<br />
1 "!:9 <br />
!3 3 <br />
<br />
& " ! !: ! <br />
! ! # # # ! <br />
! 3 "<br />
&$-/)$-/* <br />
:! <br />
<br />
% !13 @<br />
- ! # <br />
" " ! !<br />
1 , ! " ! ! <br />
! ! ! " <br />
<br />
" <br />
<br />
4#! ! " 0/- "! ! ! <br />
" 4 " )! # *<br />
! " )3 ! <br />
<br />
'
* ) <br />
*<br />
<br />
$ " 0/-3 "<br />
" !! <br />
! !3! <br />
4 " 3 2 ! 3 !<br />
! " 9 ! 0/- <br />
! 3#;&%&0-,&.%$,09
4 ! ! ! ! 3 <br />
!"= !9 &9 <br />
!5359 3" ! <br />
!" ! :! ! 3<br />
" ! 9 & 3 = <br />
!"9 9 >" <br />
!:! 2 ! <br />
<br />
MassItem<br />
id: str<strong>in</strong>g<br />
mass: decimal<br />
xcog: decimal<br />
ycog: decimal<br />
zcog: decimal<br />
bound<strong>in</strong>gbox: bound<strong>in</strong>gbox<br />
TankItem<br />
density: decimal<br />
freesurface: decimal<br />
Conta<strong>in</strong>erItem<br />
slot: conta<strong>in</strong>erslot<br />
<br />
<br />
=?+-< !9 4#9 -9 <br />
<br />
= 1 # )! 3 * <br />
4#9 !9!! 1# <br />
! ! #3 ! 3 <br />
-9 !9! ! !<br />
2 3 9,$ (E''' B 4#9 <br />
-9 5:5 " ! 9 <br />
=<br />
<br />
< ,/& "C! = FG<br />
B# - ,! 3-&%,4&;" ! <br />
" ! : ! <br />
! " ! 1 <br />
<br />
$ )<br />
<br />
4 ! ! ! <br />
! )! * ! <br />
9 ! ! ! C! <br />
) ! C , -* <br />
! ! " ! 30-,&.<br />
<br />
= ! ! 0-,&. C 3 # <br />
! # C % ! <br />
C"!!4 !3?<br />
- C< B<br />
- 0:, )C,*<br />
- CH)CH*<br />
<br />
$*( & +<br />
<br />
C< B !C )0 4*
55 ) ! * #<br />
I"-GG4" !C,
4 ! #3" " 3 CH<br />
<br />
<br />
& C,4CH3:3 ! !! 3 <br />
3 2 , <br />
3 " !<br />
-9 C 4! 3 <br />
3 !"0-,&. 1 2<br />
!) ! <br />
"*= !! 3#<br />
4! 3"CH !=A /<br />
:3C% 3 3<br />
<br />
!?4A )K9 *L<br />
K '?M)'N )K9 **"'<br />
K:'?M)'N )!KK9 K NK :*"K '*"'<br />
K'?M)'N )!KK9 K NK *"K '*"'<br />
K2'?M)'N )!KK9 K NK 2*"K '*"'<br />
)K 'K:'K'K2'*<br />
O@<br />
<br />
<br />
=7?- ! 0-,&.<br />
<br />
<br />
<br />
<br />
- <br />
<br />
<br />
CH<br />
%<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
=J&! < 9!<br />
<br />
<br />
J
& ! !3 ) #<br />
* " 4<br />
! ! ! !C 3"<br />
! <br />
<br />
, CH 3 ! ! ! " C <br />
! 4 ! ! - <br />
0-,&." C !3=74! <br />
!CH =J<br />
<br />
= "C,4CH"1 ! <br />
C !C,43"3! ! <br />
! 3 CH " <br />
!3<br />
<br />
/' 0 1<br />
<br />
= !!" ! "<br />
! !3 0/-4! !<br />
3 ! !!: <br />
<br />
<br />
& " ! ! ""!3#<br />
:! )Q ! 3*<br />
)0/-Q " * 4 ! ! <br />
C " <br />
<br />
, 33!!" 3 <br />
4 !<br />
!! ! 0/- 1# <br />
! ! <br />
<br />
<br />
&0;9,-I)*123,-44,&567<br />
4&/.9;- <br />
<br />
&/$
!"#<br />
! $ %& #<br />
<br />
"<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
#$ %<br />
<br />
'( #% )"* % #)<br />
# + , "#% # ( # <br />
##"+ -.% # %<br />
% %+ )#% # # % <br />
# ) # %<br />
# + )# , ( # # % <br />
<br />
<br />
&$ '"<br />
<br />
+ % # % # % # <br />
( #% #/ "## #<br />
#)#" &# %#<br />
% ## #%# " <br />
#### + )# ###<br />
# % 0 (# 1 % <br />
2#)#%%# %<br />
#0"%) # " (#) <br />
& ( " ) <br />
#& # #<br />
* " %# )" <br />
# " ##<br />
%#+ % )13/#%<br />
)# # % )<br />
##<br />
<br />
($ )*<br />
<br />
+%% ) ## #%#<br />
$ #% # )# "#% , # #
"# # 5 # ## % <br />
#<br />
<br />
($#$+! <br />
<br />
/# #% #) %) ,0<br />
# #,)##( #<br />
1, %6<br />
<br />
!"<br />
7<br />
<br />
<br />
= ( 7<br />
−7<br />
+ <br />
) <br />
879<br />
7+<br />
7<br />
#"<br />
<br />
<br />
<br />
+ ( :<br />
+ $:<br />
) −<br />
7<br />
− :<br />
−<br />
=<br />
<br />
<br />
89<br />
:<br />
+ $:<br />
+ 7<br />
$"<br />
$<br />
<br />
= ( −<br />
7<br />
+ <br />
− −<br />
7<br />
) 89<br />
7+<br />
$<br />
<br />
+ % #### "%6<br />
! ## #%#%#<br />
! ###%#<br />
! )89#%#<br />
<br />
0 "% ## " %# + # #) #<br />
+ # " )<br />
#+ ## ; 7 %<br />
? @ "% &! + <br />
25#%%##) ###<br />
<br />
5 # )' 7 <br />
%<br />
")#)%<br />
2 )") ) "%&!&# &! )<br />
#(%% ") ! + ")##<br />
()!") )0 #")#<br />
%<br />
<br />
+ "% % 8# 9 # ) <br />
#+ ) &! ## ) <br />
+ %##" &! ## #/%7")##<br />
4
% # )##@<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
−7<br />
$<br />
<br />
$ <br />
#)%<br />
<br />
&#<br />
$ + 7<br />
$<br />
<br />
8<br />
−7<br />
7<br />
9<br />
8<br />
−7<br />
<br />
<br />
8 <br />
<br />
7 9 9<br />
%<br />
9<br />
8<br />
7<br />
<br />
<br />
/%76$ *%<br />
/%6'( ##<br />
"<br />
<br />
/#)% ) " #"<br />
# %# ,)%0##)<br />
%#" #+ # % <br />
## % ##% <br />
## -% #.#-%.#%<br />
0 # %<br />
%# + % <br />
3 ) # + # # )# "%<br />
6<br />
<br />
")#)& #")###<br />
+ #+ %+ #<br />
%# % #+ #% <br />
)# %$")3#%&<br />
"# )% " + <br />
# % #)% %##% %<br />
<br />
<br />
% % #! # <br />
+ # % % % # <br />
##()#+ ## 6<br />
<br />
A
! 5%()# #<br />
! C%# % <br />
! @%%%## #<br />
/ % # 1 " # # #)#<br />
)#<br />
<br />
"!!#%1# <br />
##(## % + "!!## <br />
<br />
")#" )#" <br />
## % + -) .) ")) #<br />
+ ))## ) %#,%#"<br />
<br />
/%"(+ # # %+ ## <br />
% ::#)# 4+ :< )#<br />
%+ #)" %# #;:<br />
# %# 74:
%;" #% + ##<br />
(0): )%)(#% <br />
#0 :(### # "+ <br />
#B;+ 1#<br />
<br />
<br />
/%6"<br />
<br />
/%B6* %##"<br />
<br />
<br />
/%;6%#)%####/%B<br />
<br />
-$ .<br />
<br />
0 1% )""####* <br />
%# # #%(<br />
(# %)& #### <br />
"# )+ #% %<br />
## #% <br />
<br />
<br />
B7
+ # # % % #% " <br />
# ## & % # % <br />
#")#+ % "+ )#<br />
# % (% ## & #)#<br />
#)# %%#"&"#<br />
-%#. (# ) % 2!! + <br />
## #%"#)#2 # #)<br />
%") %#%"") <br />
<br />
/<br />
<br />
+ E # ) /# '# $ + % #<br />
*#%<br />
<br />
<br />
<br />
<br />
*'@'&F+ CG *0H0/ 87AA49 ' ( ) $ #<br />
$ I#<br />
<br />
&/0F8:::9; &/0FF%#FF0)%<br />
<br />
J0&@HFG/C0+87AA49* +,*5 !K%<br />
<br />
@2'00G0K@2G52H$'@@CGD0@$'@HC8:::9%% <br />
$*$%<br />
<br />
$'' &FC87AA9- 0##! <br />
<br />
$@L'J587A49 %F7B;<br />
<br />
+0K87AAM9- . 0##! <br />
<br />
<br />
<br />
B
!" # $ !%&'%<br />
( ) & ! & ! * ' <br />
#<br />
<br />
<br />
<br />
<br />
<br />
<br />
!" #$# % !# <br />
! #$$! #$!# # <br />
& '!( # $ $ !$ # <br />
$! ! ! # # <br />
!$ $ ! !$ ! $!! <br />
! ## $# $ ! !! $ ! $ <br />
!! $ # !# # !<br />
$ ! # $ ) * ! # $ <br />
#$ ! $ $ $ $ # ! $ # <br />
*!$#! $ *$ $ #$ $<br />
# ## $ !!<br />
! # *$ $ + ! , $ !$<br />
!$! #!$! <br />
$ $ # $ ! ! <br />
! $$$!$!##<br />
<br />
<br />
<br />
! $ ! $ ! !,#<br />
! !$! +(,--.*! ### !<br />
! !!$ -./$$$<br />
! ! $ # ! ! -01 <br />
(,--/*! $ #! !$ $<br />
2 ! $ $#$ # !$ $<br />
! $ $ # ! $ <br />
$ $ ! + 3 ! ! $ <br />
!$ $ #! ! ! # ! ! <br />
$ $ # $ ! ! $ ! !<br />
(/001*<br />
<br />
!<br />
<br />
! # # ! # # ! # $ $ #<br />
4 ! # #,# <br />
!! ! # #$$$!
# # $ $ ! # ! $ #*#5 ! <br />
# 5!! !!# ,$<br />
6 # ##5 $ ! ! <br />
! $ # # $,, $ <br />
! ! $ ! $ ! $ $ <br />
!#$$# # !# # !<br />
# # ! $! $# # $ <br />
$# !# $<br />
<br />
! + !# *$ $ *$ $ !<br />
6 ! $ # # 3$ <br />
!# # $ ! $#$ ! ! <br />
$ # $ ! ! ! ! # # <br />
$ $ '$ ! ! ! ! <br />
# # 6! # #$#$ !! <br />
! ,! ! * $ $ <br />
##5$#$ !!<br />
# $# $ ! ! ! <br />
! ! # $ ! $ $ ! <br />
$# $ !<br />
<br />
"#<br />
<br />
! *# $ 6 # $ !!$<br />
!$ #$ &,## 6! $$$ <br />
## $$ ! ! # $$ ! $## <br />
#$ ! % !$! ,## #$## #<br />
! !!# *! $$$7$ *! $!!$<br />
##$$ # ! ! &,<br />
## # $ $ # $ ! <br />
!, <br />
<br />
! 6 $# ! # !$ &, <br />
# $ # $ $ * $ # <br />
! #+ 6' ! ! ! $ ! ! <br />
## '! ( # ( $ #$ ! $ $ <br />
!$ #$$ ! $ <br />
<br />
6##! *! ## $! <br />
$! 8 #*## ! $ $!<br />
! $##6# !#$#<br />
$ # # '##<br />
$ ! *! $$$ $<br />
! $$ #!# $! ## $ <br />
#! $! 6 ! 5<br />
<br />
(, ##6#!<br />
!! $ $# $2!$ $ <br />
!$ !# ! <br />
$! # ,## ! # ! # <br />
$$ $! ! $#! ! # ! !<br />
!!!# *#! !! $ $
$%&#'( '<br />
<br />
6&9(49 :6 '! ( # ! &*$ $ (# &; <br />
+$ $! & $ ! (&<br />
! # !!! #$ $ <br />
# # &2 (/000* ! # ! <br />
#,!$$$,$$ # <br />
$$ $# <br />
<br />
)&<br />
<br />
! !$! 6 $ <br />
$!! # $ ## <br />
7<br />
; # !# # $ <br />
;$ $$ #$$ <br />
;$$!:#?!$;<br />
$;!# # ## #<br />
; *$ $!,:# $ *! !$!$;<br />
<br />
! $ ! ## ! $ $ ! $ $ <br />
#$$#$ $# # ! $ $! <br />
# $$$# $3! $$<br />
<br />
6&9(49 ! $ $ # $ $ $ # *! $<br />
#!!$ 3$$<br />
! $#!$ # #6 #$<br />
4 ,$ # $ $ *! $ <br />
! 6 #! $$$$! <br />
# @0 ,$ $ $ $ $ 6 <br />
$ 6 $ # $ *! $ <br />
!!! # A B!!! 6 #<br />
! #$ $! ! #<br />
$ $ *! $ # 3 <br />
3#$$$ $<br />
<br />
M o b ile A g e n t s<br />
D a t a E n t r y A p p le t s<br />
O O D a t a b a s e<br />
M id d le w a r e<br />
IN T E R N E T<br />
S u p p ly<br />
C h a <strong>in</strong><br />
S h ip o w n e r S h ip y a r d S e r v ic e S u p p lie r M a t e r ia l S u p p lie r<br />
@074!! $ $#,# <br />
<br />
<br />
/
D# ! # $ $ , $ $<br />
$ # #!$<br />
: $ ! ,,! , $ # !3 ; ! $ $ $<br />
$ #$! 3# $ #! $ $$<br />
$ $# $ $ : <br />
!$;$ *! #!$<br />
*! ! $ ! $ $ # ! $ <br />
'#!!! !$ #$ $<br />
$ # $ $! !$ 4 *$ $!,<br />
$ # ## $ # ! $ !<br />
# $ $#<br />
<br />
! # ! ! # $$!<br />
! $ 4# # ##$$ ,#<br />
! # 36# #* $#! <br />
#!, $$! $$ $$ $<br />
! $ # !$ 6 $$ ! # ,#$ <br />
$ # ! !$! ! ! $ *<br />
$!4=##6:4=6;! $ *! !! <br />
$ # ! $ 6 $$ ! D<br />
D :DD;#$ $E(F! # ! # #$ *! <br />
#<br />
<br />
'<br />
D6<br />
@0<br />
<br />
6<br />
(4&<br />
@<br />
@<br />
@<br />
( <br />
( !#<br />
@0?# !$<br />
@?$$ #$#<br />
@?#,!$<br />
@?!#<br />
@/? *$$!,<br />
@/<br />
=9<br />
<br />
@74 <br />
<br />
! ! ! #$ ## $ *! <br />
# $ # $ ! #$ $ * ! <br />
$ ! ! $ ! ! ## <br />
+,$G484$! ! ! <br />
! *#$$ $!!! 4=6! ! <br />
! # # !#$ 6! <br />
! 6 $ $ $ $ <br />
! # # ! ! :$ ! ; AB ! # ! <br />
!' # # ! !:! ; ! $<br />
! ! # ! !! ! <br />
$ $!!! <br />
C
# # ! ! # # ! #$ 6 # # <br />
$$$ ! $ $ $:! 3 $;! $!<br />
@! ! #$$$ ! $ !!! $<br />
!# $$ $!$&!$ $<br />
!! ## $ # $ *! ! ! <br />
! $ ! !6#3$ # $ ! <br />
!:!# $ $;$! <br />
! $ &! *! $ ! ! !! ! E(F<br />
!! $ !# 3 #$ ! E(F $<br />
:D6($;!! $ ! $<br />
! $ $ ! #$ $ ! <br />
# !#$$#! $!# <br />
<br />
<br />
$ <br />
6 =<br />
E(F<br />
<br />
<br />
4 <br />
)<br />
)$<br />
D&*!<br />
' <br />
D)<br />
'!<br />
% <br />
<br />
D6<br />
$ <br />
D <br />
4 +<br />
E(F<br />
<br />
<br />
#<br />
$ +<br />
6 =<br />
E(F<br />
<br />
+<br />
D6<br />
$ +<br />
D <br />
<br />
@7I ! 6&9(49# <br />
<br />
*+<br />
<br />
! $$$! +7<br />
− '! $ 7J# ! 3$ $ ! !<br />
! !$! ! ! 36 $ $! !$#<br />
$$ $ ! $<br />
− ' &# $ D D7 )$ ! ! ! <br />
!$! !!! # =$<br />
$! !$ #! =!! !$<br />
*# $ 6&9(49 $ # !3 $ ! $ <br />
! #! ## $<br />
− (, '=7!$!!+!! <br />
$ &! ! ! $ 3 6&9(49<br />
$! $$! $!<br />
!$<br />
− '! ' (7 ( ! # $ # ! $ ! <br />
! * $ ! # 6# <br />
! #$ $ ! ' # ! !<br />
! <br />
<br />
H
− &*$ $ !,9 7 $! !$! ! <br />
$#! !#! ! !:$# $# $<br />
$ # ## $ ; 4 ! $$ $<br />
## ## !<br />
<br />
,-./.<br />
<br />
! #$ $ ! *$ $ ! # ## <br />
##$ @#! !<br />
! # ! #$ $! $#$ # <br />
! *$ $ $ ,$ <br />
& J#$ , $$ # <br />
#! 8#$ $ ## <br />
$ !## D <br />
!! $ ! ! $ $ $<br />
*! # $ ! # ! # <br />
## # (& $ $ $$ $# $$ # <br />
# ! $ $ $# $# $ <br />
: $ *; ,$ $ $ $ $ # #<br />
*! I ! # $ 3 ! $ $ <br />
$ $! $# 6$$! ## $!#! <br />
$ ! !!! ! <br />
! $(! * $$ $! ! <br />
#$ *! $ # * $ !!<br />
$ ! ! $<br />
<br />
0&12<br />
<br />
! $$ $ #$ 6 ,(= !! $ % $ <br />
($= '!:I$ K&3G=$! #<br />
3$! ( # '!;"6 ,'!!$ % $<br />
'## # ! :$ $ $ "<br />
! $ $ $ ! # ! !$ # ! <br />
; I ! # # #$ ! ##<br />
! ! !! $$ @ ! ! ## ! <br />
6&9(496 ,'<br />
<br />
!$!$!! 3*5<br />
$ #+! # #$<br />
# ! !$ $ $ !<br />
(&! #$ !! $<br />
# 9 $# $ 3$$$#<br />
$ !, ! $ ! !, # ! <br />
$ $# 3# $ $!! !$7<br />
− ( #!$ $$! B#!<br />
−<br />
!# <br />
I7 ! $ * :* ! ; !$<br />
$ ## $<br />
I! + ! # ! #!#$ ! <br />
#$ * # $! $ ! <br />
(,#$! <br />
6'$ # ! (&$##!! (&<br />
$ # ! $ $ # ## 4 ! <br />
+ # ! # ! (&! & <br />
.
Ma<strong>in</strong> Shipyard<br />
= XML Interface<br />
Shipowner<br />
INTERMAR SY<br />
Plann<strong>in</strong>g DB<br />
INTERMAR SY<br />
Plann<strong>in</strong>g<br />
Application<br />
INTERMAR<br />
Collaboration<br />
Application<br />
Receive Enquiry<br />
Request Details<br />
Req. Quot.<br />
Submit Enquiry<br />
Receive Request<br />
INTERMAR S/O<br />
Plann<strong>in</strong>g<br />
Application<br />
INTERMAR S/O<br />
Collaboration DB<br />
Legacy<br />
Applications<br />
Legacy DB<br />
Legacy<br />
Applications<br />
Legacy DB<br />
INTERMAR<br />
Collaboration<br />
DataBase<br />
Subcontracted Shipyard<br />
Receive Quot.<br />
Receive Request<br />
Submit Quotation<br />
I n t e r n e t<br />
Tribon.com<br />
Receive RFQ<br />
Submit Quotations<br />
INTERMAR SU<br />
Plann<strong>in</strong>g<br />
Application<br />
Supplier<br />
Legacy<br />
Applications<br />
Legacy<br />
Applications<br />
INTERMAR S/S<br />
Plann<strong>in</strong>g<br />
Application<br />
Tribon.com<br />
Application<br />
INTERMAR SU<br />
Collaboration DB<br />
Legacy DB<br />
Legacy DB<br />
INTERMAR S/S<br />
Plann<strong>in</strong>g DB<br />
Tribon.com<br />
Global DB<br />
<br />
@7'<br />
<br />
6 ,(=$ !! &# $$ $ ! # # #! <br />
$ ! 6 # ! 6 ,(= $ ,# # $ <br />
# # = ## $ $$ # ! <br />
$ ) ,,) ! 6 ,(=##!<br />
#! 6 6 ,(=!! ##$7<br />
− =+7 ! M=+N #$ $ # ! $ $ ! <br />
&3K=!KI$ K K'<br />
− =7! M=N#$!$! $<br />
− 7! MN$ ! ! <br />
! #$<br />
<br />
3<br />
<br />
6&9(49!# # ! !,,! , # ! !<br />
$ $ # ! !$ <br />
$ !, ! M8 &N ! <br />
! #!$!# # #<br />
$ $ $# $ # $ $<br />
#3$<br />
<br />
'1<br />
<br />
D9&% 9O:110;3&4D !'<br />
<br />
(&P
! "#$$<br />
#%&' <br />
<br />
<br />
!<br />
<br />
! ! ! " " # # $<br />
%&"!#!!!" !<br />
'# !(!!<br />
' !!!"!&) #*!<br />
'#!! #""!!#'#'"!+,*!<br />
"" " " ! ## # # ! +-! "<br />
!.'!! "'# !!!##*!<br />
'#""!'#/""!$/.'#'"<br />
.'""'#!<br />
<br />
<br />
+,0*""!()*++<br />
<br />
1 ' # # ! / " ! 2 !<br />
'# ! " ! # ## ' #& "<br />
.' ! / 3 # ! '#<br />
$# 4$5 ! " !'# " / %# *! $ ! #'<br />
/!# !6$#!46$57 # '"%"'<br />
#"%# !%"'*! "!'#<br />
/%###'%"%#
,<br />
2"#<br />
$#<br />
!5:<br />
,-./8149<br />
,<br />
,<br />
A>'<br />
@'#'%<br />
><br />
"##<br />
<br />
>!<br />
* <br />
='#<br />
$##<br />
=#!>?<br />
<br />
<br />
$>?<br />
(!'#<br />
><br />
3 <br />
3 $*23 $*2<br />
B6% C<br />
C<br />
##<br />
,<br />
<br />
)<br />
<<br />
;<br />
,<br />
,<br />
,)<br />
/ 8 !:<br />
<br />
<br />
+0#!% "",4./5!/!0)***<br />
<br />
<br />
%&#'<br />
<br />
*%#7+D)%!#!+ #'<br />
&#"!!!!'#"!<br />
# *! '" % # " & #' 7 "" ! '"<br />
*!&'#! #!!!' '<br />
!!"!!<br />
<br />
*%#70""!<br />
!'# !'# <br />
E! #E ,
=<br />
{ ( ) ( )} <br />
<br />
+ <br />
<br />
!2 G/ 45<br />
/!<br />
<br />
, 4θ<br />
5<br />
4 5<br />
6 ∂η<br />
<br />
<br />
<br />
= −,<br />
{ 2<br />
} 2<br />
<br />
6 <br />
θ 4 θ<br />
+ θ<br />
5 &4 θ 5<br />
− −<br />
54θ<br />
5<br />
∂ <br />
4D5<br />
4 5<br />
<br />
{ }<br />
, 45 G,"!!'#"!3!%<br />
, = 2<br />
θ θ<br />
53<br />
<br />
%!#"!'#"+)-.4D5' #%<br />
&!!'#!%"""*!"#1 #'#%<br />
ρ <br />
1 = 4,<br />
+ 251<br />
= 4, + 25<br />
/ 4<br />
4)5<br />
<br />
2 ! " " 4 ! "# "" " # 7**>IF 1 ! "#<br />
"# !'# ! '""!!'#*!"'##'24,H25G,,<br />
<br />
<br />
+0-&#"&"#<br />
<br />
* <br />
<br />
*!!'#"/#'#%#'η !" !"!#!'#<br />
"+
*#1 %"'/0<br />
(/ 71 71 81 45<br />
#'!#"#'η "!%#η !! +<br />
'# ! ' ! # / ! # !'# 4 # # ! <br />
"5!! #%#!3 "'#!" # 0<br />
<br />
4$5>"#<br />
7 # %# # ! # &# ! ' " !<br />
/ %' ! !'# " %# *!" <br />
#" # 0<br />
JJ.<br />
−, ≤ . ≤,.<br />
−,<br />
4
*!%# #'#"##$!%"'4 5<br />
##:"#"!!'#<br />
# # ' # '# ! #" # ' *! #<br />
#!'#!*!#!"!#%<br />
%" # $!'#'!# # !!#.'#%'<br />
!/%%%#"% !:%!=#/%'0<br />
<br />
, :<br />
{ }<br />
<br />
, : = : = ⋅ &<br />
−<br />
4;5<br />
; 4<br />
5 0<br />
<br />
< <br />
;45#/"!'0 < !=#/*!&#<br />
" # =#/ " $ ! ' ! =#/ %'<br />
! !!# <br />
<br />
!7##'4 5<br />
7#'4 :( 5<br />
>'#'4 5<br />
(" <br />
#'<br />
?#"#K<br />
∆<br />
: =<br />
<br />
( <br />
) −<br />
<br />
4<br />
5<br />
'<br />
'<br />
∆: <<br />
<br />
K<br />
6<br />
!%%#<br />
%=#/"<br />
$<br />
'#'45<br />
&4<br />
− ∆<br />
:<br />
<br />
5<br />
><br />
γ<br />
K<br />
(&#K<br />
(&K<br />
A<br />
<br />
<br />
+J0$#!"'##!<br />
<br />
+J #'!#!"!$!"!!'#'#<br />
#' #'%'%!'#'#<br />
%%#%'"'4=#/%'5$ #'<br />
'#"'%"#!*!'<br />
)
456##' # !#' :( <br />
4,56'#' #'"#'###'" <br />
45>#'#"!!'#'<br />
∆: = − ∆: < <br />
7"∆:=$! #' !%%#",4∆:#<br />
4J5<br />
γ'"'%%'!#,<br />
4D5'!'' !!.'#%'<br />
4)5*!' !!"<br />
<br />
- )<br />
<br />
<br />
( ) ( )<br />
− ∆:<br />
<br />
, &<br />
<br />
<br />
(:) =<br />
> γ<br />
<br />
+,07#!'#""!'#4#"54!5<br />
<br />
<br />
+,!#!'#""!!'#!+,,! !'#"<br />
! !'# " / " # ! ! " # " !<br />
!'#+,D!'#"!7!#"!'%"<br />
#!1#! !"'*! # !!<br />
'%"#" #!#!2 !#'! <br />
%#''#"!!'#7!!!'#!"%#!<br />
"! +,,)$#!'!!#!"! !'!<br />
! # % %' ,L ! # !'# " *! !'#<br />
!!"" ''#<br />
<br />
+, ,< ! ! #! % ! / 1 " %! ! <br />
+,! !!"%!"!#!/!'#""!<br />
!'# -! # ! # #' ! " ! <br />
#!'! ! / #'# &' / # ' )<br />
*! #" # "'! # ! ! " ! +,<<br />
! %!" ! # ! / !'# " " ! 7<br />
! !"!!'#!'!! "!#<br />
!'#"#!#"#''!"!!') %'"!%#<br />
"#'#
+,,/ ! # +,/ !"<br />
4!'#5<br />
#!!4!'#5<br />
<br />
*%#77! !"" , : G1 ⋅/!" , "!'#"!<br />
#1 "!#/!'#"%!!'#!")<br />
, !'#"""
+,D0/ !#<br />
45<br />
<br />
+,)0/ !"<br />
#!!45<br />
<br />
<br />
+,0>!'4!'#5 +,!'45<br />
<br />
)<br />
<br />
$B$674,JJ,5<br />
+$*IJ,*!FFJ<br />
<br />
E$$2?-@1(M$$*-2E4,J;F5 > <br />
B#' $1'%#!<br />
<br />
E-3 7-?4-54,J;;5, ? 4 665@$(-<br />
<br />
F
! " ""#<br />
$%& #'%" #! " ""#<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! <br />
" <br />
#<br />
<br />
$ <br />
% <br />
<br />
&'(<br />
<br />
# <br />
<br />
<br />
<br />
<br />
<br />
<br />
( ) * # # # ' #* )<br />
# + # ) ) *)* )" * # # <br />
+#*,'*#)#+"( ))##<br />
)*#*)**"<br />
<br />
- +# ' *,'* # # # ..<br />
# ' *)* ) # ') #" /* # #<br />
0#'*'**,'*# ..*))##<br />
'#)"$*)*#*<br />
)*# #).,"<br />
( ##'*"(* #<br />
*##0#* *.*#* "1<br />
##,)*,# ..*<br />
"( ..*#'*)*)#)#<br />
*,'*'## )* "<br />
<br />
( '* ## ..*#*,'*#)<br />
'*"# ) )##*.'2<br />
*3"* #'*), ) )<br />
#*' * #"(*'#<br />
. *"4* *'#)"<br />
<br />
<br />
<br />
*' *' # '.#) 5"<br />
5 * #"5.
*'# #)<br />
**% &)*++,-"% <br />
##*.)%#. )*++,-"5<br />
)*% ##7 )"89,*.*"<br />
<br />
)"8:-'*)% <br />
<br />
/) ## " 4 <br />
*. '*# %* # ##" ( <br />
% & )*++,-" % # ' # ' <br />
" ( # * #%# . ##*<br />
)##* '#"(**%<br />
# #) . * # . ## # <br />
#)."<br />
<br />
<br />
<br />
(*% )<br />
# * " )" ' *#<br />
/ )*++,-"<br />
<br />
5#. +##)** <br />
''#)# )*++0-"/*#').*#)<br />
% "*#*#<br />
*" * . # )) # +)<br />
%')*#..#"(* <br />
* . )# ,*" ( ' # # '*<br />
%" $ * # <br />
*'). #%" )<br />
#) ,)*'#* #<br />
"<br />
<br />
<br />
<br />
6
)":= *% <br />
<br />
*** #'*))<br />
#) " ( ) <br />
')* *##0#"(<br />
##*#'# '#.*#'# <br />
* #" ( % ## ' <br />
"<br />
<br />
( ## % + % " <br />
'. #"(* *.*"4<br />
.*.* '##)<br />
)*) #.*'*'*)#" %<br />
* * .* * ' #. % <br />
#)# *''/<br />
)*++1-"<br />
<br />
( ) # "" % ' ) #) ' . <br />
)' )##))#')*++*-"<br />
( .* #) * ) .* " ( .<br />
)#'*#)<br />
#'*" > % ) * * <br />
) .* )'*" 4 .* # % ) <br />
*) # "4* *####'*,#"<br />
%) )* ))*'*)'<br />
* # * #'# ) 7* %*9 <br />
) # # ,) 23 4 )*++1-" ' <br />
) % ) ) " <br />
<br />
;
% ) # )"( +#*)<br />
+ * # #' # ' # # "<br />
( ) *) # '% ) #<br />
) " ( # ' '# #.) <br />
*)# ##234)*++1-"<br />
<br />
%),#*)*#%<br />
*)#*)*#)#**<br />
*'#)"<br />
<br />
!"#<br />
<br />
#%)* )*#<br />
')*++*9<br />
89 . : ( ## ' % )<br />
#<br />
9 : )#)*#)<br />
) . # *) # .* * <br />
.<br />
?9 :*<br />
#) <br />
<br />
(')##. #'*'<br />
# #*" *# ** )# #<br />
' # .* # . ) <br />
" / ** .* # *) .<br />
# #'* . " 4 )# # * # <br />
.#)"( ###) )"?"<br />
<br />
)"?:@##)<br />
<br />
<br />
1 # )'* " *)'<br />
+ # ," /*'* <br />
### ) ##*"-,#'#<br />
##*.# *)#<br />
###'"<br />
<br />
( . ) '* # *,'* '* ## <br />
* *##*7.*'*9'*<br />
) # # # * <br />
#))) #.#*&)*++5-"<br />
;8
( % ) )# # ) <br />
* # ' '* * #* # *, * ) )* #<br />
* # '*# * ##" * # #<br />
')#*"#.)/% '.**#<br />
'* ,) / 1 ' )*++*-" ( <br />
% /. # //. ' #" % * <br />
*)#*) * #* % *#)#**)<br />
*###'#))#" '# <br />
//. .* # #* ) %*# #<br />
"6"7##4)*++*-"<br />
<br />
$%#&'&&'<br />
<br />
/%/.
'* # ) ) # # . '"<br />
)* #'*# .*"( '**#<br />
##.'*)## "@<br />
*'*.)##* *##"*)#<br />
'+### .# #*#<br />
*#,"<br />
<br />
@.)..'* )##<br />
* %" 4 * * .# . ) # ." D <br />
* .#*..#))2#*3&)*++5-"<br />
<br />
)#/* *#)<br />
#)"$#'# '#) <br />
'#*'"4 # <br />
'# ')"(## .*'*#* <br />
' *# '*" '* # * .) <br />
# '#,**"(*#. .*'# #)<br />
#"<br />
<br />
* /%#/' )""(<br />
*.*"(*.* **.*7**.*9"('*.*<br />
#'#)"$ #'*. <br />
* ' <br />
*"<br />
)":/* /%#/<br />
<br />
1# %/.' '"1# <br />
//. #.#*)'*%*#/"$<br />
*# . *.* *" ' ' )#<br />
*"4'#.' "4/ **#'<br />
/ ' *"#' '.<br />
."( '%.#/.' *#"(%<br />
/. +*'* *# //."/' *<br />
' #' )" ( ' * ## " ( '<br />
#.#*##' #.#*). #.#*"4*<br />
#*' *# %)('#D #"(<br />
<br />
<br />
;?
.'*#.*### #"# )##<br />
#* * )):%)*). <br />
'" ) # # % # ' . # <br />
#'#"%)#' #.#*<br />
)# * # " % ) * %"<br />
$* % *# . * % * # #<br />
#)*)# *# "<br />
<br />
( '*):<br />
89 4* '# " ( * # . ' ## ' '<br />
'#.*'* 3 %"( <br />
) '#'5# "(.*'* <br />
%#'*#"(.*'#*#<br />
. ### )# %"('<br />
#)#%"7 )";9"('<br />
*% * #)"<br />
9 = ) ##%"4#.*,<br />
#% '**.*%.%7 )"E9"<br />
4*% '## )#' #<br />
%"(##)%)<br />
%#.."/*)*.* <br />
#"$.' *))"<br />
<br />
)";:/*8)<br />
/%<br />
<br />
<br />
)"E:/*)<br />
/%<br />
<br />
<br />
;A
# %.*# #)' <br />
) ) # ## * # #* # % #<br />
*'" ( ) ' ). # * #.#* ' <br />
' *#*'*.*.)*++5-"<br />
<br />
( %#'&%'<br />
<br />
( % . # / . ''# % ) * <br />
"(*.'#.*#'') <br />
* )# * ##.*'*"<br />
<br />
( ) # '* % /. # / . %<br />
) ' )# * " ( % ) <br />
'### 5*)"( <br />
# ' 5" * . :<br />
F" -. ')* ## # )# "<br />
% # # * # ##) " %<br />
).*%## *%"(*<br />
# # ) # '* # *' # ## ## <br />
'"1. '*. " *<br />
##" * #)# *"(#*#<br />
'####''"//.'<br />
+# #) ) #*" 1 //. ' .<br />
. # # *' *# ' % ) # * <br />
#%."<br />
<br />
(%).*)# .*#<br />
'*"() **#' )*<br />
##*%" %)%)<br />
) #'*%.#)#"<br />
<br />
)":@# * /%*5<br />
<br />
%) * '## <br />
5"4)'*# %)7 )"9"<br />
<br />
5# ' # %"(#<br />
%###* *#'<br />
* % ' ## ' % " ( 5 # <br />
#'*",5# # %<br />
#) %)"@)#%)<br />
)# * # % *# #.#* .) ' <br />
<br />
<br />
;
"5 5 %)"4<br />
%). * ""#"(#<br />
) "#)))#) <br />
%" ( % ) # * # #" ( ' <br />
*. # # # '*)) " 4 %<br />
) #* % . "" # , ) ,# <br />
## # ,# # <br />
%#*' '*#"%')#'<br />
% )" - ' . . ' '<br />
#,*#)#"()#).<br />
')# '*. #*<br />
%)"(#*).##' #*'<br />
/."# .#. . *#"(<br />
*#. *.*'*#'# #)"$<br />
)##'* %)#*.#<br />
#.* % ) # ' " <br />
#+#)%)''*)*<br />
))"-.')%# <br />
5"4 +##) .#"<br />
<br />
/*.'#.*# %)"()# <br />
%)'.#* '%<br />
*%"(##*.',#####"(#'* <br />
%)# 5 )"6"<br />
<br />
)"6:*#*/%#5<br />
<br />
)&#*+<br />
<br />
*%) ** "(% # 5<br />
%)#)"* .*.# <br />
% " ( % ) # * # #0# <br />
'*) #.* ) #" / <br />
))###'*%*#.*#### <br />
#"4))' %) #<br />
*%"*%%) %"%<br />
;;
# .% )#"(%<br />
%)#.*%'"(%' %<br />
)% %7 )"8
)"8#8? #'))(-F8,*"(-F8 <br />
)#)%)%))F*F F*F<br />
7)9" ( * ' ) '" ( ) <br />
*# ' # ) #.#* ##) <br />
+"<br />
<br />
)"8:@ '))* *<br />
<br />
<br />
<br />
)"8?:@ '))* *<br />
<br />
'*%#*'# #, # "5,<br />
*#)#')"(# )#" )"8A <br />
,#*' #.*#%"4* <br />
##)"<br />
<br />
.. .*'*'. #"4<br />
' ,**(-F8 )"8"(* )<br />
#* .*'*" *,#*.*# )<br />
*" ( .*# * %" 4 <br />
* % # ) " ( '<br />
*# %" * . ' # )<br />
)))(-FB##).*#"<br />
;
)"8A:@*'* *<br />
)"8:'*)* *<br />
<br />
#'#'*%) #<br />
)#"-'%)* <br />
## # " / )) * * # <br />
) # # # ' #<br />
#*"(@#*'#.*# .# #*# #*#<br />
" # *# # ,"#<br />
# '* # ) # ) ) # ' # # *<br />
* # # # % # # '* ')" (<br />
### #'# #.#*%##<br />
'*"(%)*##%"<br />
<br />
*%###*) #)#<br />
#*# % .# . " ( ) ' # * %"<br />
% ' *# . # ' # * #"(<br />
#'#' *# # *')#*#."<br />
<br />
,&<br />
<br />
())###+ ))"<br />
# # '* # *# " <br />
% # )# %*,<br />
###.**'*"5 ,) #.*#<br />
;6
"<br />
<br />
(% #.*# 5.) ***<br />
% " ( +) % "<br />
@) # ,# # # ##" ( <br />
).'#),##<br />
)#)#"*+)<br />
'*").''*'#. <br />
' * # # * # . ' '* '*#<br />
## "<br />
<br />
4'# * )# 5 ** %<br />
"%.#/..*'#.*#5<br />
# # * % , '<br />
#)#"4'* .#%)# )#'<br />
*#) %#*.#'#<br />
.'#"(#.) .)#)<br />
*'* # '* , % )" * * <br />
/1/%#4-,*" <br />
)'#* . *%)<br />
"4*#% *#<br />
)#%) ' 5"<br />
<br />
( %)'%.)<br />
* ') #.*#" *# % *# # ' <br />
*)%'.*#"$#<br />
%)')*#"/ ##*#'#<br />
+ # *## 5" <br />
5#*# ##% #') <br />
% #"<br />
<br />
)* # *##.*#"5<br />
#.* #) * *" ) <br />
) " ) )* * <br />
* ##.* "<br />
<br />
-<br />
<br />
-5-I"J-5@4-II">"JI=->5$"J 55-""J -4@-55$"7
D1L/"JC(-5/> -I"7
! !"! !#$<br />
<br />
%&'(")'*+, <br />
!" <br />
<br />
"<br />
<br />
<br />
<br />
<br />
<br />
! ! "#$%&' !(' %<br />
% <br />
'( <br />
&!'% ! <br />
! ! "#$ <br />
&! ) * + <br />
&!!, <br />
- ! "#$ <br />
) <br />
<br />
'(! "#$ <br />
&!<br />
<br />
-!<br />
<br />
# $$" %"""" $ " "$ <br />
" "# $ $" % %<br />
& $ " % " " <br />
% " % " % " " <br />
'$ & ( ! " !% "<br />
" )"% $ % " * $ " $ <br />
" +%$ "($ <br />
( " $ * '" $ $ <br />
$ !!%% ("" " ($ <br />
" !! % ) " " $ % $ <br />
," " % -,.#/ " " $ <br />
""<br />
<br />
'$ ( " $ " $$ " " " " <br />
" 0)10). $ $"% * . ($% <br />
" "% ",""<br />
" $ " ($% $ " ""% %+23""-2'1,<br />
/ $$ " % " % 1 $%$ $ <br />
$ """" *$ <br />
% "$ ($ <br />
" "% $ "" " + (! ."//01 % $<br />
$ $ "" )"% $ *<br />
" " " $ " " % "<br />
" $ $"% 2'1,4#/2/2%"#0."//$1<br />
#/2/2%"#3 ."//21 #/2/2%"#4 ."//$1 '$ $ '$ <br />
$ "$ $ <br />
% $ $ " "" )"" $ $
" "%$$ "" ($% !$ <br />
" " 6$ % "" " " <br />
" $$ $ ( '$ <br />
" $" %$ ""-/$$ " <br />
$ $'$ $"$ 2$0#" -20#/ $ <br />
23'0789209: 5'5$""%5 !(."//01$ <br />
" (!."//01; " %$ 0).!%$ <br />
" "" * $% " $ " $ <br />
"" " $$ " $ '$ $ <br />
""0)( ( "'"$ <br />
" $ $ & " " $ """<br />
$"% <br />
<br />
Owner<br />
Operator<br />
Shipyard<br />
Shipyard<br />
Ing.-Büro<br />
Ing.-Büro<br />
Design Agent<br />
Model Bas<strong>in</strong><br />
Supplier<br />
Classification<br />
Society<br />
<br />
6%742$" %"" <br />
<br />
, $ "2'1,"""$ 794
$ "" $ +23 " % 5 !( ."//01 2 <br />
% " " C. * $ $ $ $ <br />
%$ ),7B<br />
<br />
: $ %$ " " ""06." <br />
$ % )$ 06.!" & <br />
$%$- " 5.$ /""$<br />
$ " '$ $ % " <br />
$ $ " %" ("-" <br />
" " "" /)$ $$ $ % ("<br />
+12 .?6 " $ " $ $ % $ <br />
06.!" 6."//$1+$" "$+12.?6%<br />
" % " ($% $ """ <br />
" "$ $$<br />
"$ "<br />
<br />
'$ $ " ($% " $ " % " % <br />
- "/ " " % " " %<br />
. " $ ),78<br />
$ " )" "$ <br />
" "" %$ $ "$ "<br />
" '$ " % ($% "<br />
$ $& % ( $ " , <br />
($$"$ ""),78D " $ "$ <br />
"$" $" %*<br />
<br />
2$$ " "" ("<br />
% $"D $ " "" ($% "$ <br />
" %"+12.?6" ,"""<br />
"""% "$$ $ % "<br />
" " '$ $" $ & "<br />
%" %$$% $0).-./<br />
" )
" " $ +23 " 2( " " <br />
),E5007" %"" %$$ 0000B" <br />
007""" % $ . <br />
$ ),E5 " "),79 <br />
<br />
'$#/2/2%"/2.#77$1 $ %" $ "<br />
& $ ($% %!"<br />
5." % $"0%$(" ""<br />
$ $ 5. " " % " 1($% " % " <br />
$ $"$$ ($% "$ 5."" <br />
<br />
'$ %* ),E54<br />
," - $ " / , " "% % $% <br />
% <br />
. % $ 4: "$<br />
% : $ % #" $<br />
%6""H"<br />
) -" $ %"%$ "<br />
/<br />
3%"- /<br />
)"-""% !" <br />
/<br />
;"- "/<br />
. %"$% <br />
. %- /<br />
)!""" %$ " %$% <br />
<br />
- <br />
<br />
2 EE7 $ ), #/2/2%"#$ ."//21 "" '$ $ <br />
" ""EE5'$%" "$ " <br />
"" & "" %$$ " " " <br />
" $ $ % %"),79 %" " "<br />
" %" "%$ ,.# +<br />
$ ),G $%$ $ " $ '$ % $ <br />
" $ ($% % " % $ $ !% " '$ $<br />
$ % 0). " ,.# 6 $ * % " <br />
%6%<br />
<br />
'$ $ % $$ $ """ 4<br />
, " " "% % " <br />
<br />
,$ % $ "$$<br />
$ " % % "<br />
," " % " $$ " $ " " <br />
<br />
#% "" $ " %!"$% <br />
) -" $ %"%$ <br />
$ ""/<br />
+" " <br />
3%"- /<br />
8 % " 4.! 5.! <br />
% " " % " " <br />
" " " " $ <br />
"% <br />
<br />
A
D"% "<br />
.$ $ - "$$"%/<br />
66 <br />
,"" $%$ (" "%<br />
"" "$ $$ "+23<br />
7E5E5<br />
2"$ " <br />
, <br />
2"<br />
'<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
6%4 $ 6#/2/2%"#$."//21<br />
6%542<br />
<br />
'+40),79<br />
<br />
<br />
007I00A<br />
00BI007E<br />
0077I0075<br />
0079I007A<br />
007BI007<br />
0078I007F<br />
00E<br />
..<br />
0).J " " %J 5. " " $ <br />
%"<br />
,.#J " " % % $$ 5. $ <br />
<br />
% ""<br />
"" %( <br />
" ""<br />
%"%5.$ J <br />
" "<br />
"% "<br />
<br />
B
+),79 " " "$00E"%$ $<br />
)+#'++ 0). ( %007"006,.#<br />
00B"008"<br />
<br />
-(" ."!!<br />
<br />
--<br />
<br />
'$ %" "$ $"<br />
),79 " " -)+#/ 6%5 '$ $ <br />
-"%% /" "%$"% 0).-2!0)./ % "<br />
$"% ,.# -2!,.#/ + " $ " " < =<br />
" $ $ % % 0).! -#!0)./<br />
%),79'$ $ " "),79!)+#"<br />
" -2'1,!, 7/ -),79/ + $ $ *4<br />
2$ ) 2 , " , " " ' $ $ <br />
"% " ),79 " 4 " " <br />
""" % '$ *""" $<br />
$ $ -% "/ % $ $ !% ($% " 6$ <br />
" 4 < " " , ),79< 5 .#7741<br />
589:.#7741<br />
<br />
product_category<br />
name<br />
label<br />
*product<br />
products S[1:?]<br />
*product_related_product_category<br />
application_context_element<br />
name<br />
frame_of_reference S[1:?]<br />
label<br />
product_context<br />
*of_product<br />
*product_def<strong>in</strong>ition_formation<br />
formation<br />
frame_of_reference<br />
*product_def<strong>in</strong>ition<br />
*product_def<strong>in</strong>ition_context<br />
*application_context<br />
frame_of_reference<br />
(INV)context_elements S[1:?]<br />
application<br />
application_protocol_def<strong>in</strong>ition<br />
<br />
6%94,"+" ),79<br />
<br />
--!!<br />
<br />
+ ),79 " " " $ $ $ *<br />
-1/) $$ " "$* %<br />
; ; 6%9$$$ $ "" " "%$ <br />
1?,122'$ *; <br />
$ % K< K< K$< ) % % $ <br />
),79 " " %$ " '$ <br />
" " $ ; $$ ; %<br />
;'$ ; "$
" " -% / ) " (<br />
$$ $ ; ; <br />
<br />
--!<br />
<br />
6%A $ $ " $ ),79 " " " $ <br />
"6$$ ; ; "$$" <br />
": $$" % $ %$I<br />
$ (" + " " $ - I<br />
$/ ; ; ;;; <br />
"+$ $ ;; * K<<br />
$ $ ;; * K<<br />
<br />
<br />
*product_def<strong>in</strong>ition<br />
characterized_product_def<strong>in</strong>ition<br />
<br />
<br />
def<strong>in</strong>ition<br />
*property_def<strong>in</strong>ition<br />
characterized_def<strong>in</strong>ition<br />
<br />
<br />
represented_def<strong>in</strong>ition<br />
<br />
*product_def<strong>in</strong>ition_relationship<br />
related_product_def<strong>in</strong>ition<br />
relat<strong>in</strong>g_product_def<strong>in</strong>ition<br />
*product_def<strong>in</strong>ition<br />
*property_def<strong>in</strong>ition_representation<br />
def<strong>in</strong>ition<br />
product_def<strong>in</strong>ition_usage<br />
*assembly_component_usage<br />
used_representation<br />
*representation<br />
context_of_items<br />
(INV)representations_<strong>in</strong>_context S[1:?]<br />
*representation_context<br />
items S[1:?]<br />
*representation_item<br />
*next_assembly_usage_occurrence<br />
<br />
<br />
6%A4,"),79<br />
6%B4,", ),79)+#<br />
<br />
--!.<br />
<br />
+6%B$ " ),79%" %" " "<br />
+$$; ;$$ $ $ $%$<br />
$" " '$ ;<br />
" <br />
<br />
," " " $ % " % " $ <br />
" " $ "" 6 " % $ $ $ ( <br />
;";;" % "<br />
;<br />
<br />
6 " % " $ % $ $ $ 0).!<br />
%),79"$ "%% " $ "<br />
6$ $ $ $$ " "" <br />
" "$ -5. " <br />
/ " $ $ $ : $ $ $ % <br />
$ % " ( <br />
<br />
+ %$ $ "$ " <br />
""" +$$ <br />
;; D - $ /<br />
8
; - $ / " ; - $ / <br />
"+6%F$
: $$ 6%7E$ $ $$ 4$ "<br />
%$ "$ #0). %),79'$ <br />
$ $ %",1@+@110)'+)%$"+;<br />
$$ % % " % "<br />
+ $" " $ " " % $ $ $ <br />
-% " /$ " "$ %%<br />
%%"<br />
<br />
<br />
<br />
6%84D " "$P@),)2 >""$+.)!2; <br />
<br />
1-<br />
<br />
'$ ( $ $ $ $ ),79 " " (<br />
"$"% "+$ $ $ $ <br />
( $ % " $ $ " <br />
, $ " " $ % <br />
" ""%"" <br />
<br />
<br />
<br />
6%F42$$ " "),79"" <br />
8E
0$ $"%"$" " " % "%<br />
% $%*"" $ $%$$ <br />
< = <br />
<br />
'$ " $ <br />
#0).",.#'$ "$ )+#" $$ <br />
$ 2'1, % $ 9E $ "" ) $ " " <br />
" $ ( -%<br />
& /$ $ $" " % "$ <br />
) $ $"% "<br />
-" QR/$ % ),79$ ($" %""<br />
$" $ %"<br />
<br />
,&!<br />
<br />
'$ "$ "$ 6 " #1"2<br />
$"' $%" %E52?757'$ $"% $ "<br />
$ $ $
(<br />
<br />
H3@2)' J S@+1M+@ J T+##1#)@@ # -EE9/ ! %<br />
F $ +2,. %<br />
2$"$ 6%2-,).2/MU <br />
<br />
1#2) -EEA/ : $ 1 ! "" <br />
$4 % <br />
<br />
1@.12 #J D)21 # -7FF8/ " " , ),79 007 00J 6 .J<br />
."<br />
<br />
DVMT1M 3J M1H1 # -7FF8/ = < % ! "#$ &&3<br />
'."<br />
<br />
+.)2'1,-EEA/$4"! <br />
<br />
+23 7E5E5!E5 -7FF9/ +" " % ! ," " <br />
" ($% !,E54)40%"5." % $<br />
"<br />
<br />
+23 7E5E5!79 -EE5/ +" " % ! ," " <br />
" ($% !,794)40" $" %<br />
<br />
+23 7E5E5!7A -EE9/ +" " % ! ," " <br />
" ($% !,7A4)42$% <br />
<br />
+23 7E5E5!7B -EE5/ +" " % ! ," " <br />
" ($% !,7B4)42$" "<br />
<br />
+23 7E5E5!78 -EE9/ +" " % ! ," " <br />
" ($% !,784)42$<br />
<br />
+237E5E5!9-EE5/+""%!,"""<br />
($% !,94+%"% 4 "%<br />
<br />
+237E5E5!95-EEE/+""%!,"""<br />
($% !,954+%"% 4 <br />
<br />
+232D+,'1)#-EEA/: $ %%" %$"<br />
$4$"%"" 5$<br />
<br />
2D+@032-EEA/: $ C.*2$+"0+<br />
-2$02/$4$" <br />
<br />
2')1HM1J#+MM1HJ)S3: ,JS30D'-EE5/& &!'<br />
&!(: "# ' $%026<br />
8
!" !#$<br />
%&'() !" !#$<br />
!"#$ *+ ,!-!#" !#$<br />
%& (.%" !#$<br />
<br />
'<br />
<br />
<br />
<br />
<br />
<br />
!"<br />
#<br />
<br />
<br />
()*<br />
<br />
/!0! ! .-$0 $# !0 -! . !1<br />
$!0 0 !0 ! ! #. !#! .. ## /2 !<br />
-!!0!!#!-0#! #.!#!$!$!0<br />
$.! 3402.!0/.!-5!<br />
-!#2! <br />
<br />
3.#.!-//-$..-$0/2.- .!#.!!#!0#<br />
!#. 00 $ ! !#. -! #!0 6,0 3-7 $!<br />
! . .-$0 0 / .! 00 ! ! /2.- ! .-!0 !<br />
/2.-6,.!08!47 &99!!1!5!!# .!##-0<br />
#0! ! !0 ! -!! #! !$ . /2.- .<br />
-:#!!!#!0#-:#$#!#!$!.!$0$/.0<br />
!0 . /2.- #0! !. 0 -0 . .! $ 00 !<br />
/!!0!.0#!! ./2.-00!/0<br />
/2.-;.-/.!!00!-#!$1!$!0.-/.!<br />
!!$ .!!5.!
+)#<br />
<br />
+)()*<br />
<br />
.!#.!!##./2.- @;00#!..0 <br />
=!#.0.!/!0/!!-#!0
. /0 $ . /.#. .!0 $/ . / 0 . /2.- 3 !<br />
#-!!#/.#.0/0-#.!.!$!#20$.!!<br />
$.#.!50-- (!##!!./0#!#!0$$!-#!<br />
/!/.#.2/#!!0!./0$!0##.!!##<br />
-# /0 %. / . 0!! ! !!!$ / ! -0 .0 <br />
!!.!/0 .$!$/!!/0!0.
. 4# . -# ; (-!! @ F !#2 F E0 F . B F<br />
=!#!G E2.!!!/2!#.0.#!#20/0<br />
$ . . . !! . /0 $ / .! ! . /<br />
#.;<br />
<br />
.00-00!0!/#!.$/
..##-!#!$0!0!-0#0/#!.<br />
B.!.!$#!!.!1#4#$.0#!-!<br />
!#0-0.4##.!.!!0#@>?;<br />
<br />
C!1E2.-4#<br />
C!1,4#<br />
<br />
B;,-!#!$0##-!0-4#<br />
<br />
'!#.! G#!!$$!0 %!
-)<br />
<br />
-)()*<br />
<br />
!0$.!0.$/4#/.#...-0# 3<br />
##! #! 0 /.! / $ $:# # .! / / #! !0<br />
! /. ! - /2.- . ! ! . 2 0 /. #0<br />
!# .#00.#0!..0/$!!0<br />
-#!0!0/.!-/2.- .-$/..!/!1-<br />
-0.!!#..!1#/!/2!0#0!#!!0 <br />
<br />
.0.!2..0!!!0F/.#.0.<br />
/ / . . !$#! . #0 ! 000 $ .<br />
$#!/.!.# E/..!$0!0./$.<br />
./ <br />
<br />
3/0!/..!4#/$!!!!.!$<br />
@G?/.#.//.!#0#!L>?AAA .0/!!;<br />
<br />
1 .-!./0!0.-!.!!!!0<br />
#-0#$0$# 3.!.!.--!<br />
#! -! #! ! ! . ! ! #! ! $ -0 <br />
0# -! $#! #! ! /. ! /.#. ! #!<br />
0!$$.$!0$#$.$#.!!##.-!!<br />
.!!.. !.-!#!!<br />
$$!0 +-!!.!$#.#.!.!0-!<br />
0#$. <br />
<br />
-)+)+)<br />
<br />
/0-!#!$0;!4#!/<br />
H !!0!#.!0..4#/..<br />
! 0 . ! -!
-)+)0)%<br />
<br />
E#!/0-!#;+#%!0 !0##<br />
.-!!#0!0!##0.!/.# .<br />
!00!/.0!/$0!-!#!.<<br />
! 00!/.!$!0!!.!!#.!#$0!-! !<br />
!0#.!!0 00!.!.!2.0$0<br />
!-! <br />
<br />
-)+)1)<br />
<br />
.-!!!/#!$0.
L$/./..$!!0/!0!-5!/..<br />
/-!!;<br />
1 5!;><br />
1 $!;@><br />
1 #;$<br />
1 (!#;+#<br />
1 ,;)<br />
1 &-#;@B<br />
1 ,&;(9M<br />
1 9!;-@##<br />
<br />
<br />
L;=.<br />
<br />
.0$#!/.!!!$/.!.!!0.$<br />
B J? !0 . ! !0 . / ! . ! ! ! ! #. $ <br />
!/.!$#!$#!/0.!.!00! <br />
. ! # !0 . / # 0 - $#! . #!# . !. $<br />
$#!.00!/.#.#$#!00 ;=.&-#;@<br />
<br />
<br />
L@
E#! @>.!/!0.!
! !0 0$ /.#. #0 $ 00 !! ! ! -# !! ! .<br />
<br />
<br />
0)$<br />
<br />
!$!-/!0/.-5!#!$ .!<br />
0#.0-.0$#!!2!-!#!0<br />
/.#. ! 0!0 0 # .! $ 0-0 ! !0 ! !$<br />
-0##!!$$!0 <br />
<br />
$!0 ./ .! #- ! /. ! -5! #! $ 0 !0<br />
#!0.! !!!.!#!$$!0!#!$!<br />
. 0 !0 . /2.- .! $ 0 % / ! 0-0 ! . !!# .<br />
0 ./.$.-5! 3!!#../$#!<br />
-5 . ! ! 4# /.#. / #! . ,( .<br />
!!# -5! ! .!0 0- #-!0 /. 40 $0 .<br />
!0!04#20 &-5!!0!/$!0<br />
#-0./! <br />
<br />
!-.-5!.
!"<br />
! #!# !"<br />
" # $!$%#&' &!( ##<br />
(#&##! ) * "$<br />
% <br />
<br />
<br />
<br />
<br />
<br />
! " #$ %"#$&<br />
' ( ! %'()!& * <br />
+ , - %*,-& "#$ <br />
<br />
<br />
. %.& <br />
" <br />
<br />
! <br />
<br />
<br />
& <br />
<br />
' $ +, $$ # !# # $ + $ (! -. /01<br />
&# $ " # # $# !( # $ $ 2# " 3 4! #<br />
( # #!#& 5 -51 # . / " -6671<br />
$ #$#!+$(! ##8 # &# $# #99<br />
! $#& $ # $ # +, $! $ $!&" . !( ## #<br />
! # $ # ! + 4! $ $(! !$ # #<br />
$ ! $!4&#"<br />
<br />
' $ +, #$ & # 5 :(! % -5:%1 $ #<br />
##! $(! # !# $ + #& ( #$ # <br />
##&; (,-;1". ! # $$#!#+, <br />
$!& # # # !$ % #& # ' &!( ## (#& + +<br />
#$:(!%# #:(!&# $ $ ##<br />
:":666: 5"<br />
<br />
&& ' $%<br />
<br />
:(! (! $ 4#8 # (! +8# &# <br />
# $ (! #$ &$ ! &# <br />
##$ $!4 #" ! ?# $#(!# &<br />
, +!$ $(!!8+ +, !4$ ## #<br />
$+, ?# ! $$! $"'#$!#&#$#<br />
$(! ##!!&+## $$ &! $## $ #"<br />
<br />
$# $(! # # $ +& # 2,!& #(!& $ !& <br />
!&#+#! $4#"#+( $ $$(!( ?$# $
($ !# $ #! ## # # $ $ !& !# " # #<br />
!# 2#&#+!?$#+$(! # $$ $<br />
# !#"<br />
<br />
' #& # # !# !& ## @# # 2 $<br />
# $ $ # !#" 3# # #+ !# $ # !& ## #<br />
A # !#! #& !!& # # $ 8, # # $(! #<br />
" ' !# ! # $ # !4! + + $ #! $<br />
2$"<br />
<br />
'$$#!#5:% $. $' ! &-. /'1 <br />
4$### # #$!# 8$9 ; ( #-9;1"!&<br />
5:% $ #$ # $ ( # ! $ $ ?# (!($ #<br />
$(! # $ + $ (! ## $ $ & <br />
$! !# $ # !& " # ! # $ # <br />
$! # ## $# $ # <br />
!"<br />
<br />
5:%!$$! (#(%-%1 < #-'5:%:6
$(&"D ' !# # ## $#!$ # $$ #! #$ !&<br />
$#$+ #!&# $ !##( $$"3##<br />
###$#$ $!#!8&##?# "<br />
<br />
$#$ $$#$#!4&#(+#$ $ "<br />
$ + $$#($ #C #"D' #+$$<br />
$ ##! $ # +## #4##!# $## "<br />
<br />
&(( * -<br />
<br />
%0111& #$ # ! $# $ ##$ #$ # $#!<br />
!# " G $$ # $ #$ #& # $( $<br />
#!##$!# #+ # # #"G$$# $ $#!<br />
##$ #$ # # $# #+ # # !# # # " G !<br />
4# $$ # # !# # !$ # # $ $ <br />
$ #" . ( $ # # C! #+,D # # $ $<br />
$# $#!# #+ # # $#($ <br />
$ "<br />
<br />
3"&##! #+,E<br />
8 5 8#!#&# # 84! &+&<br />
8 , 8##! # ##,!&4# $ # #+, $!## <br />
$$(&!# $<br />
8 '#8 ### #+,#!#+ ! # ! #+,2<br />
-6671"<br />
<br />
$(!$ ; (,## !$ #<br />
$ ## !# # ! #+, ##" 3"< !## # !<br />
## $# $(! # ! #+," # ! #+, #+ $<br />
&$ # + ! $ !!$ C#"D ' ! #+ # + #<br />
F
# ! +### #+ #"'!C,D <br />
# # # !# +# ! + , $($! #<br />
$ #+ ,"<br />
<br />
&(. % <br />
<br />
# # ! ( (!($ $!& !# 4 ( # $<br />
8#& #+," # # # # ! C+ ("D<br />
;4# # + ( # ! $ # # ! ! $##$ # <br />
( #C $D# "<br />
<br />
&(.& ! %/* <br />
<br />
. (*3( / ! %0111&!8 #$ # ##<br />
#( #+, ## # # $$". (# ! <br />
$# $$($ #!,# ,!& $!&# # ! <br />
I( % 5%" . ! + ( #! +, #& $!&<br />
(!( $ +$!&$#$"<br />
<br />
. ( +# $ # ## # $ # $$ (45 ( .<br />
%0116& + !+ ! # # $##$ # ( # # +, # <br />
$ #+ #"+(!# # $$#&!+&#<br />
$#+$(! # $# # +$# $( $( #<br />
!,8#( $#&##+"<br />
<br />
<br />
<br />
3"E'. (##<br />
'! ##+(##+ 3""(($!<br />
$ #( +##(,"'(2#(,<br />
$$( $#, $# $($# #". (# ! <br />
($ #4 #$!$ $ ///3/%0114&<br />
<br />
'(,# (!9# 9(& $# # -991# <br />
#+(#!#!#(($ $($!# $<br />
(!! (" 9# ($$ & ,& +$ $ # #! ! #4# !+ <br />
( #$##+ $ ("9(&($$&#!#&#<br />
4 2 # # ( $# " 3!& # # !$ & #<br />
# ( # $ $ # (!!<br />
# #(,"<br />
<br />
H
'+)<br />
<br />
'##$##$# #$( $&# $# # $$<br />
$ # #+ 2 %0116& 5!& $$ # +& # (<br />
#$#&# $# (!($# <br />
$& # # $ $!#(!8!( "$# <br />
$##$ #+,####+ ##!#,##,"'<br />
# # # $$ #! $ " ' # $$ ($<br />
#!#& #+ # !#" $ # #+, !! !4! $<br />
"<br />
<br />
' J $ !J + # $ $# # (!! # <br />
!4 !#( ! !( " 5# #! ! # !#$ # <br />
$(& # ## #?# $ " ' ( +<br />
(+# $# ++(#!$$#+! #<br />
# $## %0116&<br />
<br />
( )<br />
9 $ ##9;; (,-;1<br />
+# $ ##### $# $$+("'!##!#&#<br />
$( $!,# ,!& $!&+# $#+ ?# "3<br />
4! 8 (! $ #?# !# # <br />
2 # ! $ ##! $ ( 2# $ # #"<br />
. ##;+, #?# $K!$ (# $<br />
(+# $!#"' # #!## ?# #$<br />
$# # $ # # (!! !# ##A <br />
(#&#"'& $#!?$$! $!# <br />
($ # $ #8 $, +!$ !?$#!##$!"<br />
<br />
' 2###9;!$E<br />
8 85# #$$ $ ( #!+<br />
#!$# !#$ ### !&#(<br />
4$"!#+# ! $##+# #$#<br />
###! &!#"##9;($<br />
#+,, $ $ $ "<br />
8 , 8 '$# ! # # # $ $ <br />
!#$2#!&##$& #$#$(! #"<br />
# 9; + $ !# ##$ $!& $ 8$ $<br />
$ # # $ # !" $ # # + ( #! <br />
!# #!#& #+ &# $ !#" 9 $ <br />
!#$#&#!!$ # # $ !&<br />
#"<br />
8 / 8 !,#2 #!($ $<br />
(#9; !#!&$ #!?$ $$##$(<br />
# # # ## $ $ & +$ (#& # $("<br />
!#!# # (, $!&$ $( <br />
$ & # #" ' ## !+ (<br />
$ $ &(!$! $&#!!#&"<br />
8 " 8#$# !& $$! $!# -/1<br />
?# $(! (!$$$ $ #!#+<br />
# $!# ". ( ##$!# ##!<br />
! &!# !&4$4#!+## $"<br />
( $ (!! $ # , # # #! (!! # #<br />
#@##"<br />
=
8 !8'$#+#(!$$ ( ## #$!#&#<br />
#!$ & (!# # # # $$ " 0# (!$# $<br />
$## -00/1 ## !+ ! # !# # # # <br />
# ( (!# # ( $ $# # !# !## $<br />
" A 9; ($ # & 4# 00/ $ ##<br />
#"<br />
<br />
(& ' <br />
<br />
';+$(!$ #+ $! +#++! ( #<br />
!#&#$!&#$!&## $$ ##$ $ #". ?$!&<br />
+, # $ # $# # $$ ## !& (!!" 2 #!& # ; <br />
# 9;!# &# $!#". (! #$ $#!$<br />
; #4 $ 4# &#+# ## #2$#<br />
$# $" # $ ## # # # # <br />
9;#;8$9;!# $# &#$ $ #"<br />
<br />
(&& *-<br />
<br />
;!& # % ?$ # ! $(& $# $ $<br />
# # $! # # 0/. #&" ' :(& $! $<br />
!# #5-:051$ ?$#!$( $<br />
( $! $ !# #! :(&8+$" ' !# + # # <br />
#&##+!$($$#!# $! $!# (!! #<br />
#:(&$ #"###&!&!##(!!/#!+##<br />
##"<br />
<br />
$(! 998$ #&#### $ $ $<br />
!& + (" ' 99 # $! $ # (<br />
# $ # $ # ! # $# ##" . ! # 99<br />
# ($ !?$##+($# (2<br />
!?$$#####E<br />
8 0!$# (# $$## (($$<br />
8 L!#& $!# #4## -" "&!# #"1<br />
8 2!# -" " 4#2 ##"1<br />
8 5+ $ !#&<br />
; # (!! #$ # ! #& " '& # C #$D<br />
#" ! # #& ( $ ( ($ # <br />
#$"';($ $##(+ !($# <br />
$(& $ # # " , #$ # # (!!<br />
($ ##!# !#!$#+ #,+ "<br />
<br />
(&( * -<br />
<br />
<br />
3"B !## # !# $ ! ; # ! #+," . <br />
$(!$ # ( # ; $ # 2 # # ! !# #<br />
?# ! $"<br />
8 !;-;1K'( !+(($##(!#<br />
!! - C !D1 $ $$ # # +# # #&" '<br />
; !#8#8&#;+ "<br />
8 ; # ; -;1 K ' ( # # ! ?# " <br />
+## ?# #; $$((!$ $#$ <br />
;;# ## "'; #;+, #!$+# !<br />
?# "$##$@### $#& #&
8<br />
<br />
# !# $ $# +# $#& # # ; #<br />
;$ !# !"<br />
9 #& ; -;1 K ' ( # #& ( ## <br />
(!! # !#! #" ; $## & ? # # +#<br />
;" $$# !& ; +# # $ #! # ( +# #<br />
;" . ?# (!($ +, <br />
$ # #& ; +, $!&$ # ! $#$<br />
#!#! !& $!#"<br />
<br />
<br />
3"BE! #+,$ # # $ $ &#<br />
<br />
#!# $!# ## $#!# !# $<br />
$#+##;+,# !#&## #"5 ## $#<br />
#&( #;$!&$&# #& $# $&##$#$<br />
#&" ! ?# # #!# # ## #9;<br />
+# # #& 9;" G+( & ?# # $ # ( !#<br />
$ ####!# #!$$ ##+"<br />
<br />
3"B + # ## #+ # $ #& ( # ;" (<br />
$(! # # !# $ !# ( +# # ; ; <br />
#&;-;1"##($(! #((!!# #<br />
#+,;"'!+## +## # $##("<br />
<br />
(&. *$%<br />
<br />
$(!$+8$#! $##!+(+ #,##! $<br />
4#(!!# ;"'# $#!!##(+ <br />
##$#,+# # &$#(#+#+(#!"( <br />
!$ ## ($ !!# $ !& $# # $# $#<br />
#&#! ## $ ! ## $ # ( <br />
$$ "<br />
8 - 8 ;4! $# ! #! ($ & +& #<br />
+# # & #+ ($ $# ( # 9;<br />
+,"<br />
8 7 + 8$(!$!##!##(( #<br />
# $$!&!$#("<br />
8 3 8$(!$! #! 8$#$(<br />
($ &+&#+## &#+#!$ $ 8, <br />
(#9;+,"<br />
<br />
B66
(&0 *+<br />
<br />
. !#,!#9; ( #(, !&$ $<br />
#$ ## $##,$ + $$#(! ####<br />
(!!("'(! ##! ## !&+&##;8$9;"<br />
+ +(#!!$ #99# ## <br />
#$##(!!"'!+!4!#& (# $#?# <br />
#### $$ #"<br />
<br />
. $(!$ ! # ## !+ # !# ( $ #<br />
$($!&"( $#$ #(! " <br />
# $$ + + !# !#! ( # " ' ($<br />
(! #+ # (" ' ( # !! <br />
(" !& $$ +,($@#"' ($!##(! $<br />
!!"' ###$ $ #& (+$<br />
!&?$"<br />
<br />
(( 1),2 3 <br />
<br />
. ! # #& #+ # # #! !# # +# <br />
# #& $ # !( # ! #+ #" <br />
# $ (!& :05 $$$ # ! & #! ! +<br />
# ## ($ $#( $ ## # $! $ !# " '&<br />
! ($ $! (!$# (# $ $## " . ! # $! <br />
4# (#&4 ( $$!##! #"<br />
<br />
'+,!$ $ $ (, ##!+ ?# <br />
#! $###+ $ " !# &# #<br />
#$! $$(!!$ (# (!$# $$## <br />
# " '! ; +, +# # # ! # ! <br />
# #2 #E<br />
<br />
'!E' ! & <br />
14<br />
*<br />
" -<br />
($!# # # #! #& 5 <br />
($! ##! #& 5 <br />
#8 ##8 # M: MM<br />
#!8($8#8 M: MM<br />
#($(&$# $## :<br />
M<br />
#<br />
# $$$. (! # #&: 5 M<br />
5 8$ $$ ## # : M<br />
!#$ $ #<br />
#N M<br />
I(N<br />
#$# : M<br />
G $#$ : M<br />
$(! #! $4# # M<br />
#$ # # : M<br />
#$ #& $# # M<br />
<br />
B6
.<br />
*-$)" - <br />
+, !!& +# # !$ % #& # ' &!( ##<br />
(#& -%1 $!& # ; ##& +, # $ #$ ! $!<br />
!&" ' !& $(!$ & % # # +, # # "" :(&<br />
##$ $ $ ##$#,"%4# (4 <br />
$ + # ! $! $ !# & $ +<br />
&#&# $&$$& "'; $!&$$ ## $<br />
## % ## $ # ###&$<br />
$####&((!!"<br />
<br />
.& + )% <br />
<br />
'#(!&## $$ ###;+##% $$(! # ( #<br />
$%+,$# ##$(!! !####!+ #E<br />
8 '! !# !&2$@#"<br />
8 ' !& 2 # # # (! $# !# $!" '<br />
!& $! $(!$ & # $ $ $ # <br />
+,## "<br />
8 ' # # # !# $ # !& # #$ #8$!<br />
$ # $ # +# # ; +, $ 4#! # # ; #<br />
!# $ "<br />
<br />
' ! ! !#$ ## # # # + #$ ! $! !& <br />
$ # @# $ # #$ # %" ' ## # <br />
$!$##!&#?###$! 3/8-
User<br />
Input<br />
Pneumatic<br />
Tube<br />
Launch<br />
Inertial to<br />
Body<br />
Centered<br />
3"7E!<br />
Air Launch<br />
Angle<br />
Conversion<br />
! #$!<br />
Water<br />
Impact<br />
Water Entry<br />
<br />
' !# # ! ! # # $! + # #$ # # #<br />
$! ( # # % # !& $ # ; +," # +<br />
#$ # $! # # $ 4 # # #$ $! $ # <br />
#+# ( #"<br />
<br />
.&& / $<br />
<br />
$ # # #( # # # # # # !&$ # # 8<br />
$!# # "'+ 3"F" &$!#%$(!$ $<br />
(!$#$ 35%'%:" ' # 2 # # ?# $ $ !!& <br />
35%'%: ! # # $(! # $!" ' $! # (#$ #<br />
#! $# (!$#$#! ##35%'%:$!# #&##<br />
!#" 5 ! $! # # #& # # #$ +# # #!<br />
!, ; ( #" # + !#$ !#$ # # # 2$ # $(! (!$#<br />
(# $# #$!##!,$! ( #"<br />
<br />
Develop<br />
FORTRAN<br />
Model<br />
Validate<br />
FORTRAN<br />
Model<br />
Covert<br />
FORTRAN<br />
Model<br />
Validate<br />
Simul<strong>in</strong>k<br />
Model<br />
Integrate<br />
Simul<strong>in</strong>k<br />
Models<br />
Run<br />
Integrated<br />
Models<br />
Develop<br />
Simul<strong>in</strong>k<br />
Model<br />
3"F8;4# $!# # <br />
<br />
.&( *" - $<br />
<br />
' $! # # +# # ; +, ! $!<br />
( 2$" $! !$ $ #$ # # ; +, $ <br />
$($! # $8! # # (!$# 2$" # $! ## $<br />
# (! # $ # ! + !# !& # #$ # <br />
#$!"<br />
<br />
Develop<br />
FORTRAN<br />
Model<br />
Develop<br />
Simul<strong>in</strong>k<br />
Model<br />
Wrap & Publish<br />
FORTRAN<br />
Model<br />
Wrap & Publish<br />
Simul<strong>in</strong>k<br />
Model<br />
Validate<br />
FORTRAN<br />
Model<br />
Validate<br />
Simul<strong>in</strong>k<br />
Model<br />
3"H8;$!# # <br />
Integrate<br />
Models<br />
Run<br />
Integrated<br />
Models<br />
'!+$!#$(!$$##!## "3# <br />
#,$( # #(# (!$# $$## # !#!$$#;<br />
<br />
<br />
B6B
( #" ( ###!#$+#(!($!$(! <br />
# $! $ $# # + # # $! (!$#$" 3"H + # $!<br />
# # +##;+,"<br />
<br />
.( <br />
<br />
'+$# !&?##!###;$##<br />
#%$!# # ( #"'+E<br />
' ; '<br />
< ! #!<br />
#!3 # !#&<br />
# (#$<br />
35%'%:##!<br />
<br />
=6G<br />
F6G<br />
;<br />
G+#+!# <br />
G+#(!<br />
<br />
=G<br />
=G<br />
'#!
'!0E#4#!#$!! $!# # <br />
# :> ' ; '<br />
< (#35%'%:$!#<br />
'$!E<br />
7$!-P6G$!1<br />
' . ##. #$<br />
3!# $. (<br />
<br />
8<br />
8<br />
8<br />
8<br />
<br />
1<br />
5 K +# # ; +, ( !$ !# $!<br />
!# $ 8$#$ ## !$ #$ $ $<br />
# #$ +## ( # ! $ +## # #&<br />
#+"<br />
* 9 K +, +# # ; +,<br />
?# !& #! $# 4# $! $ !# +##<br />
4 # !#!#&#&# "<br />
* K! $ # $$ $#!$ (!$# K<br />
#!$&(! !#&$K+##;+, !$ #<br />
(!$##$ #&$(!$+##+, $# $ <br />
# +##$(+"<br />
- K! #!&(+$$ &<br />
4#!!$8# $ !#&$8+! #!&$#<br />
!(+ $(!"<br />
! # 5 -6671 , /: * 3 5<br />
; 35.#%#:"58678
! <br />
" !" <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! <br />
<br />
#$ "<br />
<br />
# ""$ %<br />
$ $ ! $ <br />
!& ! $!" !<br />
" ' <br />
"$ ! " " " " ! ! <br />
$( "! <br />
!$ $ !" " ! ! <br />
"$ ) *!$<br />
! +,-!"*! $*! $<br />
! !$"<br />
$ .$! #!$/ $<br />
0 "$" <br />
"*! " <br />
$ $1&!"! <br />
! " % !<br />
<br />
*! ! !"$ "<br />
! 23-) *!!!")! <br />
!! !" *! <br />
! ! " ! !" <br />
" !" *! " !" <br />
$ " "4!<br />
!! "" !$ <br />
$ " " " !<br />
*! & $ ! <br />
" "<br />
<br />
%$" <br />
<br />
( !" 5!$ $ <br />
" .6778/" !.99/!!$<br />
$ $ ,!" " *! &<br />
$ !#$!! $ !$<br />
! $$ *!<br />
" ! $ )! <br />
!" !
! ! ! <br />
& $#.677:/ .98/ !<br />
!! $ $ ! ! & ! <br />
!$;& !" <br />
" " $! !$<br />
" ! ) <br />
! "!"$!5*!<br />
<br />
-! ! $ "! )"6 + < 3= > &<br />
,? 5&>!<br />
@$ .*! / " " 5& "$ > !<br />
" 4""<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
#!$<br />
'($)$(*(+<br />
)"6
!$" +3<br />
*! )"9$!0,1(?.99899/ 45<br />
"" *! " " " ! <br />
! " ! $ ! 5<br />
!!$"*! A5<br />
<br />
+<br />
$ "<br />
− " "" !"<br />
− 3"$ "<br />
− &" $ ! $ !" $$<br />
<br />
− " " " ! <br />
!" "*! <br />
− !$<br />
<br />
""! ! 5"$-"! <br />
!!" " $" " "$!<br />
." !" ! <br />
!/ 5 . "& <br />
$ " / ; *! ! <br />
"!"5$ ! $ <br />
(+ $ $ ! $ <br />
!$ $ " <br />
" ! ! <br />
*! # <br />
" > $ !<br />
<br />
7
$)! !! $ ! !<br />
& !"$$ $!" <br />
! (+ "!$"H!" H!<br />
! $ 3!" " 5 " $ <br />
! ! )" & ! $ "<br />
<br />
<br />
6'3*-)'121'127/-/<br />
)8/7/*($<br />
(3(//1+73)$.)'('*/<br />
9(*(+$-'1*-)'<br />
1'9<br />
(8-'-*-)')8/7/*($3)$.)'('*/<br />
86'3*-)'/<br />
7/*($/*+63*6+(<br />
/(2(3*-)'<br />
)+$621*-)')8$1*,($1*-312<br />
$)9(2)8/7/*($86'3*-)'-'0<br />
:')42(90(;1/(8)+$-'0<br />
5 & $$$ $ <br />
! " " $ "! (+ ' <br />
! <br />
$! .98/ & !" <br />
" !"!$"<br />
, " ! $"%<br />
&! )! ! .6729/<br />
% .677/<br />
<br />
""" "$!<br />
" ! " <br />
! $ " , <br />
6
! 5"$ <br />
! !!*!! $" !" <br />
!" ! ( <br />
$ *!! !" ""! "!<br />
!! <<br />
<br />
C IJϕ 6 .&6&9/ϕ 9 .&6&9/K <br />
<br />
.&6&9/"!Jϕ 6 ϕ 9 K $ !<br />
&6&9 $ "! L$ C<br />
.&6&9/ L !<br />
ψ. $!/ <<br />
<br />
& = & + ϕ<br />
<br />
<br />
6<br />
6<br />
6 9 <br />
& = & + ϕ <br />
9<br />
6 6 9<br />
<br />
# " I .& 6 & 9 / " ." <br />
I.& 6 & 9 / ! <br />
<<br />
<br />
<br />
<br />
<br />
' . /<br />
= 6<br />
<br />
4! & $ ! & <br />
!& !!<<br />
#<br />
!&?<<br />
<br />
=<br />
<br />
−<br />
= ! ⋅ ⋅ <br />
6<br />
<br />
9<br />
=<br />
6<br />
4.& 6& 9/ &; !) <br />
! ϕ 6 ϕ 9 <<br />
ϕ6.<br />
6<br />
9<br />
/ = #.<br />
/α<br />
<br />
ϕ . / = #.<br />
/α<br />
9<br />
6<br />
9<br />
9<br />
= . α = <br />
−<br />
−<br />
<br />
6<br />
9<br />
<br />
9<br />
9 9<br />
6<br />
/ + .<br />
9<br />
/ <br />
!& ! )"8<br />
<br />
$"%&! $.!!/ $<br />
#! $5 $&$!"<br />
$" & $&#.&/<br />
!$" $ 5*!<br />
$ & & <br />
.5656/ !)"E <br />
M "N " ! + MN<br />
&5 #$ ! ! !$ *! ! <br />
& ! M! N " <br />
M! N#$""!! +! !$<br />
! ' " ! <br />
"*! $! & & <br />
% ! !" #!$<br />
<br />
<br />
<br />
<br />
<br />
6<br />
<br />
9<br />
<br />
66
5 <br />
*! " ! ! & " &<br />
# $ H " ! HH " H <br />
4*!$ $$ #$B <br />
& "$! &<br />
<br />
<br />
<br />
<br />
)<br />
)-<br />
<br />
)"8C& !)! <br />
(. (/)/<br />
-. (*0-<br />
(<br />
<br />
)<br />
)-<br />
(. (/1,<br />
-. (,02<br />
(<br />
-<br />
),<br />
-<br />
),<br />
)+<br />
)+<br />
)*<br />
(<br />
<br />
( )* )+ ), )- )<br />
<br />
(- ( )* )+ ), )- ) )-<br />
<br />
)"E- #&<br />
<br />
($)"!<br />
<br />
O 5"$ %$5"<br />
! )! $!5""<br />
$ ( ! " & & <br />
#!!& $ %$+$ & <br />
!" ' !<br />
$$ " $!$!"<br />
& " ! O & "<br />
!! "$<br />
<br />
*<br />
<br />
(;?(,-O @+ > (;?(,-O @+ > PO +1?P1O +4P P .98/ # <br />
"6<br />
(!4 8E587<br />
<br />
;1??(,C.9/3 4 9 4-,+'!" 965E<br />
)*<br />
(<br />
<br />
69
;1??(,C.9/ ;! 5'<br />
<br />
)(QQ1?O >'13>,+'O >'(0=.677:/$ !<br />
99 $ 0'$$O "34 :59<br />
<br />
'1QQ?-,.9/ 9 4-,+'!" 965<br />
E<br />
<br />
'13>)(QQ1?O >(,,1101>)4.98/5! <br />
4-,+"! 85E<br />
<br />
,(+(@ F .99/ ,<br />
? ,+!)<br />
<br />
014'(1C R+> 31=R(?1C (; .96/ 6 7 <br />
'R3?-0(C%9668 +4<br />
'$$ 3"5,$#<br />
<br />
-0?R(=+0 Q> ;-QR(0@++ C=> =(,@?1Q+3P1 ?C> ,+'4'10@- 1)<br />
.6729/" 8 .$@0"QO 0!<br />
/F O $<br />
<br />
1+),>F('(0;(@''1.98/5! 4-,+<br />
"! 8E75827<br />
<br />
P1Q(0=+1O +4P .677E/ 9 <br />
<br />
<br />
P1Q(0=+1O +4P.6778/ 7 <br />
, "$E 865E<br />
<br />
PO +1?P1O +4PP.677/" <br />
,$<br />
<br />
6
!"#<br />
<br />
!" !" #<br />
# $ <br />
" % % " &" <br />
'#"# &( )# "<br />
" ( ( % <br />
*% "#" ""<br />
(##'# "<br />
) " % + # "<br />
" '##" ##" ##<br />
" ," #" ## <br />
" # # " ( # -# #<br />
"## #& <br />
# " ( # .)-) /00 *123 )34 ) #<br />
# " '# % <br />
"# +%( # % ""<br />
" " # " '# " #<br />
" % # # #<br />
#'#"#5'5'55'<br />
%# &# " '#&<br />
# " " #" <br />
<br />
$"%<br />
<br />
'# " "" " ( #<br />
#"" ##"#" <br />
5'5'55'" #3 "### <br />
""&" ( %6 ## #5'<br />
5'55' "# " # #7<br />
'##" ##( %$<br />
o " ," " <br />
o "" " " <br />
o " " ( ###<br />
o ( ##" <br />
'# &# " #,"#" ( #<br />
# # #8 # '# # <br />
" " ##
'#(## # # "'#%<br />
% " " # " (<br />
# # <br />
<br />
: ( " # " <br />
#" ##"" 7 # <br />
!"##$%'## " <br />
""". 4'#(## # #" #<br />
#: ## #<br />
### " '#( <br />
$#( ( # " '#<br />
&" % # " (% <br />
( # ." <br />
7 4& . "7<br />
42"##"#&" " <br />
" #" 7 <br />
<br />
'%"# + " % <br />
"#; '##" % <br />
" # )3.)" 3 7<br />
45'55'&'!())(%*!())+%<br />
<br />
<br />
&<br />
'(()<br />
'(()&<br />
<br />
'(()<br />
<br />
; $#" <br />
<br />
<br />
'(# $<br />
• ""<br />
• "" (#<br />
• ##," " <br />
• # " " <br />
• #<br />
• %( #" #&"<br />
• #" #<br />
• " ( ## <br />
<br />
$"<br />
<br />
'# ##"# . ( #%(7<br />
4#"7<br />
( ## '# #" ( ##" 7<br />
<br />
9
) ( ## " # "##<br />
" ; =$<br />
7," " <br />
7 " " <br />
7" ( # ##<br />
( ##" <br />
*# # " $ <br />
<br />
<br />
<br />
<br />
-,.<br />
,<br />
<br />
<br />
-#,<br />
.<br />
*+,<br />
<br />
<br />
/,<br />
,&<br />
*+,<br />
<br />
*+,<br />
<br />
,<br />
<br />
<br />
; =$3 # <br />
<br />
• 3 $ '# # " ," "" # <br />
'#%"# # <br />
% "# &"<br />
# <br />
• 0 $3% " 7<br />
" # ( # #<br />
<br />
• 1 $3% #" (# "#<br />
&# (% "" " . 47<br />
" . 4'# "7<br />
"(## #&". 4 <br />
# '# # ( # #<br />
># <br />
<br />
0" <br />
<br />
'# # "% #<br />
6 # # # ( ( ","<br />
# ) <br />
,!"###%,!())+%3("## <br />
""###( <br />
# ( 7 " ""<br />
'# # # <br />
" % * -' ./ !"##$% * ' <br />
!())+())0% # # "" <br />
(#<br />
<br />
'# # ," # #<br />
" "# @A .@ %" A " 4<br />
6 % 3 %" A " .3A4 # @A<br />
" # " 1 , !())(%<br />
, !())(% 0 ( # 3A <br />
&# <br />
<br />
1" ( # # " <br />
&# " # " '# " "#<br />
" $<br />
• &# <br />
• (% "" " . 4<br />
• " " . 4<br />
&''!())$%# #" <br />
"# &# <br />
(#'## #"7 <br />
" #3A<br />
<br />
1" <br />
<br />
6 # . 4 " # # #<br />
( +#" $<br />
4 <br />
=4 " 8<br />
4 ># "<br />
."#" 48<br />
4 " . #"""<br />
4<br />
94 &" ". 4#"" <br />
# "<br />
2"<br />
<br />
'#" # #7<br />
*# ###( "$<br />
• . ( # %( <br />
4<br />
• " ( ## <br />
• ( ##" <br />
: # # <br />
%( '# ##%( $<br />
• "," <br />
• " <br />
• (%"". 4<br />
• " " ( ## <br />
) # 1 " (% A)0<br />
) " # " '# " <br />
( "( ##"#'13" <br />
<br />
'# # " # "" #( 1#= )<br />
#" ( #BA ! " <br />
'# # B " " '# <br />
7 # "# " # # " <br />
(" '# "" #DEA = .D( E"<br />
" A " 4# '# <br />
#3A-# " 7 <br />
<br />
4<br />
4<br />
x [Nm]<br />
6<br />
5<br />
4<br />
3<br />
2<br />
1<br />
0<br />
ψ A<br />
= 15 [°]<br />
V A<br />
= 10,5 [kn]<br />
ψ B<br />
= 270 [°]<br />
V B<br />
= 10,5 [kn]<br />
4<br />
TCPA [s]<br />
4<br />
CPA [Nm]<br />
1000<br />
3<br />
2<br />
1<br />
0<br />
0 500 1000 1500 2000<br />
t [s]<br />
0<br />
-1000<br />
0 500 1000 1500 2000<br />
t [s]<br />
<br />
<br />
0 1 2 3 4 5 6 7<br />
y [Nm]<br />
<br />
H AB<br />
[°]<br />
200<br />
0<br />
0 500 1000 1500 2000<br />
t [s]<br />
; $ ' # 7 " " $ 4 # ># )<br />
( ( # # " # # " 7 # " 8<br />
.&4FFGH 841)84'1)84# !##I<br />
<br />
; #( & " # " (# # <br />
# # '# # <br />
( " # 00 !"##0% 6 6 !"##$% '#<br />
<br />
BA =FF=#$ "<br />
= DEA I 1"=FF=#$ (((""%,<br />
<br />
C
" (% 78296!"#:(% 3<br />
# ""# ## "# ) # " <br />
# " #)" (#I#)%<br />
"0 ##"# "# <br />
"<br />
<br />
'#( ('# " ( 7<br />
" '##) #I # " (<br />
.#)4" "#)#<br />
" " #" "&'!())+())0%3<br />
#I #%#"# #)7<br />
#I# <br />
#)'# # "#)%( #)<br />
#"> " ; <br />
<br />
3" <br />
<br />
'# # " # &" #<br />
'# &" # #" # (% 3 <br />
" " " # " )3 " 5!; <br />
" 2"# # #" 5!;#"<br />
5!;.9 <br />
> 1:A-*/ .# 1:A-*/ <br />
#" 4<br />
<br />
4"<br />
<br />
'# " # " <br />
"# #('# " <br />
( ## $ : (<br />
# '# <br />
# " ( # # (% # <br />
'##" (" <br />
3 '.3'4" " <br />
<br />
<br />
<br />
5<br />
<br />
1:A-*/.J?=471&7 -3:<br />
<br />
00.JJ4; <br />
( + <br />
&8281?
1)8/)A:-6 .JJ941 <br />
< # 3 '#1<br />
'* . #4<br />
<br />
3K*0/)- -D8 )AK)0D)- - .=FF=4 1 <br />
2 /. =FF=4A) <br />
<br />
D:)1LL8:-/)M6 8-I)ND3B.=FF41@1<br />
*"B0 =97 =FF<br />
'# '09D( 997' <br />
C # 3 1 R)*D'K I*L3*1L*N'6 ) 0)6 :20*/:3<br />
:26 :20*/::-)LA:'S6 0)2:-L*T6/. 4<br />
<br />
-*AB80:-53/.JJ941 , !<br />
<br />
')'*1L0K )8 -I)ND3 B .JJ
!"#$%&' $<br />
! "#$ #%<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
!"<br />
## <br />
$ <br />
%<br />
#&"'!<br />
()"(*) <br />
( () <br />
+, <br />
$ +, <br />
()$+,<br />
<br />
() *+<br />
<br />
()( , *<br />
<br />
&&'( %"<br />
&' &% & %&<br />
%%& %<br />
& & &" % ) % ) <br />
&&%*+,#%&%<br />
) *+,-.<br />
• %/010%) &&<br />
#&-)#(2 %<br />
• %%<br />
• %2&&(34&<br />
&<br />
• &,005/,51500607/5671&" <br />
8#&"'& )<br />
0 *&#&%&.<br />
• )% &&&0&<br />
• &<br />
• %& %<br />
• %% %#<br />
&<br />
* &4 & & 2%<br />
&4#&&&%%/3)#41<br />
<br />
()- .* ,<br />
<br />
9) *: && *+, & % & 0) %<br />
;*&%&%%&%&:8:8,%&&
& 8 % *+, & &<br />
#).<br />
• *+, & %<br />
% :8 / #0 <br />
&%1 & % <br />
-$.&/000'<br />
• 0 %0) #<br />
#0%)&&%%%&(<br />
/1#<br />
8&%%&,7<br />
* %;&/,7*;1*+,:81&/000'<br />
# % :8 % <br />
:8%:8:8< !2&3445'<br />
% # 2 *+, <br />
% & % <br />
16&3445'<br />
<br />
; % ;= &<br />
% 7< % & %<br />
& 0 # & & ) <br />
* #0 & :8 % % - <br />
%&*+,&&% & >&&<br />
>,78&&7/871 87'%<br />
# && % %% :8 && # *+, && <br />
)#& 0& <br />
<br />
*%#,7*;#))#-<br />
&& ##%#%,&;=&"<br />
,7*; & *: 2% <br />
# 7< % ,7*;<br />
&% && # ) ,7*; -<br />
<br />
<br />
-) /0123 <br />
<br />
? 0) *: &% <br />
4>&*8/@>* 1%0<br />
AA 35&
%#% %%%%<br />
% % & / & >,7 &%<br />
&" 9 &3445'1 # #) %&% && <br />
*+, )# %% & <br />
& >%00D&%<br />
− : % & &% C 9)# <br />
%& #&#)#+,"#<br />
%%&#<br />
− 7
0>#'&%%#)<br />
% & & %% &% %00%)<br />
'E %& % <br />
&& 5&&"<br />
% %% &% ,& & # <br />
%)0& 00<br />
<br />
&"&&&&<br />
% %% # <br />
2% +2% %#) <br />
& & / &<br />
%1<br />
<br />
+& 7 # # <br />
7
• % %& & /1 C & # ) <br />
%& ,7*; & % & &<br />
-%&&>7$&/00?'<br />
>&%%&&<br />
&&+&%&)<br />
& %,7*;%&&<br />
<br />
• & /1 C % & 2 '&<br />
' (: % <br />
, ,7*; # &0 #<br />
% &&4 & 5 34<br />
& / &&1 & % & <br />
&,7*;<br />
&&%%%),7*; 5+,&<br />
%& &<br />
# & :8 % & && % % <br />
#)&+,%%%&<br />
#,7*;0%0%<br />
% <br />
<br />
,&&-&'&,7*;0)<br />
#,7*;#&&*&)#<br />
&%,7*;0)%&0)8*+, ,7*;<br />
# % <br />
&&!2&3445'>&&4,7*;<br />
3%&%4C ##)<br />
%#%#),7*;%<br />
<br />
6)4 7<br />
<br />
8*+,&%-/0%)<br />
1 %% & ,7*; %&<br />
).<br />
• <br />
� <br />
> <br />
& %& & & % <br />
& # & % & %) & <br />
%& & %#'&%'<br />
<br />
8 &*+,<br />
&%%%&*+,&"0#<br />
%%&<br />
• -&&<br />
8 ,7*; %& &% - <br />
&% &<br />
#& # ;E &% & & &% <br />
- % % & # ,7*; & <br />
-,7*;% <br />
• %0&&<br />
8 &% & % &<br />
%&&%&)&<br />
)&&%-%)%4%<br />
%#4%<br />
<br />
6)6 85+.<br />
<br />
8%# %(<br />
%%&&'&%<br />
%"2%& <br />
<br />
6): &,9 ,<br />
<br />
? ,7*; 2 - & <br />
%&%,7*;#%" <br />
*&%%%&&<br />
&%&%,7*;<br />
+,7*;0) <br />
%&%7
• & %%&0% <br />
# %%%%<br />
&%&)#%&%% *#<br />
%/1%&%&%% <br />
• & 0% 0 %0 # &( <br />
&&%<br />
&%%<br />
• 2%0 &&&%<br />
%&% &<br />
&<br />
• % % 3%4 " & -. %& &<br />
% ) F :%& /& %1 /&<br />
%1%#"&8<br />
& 2% %%&<br />
8 & % - %&% &<br />
% % & && 8 ) &<br />
%&%%&8<br />
0&%&&& <br />
• %&#%"&&&F <br />
8% #2%%<br />
&%&%#& %%)&<br />
%&%%<br />
<br />
Product model<br />
item<br />
Function Document<br />
Set Component<br />
Spatial region<br />
Socket<br />
Connection<br />
F.>%%"&5%&&5
:)- &!<br />
<br />
+ %% - % % 0<br />
- & %<br />
% % % & &" , 3<br />
&%&43#&%&43&%&40&& <br />
% >#>&8%#<br />
&%&%&+& % &%&#&<br />
&%/3 %43%418<br />
/N%1 /& %1 " & & /%&1 <br />
"&#"& %&"&/1<br />
30&-04 F"&&&&<br />
# & & # " & * -<br />
%%&&%%<br />
34#7% 3)4<br />
"&8%-<br />
& & % #<br />
&(%3#4<br />
#3&&4%%"& ,79 2 <br />
* %&% %&0<br />
>,79% <br />
<br />
:)4 <br />
<br />
5< %% % - & >,7<br />
- 21 &/000'%%<br />
0&%% % & &<br />
#)&&#) &)%5<<br />
% *&%&) 8& % <br />
&>,7872 &3444'%&%#<br />
%&>,7%%#%)%<br />
C &&%&%#<br />
&0*: *+,&&<br />
& <br />
<br />
;) !,93 !9 <br />
<br />
7
• &&<br />
• %&2&&<br />
:% &7&<br />
O:88:&%<br />
<br />
;)( ,<br />
<br />
* & ,7*; % 7< # <br />
%% % % % 7< % % % <br />
%% & % % %- : <br />
)5"5/551&&&"<br />
GAA %& 0 <br />
%&: 8&% %<br />
%0%&%%%&%%<br />
%%?>#%<br />
& && % 8 # <br />
&% & " %& # <br />
7
" 7< & 9) 7<<br />
%:-&&%&&,<br />
C&#+#8<br />
00 , # " & <br />
2#0& 80&&&<br />
&3%4(&%&&%&0%#<br />
& &"0# FG # : % % <br />
%&&8+#&&%&<br />
+ *+&83)4&&<br />
%+ :"&%%<br />
2% &"0#>%&"0# <br />
&3&"4&&&:(34&<br />
% % % % % 8 # % 7< %<br />
&"%34##5"<br />
)0)0 %&<br />
)0&"0 5"&<br />
& %0 # %( %4<br />
%: #)#%% (#)#<br />
& <br />
<br />
Ship (part)<br />
Ship (part)<br />
Situation A<br />
M<br />
Rev<br />
A<br />
Situation B<br />
(Propulsion<br />
system + ship<br />
have been<br />
revised)<br />
Rev<br />
A<br />
M<br />
Rev<br />
B<br />
Propulsion (part)<br />
Navigation (part)<br />
Propulsion (part)<br />
Navigation (part)<br />
M<br />
M<br />
M<br />
M<br />
Rev<br />
A<br />
Rev<br />
A<br />
Rev<br />
A<br />
Rev<br />
B<br />
Rev<br />
A<br />
<br />
FG.%:&&%&<br />
<br />
;)4 ,<br />
<br />
:#&7
?&O:8%%&:&<br />
#:%"&)<br />
# : 7< % % :8 <br />
3)47 *&&&%#:87* <br />
0 && E ) % 0 & %<br />
&%0&2/&1&<br />
: 7< % && ?> C & :<br />
%# ?>%- <br />
<br />
;): !7<br />
<br />
&-:7 & % )# 2%( & <br />
& && # & 5& && 2% :<br />
- %& % ?> %0 <br />
#)&%&<br />
<br />
;); 93 !<br />
<br />
5) % :%<br />
&&>)<br />
&& && 30004 : &% <br />
- * )# # # :4 % %<br />
%- 87* # & F %:<br />
,&7% % >%%7%% & ( % %<br />
%& % 87*<br />
* % >>8 79< %& & &<br />
2%%( %<br />
%&2%-
% % % %& 7< # & <br />
&.<br />
• #%%&&&>+&"<br />
• %& %&�<br />
• %% / #1 %0 <br />
( # % " 7< %<br />
<br />
• & 0 %& <br />
& ( %" %- 2 %& <br />
&
,;*6>8 > 5 ( F8P* , /MMM1 !-( <br />
A * : :%& 8&& <br />
>&/*::8>1:%<br />
<br />
?*>:P,; + ( 68>>,9 + ( 98R5 7 ( E 55 ; ( E L
!<br />
" <br />
<br />
<br />
<br />
<br />
<br />
<br />
!!<br />
" # ! $<br />
! %&'"<br />
!!!#!!<br />
# $ ! #! "<br />
"("#!)<br />
" "!! <br />
*# + $*<br />
!" ) " * * !!!<br />
#!,<br />
,!"#!!-!<br />
! ." (! (! (" '!! <br />
/! $ # (! 0# !## # %12(& <br />
) ) 2 ! #!<br />
3! )<br />
<br />
! <br />
<br />
$3!#%4$& ##<br />
!! !! ." ! ! " <br />
#)"!#" #!<br />
#! "!# " ) # !! " <br />
! ! ! ! # ) ! "<br />
!"")*0#5#<br />
! *##,,, %-$(& ! # "# $ 4<br />
##!#)"#<br />
<br />
#! $%&'(()!<br />
* ! + # <br />
!+ <br />
!<br />
+
4 # # ) !# ,) #<br />
!###$") <br />
3 !## # %12(& ! <br />
!! # "# #!3 $ # #!" !!<br />
#! 3! 12( #!" ! <br />
!!!#"""#<br />
#! $ / !) <br />
!! #" # 0 7 <br />
" ! #! # , # <br />
# !) 1 ) #!,!) <br />
##!"<br />
/#!3)/,<br />
!" , ! ! <br />
" " # " $ " <br />
#!#" <br />
#!2#") <br />
! 7# # !<br />
# # #! # ,<br />
! 5"!#!# <br />
"$) 3#!!!<br />
) )<br />
<br />
$ 48$!292#(! ("#!#12(<br />
,<br />
,-,<br />
1!<br />
##!! !!)!<br />
(<br />
(<br />
."' 2(:#!4;(#<br />
1)5# 2 # )<br />
0 /!"<br />
4 5 5 # # # " ) !<br />
'#<br />
7(<br />
(! <br />
78$51!5<br />
<br />
1!###!##<br />
* * * ) * * ) * # " <br />
# 7 .) # # # !" " # )<br />
"!###! ))<br />
" # # #= " # ! <br />
!= # # ,) # '<br />
/ 3#! # # ) ) )/! #<br />
!!$)/!) !"<br />
###!0!#$ #<br />
# ) #! # # ! 3<br />
! ! # <br />
$ 44!"<br />
<br />
$ 448$
• !!2#<br />
" #<br />
<br />
4#"# !/)"!"<br />
$"#?)!)#)<br />
# !! ) ) ! !! $ ! -$(,<br />
# # ) # !! ) " <br />
$#!#*! #<br />
#! 4 -$( *) !*<br />
%(-1& ! # " ! " ' <br />
!# ) !/-$((-120""<br />
0!"<br />
<br />
D 42 EB+ 7 !## ,<br />
#!!"##!,<br />
''142EB+)*,*<br />
!! !! # $ # <br />
!!"##)# <br />
*,* ' ! # # "# ?" <br />
# # " !! # " #<br />
<br />
<br />
0##" !<br />
# !! '! <br />
! 3 # !" #? 7 <br />
!!!42EB+# "#<br />
<br />
<br />
<br />
$0#))"$1## <br />
2(#%12(&2!")<br />
)0#$0#" "!CCB*<br />
<br />
7 0!# !" # # <br />
!)8<br />
• ()9!!<br />
• () # # %! ! 3 ) <br />
!!!)&<br />
• 7!<br />
• @(<br />
<br />
$%&&<br />
<br />
$ ! !! ) ) #!<br />
, )"
#!!!#<br />
<br />
7# ) ! !! <br />
!# 0# ! !! <br />
#!)$)0!! 4(-42CBBB<br />
$ " !!" ) # " #! )<br />
0#<br />
<br />
#'(<br />
<br />
-#?#12(0#" !*2<br />
("*#3%*2("*<br />
!!!)"0!#&#<br />
* #* # " ) !" <br />
# " $ <br />
# 8<br />
%& )!" ! ! # ! % <br />
3!&<br />
%& !"##)#"%<br />
#) !)&<br />
$" "#3)<br />
$##"<br />
" #D 12(#!,!*<br />
" 12( ) 0 0# *<br />
()#(#7><br />
<br />
* () # '# %('& ,! !<br />
# # ) ! ! $<br />
## ) "!<br />
!)!$"#<br />
4(-CBB4(-CBBB,4222+>C4# #<br />
!"!#"!#<br />
" ## # !" <br />
)!$('#3#3 !")<br />
0 $##!"#)<br />
!#('!""!#<br />
!"!#)!!"!!<br />
<br />
$ ! !" 12( ) # " <br />
! " # "" # # <br />
#!#*!#!)<br />
! 0#-!#!!"<br />
#" ! #('!!"!###<br />
) ## # )<br />
#!!#!)##!#<br />
#!!##<br />
<br />
")#&*%<br />
<br />
'"#*)/ !! 4<br />
# ) 7 '! $ %7'$& # ! ! <br />
#!#7'$#/!"<br />
!!*#
4)<br />
##!)#!3#<br />
5 # %! ) !& <br />
"#"<br />
7'$!! <br />
) 0# <br />
!<br />
<br />
+,$*<br />
<br />
- # ## 0 <br />
" ! ** () <br />
# ! ! !<br />
0 ) D !# <br />
) !!" " <br />
! ) 0<br />
###<br />
<br />
$ # !## # " / <br />
## $ # ! <br />
"#"#/!<br />
'#!!)#<br />
"&%.*%#<br />
<br />
"! &&<br />
<br />
$ ! " !" # ) " " <br />
! "#!#!)!! <br />
0# @)" ! "* ) "<br />
12(,!!<br />
)/##3#!# 74<br />
"##! ##<br />
<br />
<br />
78 (! # ! ! ## # %$<br />
#!! )"#<br />
##"#0<br />
#!&<br />
<br />
$ # !! " #<br />
# !!"!$<br />
0# !!" ! # ) 0" # # #!<br />
'! #)/ "# )<br />
" ! # #? #" ) #<br />
*#* !!" #! !!" <br />
"!<br />
<br />
(!#!" " )/!<br />
/""$#?!!"#12(<br />
"!#! !$!)"!#<br />
"" " !! 0# ! (!!* )<br />
"!#!#!!<br />
<br />
F
#<br />
D # #!!#<br />
0# 3 ) ) ! !" !<br />
" 7 # #!3 # ) "!# , )<br />
"!#-$(#!! #<br />
0#! (!#""!<br />
)*#!*!!!!"#!<br />
! " # # <br />
!!##!#!) ,#<br />
<br />
4 ! ! #<br />
#!!0#""#<br />
" ! ! '" # #)/ <br />
# / 4 5#<br />
- %45-& " # " ) !# <br />
12(#!<br />
<br />
<br />
5?!""!#12(<br />
! ! # !) <br />
!$##12()"<br />
0# # 0 # "! / # $<br />
!#### )/<br />
! # 5 " # <br />
@54!$0#" <br />
"!!0#4#!!!<br />
#! =4;(#"#"=<br />
471)3#<br />
$")!)#<br />
!!!!)* 0#! "!<br />
#!)0#<br />
<br />
"/ #!# <br />
<br />
# #""#!)!<br />
# !! 7F $ #! ) # <br />
#! ) <br />
!""#0!#()"#<br />
"#!/)#!3)"<br />
# 4 # # ) <br />
!" 0<br />
<br />
<br />
7F8AKAK##))<br />
( ! ! ,. '56(3 + <br />
FF
(#!/##8<br />
• 4" #! # "! # ) # <br />
0# # " 3" 3!" <br />
"!#0 !!<br />
""0#!<br />
• 4!!# ! #)<br />
!/ )/3!")/<br />
##" #<br />
• D ! # "" #! #!# "<br />
0 #!3 # <br />
"!#<br />
• $ # # !) ? <br />
!! 0#<br />
• $##!)#0#<br />
! # <br />
• $#! "#! ))#<br />
<br />
<br />
""*01#%<br />
<br />
$"!"# #!"#!<br />
"3!)#! !#<br />
4 " #? !" 3# " # <br />
" ! 5 !! 0 <br />
0<br />
<br />
D 4(-42CBBB*#!#*0<br />
"!!*!/!#<br />
$ 0# " !" ! <br />
"!# #!3 # # ! <br />
!!!#(!! 45-)<br />
#! <br />
<br />
"+ %2<br />
<br />
5#)/! )"#<br />
/#!-$((-1'#-$(" <br />
) D , 3 # 0 ) # <br />
#!# 1# <br />
) #!"# #!!"<br />
/ ! # / "! # )<br />
0#<br />
<br />
+ #&<br />
<br />
+0& #<br />
<br />
$4'(%4'(&!!<br />
0#12( ###0### )<br />
!! 3#!3#4!<br />
#"!30#)42<br />
## %42& " !" " ) !! $ 4<br />
5#%45-&!#12( <br />
<br />
FE
0### "#!!<br />
+#&&$* <br />
<br />
$ 2! %2& '$-5-( !? %'" $ -!#<br />
5# -! (& # #!# 2* ##<br />
$!1#"!# <br />
#!"# 0 ,# " " 0 ,<br />
! ! #!" " <br />
!! $ !? ! " # # # <br />
*-"!###!?<br />
) !! "!# # 12( #<br />
!!,!$!!!"#<br />
#/#512(# !"<br />
#3$ 444)!!$)!!"<br />
"! ,. '56(31 , 4 & <br />
4+ <br />
<br />
<br />
$ 4448 : !! "!# !## # <br />
#!!,.'56(3<br />
7,<br />
$12, # "/!34<br />
"## ##$)<br />
)12(!$) " )<br />
!! 12(#<br />
7 7,<br />
&$12( #! /#!"#<br />
>&4"12(#"<br />
&$12(!")#<br />
F&7 !!! )12(<br />
E&$12( !<br />
&$12(#!"#)<br />
!!"#<br />
6&12( !"<br />
+&$12( ! !!"!!<br />
<br />
C&$!12( !3!<br />
!<br />
B&$ )12( <br />
&$12(!!!!#!#<br />
7,<br />
$! #12(0#!!<br />
12($/0#!!)##!!!"<br />
)<br />
&' 12(" !##<br />
>&$0" !!!"<br />
&@#," #!<br />
F&A"" #!<br />
E&' !""""L5#(#<br />
&230### /<br />
6&( # !12(" !#"<br />
+&1)"!" #!<br />
!<br />
C&$12( <br />
<br />
F
$#'$-5-("!!!)##<br />
"!#! 12(! $"!#<br />
#" )# ,./&<br />
'82661 , 1 , 7 ,. ')3051 9 & :<br />
7 & ;'806 $ !! 0 / ""#<br />
#)!!)#/#<br />
#"#<br />
<br />
$!!12(!) #0!#<br />
3 # / ! ! "#<br />
7##!)" !<br />
!$!!012( )) )<br />
/ " " 12( "!# # <br />
## ,",.'56(3!<br />
" # ## # # *3 *<br />
# 12( ! 7E 3 * /<br />
0!#%))#&!"#)<br />
!*$#)/3/#<br />
#!!)<br />
<br />
<br />
7E8#!#,.'56(3<br />
<br />
D ! 3 " / ) <br />
!12($)!!!"!# )<br />
!"12() "! $"#0<br />
4(- 6+CF #)/ ## ) 12( / $! 3#! /<br />
##12()!!"##)<br />
/12(#!##)/! !12(<br />
!!!3""<br />
<br />
$#)/)#!!##/<br />
#!3##! #!)<br />
!#!#))!!!!<br />
!#'!!#!!!" <br />
)<br />
##"#""#!%<br />
###/&12(,.'56(3<br />
<br />
F6
-)<br />
<br />
-& !<br />
<br />
4 ! ) / # " 4(- 6+CF <br />
!"#"#!312(!!$<br />
) !" #)/ ! <br />
"!#4)0#)<br />
# / !! !! <br />
#)#<br />
<br />
-�<br />
<br />
D !! !! # ! ! ) <br />
/ #!!!3#!<br />
)##!! ,<br />
!! !! # / 0<br />
!!!!#*!*!7F<br />
!!!)( #""<br />
"!#'*)"!<br />
!
!<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
! <br />
<br />
<br />
<br />
<br />
"# $%#<br />
# $&'$($) <br />
&' ) % $<br />
&#'# # $ $) '<br />
#&* !$+ & ) !##&' & &'<br />
'$ $$$ && & <br />
& , $$$'#) <br />
&&'#&&''!<br />
$$ &#&##&#<br />
$$&<br />
<br />
& +$# '<br />
$-# '' #) <br />
$ ' &( <br />
$# & '#$$ $. $ <br />
# #&-$$$<br />
## $ $ & $ <br />
' / #$ #$ $<br />
& & 0 # # +<br />
# $&'<br />
& $ $ ' $ & # $<br />
<br />
<br />
! "<br />
<br />
0 $ 123$$&$ $'<br />
! $' ' $ $ ) <br />
4*56 '$&<br />
$ % $'' $& <br />
'& .7 #<br />
'$' '.4*5$&
) $ $ %$$9&:&<br />
' &&+ #& <br />
) <br />
<br />
*$&&$ &$$ <br />
4*5 #$$ $) '<br />
& & %$ <br />
# $<br />
&$ ##&'#' $<br />
. $ ) && <br />
& & $$<br />
'&;'&'$ <br />
$ '$ <br />
$$&<br />
<br />
%#'###
# && & <br />
<br />
,&'$&# $) <br />
$ $$9 #:#% <br />
$& ' $&#<br />
$$&;'&$<br />
) & $# &#<br />
, ' + !'$$ <br />
$7) $+#! &<br />
<br />
# & "!$<br />
<br />
$$ & + $<br />
$$$$$ &% & '<br />
$-$'$'<br />
& $ $# #$<br />
#9! !: &$+ 7$<br />
#' ' $$$$# &<br />
'! . & & $ & '! <br />
) + $$ $$ &<br />
'! %#$ $) & <br />
$#>+ <br />
<br />
#"& ) & & '( <br />
<br />
. $& ' $$# $<br />
$#$ #$ &+&#-<br />
$'$# $<br />
$$%$$!'6'<br />
$ #*$$$ &<br />
$ $'$$ <br />
& "#$%&'()7 & #$ <br />
$#$$ $$$<br />
#$7$$# <br />
#$$& '#$ + $<br />
<br />
## & & & & * <br />
<br />
*$ $$ #$$<br />
$ $$$) #!) #<br />
$$&# $; $$$& <br />
# $ $ $ $ $ = <br />
$' & $) *$<br />
&#$4*5$$$' $$' <br />
# $ '') $ <br />
' ### $ $ $$ <br />
# $ ' ! $# # $ #<br />
) $ $ $' $+<br />
'> 4*5' <br />
.&'$ #4'!'$4*5<br />
& ) # #,5#<br />
<br />
81
% #$ $) $$$$$ <br />
& ' '#$''<br />
;'& $&#$$&#$$ <br />
$$ '' # $'&$<br />
&%$& '$ <br />
$' $ # $$'<br />
&5 $# $ '#<br />
# 6 & $$ & <br />
) $'&$ $<br />
=$ $'#$$<br />
#$'&#'$<br />
$, $ #) # <br />
$!) $$ . ' <br />
$) '& ;'&<br />
&#$$ $$' &<br />
$ $$$' #<br />
$ &) #&$ ) $<br />
$ ) <br />
<br />
#) $& $$ &* <br />
& & +<br />
) & '!$# ) #<br />
' & $ && ' & $ <br />
$<br />
<br />
+ ! ! & ! , %<br />
<br />
%& $' & $) $<br />
&## $.<br />
& ' & $ ' &<br />
) '$ $ $-#<br />
& !$ $'!$$ $ $#<br />
?<br />
• % ##@' $<br />
• % #$+$ ) <br />
• %$# '! 9+ $.7<br />
:<br />
<br />
.'& $ $'& '+<br />
& ' $ .7 # & ! ,1 # <br />
=$ !'&# # # !#$<br />
! !$ ) #<br />
#'A$B$$#$ $<br />
,
,1?*;5!$5; ,<br />
<br />
,
$$$# &'$ ') #<br />
$$;&$ $ '&$<br />
$+) $$ #+<br />
! $'$' #<br />
<br />
-"! &% '( & "<br />
<br />
C &' '& ') ) #<br />
$# '#'&<br />
'' ?<br />
<br />
/ 5 $ ? % # $$ $ ) <br />
.7 # 4*5#$ &$<br />
$$'$$#- #$ <br />
$# <br />
<br />
.5 $,+#?"+ $ <br />
' # '# $ <br />
#+$ . $&<br />
# ) # &' # $<br />
%## # $ <br />
$ # #&#'$+$$<br />
$+#) <br />
<br />
$#?0 & ) $ '<br />
$+ ' ' $%$' <br />
$' $# &&$$+#<br />
# #' $$'!$'-# & <br />
$# &$!$ 5 $ #'<br />
$$) &$ <br />
'& $$$ 4 #<br />
+#'-$ $$ <br />
($D<br />
<br />
6 5) ?- #$$& @<br />
$ $''$$0 <br />
! $$$'#!'<br />
' ! $$# $$<br />
<br />
-'=%?0 #' &AB <br />
$#$$ %$# $<br />
'+'# # <br />
'$ #- +$<br />
$ $ -"57E,' $ $& $5%"6<br />
E/$'#<br />
<br />
8
./0 * (! ! "! /<br />
<br />
-+ $' &<br />
$$ % # $<br />
!#$;'&'<br />
$& ' ) $<br />
$) && $') <br />
#$ '#$#$ >*#<br />
$'!$' $$$'<br />
'$$ $$<br />
&$$$ $ $$'?<br />
<br />
• $ #$$! '$$<br />
$$ $<br />
• /& $ ' $<br />
• , $# $ $' +' <br />
$ &<br />
• % & ' # #$ <br />
#$$<br />
<br />
. & & & &<br />
<br />
- $!$ $'$<br />
$#$ $$$ '<br />
# $'!# # <br />
&$' &$ <br />
#$ $%' <br />
' ' & $ # # # $ <br />
$) #& &' <br />
$ #<br />
<br />
. &! &<br />
<br />
#$ $#$ <br />
#$$'- ) <br />
#$ $ +$ $<br />
! &# $$ ' ;& # $<br />
& $ <br />
# &#$#'$ $<br />
$ ' & # # 6 ' <br />
$'!$-& $<br />
$$$## #5 ) #$$<br />
&!' # '$$$ <br />
$<br />
<br />
.# & 1 * ! &<br />
<br />
% $$ $ !$) <br />
$$'$;'&##<br />
&#&$$ $!$$<br />
<br />
88
$$$ ! $.$'!$$&#<br />
$$& ''$$<br />
$) ##$#$$*#<br />
$ $ #$$ $) $<br />
$'#(#$#&<br />
$#$$ ' &!&<br />
$$<br />
<br />
2 & 3 <br />
<br />
%#'$# $ #<br />
+$ &$' $$# & ) -<br />
; * $+((+)'& $$ $ <br />
4*5$$ $'' -; <br />
$ & $'?<br />
<br />
2*(4<br />
<br />
.& & $ $ %<br />
&$$#$ &$'4*5$ #<br />
%$ &% &'<br />
$$$ $ $$ &<br />
% $'! # & $ # 9<br />
: <br />
<br />
<br />
,?1 &) <br />
$$G# #<br />
<br />
,8?,#$$ <br />
$ $<br />
<br />
<br />
2 % <br />
<br />
* &' & <br />
$# $ .$,F &''<br />
# &'&<br />
' &<br />
8F
: $<br />
: $<br />
9! ! :<br />
<br />
: <br />
$<br />
<br />
<br />
: <br />
<br />
<br />
: <br />
$+9$ <br />
#:<br />
,F? -; <br />
<br />
2# "& &! &<br />
<br />
.$$ $ <br />
$'##$ $ <br />
# $,H.$ $ <br />
#$ $$##0 <br />
& $'!&! $"+ $$#<br />
$$$ $& &'<br />
&<br />
<br />
%$ $ #$#' &<br />
0 $+ $ ' $#<br />
&#$#&+$$-<br />
'&&$&&' $<br />
& & =$ <br />
#&# #<br />
& # ' <br />
# # & $ # $+ & $ <br />
&!<br />
<br />
;# &$ $$$$'! <br />
%$'!$$ <br />
'9& : &'' #(' &<br />
$ $ #$'!&#$&<br />
<br />
8H
& $*-$) <br />
*&<br />
<br />
Parallel Deck<br />
LOA<br />
Parallel Deck Aft (PDA)<br />
Parallel Deck Fwd (PDF)<br />
Displacement (DISP, CP, CB, LCB)<br />
Parallel Middle Body Aft (PMBA)<br />
Parallel Middle Body Fwd (PMBF)<br />
Parallel Middle Body (PMB)<br />
LBP<br />
Breadth (BWL)<br />
Depth (D)<br />
Draught (T)<br />
<br />
,H?% # $ $<br />
<br />
2+1 & &<br />
<br />
$$ #$$<br />
6%918:$&$' $<br />
$ &$# $! #$<br />
'&$ #$$! # #$<br />
$,2. $+#$$<br />
$ $$%-; <br />
$ $# # <br />
#$'!<br />
<br />
<br />
:-$#&<br />
<br />
<br />
:$ <br />
#<br />
$# &<br />
<br />
:/& <br />
$<br />
<br />
,2?. $ #<br />
<br />
:; <br />
<br />
5) & &<br />
<br />
%$'$) > $ $$<br />
% $ + ' ' $ % #<br />
! &, '$<br />
# &,13% &' $$'$<br />
$ #,11%&+9$$:<br />
82
&& &, <br />
$$$ &'%' &$<br />
'# #+IJ>$' & $ <br />
% $$# <br />
$ &. $ $# $<br />
$$&$ #',# #$<br />
# #$$) & #,1<br />
<br />
<br />
,?5! &'<br />
<br />
<br />
,13?$ <br />
) + <br />
<br />
,11?. & #<br />
!9: & <br />
<br />
,1
$$ $&$$$<br />
$) ##<br />
'13K$*$3K<br />
% + $ # $ <br />
'$ &' $$$ 7<br />
$$ $' <br />
<br />
<br />
,18?/'$ <br />
$+ <br />
-; <br />
,1F?/56$* !<br />
$5$<br />
-; <br />
<br />
6 "<br />
<br />
6 & ( $'!<br />
'' $$ #<br />
%# ## '<br />
#''' &#' <br />
+ ! 96: $ ' $ -<br />
:-; 5 $ <br />
:*5.5/%# <br />
' <br />
<br />
<br />
:% <br />
: ' <br />
<br />
,1H?L $$$$6+ <br />
%-? $& && $<br />
$#' '<br />
Initial Design Change Beam Change LBP Change Parallel Change Cb<br />
Change LCB<br />
Middle Body<br />
Change 27m -> 28m 161.9m -> 171.9 13.3m -> 18.3m 0.62 -> 0.64 77.1m -> 75.1m<br />
Limit<strong>in</strong>g KG - Even Keel (m) 8.81 9.10 8.97 8.78 8.68 8.63<br />
Structural Weight (Tonnes) 3518 3607 3783 3534 3553 3505<br />
Effective Power @ 18 kts (MW) 39.38 40.29 42.95 39.89 40.32 38.97<br />
<br />
<br />
% + #'# & <br />
$ &#$# $ $'6<br />
$$& & "#57 <br />
9&# : , * *! # *# <br />
) #@$$& $ <br />
$) 9 #:# <br />
$ $) 9:5 ) # $' <br />
6 $$$# $ $'-<br />
<br />
F1
; & '$$ &,12 <br />
$ & * *!/# # <br />
$#-; - #$ <br />
$&<br />
<br />
<br />
<br />
,12? $ -; $% <br />
!$ ) '<br />
$ ' ' <br />
6 7 4<br />
<br />
: . & '! $ #<br />
# $ $<br />
!* &<br />
<br />
:5 $$9:<br />
<br />
<br />
: 5 $ 5 $ ; <br />
$ ' & <br />
$ ' <br />
# <br />
,1?5#$ #) & $-; ) <br />
<br />
0 -; $# +<br />
%'!#&<br />
& &'&#& #$$ <br />
$ $&$$$'' <br />
$# $!&!$$ $#$<br />
#5 $$# $$ $$<br />
) $$ #$ $$<br />
, ) #&#$ & $ #$<br />
#'$$ <br />
$$$,'$+ -; &'#+<br />
# 4*5 ' #<br />
'! ! <br />
%$$'#$ '$$<br />
$'!$ & & #+4*5 &<br />
F
0 ! & # $ <br />
&&& $) '$$ '!$ &<br />
'&+#$$. <br />
$$ + $ $#$<br />
$ #,1<br />
<br />
8 /4<br />
<br />
5 & $$'& ' (<br />
& &0 #<br />
$#&& $ $ #<br />
$$ %#&' $<br />
& $ # $ ) & #<br />
+# &&) '<br />
$9 #:$ # &$''$$&<br />
$ $$ . <br />
'& $##''<br />
$ ' #' $ <br />
#+$,#' # $<br />
&'$$$$ + $ $ <br />
$$$$!<br />
<br />
/&<br />
<br />
*="/9
Modell<strong>in</strong>g and Simulation of<br />
Shipboard Environment and Operations<br />
Jayanta Majumder, Dracos Vassalos, Univ. of Glasgow and Strathclyde, Glasgow/UK,<br />
j.majumder@strath.ac.uk, d.vassalos@na-me.ac.uk<br />
Luis Guar<strong>in</strong>, Guro C. Vassalos, Safety at Sea Ltd., Glasgow/UK,<br />
l.guar<strong>in</strong>@safety-at-sea.co.uk, guro.c.vassalos@na-me.ac.uk<br />
Abstract<br />
This paper presents HELIOS, a software tool and framework that <strong>in</strong>corporates simulation models<br />
of various aspects of shipboard operations and facilitates quick development of new models. The<br />
core of HELIOS is a spatial database represent<strong>in</strong>g hull, subdivisions and general arrangement.<br />
Numerical models and simulations of various aspects of shipboard operations are built around<br />
the spatial database. These models are controlled <strong>in</strong>teractively and visually as far as possible.<br />
The numerical models <strong>in</strong>tegrated so far <strong>in</strong>clude crowd flow, vehicle load<strong>in</strong>g, fire/smoke spread,<br />
spatial coverage, and hydrostatics.<br />
1. Introduction<br />
Simulation-based ship design progresses rapidly. Initially, simulations were classical mechanical<br />
models of hydrodynamics or structure analysis. More recently, computer based modell<strong>in</strong>g and<br />
simulation of people onboard the ship has opened up new opportunities. Most importantly,<br />
it allows evaluation of a ship’s layout with respect to ease of evacuation and emergency muster<strong>in</strong>g.<br />
Crowd-flow modell<strong>in</strong>g has been an active area of study for a long time, specially <strong>in</strong><br />
urban plann<strong>in</strong>g and design. Also, there have been some statutory guidel<strong>in</strong>es about manag<strong>in</strong>g<br />
civic crowd scenarios. More recently the concern has reached the maritime sector. A recent<br />
statutory requirement that every new Ro-Ro passenger ship must be evaluated for evacuation<br />
performance, has generated new waves of <strong>in</strong>terest <strong>in</strong> crowd dynamics <strong>in</strong> the maritime sector. In<br />
response to this regulation, several corporations and research organizations have come up with<br />
their own method and software for modell<strong>in</strong>g evacuation. Out own simulation tool HELIOS (Human<br />
Environment Local Interactions onboard Ships) took some ideas from our well-established<br />
evacuation simulation software EVI, http://www.shipevacuation.co.uk, and is meant to be an<br />
extensible framework to accommodate modell<strong>in</strong>g of various aspects of shipboard operations.<br />
Modell<strong>in</strong>g of shipboard occupants is not much different from modell<strong>in</strong>g of crowds <strong>in</strong> build<strong>in</strong>gs,<br />
by and large it is a sound assumption that behavioral model of people onboard ships is no<br />
different from that <strong>in</strong> a hotel build<strong>in</strong>g of similar size. However there are two aspects <strong>in</strong> which<br />
the shipboard emergency differs from a civic build<strong>in</strong>g scenario.<br />
– The ship motion and tilt can have additional effect on the people.<br />
– The usual command to the occupants <strong>in</strong> case of a shipboard emergency is not as simple<br />
as ’desert the compound’. It usually <strong>in</strong>volves dynamically decided muster<strong>in</strong>g based on the<br />
location and nature of the emergency.<br />
It may be tempt<strong>in</strong>g to equate crowd flow with fluid flow, but crowd dynamics is fundamentally<br />
different fluid dynamics, s<strong>in</strong>ce we deal with much smaller collections of people than would be<br />
required to cancel all <strong>in</strong>dividuality effects and be amenable to cont<strong>in</strong>uum formulations. People<br />
exercise significant free choice and do not follow conservation of momentum. This makes it<br />
impossible to make a sound macroscopic model of crowd dynamics on the l<strong>in</strong>es of fluid dynamics.<br />
Microscopic simulation is a more reasonable model for crowd dynamics because it can capture<br />
the complex behavior of a crowd as an emergent product of actions of its <strong>in</strong>dividual members.<br />
364
As a result most of the models <strong>in</strong> the literature are microscopic. The many microscopic models<br />
reported <strong>in</strong> the literature till date differ from each other <strong>in</strong> terms of the microscopic rules,<br />
dynamics and algorithms. In this situation it is quite natural to expect different simulators to<br />
differ <strong>in</strong> the details of crowd behavior. However, for a given scenario, the large-scale observations<br />
<strong>in</strong> all the simulators should be the same, or else the situation calls for a concerted standardization<br />
effort. There has not been any effort to harmonize the differences between different models<br />
ma<strong>in</strong>ly due to the expensive and commercial nature of the implementations and non-disclosure<br />
of the model details. One way forward <strong>in</strong> address<strong>in</strong>g this landscape of confus<strong>in</strong>g multitude of<br />
models is to put forward an open source implementation with fully published details, which can<br />
come <strong>in</strong> totality under public scrut<strong>in</strong>y and can grow freely with contribution from students and<br />
researchers worldwide. We have similar plans for HELIOS and a public licensed open version<br />
may be released eventually, but before that we need to to draw up the long-term road-map, and<br />
evaluate whether the project has the potential to be a susta<strong>in</strong>able open-source project.<br />
2. Def<strong>in</strong><strong>in</strong>g the Requirements of the Shipboard Environment Model<br />
HELIOS was developed to help <strong>in</strong> the problems of shipboard operations plann<strong>in</strong>g and layout<br />
design. The first problem was to def<strong>in</strong>e a faithful representation of the shipboard environment.<br />
A 3D virtual reality model of the ship may be a good visual representation, but conta<strong>in</strong>s just<br />
graphic primitives. For a simulation, a model must have some sort of mean<strong>in</strong>g (semantic <strong>in</strong>formation)<br />
associated with geometric entities. E.g., it should be possible to mark a region and give<br />
it a name and later refer to it by that name, or to mark a po<strong>in</strong>t <strong>in</strong> the environment and enquire<br />
about the occupancy or volume or material composition of the compartment that conta<strong>in</strong>s the<br />
po<strong>in</strong>t. The entire environment is a large entity, and the occupants (people, cars etc.) are much<br />
smaller. The locality of each occupant is a small part of the entire ship, yet these small parts<br />
must be bound together by some topological relationship so that a plan to go from location of<br />
the ship to another may be seen as a long sequence of small transitions from one small locality to<br />
another. It should also be possible to distribute values over parts or whole of the environment,<br />
e.g., the data from a fire/smoke transport simulation program (e.g. CO, CO 2 ,O 2 , and toxicity<br />
distributions) should be possible to associate with the environment model <strong>in</strong> such a way that<br />
the local and <strong>in</strong>stantaneous value can be made to affect the occupants. Another aspect of the<br />
requirements of an environment model comes from the fact that the exist<strong>in</strong>g data of shipboard<br />
layout comes <strong>in</strong> the form of l<strong>in</strong>e draw<strong>in</strong>gs, i.e. general arrangement (GA) draw<strong>in</strong>gs. This does<br />
convey the <strong>in</strong>formation <strong>in</strong> a human <strong>in</strong>terpretable form, but it is not directly usable as a representation<br />
of the topology of the ship. The l<strong>in</strong>e draw<strong>in</strong>g is merely a set of l<strong>in</strong>es, spl<strong>in</strong>es and arcs.<br />
But we need <strong>in</strong> our computational representation the geometry, connectivity and the semantics<br />
of sub-regions of the shipboard environment. One way to create our requisite representation is<br />
to model it from scratch, but it would be all the more economical if it can be mechanically generated<br />
from the l<strong>in</strong>e draw<strong>in</strong>gs. Semantic <strong>in</strong>formation may be <strong>in</strong>cluded manually <strong>in</strong>to the draw<strong>in</strong>g<br />
<strong>in</strong> the form of special graphical symbols, but the process of creation of the environment model<br />
should be largely mechanized. The environment model should have a visual representation that<br />
looks like the actual ship allow<strong>in</strong>g display of simulation and user <strong>in</strong>teraction. Thus, the ideal<br />
solution appears to be a spatial database support<strong>in</strong>g object localization and local annotation.<br />
3. Generation of the HELIOS Spatial Database<br />
In HELIOS, the simulated objects are embedded <strong>in</strong> a spatial database. The spatial database<br />
is a data structure represent<strong>in</strong>g the shipboard layout and support<strong>in</strong>g efficient computation of<br />
queries like po<strong>in</strong>t localization, range search<strong>in</strong>g etc. It is a two-manifold composed of triangular<br />
faces. The GA draw<strong>in</strong>g goes through a process<strong>in</strong>g pipel<strong>in</strong>e, Fig.1. The GA draw<strong>in</strong>g<br />
is a piece of 2D vector graphics consist<strong>in</strong>g of l<strong>in</strong>e segments and algebraic curves like arcs,<br />
circles and spl<strong>in</strong>es, Fig.2. HELIOS accepts such draw<strong>in</strong>gs <strong>in</strong> the widely used DXF format,<br />
http://www.autodesk.com/techpubs/autocad/acad2000/dxf/. The first stage <strong>in</strong> the process<strong>in</strong>g<br />
pipel<strong>in</strong>e <strong>in</strong>volves replac<strong>in</strong>g the algebraic curves by piecewise l<strong>in</strong>ear approximations to convert<br />
365
the GA data to a set of l<strong>in</strong>e segments (we call this stage ’l<strong>in</strong>earization’).<br />
Draw<strong>in</strong>g<br />
L<strong>in</strong>earization<br />
Cross−Snapp<strong>in</strong>g<br />
&<br />
Duplicates removal<br />
Cluster<strong>in</strong>g <strong>in</strong>to levels<br />
Symbol recognition<br />
Staircase/Ramp<br />
generation<br />
Triangulation<br />
and Extrusion<br />
Fix<strong>in</strong>g degeneracies<br />
Recognition by<br />
shape analysis<br />
Spatial<br />
Database<br />
Fig.1: Process<strong>in</strong>g pipel<strong>in</strong>e used to prepare the spatial data from GA draw<strong>in</strong>gs<br />
Fig.2: General arrangement (GA) (left) and 3D spatial database generated from it(right)<br />
The next stage is cross snapp<strong>in</strong>g and removal of co<strong>in</strong>cident l<strong>in</strong>e segments. Cross snapp<strong>in</strong>g<br />
<strong>in</strong>volves merg<strong>in</strong>g of end-po<strong>in</strong>ts that are close enough. This is done efficiently by successive<br />
<strong>in</strong>sertion of end po<strong>in</strong>ts <strong>in</strong> a BSP tree, Berg et al. (2000). So far, there is no explicit <strong>in</strong>formation<br />
on which groups of l<strong>in</strong>e segments belong to the same deck. This <strong>in</strong>formation is generated by<br />
cluster<strong>in</strong>g the l<strong>in</strong>e segments with the follow<strong>in</strong>g algorithm:<br />
Data: A Set S of l<strong>in</strong>e segments<br />
Result: A set of disjo<strong>in</strong>t sets of l<strong>in</strong>e segments, each represent<strong>in</strong>g the cluster for a deck<br />
Let D be the set (of sets) with elements {s} for each s ∈ S;<br />
repeat<br />
n ←− ||D||;<br />
Create and <strong>in</strong>itialize a 2D segment tree T ;<br />
for each d ∈ D do<br />
end<br />
Create a slightly padded box b(d) bound<strong>in</strong>g all segments <strong>in</strong> d;<br />
Add b(d) to T<br />
for each d ∈ D do<br />
end<br />
for each δ ∈ D with b(δ) overlapp<strong>in</strong>g b(d) (us<strong>in</strong>g T ) do<br />
end<br />
UNITE (d, δ)<br />
D ← The new set of disjo<strong>in</strong>t sets of segments, as result<strong>in</strong>g from the UNITE operation;<br />
366
until n == ||D||;<br />
Return with result D;<br />
In this algorithm, the set of disjo<strong>in</strong>t sets of segments is ma<strong>in</strong>ta<strong>in</strong>ed us<strong>in</strong>g the UNION/FIND<br />
algorithm with path compression (see Ch.22 of Cormen et al. (1990)), and the UNITE operation<br />
is that of a standard disjo<strong>in</strong>t set union. In practice the Repeat. . .Until very few times - twice<br />
or so. After this process<strong>in</strong>g is complete we get groups of l<strong>in</strong>e segments each belong<strong>in</strong>g to one<br />
deck. As regards order<strong>in</strong>g the decks, we use a convention. Usually the decks <strong>in</strong> the GA draw<strong>in</strong>g<br />
are laid out <strong>in</strong> ascend<strong>in</strong>g order of height from bottom to top. If this is not the case <strong>in</strong> a GA,<br />
this has to be manually done.<br />
The next stage is extraction of symbols. The symbols are added to the GA draw<strong>in</strong>g to convey<br />
<strong>in</strong>formation about the layout. Different symbols have different mean<strong>in</strong>gs, the location and pose<br />
of the symbols are determ<strong>in</strong>ed and appended to the data pass<strong>in</strong>g through the process<strong>in</strong>g pipel<strong>in</strong>e.<br />
E.g., a ’reference po<strong>in</strong>t’ or ’orig<strong>in</strong> of coord<strong>in</strong>ate system’ for each deck tells that the decks are<br />
to be so aligned that the reference po<strong>in</strong>ts fall on the same vertical l<strong>in</strong>e <strong>in</strong> 3D. The symbols are<br />
themselves simple draw<strong>in</strong>g objects or entities available to the user as a library and meant to be<br />
copied and pasted at appropriate locations <strong>in</strong> the draw<strong>in</strong>g to convey the <strong>in</strong>formation. E.g., the<br />
symbol for a reference po<strong>in</strong>t has the appearence of . The symbol can be <strong>in</strong>serted <strong>in</strong>to the<br />
draw<strong>in</strong>g us<strong>in</strong>g any of the common draw<strong>in</strong>g editors like autocad, qcad, <strong>in</strong>tellicad etc.<br />
The recognition of the symbols is done us<strong>in</strong>g the ’aff<strong>in</strong>e <strong>in</strong>variant’ match<strong>in</strong>g, <strong>in</strong> which the symbol<br />
is compared with with parts of the draw<strong>in</strong>g <strong>in</strong> a way that is <strong>in</strong>dependent of scale, translation,<br />
rotation and mirror flipp<strong>in</strong>g as follows:<br />
Data: A symbol Σ as a set of l<strong>in</strong>e segments<br />
The relevant part of the draw<strong>in</strong>g given as a set of l<strong>in</strong>e segments, ∆<br />
Result: Location of the symbol <strong>in</strong> the draw<strong>in</strong>g;<br />
l s ←longest segment <strong>in</strong> Σ with end po<strong>in</strong>ts (x 0 , y 0 ) and (x 1 , y 1 );<br />
Create a 2D segment tree T of the bound<strong>in</strong>g boxes of each of the l<strong>in</strong>e segments <strong>in</strong> ∆;<br />
startpt[0] = (x 0 , y 0 ); endpt[0] = (x 1 , y 1 );<br />
startpt[1] = endpt[0]; endpt[1] = startpt[0];<br />
( )<br />
( )<br />
1 0<br />
1 0<br />
flip[0] ← ; flip[1] ←<br />
;<br />
0 1<br />
0 −1<br />
Create an empty set of matched locations M;<br />
for each δ ∈ ∆ do<br />
for each i ∈ {0, 1} do;<br />
for each j ∈ {0, 1} do;<br />
p 0 = (X 0 , Y 0 ) and p 1 = (X 1 , Y 1 );<br />
p 0 = (X 0 , Y 0 ) and p 1 = (X 1 , Y 1 );<br />
Get an Euclidean operator E that transforms (flip[i] × startpt[j]) to p 0<br />
and (flip[i] × endpt[j]) to p 1 ;<br />
Transform a copy of the symbol and its bound<strong>in</strong>g box first by flip[i] and then by E,<br />
let the set of segments <strong>in</strong> the transformed symbol be Σ δ ;<br />
Get all the l<strong>in</strong>e segments <strong>in</strong> ∆ that fall with<strong>in</strong> bound<strong>in</strong>g box of Σ δ call it ∆ δ ;<br />
Boolean match ← TRUE;<br />
for each σ ∈ Σ δ do<br />
end<br />
Look for a match of σ <strong>in</strong> ∆ δ , if not found set match ← FALSE<br />
367
end<br />
end<br />
end<br />
if match is TRUE then<br />
end<br />
Add the tuple (i, j, E) to the matched locations set M;<br />
The step ’Get an ⎛ Euclidean operator ⎞ E . . .’ may be expanded <strong>in</strong>to the follow<strong>in</strong>g. The operator E<br />
C −S T ( ) ( ) ( )<br />
x<br />
⎜<br />
⎟<br />
xa xb Xa<br />
is of the form ⎝ S C T y ⎠, which transforms two po<strong>in</strong>ts and to<br />
y<br />
0 0 1<br />
a y b X a<br />
( )<br />
Xb<br />
and respectively by rotation, isotropic scal<strong>in</strong>g, and translation. This leads the equations<br />
Y b<br />
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞<br />
x a X a<br />
x b X b<br />
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟<br />
E × ⎝ y a ⎠ = ⎝ Y a ⎠ and E × ⎝ y b ⎠ = ⎝ Y b ⎠, to be solved to determ<strong>in</strong>e the elements<br />
1 1<br />
1 1<br />
of E. The symbols are restricted to specially designated layers of the draw<strong>in</strong>g to reduce the<br />
computational cost.<br />
The next stage <strong>in</strong> the process<strong>in</strong>g pipel<strong>in</strong>e reconstructs the geometry of staircases <strong>in</strong> the environment.<br />
A general arrangement draw<strong>in</strong>g or a build<strong>in</strong>g floor-plan is the plan view of the layout,<br />
and the plan view does not conta<strong>in</strong> the aspects of the geometry that would appear on a lateral<br />
projection. Thus much of the description of a staircase or a ramp between two levels is absent <strong>in</strong><br />
a s<strong>in</strong>gle planar projection. To reconstruct this <strong>in</strong>formation, <strong>in</strong>stead of seek<strong>in</strong>g what the geometry<br />
actually is, we would f<strong>in</strong>d out what the best geometry should be, and hope that the designer<br />
must have designed it reasonably close to the best geometry. We must formulate as to what<br />
constitutes the quality of a staircase or ramp. A fairly reasonable and simple answer is that the<br />
least steep is best. If we m<strong>in</strong>imize an aggregated value of steepness of the staircase, it should lead<br />
to a reasonable reconstruction. We optimize a steepness functional over the projected doma<strong>in</strong><br />
of the staircase us<strong>in</strong>g the f<strong>in</strong>ite element method. If the vertical coord<strong>in</strong>ate over the projected<br />
doma<strong>in</strong> of the staircase or ramp is given as a function z(x, y), we use the steepness functional<br />
∫ ∫ (<br />
∂z<br />
∂x )2 + ( ∂z<br />
∂y )2 dxdy. The <strong>in</strong>tegrand is the square of the greatest slope at a po<strong>in</strong>t (x, y), and<br />
an area weighted sum of such local greatest slopes is what we m<strong>in</strong>imize us<strong>in</strong>g the Ritz method,<br />
Farlow (1993). We generate a mesh over the doma<strong>in</strong> and describe the distribution of z over the<br />
projected doma<strong>in</strong> as a function of <strong>in</strong>dividual z values at the mesh nodes, and make a discretized<br />
formulation of the steepness functional. For the value of the functional to be m<strong>in</strong>imum, the<br />
derivatives of discretized steepness with respect to z values at mesh nodes must vanish. Some<br />
z values (where the stair-sheet meets the flat levels) provide additional equations, similar to<br />
boundary conditions <strong>in</strong> physical problems. We reconstruct the staircase geometry by solv<strong>in</strong>g<br />
these equations.<br />
Fig.3: Generation of a spiral staircase<br />
Fig.3 illustrates the generation of the geometry of a spiral staircase.<br />
The center figure<br />
368
shows a mesh with z function displayed by a color scale. Fig.4 shows two more examples<br />
of reconstructions made us<strong>in</strong>g the variational method described. M<strong>in</strong>imiz<strong>in</strong>g our <strong>in</strong>tegral<br />
∫ ∫ (<br />
∂z<br />
∂x )2 + ( ∂z<br />
∂y )2 dx dy is equivalent to ensur<strong>in</strong>g that the divergence of the gradient ( ∂z<br />
∂x , ∂z<br />
∂y )<br />
is zero everywhere <strong>in</strong> the <strong>in</strong>terior of the doma<strong>in</strong>, i.e. there is no unnecessary <strong>in</strong>ternal source of<br />
slope. Staircase steps can be generated by quantiz<strong>in</strong>g the surface at iso-z contours, but we have<br />
left that for a later release of HELIOS.<br />
Fig.4: Generation of staircases<br />
The data at this stage of process<strong>in</strong>g represents a build<strong>in</strong>g or ship with multiple horizontally laid<br />
out levels, but the <strong>in</strong>ternals of each level consist merely of l<strong>in</strong>e segments (stand<strong>in</strong>g for walls and<br />
boundaries of obstacles). The whole environment needs to be seen as a connected network of<br />
subregions, primarily for the purpose of route plann<strong>in</strong>g and reason<strong>in</strong>g about the space. There<br />
are several ways of gett<strong>in</strong>g a topological structure suitable for route plann<strong>in</strong>g from a set of<br />
l<strong>in</strong>e segments and polygons, Fig.5 shows some Berg et al. (2000). The structured grid cell<br />
partition - the last two <strong>in</strong> figure 5, although commonly used, is a big compromise s<strong>in</strong>ce it entails<br />
anisotropy (i.e. differences <strong>in</strong> speed along different directions) of motion. Of the rema<strong>in</strong><strong>in</strong>g,<br />
we chose constra<strong>in</strong>ed triangulation for HELIOS because it enables efficient solution to three<br />
most important problems relevant to our model: route plann<strong>in</strong>g, po<strong>in</strong>t location, and visibility<br />
predicate.<br />
Fig.5: Network and partition schemes for location and navigation <strong>in</strong> the presence of three obstacles<br />
shown as solid black polygons; from top left to bottom right: trapezoidal map, vertex<br />
visibility graph, convex connectivity graph, constra<strong>in</strong>ed triangulation, Voronoi diagram, structured<br />
grid-cell graph (squares and hexagonal version).<br />
This stage of process<strong>in</strong>g computes a ’Constra<strong>in</strong>ed Delaunay Triangulation’, Chew (1987), of the<br />
l<strong>in</strong>e segments on each level. This essentially is a partition of each deck <strong>in</strong>to triangular pieces<br />
of floor space such that the orig<strong>in</strong>al l<strong>in</strong>e segments only form some of the sides of the triangles<br />
and not <strong>in</strong>trude the <strong>in</strong>terior of any of them. The triangles on the floor area of Fig.7(B) show<br />
an example of such a partition. The extrusion stage, which immediately follows triangulation,<br />
369
merely assigns a height value to each of the walls. Constra<strong>in</strong>ed triangulation of an arbitrary<br />
set of l<strong>in</strong>es can lead to degenerate triangles that need to be excluded for the sake of robustness<br />
and correctness of subsequent computations. Here degeneracies are overly sk<strong>in</strong>ny triangles that<br />
topologically represent a connection but geometrically the connection is <strong>in</strong>feasible. Fig.6 (bold<br />
l<strong>in</strong>es are walls, dashed l<strong>in</strong>es open sides), the triangle ABC illustrates a class of degeneracy<br />
that appears <strong>in</strong> a constra<strong>in</strong>ed Delaunay triangulation. However all sk<strong>in</strong>ny triangles are not<br />
topologically degenerate as they may as well represent a significant connection, e.g. the triangle<br />
PQR <strong>in</strong> Fig.6. As degeneracies are excluded, new degeneracies become detectable requir<strong>in</strong>g<br />
iterations.<br />
R<br />
A<br />
C<br />
B A<br />
C B<br />
P Q<br />
Fig.6: Illustrat<strong>in</strong>g degeneracy of triangles<br />
(A)<br />
Fig.7: (A) Compar<strong>in</strong>g paths: based on adjacent cell sequence (zigzag), and based on furthest<br />
visible way po<strong>in</strong>t (3 segments) (B) Simulated people f<strong>in</strong>d<strong>in</strong>g their way on triangle mesh.<br />
The environment model so generated is ready as a spatial database to support annotation and<br />
simulation. However, a few more operations are supported for recogniz<strong>in</strong>g shapes of regions.<br />
Regions have shapes and shapes have have certa<strong>in</strong> signatures that helps group<strong>in</strong>g similar shaped<br />
regions together. This is done by statistical cluster<strong>in</strong>g <strong>in</strong> the feature space of the shape/size<br />
signatures. This is important <strong>in</strong> the context of ships to automatically identify cab<strong>in</strong>s and public<br />
spaces.<br />
4. Modell<strong>in</strong>g of Human Occupants<br />
Modell<strong>in</strong>g the mechanisms of human response to extreme stimuli like evacuation alarm onboard a<br />
ship appears to be a formidable task, if at all possible <strong>in</strong> the framework of our present scientific<br />
understand<strong>in</strong>g. However, a behavioristic model may be built relatively easily. We take that<br />
pragmatic approach. Rather than try<strong>in</strong>g to address the complexities of human physiology and<br />
psychology, we treat them as computational objects with some built-<strong>in</strong> behavior and capability<br />
of be<strong>in</strong>g <strong>in</strong>structed and programmed to move around <strong>in</strong> the ship environment. These are heavily<br />
attributed entities, with attribute values decid<strong>in</strong>g their behavior. E.g., each one has age and<br />
gender attributes. There is some data about dependence of walk<strong>in</strong>g speed on age <strong>in</strong>corporated<br />
<strong>in</strong> HELIOS. Yet when the speed distributions are explicitly specified, this specification overrides<br />
the values deduced us<strong>in</strong>g the built-<strong>in</strong> relationships. To model a shipboard emergency operation,<br />
the modelled humans (’agents’) may be <strong>in</strong>structed or assigned a task of the follow<strong>in</strong>g types:<br />
(B)<br />
– Go<strong>in</strong>g to an <strong>in</strong>dividually designated location.<br />
– Given a set of locations, go<strong>in</strong>g to a convenient one of them.<br />
370
A Wall<br />
– Go<strong>in</strong>g to one or more locations successively follow<strong>in</strong>g a convenient route.<br />
– Follow an explicitly specified path.<br />
– Stay<strong>in</strong>g at one po<strong>in</strong>t for a specified period of time.<br />
– Carry<strong>in</strong>g out a specified action with<strong>in</strong> a specified time of an event.<br />
– Carry<strong>in</strong>g out a specified action on another agent when visible and nearby.<br />
– Carry<strong>in</strong>g out an action when <strong>in</strong> a particular location.<br />
– Execute a compound agenda with several tasks of the above assigned with various priorities.<br />
These task types give immense modell<strong>in</strong>g power. The computational operations performed<br />
by an agent may be seen as subdivided <strong>in</strong>to perception, decision, and action. Perception<br />
<strong>in</strong>volves collect<strong>in</strong>g local (<strong>in</strong> space and time) <strong>in</strong>formation as may be relevant, decision <strong>in</strong>volves<br />
figur<strong>in</strong>g out the th<strong>in</strong>gs to do (eg. f<strong>in</strong>d<strong>in</strong>g out the direction to head for etc.), and action <strong>in</strong>volves<br />
mak<strong>in</strong>g the changes as may be relevant (eg. stepp<strong>in</strong>g to the next position). This dist<strong>in</strong>ction<br />
<strong>in</strong> HELIOS really aims at organiz<strong>in</strong>g the program structure and does not relate to modell<strong>in</strong>g<br />
<strong>in</strong>ner mechanisms of human m<strong>in</strong>d. The most noteworthy aspects of procedural implementation<br />
of agent behavior is navigation. Tasks that are def<strong>in</strong>ed <strong>in</strong> terms of a dest<strong>in</strong>ation location need<br />
to be executed by f<strong>in</strong>d<strong>in</strong>g the way to the dest<strong>in</strong>ation and follow<strong>in</strong>g a route that leads to the<br />
dest<strong>in</strong>ation. The computational procedure for this is named navigation. The navigation procedure<br />
<strong>in</strong>volves two ma<strong>in</strong> sub-procedures - topological plann<strong>in</strong>g and geometric plan-follow<strong>in</strong>g.<br />
Topological plann<strong>in</strong>g <strong>in</strong>volves creation of a path based on connectivity of the triangular cells,<br />
<strong>in</strong> a loose sense the path is a sequence of triangular cells of the triangulation that need to be<br />
passed through to reach the dest<strong>in</strong>ation. This path is computed by process<strong>in</strong>g the connectivity<br />
structure of the triangular cells. Let us consider the dual of the triangulation, which is a graph<br />
<strong>in</strong> which there is a node for every cell. For every pair of adjacent cells that share a non-wall<br />
side, there is an edge or arc between their nodes correspond<strong>in</strong>g to the shared non-wall side.<br />
Fig.8 shows an example of the dual graph of a triangulation <strong>in</strong> which only one connected<br />
component of the graph is shown. The paths followed by agents are based on graph paths<br />
on this navigational network. A path plan is more than just one path, it is a bunch of paths<br />
clubbed together <strong>in</strong> such a way that for any cell it gives a path from the cell to the dest<strong>in</strong>ation cell.<br />
To FVW<br />
007<br />
Fig.8: Navigational Network<br />
Fig.9: Free configuration space<br />
It would be unrealistic model if the agents actually followed the zigzag paths <strong>in</strong> Fig.8. This is<br />
addressed <strong>in</strong> the plan follow<strong>in</strong>g sub-procedure. A basic premise <strong>in</strong> the plan follow<strong>in</strong>g procedure<br />
is that an agent always knows the cell that it is <strong>in</strong>. On a triangulation-based environment,<br />
this <strong>in</strong>formation is much easier to ma<strong>in</strong>ta<strong>in</strong> and recompute than <strong>in</strong> a convex segments based<br />
environment like Vassalos et al. (2001). Know<strong>in</strong>g its current cell, the agent can immediately<br />
retrieve a sequential path to the dest<strong>in</strong>ation cell. Each element <strong>in</strong> the sequence of cells (path),<br />
is a ’way-po<strong>in</strong>t’ lead<strong>in</strong>g to the dest<strong>in</strong>ation. However, agents cannot go through walls or floors.<br />
371
An agent will then head for the furthest way-po<strong>in</strong>t accessible by a straight l<strong>in</strong>e, i.e. the furthest<br />
visible way-po<strong>in</strong>t (FVW). Fig.7(A) shows an example of a path go<strong>in</strong>g through successive furthest<br />
visible way-po<strong>in</strong>ts. As the agent moves on, new way po<strong>in</strong>ts may become visible, and the direction<br />
of approach might change. Observ<strong>in</strong>g human navigational behavior suggests that most people<br />
don’t re-evaluate their direction of approach very frequently. This is <strong>in</strong>corporated <strong>in</strong> HELIOS<br />
by some parameterized and randomized deferment of re-evaluation. The size of the navigational<br />
network can have possibly millions of nodes. So it is important to use fast algorithms for<br />
plann<strong>in</strong>g paths over them. The fact that the navigational graph is planar makes a hierarchical<br />
s<strong>in</strong>gle source shortest path algorithm computable <strong>in</strong> l<strong>in</strong>ear time, Kle<strong>in</strong> et al. (1994). Also<br />
s<strong>in</strong>ce the <strong>in</strong>ter-cell distance values may be reasonably quantized and scaled to range with<strong>in</strong> a<br />
small range of <strong>in</strong>tegers, the bucket based priority queue approach to Dijkstra algorithm, Thorup<br />
(1999), makes the path computable <strong>in</strong> l<strong>in</strong>ear time. We have tried a recent implementation,<br />
http://www.avglab.com/andrew/soft.html, of this idea and it measured adequate. We also<br />
often use the BFS tree, Cormen et al. (1990), as a path plan, as it is easy to compute and<br />
results <strong>in</strong> acceptable paths <strong>in</strong> most cases.<br />
Plan follow<strong>in</strong>g gives the agent a reasonably nice geometric path that leads to the dest<strong>in</strong>ation,<br />
respect<strong>in</strong>g walls, but not yet the bodies of other agents. We call the procedural element that<br />
ensures this ’dynamic avoidance’. In dynamic avoidance the agent tries out a number of directions<br />
centered around the direction towards the FVW, and chooses the direction that makes<br />
the most advance along the desired direction without violat<strong>in</strong>g the extents of any other fellow<br />
agent. The dynamic avoidance primarily <strong>in</strong>volves comput<strong>in</strong>g the maximum component of displacement<br />
along a particular direction (which usually is the direction towards the FVW) that<br />
can can be achieved by a displacement <strong>in</strong> any direction. The extent of the neighbor<strong>in</strong>g agents<br />
is modelled by a rectangle - which is equivalent to four l<strong>in</strong>e segments. Walls and obstructions<br />
too need to be avoided. When try<strong>in</strong>g to compute dynamic avoidance for an agent (e.g. agent<br />
007 <strong>in</strong> Fig. 9, where the agents are be<strong>in</strong>g shown as crossed squares), we consider the extents<br />
of other agents or walls so offset from their actual geometry that the region outside such offset<br />
regions is free for the center of 007 to move. This free region is the ’free configuration space’<br />
for agent 007, shown as white <strong>in</strong> Fig. 9. The boundary of the free configuration space consists<br />
of l<strong>in</strong>e segments and a circle. It can be shown that the optimal displacement is on one of the<br />
<strong>in</strong>tersection po<strong>in</strong>ts of the configuration space boundary. The proof of this fact is similar to that<br />
of the fundamental theorem of l<strong>in</strong>ear programm<strong>in</strong>g, with a m<strong>in</strong>or addition due to the presence<br />
of a circle. Us<strong>in</strong>g this observation, scann<strong>in</strong>g through the <strong>in</strong>tersection po<strong>in</strong>ts on the boundary<br />
of the free configuration space of an agent is used <strong>in</strong> f<strong>in</strong>d<strong>in</strong>g the optimal displacement. This is<br />
computed by transformation to a canonical position and orientation, for efficiency’s sake.<br />
For the visual representation of the simulation to be conv<strong>in</strong>c<strong>in</strong>g, it is important to animate<br />
the motion realistically. To display the walk<strong>in</strong>g of agents, a skeletal animation<br />
is used. Human gait data was extracted from the publicly available implementation <strong>in</strong><br />
http://www.mathworks.com/moler/chapters.html. In addition, we have recently <strong>in</strong>corporated<br />
a technology used for game development, <strong>in</strong> which the animation can be designed us<strong>in</strong>g a class<br />
of software tools called animation studios (e.g. 3D-studio MAX). This will enable us to <strong>in</strong>clude<br />
more types of agent postures and movements <strong>in</strong> HELIOS than just stand<strong>in</strong>g, walk<strong>in</strong>g and runn<strong>in</strong>g.<br />
Quality of the graphic objects, which is directly related to the number of polygon faces used,<br />
costs process<strong>in</strong>g time. Naturally this leads to a limit to the best quality of agent graphic<br />
that may be displayed given the contemporary computer performance. However, there is some<br />
scope for tricks to stretch this limit. One example is avoid<strong>in</strong>g draw<strong>in</strong>g of agents that are<br />
not visible, another is to draw the agents <strong>in</strong> low quality when they are far away from the<br />
camera. The lalter method is called ’level of detail’ (LOD), Luebcke et al. (2002). There<br />
are techniques for dynamic computation of LOD variations. To keep th<strong>in</strong>gs simple, we have<br />
372
used <strong>in</strong> HELIOS discrete LOD, <strong>in</strong> which the different LODs are fully pre-computed graphic<br />
objects (<strong>in</strong> the form of OpenGL display lists), that are switched between as the camera<br />
moves nearer or further. The apparent dimness near the top right corner of the screenshot<br />
of Fig.10 (left) is actually due to switch<strong>in</strong>g to a lower level of detail for distant agents.<br />
Fig.10: Some glimpses of evacuation/muster<strong>in</strong>g simulation<br />
5. Models of Other Shipboard Operations<br />
The spatial database <strong>in</strong> HELIOS is a data structure support<strong>in</strong>g the queries and operations<br />
required by applications that have objects requir<strong>in</strong>g localization, local annotation, and geometry<br />
based calculations. It serves as a support framework, and eases development of applications<br />
around it. We have developed a few such applications and are <strong>in</strong> the process of develop<strong>in</strong>g more<br />
of them. We describe a few of them <strong>in</strong> the follow<strong>in</strong>g subsections.<br />
One of the applications is an <strong>in</strong>put model generator for a fire/smoke transport simulator called<br />
Raeume, Shigunov (2004). The <strong>in</strong>put model requires <strong>in</strong>formation on compartments, their<br />
connectivity, material composition, ventilation details, fire location etc. It was quite easy add<br />
some annotation functions <strong>in</strong>to HELIOS, to enable the user to annotate the 3D geometry and<br />
generate the <strong>in</strong>put for Raeume. As a side effect, this ensured that fire/smoke simulation results<br />
from Raeume could be easily imported <strong>in</strong>to HELIOS. This, <strong>in</strong> conjunction with a model of effect<br />
of heat and smoke on people, when <strong>in</strong>corporated <strong>in</strong>to HELIOS agents, gave rise to the ability<br />
of simulat<strong>in</strong>g complex scenarios of evacuation react<strong>in</strong>g to a fire/smoke spread. A study of this<br />
capability has been described <strong>in</strong> Guar<strong>in</strong> (<strong>2005</strong>). Figs.10 and 11 illustrates the functionality of<br />
simulat<strong>in</strong>g people <strong>in</strong> conjunction with fire and smoke.<br />
Fig.11: Heat/smoke affect<strong>in</strong>g people<br />
Fig.12: Load<strong>in</strong>g of cars <strong>in</strong>to ferry<br />
As an extension to this direction of development, we envision that with maturation of our fire<br />
model, it would be possible to not just react to the fire simulation data but <strong>in</strong>teract with it (e.g.<br />
simulat<strong>in</strong>g our agents fire-fight<strong>in</strong>g). Another model that would be easy to <strong>in</strong>corporate but add<br />
373
great value is that of motion sickness. The motion sickness model itself is on the anvil <strong>in</strong> our<br />
research center, and once completed, it may be <strong>in</strong>tegrated <strong>in</strong>to HELIOS to achieve greater levels<br />
of <strong>in</strong>teractivity and ease of use. It was easy to <strong>in</strong>corporate mobile entities <strong>in</strong>teract<strong>in</strong>g with the<br />
environment. We did an application on load<strong>in</strong>g vehicles <strong>in</strong>to ships. It is a simple extension <strong>in</strong><br />
which users can create vehicles at any location <strong>in</strong> the environment, then specify their paths and<br />
watch them follow their paths. Fig.12 shows a screenshot of this feature. Frivolous as it may<br />
seem, this application can be useful <strong>in</strong> design<strong>in</strong>g facilities for load<strong>in</strong>g and unload<strong>in</strong>g of cars <strong>in</strong><br />
ferries. It has actually been used <strong>in</strong> an undergraduate thesis <strong>in</strong> our university to study design<br />
alternatives for the Ardrossan car ferry term<strong>in</strong>al, Scotland.<br />
An application on hydrostatics calculation has been developed that uses the geometric data subsumed<br />
by HELIOS. It consists of a module that imports the hull and the subdivisions exported<br />
by programs like NAPA. The essence of its work<strong>in</strong>g <strong>in</strong>volves calculation of sectional properties<br />
from the hull and computation of volumes of the compartments. The former is done by an added<br />
functionality and the later was ord<strong>in</strong>arily supported by the geometric data structures <strong>in</strong> HELIOS.<br />
A case based reason<strong>in</strong>g program has been plugged <strong>in</strong>to this setup, that makes counter-flood<strong>in</strong>g<br />
suggestions when a flood<strong>in</strong>g scenario is presented to it. This case based reason<strong>in</strong>g program was<br />
developed as a deliverable to an European Union funded project titled COMAND. Fig.13 shows<br />
some screenshots of the counter-flood<strong>in</strong>g advice function developed us<strong>in</strong>g HELIOS. This feature<br />
of HELIOS makes it useful as an onboard load<strong>in</strong>g analysis program.<br />
Fig.13: The hydrostatics functions of HELIOS<br />
Visibility coverage as def<strong>in</strong>ed <strong>in</strong> the context of HELIOS, is the extent or part of the environment<br />
that can be swept by a straight ray rotat<strong>in</strong>g about a po<strong>in</strong>t but is obstructed by an obstacle.<br />
Intuitively, it is the area that can be guarded by a person stand<strong>in</strong>g at a po<strong>in</strong>t but free to look<br />
around. One can def<strong>in</strong>e variants like distance (or angle) limited visibility, <strong>in</strong> which the region<br />
beyond a specified distance (or angle) is considered <strong>in</strong>visible. We can also have both angle and<br />
distance limited visibility coverage. Another <strong>in</strong>terest<strong>in</strong>g form of visibility coverage would be<br />
locus visibility coverage, which is the region that would be under surveillance as a guard goes<br />
along a specified path. All these forms of visibility coverage can be evaluated us<strong>in</strong>g HELIOS.<br />
This can be useful <strong>in</strong> decid<strong>in</strong>g positions and loci of of guards, crew and surveillance cameras. It<br />
may also be useful <strong>in</strong> evaluat<strong>in</strong>g and grad<strong>in</strong>g positions of adverts <strong>in</strong> a pedestrian area. Fig.14<br />
shows some screenshots of this module.<br />
374
Fig.14: Visibility Coverage<br />
Distributed ubiquitous monitor<strong>in</strong>g with sensor networks is slowly but surely catch<strong>in</strong>g on as<br />
a viable measure for ensur<strong>in</strong>g safety and security. Too many accidents have been caused or<br />
aggravated due to absence, <strong>in</strong>completeness or <strong>in</strong>accuracy of monitor<strong>in</strong>g. With modern technology<br />
of t<strong>in</strong>y and <strong>in</strong>expensive processor-cum-transponder units (sometimes referred to as smart dust),<br />
it is possible to set up a pervasive monitor<strong>in</strong>g system that keeps track an <strong>in</strong>door or outdoor<br />
environment with a very f<strong>in</strong>e granularity and high reliability. The overal setup of such system<br />
requires a spatial database support for associat<strong>in</strong>g sensed data with spatial semantics. A spatial<br />
database similar to HELIOS would be <strong>in</strong>evitably required <strong>in</strong> organiz<strong>in</strong>g the sensed data and<br />
for automated reason<strong>in</strong>g about the sensed environment. Rang<strong>in</strong>g from security to <strong>in</strong>ventory<br />
track<strong>in</strong>g, from structural to power monitor<strong>in</strong>g, from fire safety to personal navigational aids,<br />
from process automation to smart build<strong>in</strong>gs - the potential applications are numerous and the<br />
impact is go<strong>in</strong>g to be big. The geometry savvy spatial database of HELIOS already supports<br />
several classes of queries that we envision as useful to such sensor fusion scenarios, more can be<br />
developed as we <strong>in</strong>vestigate further along these l<strong>in</strong>es.<br />
6. Software constituents of HELIOS<br />
HELIOS gratefully leverages great pieces of software without which it would have been too<br />
much work to even contemplate - the usual and <strong>in</strong>dispensable ’shoulder of giants’ phenomenon<br />
of software development. In the development of HELIOS, we have deliberately not given <strong>in</strong><br />
to dependence on any proprietary or closed-source component. It is written <strong>in</strong> ansi C++,<br />
compiled with gcc and Microsoft Visual Studio .NET. It uses OpenGL and GLUT for 3d graphics,<br />
Tcl/Tk for graphics widgets and a script<strong>in</strong>g <strong>in</strong>terface, swig for generat<strong>in</strong>g the script<strong>in</strong>g<br />
<strong>in</strong>terface from a C++ header file, libSDL for audio and for load<strong>in</strong>g texture bitmaps, L A TEX,<br />
fig2dev, and ImageMagick for automated report generation and numerical recipes textbook,<br />
Press et al. (1993), code for l<strong>in</strong>ear algebra. The aforementioned jargon are household names<br />
to developers and it is easy to f<strong>in</strong>d out about them by a google search, so omitt<strong>in</strong>g the references<br />
to them. The geometry library CGAL (http://www.cgal.org) and the tool Triangle<br />
(http://www-2.cs.cmu.edu/∼quake/triangle.html) have been used with<strong>in</strong> separate programs <strong>in</strong><br />
the GA process<strong>in</strong>g pipel<strong>in</strong>e.<br />
7. Conclusion and scope for future work<br />
A well designed framework should significantly relieve the burden of work from the class of<br />
activity it caters to. The HELIOS effort seems to be head<strong>in</strong>g correctly <strong>in</strong> that direction.<br />
We shall cont<strong>in</strong>ue to add value to the framework and hopefully it will become an even more<br />
efficacious tool <strong>in</strong> near future.<br />
References<br />
BERG, M. de; KREVELD, M. van; OVERMARS, M.; SCHWARZKOPF, O. (2000), Computational<br />
Geometry, Spr<strong>in</strong>ger<br />
375
CHEW, P. (1987), Constra<strong>in</strong>ed Delaunay triangulations, Symp. Comp. Geometry, pp.215-222<br />
CORMEN, T.H.; LEISERSON, C.E.; RIVEST, R.L. (1990), Introduction to algorithms, MIT<br />
Press<br />
FARLOW, S.J. (1993), Partial Differential Equations for Scientists and Eng<strong>in</strong>eers, Dover Publ.<br />
GUARIN, L.; MAJUMDER, J.; SHIGUNOV, V.; et al. (2004), Fire and flood<strong>in</strong>g risk assessment<br />
<strong>in</strong> ship design for ease of evacuation, Int. Conf. Design for Safety, Osaka<br />
KLEIN, P.; RAO, S.; RAUCH, M.; SUBRAMANIAN, S. (1994), Faster shortest-path algorithms<br />
for planar graphs, 26 th ACM Symp. on Theory of Comput<strong>in</strong>g, pp.27-37<br />
LUEBKE, D.; et al. (2002), Level of detail for 3D graphics, Morgan Kaufmann<br />
SHIGUNOV, V. (2004), Fire development and smoke propagation <strong>in</strong> ship compartments, Ship<br />
Technology Research 51(1), pp.9-20<br />
THORUP, M. (1999), Undirected s<strong>in</strong>gle-source shortest paths with positive IntegerWeights <strong>in</strong><br />
l<strong>in</strong>ear time, J. ACM 46, pp.362-394<br />
VASSALOS, D.; KIM, H.; CHRISTIANSEN, G.; MAJUMDER, J. (2001), A mesoscopic model<br />
for passenger evacuation <strong>in</strong> a virtual ship-sea environment and performance-based evaluation,<br />
Pedestrian and Evacuation Dynamics, Duisburg<br />
376
!" <br />
<br />
<br />
# <br />
<br />
<br />
<br />
! " <br />
##$<br />
" <br />
$ % % " <br />
&""%<br />
" # <br />
<br />
$% <br />
<br />
!"""#<br />
$##" #<br />
%#$& <br />
# ##'%"% " <br />
" # $ #<br />
("<br />
# " " <br />
#%<br />
<br />
<br />
<br />
" # <br />
<br />
<br />
<br />
<br />
Virtual Knowledge<br />
Base<br />
<br />
<br />
<br />
! <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
" # ! <br />
<br />
<br />
")*+" !,-.
&% ' <br />
<br />
!0 ##1"#<br />
" <br />
$ , % # <br />
% (<br />
#"")<br />
<br />
(%# <br />
<br />
'"" #&<br />
"2%'"3<br />
& "" #&<br />
" " ! " #" # #<br />
%%%<br />
1"0##<br />
%#% !<br />
#" """"#"<br />
" " " " & " '<br />
% " " " 4# <br />
#5 5 " "" " +" <br />
!%"0#%%5 %"#"<br />
"<br />
<br />
<br />
"2* !'""<br />
<br />
"* !#"<br />
<br />
/
" "7#5 (" <br />
, " % 8$ #<br />
<br />
<br />
"7* !!0 <br />
<br />
! " <br />
( 8 ! " % <br />
## ""&'%"<br />
(( " 0"<br />
'()**+,<br />
• 9%"#""# 5 " <br />
<br />
• +" 8%<br />
" " "<br />
• "#<br />
• """" "<br />
<br />
!"""5#" <br />
"#""#" &'%""<br />
5" ' " '% ' "<br />
&'()**+ :<br />
" 884#;<br />
"%'<br />
<br />
<br />
<br />
"
'<br />
<br />
:.?@3941?3A@1:.?+43341:B!3C:)666-"%./<br />
. 0 0 -"/1 7<br />
4""4 8"<br />
<br />
D-B4@-@1:.?@3941:.?+43342>><br />
@8""D<br />
<br />
D-B4@-@1:-@4E 1:.?@3942>>2#,<br />
F .!,:4"::G'<br />
2)6822<br />
<br />
/>
Knowledge Model<strong>in</strong>g <strong>in</strong> Ship Design<br />
us<strong>in</strong>g Semantic Web Techniques<br />
Robert Bronsart, Michael Zimmermann, University of Rostock, Rostock/Germany,<br />
{robert.bronsart,michael.zimmermann}@uni-rostock.de<br />
Abstract<br />
With<strong>in</strong> the research project KonSenS, the prospect of knowledge based eng<strong>in</strong>eer<strong>in</strong>g applied to<br />
ship design is analyzed. An improved <strong>in</strong>tegration of knowledge, geometry and manufactur<strong>in</strong>g<br />
<strong>in</strong>formation can achieve an improvement <strong>in</strong> product quality and a reduction of design time as<br />
well as manufactur<strong>in</strong>g costs. Semantic web techniques are one means for knowledge model<strong>in</strong>g. A<br />
consolidated <strong>in</strong>formation server is used to manage documentation and standards. An <strong>in</strong>tegration<br />
<strong>in</strong>to a CAD system is used to improve standards compliance. Decision support is given for<br />
standardized tasks hereby eas<strong>in</strong>g the workload of the designer.<br />
1. Introduction<br />
The development process <strong>in</strong> the maritime <strong>in</strong>dustries is characterized by a complex <strong>in</strong>teraction of<br />
many partners work<strong>in</strong>g <strong>in</strong> parallel. From the <strong>in</strong>itial concept up to the f<strong>in</strong>al design diverse and<br />
often conflict<strong>in</strong>g requirements need to be taken <strong>in</strong>to account. This requires a constant exchange<br />
of a significant amount of <strong>in</strong>formation.<br />
Today, <strong>in</strong> the CAD/CAM environment there is no tight <strong>in</strong>tegration of the knowledge and <strong>in</strong>formation<br />
needed to develop the f<strong>in</strong>al design with the geometry and manufactur<strong>in</strong>g <strong>in</strong>formation<br />
used for production. While the later is handled with sophisticated CAD/CAM solutions, for the<br />
former a diverse range of applications and methods is used.<br />
As a result, for the designer the <strong>in</strong>formation needed is available <strong>in</strong> different formats and from<br />
different sources. Often, media breaks require that the <strong>in</strong>formation given <strong>in</strong> e.g. a document<br />
is <strong>in</strong>terpreted by the designer and applied <strong>in</strong> the appropriate context. With such procedures<br />
there exists not only the <strong>in</strong>creased probability of mis<strong>in</strong>terpretation. Also knowledge reuse <strong>in</strong> the<br />
design process as well as automatization is prevented.<br />
Knowledge based systems are applications that capture the knowledge available for a problem<br />
doma<strong>in</strong> and represent this knowledge <strong>in</strong> a mach<strong>in</strong>e <strong>in</strong>terpretable way. In ship design these<br />
systems can be used to e.g. document best practice recommendations or for standardization<br />
and automatization purposes. Also, knowledge can be extracted from key persons and made<br />
available to other members of the team.<br />
For the world wide web, the development of ’semantic web techniques’, Daconta et al. (2003),<br />
is geared towards the def<strong>in</strong>ition of mach<strong>in</strong>e readable and more powerful relations between bits<br />
of <strong>in</strong>formation. Such relations allow for a tight <strong>in</strong>tegration of all pieces of <strong>in</strong>formation needed<br />
for ship design. Once accepted by the public this allows for a more precise search and retrieval<br />
of <strong>in</strong>formation; automatic deductive reason<strong>in</strong>g is also supported.<br />
In the collaborative research project KonSenS the University of Rostock is work<strong>in</strong>g on the problem<br />
of model<strong>in</strong>g requirements and algorithms needed <strong>in</strong> ship structural design. With the help of<br />
knowledge based methods the objective of this project is to achieve an enhanced automatization<br />
and a higher quality of the f<strong>in</strong>al product.<br />
In this paper the use of semantic web techniques for ship design is analyzed. Based on a<br />
formal <strong>in</strong>troduction to knowledge model<strong>in</strong>g a description of the techniques available and their<br />
capabilities is given. The application of semantic web techniques <strong>in</strong> ship design is described.<br />
381
2. Knowledge<br />
As <strong>in</strong> all other eng<strong>in</strong>eer<strong>in</strong>g processes, <strong>in</strong> ship design one factor that determ<strong>in</strong>es the result is<br />
knowledge. Today, with the focus placed on high quality, low costs and a short time to market,<br />
a thorough and extensive understand<strong>in</strong>g of all possible aspects is a key factor for success. A lack<br />
of understand<strong>in</strong>g may lead to cost <strong>in</strong>tensive effects at a later design stage or <strong>in</strong> operation. But<br />
what is knowledge?<br />
Data, Information, and Knowledge<br />
Knowledge is comprised of <strong>in</strong>dividual pieces of <strong>in</strong>formation called facts, Franken and Gadatsch<br />
(2002). A fact is an atomic entity, i.e. an entity that is self supportive and not related to other<br />
entities. Information is def<strong>in</strong>ed as a collection of facts with additional relationships and rules<br />
between facts. This network of facts def<strong>in</strong>es the <strong>in</strong>formation base.<br />
Us<strong>in</strong>g deductive reason<strong>in</strong>g and experience the essence of the <strong>in</strong>formation is extracted from the<br />
<strong>in</strong>formation base lead<strong>in</strong>g to knowledge. Here, experience represents knowledge the designer has<br />
ga<strong>in</strong>ed from e.g. previous reason<strong>in</strong>gs, experiments or observations. Therefore, the extraction of<br />
knowledge is always bound to the background of the person. Differ<strong>in</strong>g conclusions might be<br />
drawn from an identical <strong>in</strong>formation base.<br />
Types of Knowledge<br />
From a theoretical standpo<strong>in</strong>t different types of knowledge can be identified, Riempp (2004).<br />
Structured knowledge is given <strong>in</strong> a predef<strong>in</strong>ed format, i.e. the semantic of the knowledge is<br />
given as meta-<strong>in</strong>formation. Examples <strong>in</strong> ship design are parts libraries or class rules. For<br />
unstructured knowledge no meta-<strong>in</strong>formation is given. Evaluation of this type of knowledge is<br />
only possible if <strong>in</strong>formation about the context is available, i.e. additional constra<strong>in</strong>ts apply but<br />
are not def<strong>in</strong>ed explicitly. In ship design the experience of the designer, some <strong>in</strong>formal notation<br />
or comments enables the eng<strong>in</strong>eer to abstract the required knowledge; for automatic evaluation<br />
the determ<strong>in</strong>ation of the context and constra<strong>in</strong>ts is difficult.<br />
Explicit knowledge is knowledge that can be described us<strong>in</strong>g facts and relations. The transfer of<br />
this k<strong>in</strong>d of knowledge is possible with only little danger of mis<strong>in</strong>terpretation. Implicit knowledge<br />
is knowledge that is not def<strong>in</strong>ed explicitly but nevertheless available. Often, implicit knowledge<br />
forms one foundation a team or society is based on. In ship design, experience and general<br />
agreements are examples of implicit knowledge.<br />
F<strong>in</strong>ally, <strong>in</strong> most companies some knowledge is bound to certa<strong>in</strong> persons as <strong>in</strong>dividual knowledge<br />
and not available to all employees of the company. If such an employee leaves the company the<br />
knowledge is no longer available. In contrast, social knowledge is shared by a team or society,<br />
i.e. all members of the team are able to understand and apply this knowledge.<br />
Knowledge Exchange and Management<br />
In a collaborative environment the constant exchange of <strong>in</strong>formation and knowledge is mandatory.<br />
As mentioned above knowledge is extracted from an <strong>in</strong>formation base. Hereby, the <strong>in</strong>formation<br />
given is set <strong>in</strong> relation to already exist<strong>in</strong>g knowledge, the context and the problem<br />
at hand. Therefore, for each person knowledge is represented as a different mental model. For<br />
knowledge exchange to take place the mental model is transformed to a set of <strong>in</strong>formation that<br />
is transferred to another person. Then, the other person uses its background to abstract the<br />
knowledge from the <strong>in</strong>formation depict<strong>in</strong>g a new mental model.<br />
For knowledge exchange to take place the partners have to agree on a common language and<br />
have to share a basic set of already exist<strong>in</strong>g common knowledge. Or from a more theoretical<br />
standpo<strong>in</strong>t, the transfer of knowledge can only take place if a common syntax and semantics is<br />
agreed upon.<br />
382
For an effective exchange of knowledge <strong>in</strong> a company precautions should be taken to optimize<br />
a formalized representation of <strong>in</strong>formation required for operation. This enables the employees<br />
to search for and acquire knowledge <strong>in</strong> a structured way. Structured knowledge with additional<br />
meta-data gives a mach<strong>in</strong>e <strong>in</strong>terpretable representation. Also, the extraction of implicit <strong>in</strong>dividual<br />
knowledge from key persons and its preparation leads to a reduced dependency from these<br />
persons.<br />
In the mar<strong>in</strong>e bus<strong>in</strong>ess a formalized representation of knowledge used for ship design can also<br />
be used to simplify the reuse of proven solutions and to document quality standards.<br />
3. Semantic Web and Semantic Web Techniques<br />
3.1. Standards<br />
For a mach<strong>in</strong>e <strong>in</strong>terpretable representation of knowledge not only is it necessary to store <strong>in</strong>dividual<br />
facts electronically but also to add additional <strong>in</strong>formation about the relationships between<br />
these facts. Here, relationships can either describe simple dependencies between facts but can<br />
also impose more complex constra<strong>in</strong>ts def<strong>in</strong>ed by card<strong>in</strong>ality constra<strong>in</strong>ts or rules.<br />
For an IT based evaluation of relationships and facts applications need to know about the<br />
problem doma<strong>in</strong>, i.e. some basic def<strong>in</strong>ition of the syntax, semantics and concepts used <strong>in</strong> a<br />
problem doma<strong>in</strong> needs to be available. For knowledge model<strong>in</strong>g the syntax and semantics is<br />
mostly given by the model<strong>in</strong>g standards used. For each given problem doma<strong>in</strong> the concepts and<br />
the relationships between these concepts are def<strong>in</strong>ed <strong>in</strong> a taxonomy or ontology.<br />
Taxonomies and conceptual models spawn a loosely coupled mesh of concepts and creates parentchild<br />
relations between concepts The type of relationship between concepts is not identified. An<br />
ontology takes the approach used for taxonomies to a higher level and adds e.g. support for<br />
relationship types and constra<strong>in</strong>ts between concepts.<br />
The semantic web work<strong>in</strong>g groups of the World Wide Web Consortium (W3C) were established<br />
to work on standards for the doma<strong>in</strong> and application <strong>in</strong>dependent model<strong>in</strong>g of knowledge. The<br />
objective of the work<strong>in</strong>g groups is to create the means to structure <strong>in</strong>dividual facts and to put<br />
them <strong>in</strong>to context. Also, improved search capabilities, automatic deduction and reason<strong>in</strong>g based<br />
on <strong>in</strong>formation available on the <strong>in</strong>ternet are addressed.<br />
For this purpose the W3C published a series of XML-based standards, Fig.1. Based on XML and<br />
XML Schema for the def<strong>in</strong>ition of content and syntax the semantics are described us<strong>in</strong>g the Rich<br />
Description Framework (RDF) or Topic Maps (TM). For ontology support the Web Ontology<br />
Language (OWL) standard is available. Upon these models <strong>in</strong>ference eng<strong>in</strong>es or description logic<br />
(DL) can operate to query the <strong>in</strong>formation model and to extract additional <strong>in</strong>formation us<strong>in</strong>g<br />
deductive reason<strong>in</strong>g or first order description logic.<br />
DL<br />
OWL<br />
RDF / TM<br />
XML<br />
Fig.1: Semantic web standards<br />
383
3.2. Different Levels of Information Models<br />
The standards developed for the semantic web cover different aspects of knowledge model<strong>in</strong>g and<br />
representation. From the basic document <strong>in</strong> XML over relational model<strong>in</strong>g us<strong>in</strong>g topic maps<br />
or RDF to full fledged description logic different levels of complexity needed for <strong>in</strong>formation<br />
model<strong>in</strong>g are covered. In the follow<strong>in</strong>g some aspects relevant for knowledge management are<br />
expla<strong>in</strong>ed.<br />
Topic Maps and RDF The Topic Map and the Resource Description Framework, Powers<br />
(2003), standard orig<strong>in</strong>ate from two standards organizations, ISO and the W3C respectively.<br />
Because these standards cover similar aspects and a conversion from TM to RDF or vice versa<br />
can be achieved, Garshol (NN), <strong>in</strong> the follow<strong>in</strong>g only topic maps are exam<strong>in</strong>ed. Fig.2 shows an<br />
example for a graphical representation of a topic map that can be used <strong>in</strong> ship design. With<br />
this simple knowledge model the relations between stiffeners, cutouts and clips are def<strong>in</strong>ed.<br />
Attributes with resource identifiers (URI) are used to l<strong>in</strong>k to objects not stored <strong>in</strong> the topic<br />
map.<br />
Cutout:<br />
Type A<br />
For Profile 140 mm<br />
Clip<br />
COA140<br />
subtype of<br />
requires<br />
<strong>in</strong>stance of<br />
more <strong>in</strong>fo<br />
Basic<br />
Document<br />
Feature<br />
subtype of<br />
Cutout<br />
subtype of<br />
Cutout A<br />
<strong>in</strong>stance of<br />
subtype of<br />
subtype of<br />
Stiffener<br />
is subtype of<br />
Bulp<br />
COA120<br />
Text<br />
Document<br />
<strong>in</strong>stance of<br />
has<br />
URI<br />
120<br />
is<br />
Height<br />
has<br />
STB120<br />
def<strong>in</strong>ed by<br />
ISO Norm<br />
Fig.2: A simple topic map model for the relations between stiffeners, clips and cutouts<br />
Stemm<strong>in</strong>g from the work on electronic <strong>in</strong>dexes topic maps are a subject-based classification<br />
technique. As the name implies topic maps are organized around the notion of topics, which<br />
usually represent primary concepts important for a specific problem doma<strong>in</strong>, the ’subjects’.<br />
Subjects are described us<strong>in</strong>g the basic constructs names, occurrences and associations. In a topic<br />
map, every object itself is a topic with one or more given names. The name is used to create an<br />
unambiguous identification and can be used for human <strong>in</strong>teraction or mach<strong>in</strong>e <strong>in</strong>terpretation.<br />
For classification purposes topic types, which itself are also topics, can be assigned to topics.<br />
This makes it possible to group certa<strong>in</strong> topics relevant for a specific context.<br />
Information objects like documents, class standards or parts of a CAD/CAM model are not<br />
stored <strong>in</strong> the topic map but a topic may l<strong>in</strong>k to an <strong>in</strong>formation resource us<strong>in</strong>g the occurrence<br />
construct. This enables a software designer to store <strong>in</strong>formation <strong>in</strong> the most suitable way for<br />
a given task. CAD/CAM objects can be stored <strong>in</strong> the CAD/CAM system, documents <strong>in</strong> the<br />
document management system.<br />
With associations relationship between topics are def<strong>in</strong>ed. While associations only express that<br />
some relation between two topics exist, association types are used to classify and group the<br />
relationships accord<strong>in</strong>g to their type.<br />
384
One advantage of topic maps for software development is that it allows the development of<br />
applications <strong>in</strong> which only an <strong>in</strong>complete application data model is known at design time but is<br />
configured or extended by the user.<br />
Ontology Support<br />
Topic Maps and RDF are tools for relational model<strong>in</strong>g. But similar to the mostly keyword or<br />
full-text based search used <strong>in</strong> search eng<strong>in</strong>es on the <strong>in</strong>ternet, without a limited and predef<strong>in</strong>ed<br />
vocabulary the quality of results is poor. Also, for knowledge based applications the mean<strong>in</strong>g<br />
of the relations and relation types need to be known.<br />
With the OWL standard so called ontology based vocabularies are used. Applied to knowledge<br />
model<strong>in</strong>g with semantic web techniques an ontology def<strong>in</strong>es a basis of permissible relations and<br />
types of relations. Roles can be restricted to apply e.g. only for certa<strong>in</strong> relation types; card<strong>in</strong>ality<br />
constra<strong>in</strong>ts can be given. Additional knowledge is encoded <strong>in</strong> the ontology model with basic<br />
logical operators like AND, OR or IS NOT.<br />
For ship design an ontology def<strong>in</strong>ed for a specific problem doma<strong>in</strong> can be used to represent<br />
knowledge about this doma<strong>in</strong> as structured knowledge. For example, a basic ontology that<br />
def<strong>in</strong>es the knowledge about brackets, stiffeners and cutouts can be used to def<strong>in</strong>e a set of<br />
permissible comb<strong>in</strong>ations. An extended ontology allows also to describe the context valid for a<br />
certa<strong>in</strong> part.<br />
Reason<strong>in</strong>g<br />
With RDF, topic maps and OWL a static yet extensible encod<strong>in</strong>g of doma<strong>in</strong> specific knowledge<br />
can be achieved. On this basis the process of reason<strong>in</strong>g uses first order description logic to <strong>in</strong>fer<br />
additional knowledge from the knowledge base or to apply the knowledge base to a specific given<br />
problem.<br />
While first order description logic provides the means to validate a given knowledge model and<br />
to compare a given set of <strong>in</strong>formation with this model it does not give a mach<strong>in</strong>e <strong>in</strong>terpretable<br />
representation of the mean<strong>in</strong>g of constructs like relationship types, documents or other data<br />
objects. Therefore, the mean<strong>in</strong>g of the different concepts def<strong>in</strong>ed <strong>in</strong> a base ontology need to be<br />
implemented as part of the application software applied. For a complex problem doma<strong>in</strong> this<br />
can be a significant task to complete.<br />
4. Ship design and Semantic Web Techniques<br />
Knowledge model<strong>in</strong>g and semantic web techniques offer the possibilities to apply the concept of<br />
knowledge based eng<strong>in</strong>eer<strong>in</strong>g to the process of ship design, Krüger (2003). With<strong>in</strong> the research<br />
project KonSenS, an application server is developed, Fig.3. The system offers an <strong>in</strong>formation<br />
server that provides <strong>in</strong>formation about design procedures, standard and documentation. Us<strong>in</strong>g a<br />
platform and system <strong>in</strong>dependent design the client-server solution offers a flexible and centralized<br />
<strong>in</strong>frastructure. With the help of adapter <strong>in</strong>terfaces to CAE-Systems or other applications present<br />
<strong>in</strong> the heterogeneous environments found <strong>in</strong> many shipyards it is possible to consolidate and<br />
centralize the management and configuration of different applications <strong>in</strong>to the system KonSenS.<br />
A stand-alone client can be used to access the application server directly.<br />
Interoperability and Integration<br />
With a diverse software landscape as present <strong>in</strong> today’s shipbuild<strong>in</strong>g <strong>in</strong>dustry <strong>in</strong>tegration and<br />
<strong>in</strong>teroperability are important issues.<br />
Here, <strong>in</strong>teroperability guarantees that data can be exchanged with other applications. In most<br />
cases a file based exchange us<strong>in</strong>g import and export functionality is chosen. While such a<br />
solution is helpful to reuse data <strong>in</strong> other applications the issue of multiple data storage persists.<br />
Merg<strong>in</strong>g issues and synchronization problems arise. Similar to STEP physical files the XML<br />
385
ased standards offer the means for file based data exchange. The format of the exchange files<br />
can be described with the help of so called schema’s. Checks for syntactical correctness are<br />
possible.<br />
Fig.3: Conceptual model of the application server and the client-server <strong>in</strong>teraction<br />
One objective of the project KonSenS is <strong>in</strong>creased <strong>in</strong>tegration, i.e. the concurrent access of<br />
diverse applications to one or multiple unified data repositories. This enables the user to work<br />
with up-to-date data and removes the need for synchronization. For a seamless <strong>in</strong>tegration of<br />
applications an ontology based <strong>in</strong>formation model can be applied. With such an ontology not<br />
only the syntax but also the semantic and logic of the data can be def<strong>in</strong>ed.<br />
Because of the network centric orientation of the committee develop<strong>in</strong>g the XML based standards<br />
distributed storage and retrieval can easily be implemented. Platform <strong>in</strong>dependence, a large user<br />
base and the availability of the appropriate tools ease the transition to these standards.<br />
Flexibility<br />
With STEP the <strong>in</strong>formation model for data storage and exchange is fixed upon system design.<br />
This means that custom extensions to the capabilities of the data model can only be achieved<br />
<strong>in</strong> cooperation with the software developers. Customization and extension of a data model at<br />
runtime is not an option.<br />
With semantic web techniques the knowledge encoded <strong>in</strong> topic maps or RDF structures an<br />
<strong>in</strong>crease <strong>in</strong> flexibility is given allow<strong>in</strong>g for a yard specific configuration of applications. As<br />
long as the underly<strong>in</strong>g ontology is sufficient the <strong>in</strong>formation model can be extended without<br />
loos<strong>in</strong>g the power of automatic computer-based <strong>in</strong>terpretation. Still, limits are present if the<br />
base ontology does not offer sufficient means for the description of upcom<strong>in</strong>g requirements.<br />
Due to the <strong>in</strong>tegration capabilities offered by the network centric design the flexible comb<strong>in</strong>ation<br />
of different services available <strong>in</strong>to one homogeneous <strong>in</strong>formation model is possible. Multiple<br />
applications can be <strong>in</strong>tegrated.<br />
5. Use Cases<br />
For the maritime <strong>in</strong>dustry or, more specific, <strong>in</strong> ship design the application of semantic web<br />
techniques can be used to ease the passage to knowledge based applications and to knowledge<br />
based eng<strong>in</strong>eer<strong>in</strong>g. In this chapter possible use cases are shown.<br />
386
5.1. Knowledge Model<strong>in</strong>g<br />
For an organization operat<strong>in</strong>g <strong>in</strong> the maritime <strong>in</strong>dustry knowledge is an important asset for the<br />
successful acquisition of and work on contracts. In the last years the importance of this role<br />
has <strong>in</strong>creased so that corporations actively seek to extend the benefits from all the knowledge<br />
available. With the knowledge types def<strong>in</strong>ed above, the objective of corporations is to make<br />
<strong>in</strong>formation available to the employee as needed, Nieuwenhuis and Nienhuis (2004). In ship<br />
design explicit knowledge is available <strong>in</strong> the form of standards, that i.e. regulate the use of parts<br />
etc., or as other documentation, mostly available as text.<br />
Documentation<br />
As part of the application server a documentation component offers the means to store documentation<br />
that hereby is made available to all employees. These documents are l<strong>in</strong>ked to a topic<br />
map or RDF representation founded on a base ontology of topics and topic types. This allows<br />
for a consistent classification and reduces the ambiguity of freely def<strong>in</strong>ed keywords.<br />
The search capabilities offered by semantic web techniques do not restrict <strong>in</strong>formation retrieval<br />
to full text and keyword search, but also enables an application or user to search by context or<br />
with even more complex logical operations. This leads to a reduced time needed for <strong>in</strong>formation<br />
retrieval.<br />
Also, for new employees or people not familiar with certa<strong>in</strong> specifics a consistent knowledge<br />
model provides these people with a tool that enables them to research and learn the procedures<br />
common <strong>in</strong> a company. The problem of different work techniques or solution strategies can be<br />
m<strong>in</strong>imized.<br />
Standards<br />
In this paper, standards def<strong>in</strong>e predef<strong>in</strong>ed solutions or parts. A standards component is <strong>in</strong>tegrated<br />
<strong>in</strong>to the server. This component enables the adm<strong>in</strong>istrator to store standard parts and<br />
comb<strong>in</strong>ations of part. For constra<strong>in</strong>ts, context <strong>in</strong>formation and requirements a knowledge model<br />
is used. While no <strong>in</strong>ternal shape representation is implemented, the relevant dimensions are<br />
stored so that they can be exported to different CAD systems.<br />
The standard catalog is l<strong>in</strong>ked <strong>in</strong>to the base ontology for ship structural parts. While a paper<br />
based catalog only provides a s<strong>in</strong>gle, mostly shape-oriented view onto the standard parts, with<br />
an ontology-supported relational representation of e.g. a parts catalog additional views can be<br />
derived. These can be views with respect to areas of application, strength or manufactur<strong>in</strong>g.<br />
With these views additional support for a decision is given at the hand of the eng<strong>in</strong>eer.<br />
Knowledge and the CAD Model<br />
In the maritime <strong>in</strong>dustry the CAD model of e.g. a ship comprises the result of the application<br />
of the eng<strong>in</strong>eers’ knowledge on the shipyard or network.<br />
If life-cycle aspects, modifications or rebuilds of a design are taken <strong>in</strong>to account the search for the<br />
reason a certa<strong>in</strong> solution was developed <strong>in</strong> a given way and which standards were used becomes<br />
a serious obstacle. Media breaks play an important role <strong>in</strong> this issue. With an electronical<br />
representation of knowledge the knowledge used for certa<strong>in</strong> decision can be documented. This<br />
allows an eng<strong>in</strong>eer to reproduce and validate the design if needed.<br />
Also, the project KonSenS strives for a tight <strong>in</strong>tegration of electronically encoded knowledge -<br />
this might be standards or other documentation - and the CAD model. With the help of ontology<br />
model<strong>in</strong>g techniques, for a specific problem the correspond<strong>in</strong>g documentation and standards can<br />
be presented to the eng<strong>in</strong>eer with<strong>in</strong> the CAD system where appropriate. With such an approach<br />
the system provides decision support.<br />
387
5.2. Context Sensitive Design<br />
Standards like parts catalogs or list<strong>in</strong>gs of preferred solutions are stored <strong>in</strong> the application server.<br />
Here, with relational model<strong>in</strong>g based on an ontology these parts can not only be l<strong>in</strong>ked with<br />
documentation items but also can be put <strong>in</strong>to relationship to other parts. A classification of<br />
the type of relationship is possible; constra<strong>in</strong>ts can be imposed. This allows to def<strong>in</strong>e possible<br />
solutions, e.g. a certa<strong>in</strong> comb<strong>in</strong>ation of parts. In the knowledge model such a solution can then<br />
be l<strong>in</strong>ked to context <strong>in</strong>formation. This can be seen <strong>in</strong> Fig.2.<br />
With qualified l<strong>in</strong>ks from a solution to the context the def<strong>in</strong>ition of a permissible doma<strong>in</strong> for this<br />
solution can be achieved. I.e. it is possible to specify the requirements a design solution needs<br />
to fulfill. Conversely, if the context is known the server can supply possible solutions. This eases<br />
the workload of the designer.<br />
As part of the CAD model additional meta data can be stored to add further <strong>in</strong>formation<br />
about the context. This meta <strong>in</strong>formation is then used to constra<strong>in</strong> solutions or further quality<br />
constra<strong>in</strong>ts. For example a bulkhead adjacent to a tank is marked as watertight. For subsequent<br />
operations on this bulkhead this context <strong>in</strong>formation is used to e.g. deny the use of non-watertight<br />
clips or of holes.<br />
For the <strong>in</strong>teraction of the system with the eng<strong>in</strong>eer or the CAD system the notion of tasks is<br />
def<strong>in</strong>ed. A task represents an action performed frequently by the designers <strong>in</strong> ship structural<br />
design. For each task the work-flow, i.e. the order of necessary action to reach a solution,<br />
from the problem to a solution is def<strong>in</strong>ed; <strong>in</strong>formation needed as part of the solution process is<br />
identified.<br />
Similar to a design process based on conventional CAD system a knowledge enabled CAD system<br />
offers the eng<strong>in</strong>eer an <strong>in</strong>terface to select the action to perform. With an analysis of the context<br />
additional <strong>in</strong>formation is derived from the geometric model present <strong>in</strong> the CAD system. With<br />
this <strong>in</strong>formation the knowledge model stored <strong>in</strong> the knowledge base is queried and possible<br />
solutions are presented to the end user. This leads to a significant reduction of the amount of<br />
<strong>in</strong>formation that needs to be entered by the end user. Hence, a reduction of design time can be<br />
achieved.<br />
In the future an additional assessment of the solutions provided by the application with respect<br />
to manufacturability, cost, strength and other factors is planned.<br />
5.3. Standardization, Quality Control, and Automatization<br />
The def<strong>in</strong>ition of an electronic standards catalog can guarantee that only correct and unequivocal<br />
standards are published. For this purpose first order description logic can be used to validate<br />
an <strong>in</strong>formation model and to warn about ambiguities or contradictions. As a result a concise<br />
set of standards avoids differ<strong>in</strong>g <strong>in</strong>terpretations of a standard that may lead to design errors.<br />
With standards def<strong>in</strong>ed <strong>in</strong> such a way the eng<strong>in</strong>eers are not only supported <strong>in</strong> their decisions<br />
but also the validity of exist<strong>in</strong>g solutions can be tested for. The validation of the CAD model<br />
can detect errors at an early stage <strong>in</strong> the design process. This shows the ability to improve the<br />
process of quality control and to guarantee that only applicable solutions are used. An improved<br />
overall quality of the f<strong>in</strong>al product can be achieved.<br />
In ship structural design series effects play an important role with respect to manufactur<strong>in</strong>g<br />
costs. The def<strong>in</strong>ition of standard solutions can help to reduce the diversity of parts used and to<br />
<strong>in</strong>crease the number of parts needed of certa<strong>in</strong> types. This allows to explore cost sav<strong>in</strong>g series<br />
effects.<br />
With an extended version of the system the knowledge model available might help to improve<br />
automatization and reduce the design time needed. As possible scenarios for application the fully<br />
automatic creation of e.g. holes or of notches could be achieved if the context is fully def<strong>in</strong>ed.<br />
388
Also, the def<strong>in</strong>ition of additional task groups can be applied to describe the def<strong>in</strong>ition of complex<br />
parts like bulkheads or girder. Here, the design process is split <strong>in</strong>to <strong>in</strong>dividual tasks; for a given<br />
part the order these tasks are executed is def<strong>in</strong>ed. As an example, Fig.4 shows the tasks as they<br />
might be applied to a girder.<br />
5.4. Design Agents<br />
Fig.4: Task based design process<br />
In the maritime <strong>in</strong>dustry often design agents work as subcontractors on projects of one or multiple<br />
shipyards. For each project the design agent’s CAD system needs to be configured accord<strong>in</strong>g<br />
to the shipyards requirements. Also, the eng<strong>in</strong>eers at the design agent need to familiarize themselves<br />
with the appropriate standards and regulations.<br />
Fig.5: CAD system configuration and knowledge base access<br />
With the system, the access to documentation and standards can be granted, Fig.5. Furthermore<br />
the advantages of the system can also be used by the design agent via network connection.<br />
With the standards stored on the server <strong>in</strong> a platform <strong>in</strong>dependent way it is possible to create<br />
<strong>in</strong>structions for automatic preconfiguration of a CAD environment.<br />
389
Because standard catalogs are stored platform and system <strong>in</strong>dependently <strong>in</strong> the knowledge base<br />
these standards can be used to configure the CAD system of the design agent accord<strong>in</strong>g to the<br />
specification provided by the system, i.e. to the requirements of the shipyard. For this purpose<br />
the <strong>in</strong>formation server is able to export the configuration to a platform specific format that is<br />
then used to configure the workspace. This enables the design agent to reproduce the sett<strong>in</strong>gs<br />
used on the shipyard with little effort.<br />
Also, with the networked solutions chosen for the system access to documentation and standards<br />
used on the shipyard can be given to the design agent. The decision support provided enables<br />
the design agent to build accord<strong>in</strong>g to the standards of the shipyard. The time needed to get<br />
familiar with the specifics of a shipyard is reduced.<br />
6. Summary and Conclusion<br />
With semantic web techniques standards for the creation, management and use of mach<strong>in</strong>e<br />
<strong>in</strong>terpretable knowledge models are available. Such models are comprised of <strong>in</strong>dividual facts<br />
l<strong>in</strong>ked by qualified relations; constra<strong>in</strong>ts can be used to capture additional <strong>in</strong>formation.<br />
With knowledge model<strong>in</strong>g and semantic web techniques the <strong>in</strong>tegration of CAD data and other<br />
<strong>in</strong>formation can be achieved. Knowledge and <strong>in</strong>formation can be made available to a team of<br />
eng<strong>in</strong>eers. The def<strong>in</strong>ition of standard solutions and parts is possible. An ontology base allows<br />
for an unequivocal classification of facts with respect to the problem doma<strong>in</strong>.<br />
With the software components described the knowledge base can be used for context sensitive<br />
design. For a given problem only applicable solutions are presented to the user hereby reduc<strong>in</strong>g<br />
the probability of design errors. Quality control can be achieved with validation of solutions<br />
based on the knowledge available; series effects can be explored. Design agents can use the<br />
knowledge model to understand the solutions preferred on a specific shipyard. The standards<br />
available can be used for the configuration of the design agent’s CAD system.<br />
An improved quality assessment of solutions presented with respect to multiple criteria would<br />
help the eng<strong>in</strong>eer to choose correct solutions. The application of the knowledge model for<br />
advanced automatization and for the def<strong>in</strong>ition of complex parts needs to be explored. F<strong>in</strong>ally,<br />
the development of a suitable base ontology that captures the relevant concepts for ship<br />
structural design is a significant task not completed yet.<br />
References<br />
DACONTA, C.; OBERST, L.; SMITH, K. (2003), The Semantic Web, Wiley<br />
FRANKEN, R.; GADATSCH, A. (2002), Integriertes Knowledge Managment, Vieweg Verlag<br />
GARSHOL, L. (NN), Liv<strong>in</strong>g with topic maps and RDF, www.ontopia.net<br />
GARSHOL, L. (NN), Metadata? Thesauri? Taxonomies? Topic Maps!, www.ontopia.net<br />
KRÜGER, S. (2003), The Role of IT <strong>in</strong> Shipbuild<strong>in</strong>g, Compit’03, <strong>Hamburg</strong>, pp.525-533<br />
NIEUWENHUIS, J.; NIENHUIS, U. (2004), Knowledge and Data Reuse <strong>in</strong> Ship System Design<br />
and Eng<strong>in</strong>eer<strong>in</strong>g, Compit’04, Siquenza, pp.190-203<br />
POWERS, S. (2003), Practical RDF, O’Reilly<br />
RIEMPP, G. (2004), Integrierte Wissensmanagment-Systeme, Spr<strong>in</strong>ger<br />
390
! "#<br />
<br />
<br />
<br />
<br />
<br />
<br />
!<br />
!<br />
"#<br />
<br />
$<br />
% <br />
!<br />
$<br />
<br />
<br />
<br />
& $''<br />
<br />
<br />
<br />
<br />
$ " # %& ' # ""<br />
# "# ( ) "" "# ) *<br />
# # "<br />
* " "# " <br />
<br />
+ , "# " # <br />
"# # #* ##"## #-<br />
## * " " "# <br />
" # ."# # #<br />
" / * # -/ " # # <br />
#*"#<br />
<br />
<br />
<br />
/ 0" '# #*" #<br />
"0 # #" # #1<br />
# # "" + ' * <br />
) 2 #*"# ' # <br />
/)"#%/&/) '* ##<br />
# "# " # "" / / "# " #<br />
#" # / " #)# #/ <br />
#"#) #" #1"<br />
<br />
+/ # * #"#"#*" # <br />
"# " # " #<br />
* # # 3 % " 3 -#&<br />
()**+ $" " " " #<br />
4#')5# / ( * #6 #
3/ %3/&/" #*" /<br />
) ''0 "" #13 <br />
/" '#*" "'" #""1<br />
' * "# " <br />
<br />
SimCoMar<br />
FSG<br />
NSWE<br />
<strong>TUHH</strong><br />
TU-Delft<br />
CMT<br />
Ma<strong>in</strong>tenance of the<br />
Simulation Toolkit<br />
Tool<br />
development<br />
Simulation<br />
application<br />
Research<br />
Simulation and<br />
optimisation<br />
Module<br />
development<br />
Influence on software<br />
development<br />
Support of<br />
implementation<br />
Distribution of<br />
simulation<br />
Know-how<br />
exchange<br />
Distributed<br />
simulation<br />
Simulation and<br />
Virtual Reality<br />
Simulation <strong>in</strong><br />
outfitt<strong>in</strong>g<br />
. #)3<br />
<br />
<br />
<br />
$## 8# 8 " <br />
" "7-" ""# ""##<br />
# # /" ##" ##1<br />
#"# -"* "9<br />
" # "# ))#' <br />
<br />
Strategic plann<strong>in</strong>g Tactical plann<strong>in</strong>g Operative control<br />
Def<strong>in</strong>ition of production flow<br />
and milestones<br />
Verification of plan<br />
Adjustment of resources<br />
Reaction on disturbances<br />
or deviations<br />
<br />
Production start Time<br />
7./""# " <br />
<br />
" # ') " <br />
" "/" *#" "' # <br />
## #"# / , #0<br />
" <br />
<br />
/*#* $88.* " # #<br />
# ' "# #"# $*" #<br />
# *# ,) " ## $ #* <br />
"* " / " ' ') <br />
" #" "# ' "$ <br />
" " #+ ##*## ##<br />
"# # ') 1*# <br />
/"*## *# "/0<br />
7
" # $88 / )# # #<br />
"# /" #"# "#"# <br />
#" *#,)<br />
<br />
Plann<strong>in</strong>g<br />
Production<br />
Design /<br />
Eng<strong>in</strong>eer<strong>in</strong>g<br />
Activities and<br />
plan structure<br />
Production<br />
status<br />
Assembly<br />
tree<br />
Part<br />
geometry<br />
Simulation<br />
model<br />
Verified<br />
plan<br />
<br />
.6'$88<br />
<br />
/ #" " "##<br />
- " " #" * # <br />
## # / "1 # :63; ' # ' <br />
0" <br />
# "$88 0" <br />
* " **# / "" " * #*"# <br />
" " " # ' <br />
<br />
<br />
< "* # "3)#' <br />
'))+# ## #<br />
0 /"# *#,) #<br />
$88<br />
<br />
<br />
<br />
'#" *"" $88* #*"#<br />
# "# =/"" ##9# #" #<br />
' * # # / "" # "<br />
' #"*" # # 8 <br />
# """ / " *<br />
**# *# ' #/"*#<br />
# ) '# # # / # <br />
" " # "<br />
<br />
- " ## " # /' "" <br />
" # " # " <br />
# ')" / #* # " <br />
' > ' #* "<br />
9 ##'*9#<br />
/##1, #* ' <br />
9#
Part fabrication<br />
Mach<strong>in</strong>e plann<strong>in</strong>g<br />
Pre-production<br />
Program plann<strong>in</strong>g<br />
Schedul<strong>in</strong>g and<br />
personnel plann<strong>in</strong>g<br />
Profile fabrication<br />
Mach<strong>in</strong>e plann<strong>in</strong>g<br />
Section and<br />
block assembly<br />
Program plann<strong>in</strong>g<br />
Schedul<strong>in</strong>g and<br />
personnel plann<strong>in</strong>g<br />
<br />
Logistics<br />
Simulation based transport<br />
and storage plann<strong>in</strong>g<br />
Hull erection<br />
Schedul<strong>in</strong>g and personnel plann<strong>in</strong>g<br />
Pa<strong>in</strong>t<strong>in</strong>g<br />
Simulation based<br />
plann<strong>in</strong>g of pa<strong>in</strong>t<strong>in</strong>g<br />
=.$88"" <br />
<br />
<br />
>. ".##*" <br />
<br />
<br />
?.
*#99 # 9#?-"#'<br />
') # "# " # # + 9 9 # <br />
##'<br />
<br />
$" ## #"# #<br />
# 3$@ ( "# # 3$@<br />
"""'"* # <br />
""<br />
<br />
/""# #"" # '$88"" <br />
7AA7" # # # ##-<br />
" " , # # ## * /<br />
*#1#') # "# / <br />
')' ##, ##* *')/<br />
#*" ""# "# 3# <br />
'" #*##8 #" #<br />
" "/"" ""# . #*<br />
# " ' "# <br />
' /*' " ##<br />
9 * #1 "# #<br />
<br />
- #) *# #" "/#<br />
' " # ##<br />
*# "# #"<br />
B 2 ' #*"# # " # / ' 0<br />
#" #- """<br />
##*# #<br />
<br />
<br />
C. # #) #" <br />
<br />
$ " $88"" "" '"<br />
) 2"# - " " <br />
"# #<br />
• *" #)2<br />
• #*#", )<br />
• #)<br />
• * # # #<br />
• ' <br />
<br />
>
D. #"<br />
<br />
/ "# , ' "# " <br />
#""" ## < <br />
# # " / <br />
#"*#"" "#"9 <br />
" #"* "#/' <br />
' ) # ' " ### ) ' <br />
'"#<br />
<br />
. " """ <br />
<br />
/ # ')# "' <br />
#*" / ' " ## <br />
"" $ " # *## ' $88 / <br />
"" ' ')29 <br />
<br />
?
A. # "' <br />
<br />
/ # ') <br />
"# - # ' " ) <br />
<br />
<br />
5" " " # ) " ) # <br />
$ # "*) # ) "## #<br />
' "' / ## <br />
" #", <br />
') ( )**+ # "# #<br />
"*# " " * <br />
/ " # " <br />
# # ") # #"'#<br />
")# " # #*# 5" " # " )<br />
#* ' " ##' <br />
<br />
!" #<br />
<br />
" # ' " $88 <br />
" #*#-# # <br />
## ## *# #<br />
"# # ## " ( $88 # <br />
#"# E " $88""# * <br />
" # ## AF )**, - "<br />
" '"')#, 9#"# <br />
/##*## $ "<br />
"# ####$ '"" '"## <br />
/" '" "# '*#<br />
"+ )## *# ' # , / <br />
*#"#"<br />
<br />
<br />
C
$%<br />
<br />
/ 0"$88'" #9*" #<br />
- "# # " ' *# #<br />
" 0 # @# <br />
#*"#" ## #35# #@*<br />
" # / $ , # # ' * <br />
')# #'"##B632*<br />
" #'"" " ## "# "<br />
' " / ' "# " ' # <br />
0 ' " # # " / " <br />
"# " # / "* " <br />
### + "9 # <br />
*" /) #"9 <br />
' 1#*" <br />
<br />
<br />
"#<br />
<br />
/5-4$(5@ 6 %7AA& - ! ' . <br />
7 #
The Potential of Free Software for Ship Design<br />
Bastiaan N. Veelo, NTNU, Trondheim/Norway, Bastiaan.N.Veelo@ntnu.no<br />
Abstract<br />
This paper considers an unorthodox approach to software development and licens<strong>in</strong>g, called Free Software,<br />
Libre Software or Open Source Software (recently abbreviated as FLOSS), with a focus on its<br />
potential <strong>in</strong> naval architecture. It is argued that this approach will <strong>in</strong>crease productivity of programm<strong>in</strong>g<br />
efforts related to ship design, and <strong>in</strong>crease the use/development ratio of the result<strong>in</strong>g software. Advantages<br />
and disadvantages of the approach are discussed, and an <strong>in</strong>dication of how FLOSS for ship design<br />
can be <strong>in</strong>itiated is given.<br />
1 Software for a Special Purpose<br />
Maritime <strong>in</strong>dustries put special demands on computer applications, as the existence of this conference<br />
illustrates. Advanced general purpose eng<strong>in</strong>eer<strong>in</strong>g design software (covered by the term “Product Lifecycle<br />
Management”, PLM) is successful <strong>in</strong> many other branches of eng<strong>in</strong>eer<strong>in</strong>g. But, despite their level<br />
of sophistication, they often fall short of our demands. Of course, computer applications exist that address<br />
some of our special needs, but because the user-base, or market, is relatively small, these products usually<br />
have one or more of the follow<strong>in</strong>g disadvantages:<br />
1. Limited functionality; especially functionality that is common <strong>in</strong> PLM applications may be lack<strong>in</strong>g.<br />
2. The user <strong>in</strong>terface that we are presented with may not be as productive as the <strong>in</strong>terfaces that most<br />
of our colleagues <strong>in</strong> general eng<strong>in</strong>eer<strong>in</strong>g enjoy.<br />
3. High licence fees for the user.<br />
4. Low profitability for the producer.<br />
Because of these disadvantages and the shortcom<strong>in</strong>gs of ma<strong>in</strong>stream CAD systems, it happens that custom<br />
software is developed <strong>in</strong>-house by an able naval architect, probably to address a small number of<br />
issues <strong>in</strong>itially. An important advantage of such an effort, besides now be<strong>in</strong>g able to address particular<br />
issues, is access to the source code. Let us have a quick look at the advantages of source access.<br />
1.1 Access to Source Code<br />
Source code is of course the textual orig<strong>in</strong> of software, that can be “understood” by both humans and<br />
mach<strong>in</strong>es. It is what designers and implementors of software (the programmers) produce, and what mach<strong>in</strong>es<br />
can compile <strong>in</strong>to a b<strong>in</strong>ary format, which then can be run on computers to execute certa<strong>in</strong> tasks.<br />
The b<strong>in</strong>ary format is mach<strong>in</strong>e-readable only, so the <strong>in</strong>ternal work<strong>in</strong>gs of the software are hidden from<br />
humans. Most software that you buy does not actually become your property, but you buy a license to use<br />
the software <strong>in</strong> its b<strong>in</strong>ary form. Any attempt to “reverse eng<strong>in</strong>eer” its <strong>in</strong>ner work<strong>in</strong>gs is often explicitly<br />
prohibited. The knowledge that is represented by the source code of the program is a secret that is fiercely<br />
protected by its licensor, and as a licensee you cannot extend or evolve the software.<br />
Access to source code, on the other hand, means you have access to the <strong>in</strong>ner work<strong>in</strong>gs of the program.<br />
Besides the fact that this <strong>in</strong>formation is valuable <strong>in</strong> its own right, it also gives you various privileged<br />
abilities, provided you have the required technical skills (which you are likely to have if you produced<br />
the source):<br />
To fix bugs. If you discover a bug <strong>in</strong> some software to which you only have access <strong>in</strong> b<strong>in</strong>ary format,<br />
the only th<strong>in</strong>g you can do is report the bug to the vendor, wait for the next version to be released<br />
399
and pay an upgrade license. This new version may or may not fix the bug, and almost certa<strong>in</strong>ly<br />
<strong>in</strong>troduces new ones. With access to the source code you can track down the problem yourself and<br />
most probably fix it yourself, here and now.<br />
To <strong>in</strong>spect it. With access to the source code, you can assess the correctness of its algorithms, and with<br />
it, the validity of the results that it produces.<br />
To customise it. The software may have been written to assist <strong>in</strong> the solution of one specific problem.<br />
At a later po<strong>in</strong>t <strong>in</strong> time, you may encounter a problem that has similarities with the earlier one.<br />
With access to the source code, you can adapt the software to also assist <strong>in</strong> the solution of this new<br />
problem.<br />
To extend it. With access to the source code of a program, you have the possibility to extend it so that<br />
it can perform more diverse tasks for you.<br />
To evolve it. No non-trivial computer program is perfect. There is always room for improvement, e.g., <strong>in</strong><br />
the user <strong>in</strong>terface, <strong>in</strong> the efficiency of its algorithms, <strong>in</strong> the accuracy of its output, to support new<br />
hardware or to make use of performance features of new processors. Access to the source code<br />
gives you the opportunity to make general improvements, or even experiment with alternative<br />
approaches to the problem.<br />
If you lack the technical skills to make use of these privileges, you can pay someone skilled, and still<br />
profit from the privileges. These privileges are so valuable that they sometimes are the ma<strong>in</strong> motivator<br />
for an <strong>in</strong>-house programm<strong>in</strong>g effort (Mutu, 2003).<br />
1.2 Parallel Efforts<br />
Raymond (1999) gives “empirical evidence that approximately 95% of [source] code is still written <strong>in</strong>house”,<br />
and one may not be exaggerat<strong>in</strong>g much by say<strong>in</strong>g that all ship design software available today<br />
started out as an <strong>in</strong>-house or even personal project. However, few of these <strong>in</strong>itiatives grow out to produce<br />
a tool that others are will<strong>in</strong>g to pay for. As software development usually does not belong to the core<br />
activities of a design office, there are not enough resources to raise the software to production quality 1 ,<br />
nor for support and ma<strong>in</strong>tenance that commercial distribution of the software would require.<br />
Therefore, there is reason to believe that, on a global basis, there is a considerable programm<strong>in</strong>g effort<br />
that is not exploited to its full potential. Now, if you are responsible for some of this effort, what can you<br />
do to make it more worthwhile?<br />
1.3 The Conventional View Reviewed<br />
One way of see<strong>in</strong>g it, is that <strong>in</strong>-house developed software is an asset of very high value. Indeed, the<br />
<strong>in</strong>vestment <strong>in</strong> programm<strong>in</strong>g effort may be colossal <strong>in</strong> relation to your sell<strong>in</strong>g bus<strong>in</strong>ess. But s<strong>in</strong>ce you are<br />
not licenc<strong>in</strong>g the software to others, it does not generate direct <strong>in</strong>come. Obviously, the software is of<br />
great value to you <strong>in</strong> the work that you do. You may then conclude that, because of the <strong>in</strong>vestment and<br />
of its value to you, it is important to keep the software to yourself. In other words, you assume that the<br />
secrecy of your software <strong>in</strong>creases the competitive position of your sell<strong>in</strong>g bus<strong>in</strong>ess. However, chances<br />
are that your competitors also use <strong>in</strong>-house developed software, and that the real advantage of mutual<br />
secret efforts is relatively limited.<br />
So what are the alternatives?<br />
Give the software away <strong>in</strong> b<strong>in</strong>ary form. As you are not mak<strong>in</strong>g any money on the software, and its<br />
value for your competitive position is debatable, one option could be to give it away. You can try<br />
1 “Production quality” <strong>in</strong> this context is a term to <strong>in</strong>dicate a level of quality that makes software sell-able and <strong>in</strong>volves<br />
stability, efficiency, ease of use, user manuals etc. In-house developed software can be used for production <strong>in</strong>-house, even if it<br />
is not of production quality.<br />
400
to create a small revenue stream by mak<strong>in</strong>g it “share-ware”, where people can pay a fee to make an<br />
advertisement notice go away. But, as the state of the software is below production quality, it is not<br />
likely to be of much use to anybody outside your department, not without the support that you are<br />
not <strong>in</strong>terested to provide, and few will th<strong>in</strong>k it is worth pay<strong>in</strong>g the share-ware fee. The benefits for<br />
you are practically zero, unless you manage to turn it <strong>in</strong>to a market<strong>in</strong>g stunt and attract attention to<br />
your sell<strong>in</strong>g bus<strong>in</strong>ess, and for others the benefits are close to zero. M<strong>in</strong>d that you should disclaim<br />
any warranty for the correctness of the software and the validity of its output, which you cannot<br />
afford without the software creat<strong>in</strong>g revenue, so serious application of the software is not attractive<br />
for third parties.<br />
Publish the source code. This option <strong>in</strong>cludes the former one, because anyone will be able to compile<br />
the software <strong>in</strong>to b<strong>in</strong>ary form. It is important to note that source code is automatically protected<br />
by <strong>in</strong>ternational copyright law, as per the Berne Convention of 1886, mean<strong>in</strong>g that no registration<br />
is required, nor is the <strong>in</strong>clusion of a copyright notice. No-one is allowed to make copies of source<br />
code or make changes without permission of its orig<strong>in</strong>ator.<br />
With the availability of the source code, third parties can now <strong>in</strong>spect it and assess the correctness<br />
of the software, and theoretically decide to use it to solve real-world problems. However, source<br />
code analysis requires a considerable effort, and copyright law prohibits third parties to make use<br />
of any of the other advantages mentioned <strong>in</strong> section 1.1, mean<strong>in</strong>g that they will have to accept a<br />
low level of quality and lack of vendor support. The protected advantages may still be important<br />
enough for others to engage <strong>in</strong> parallel programm<strong>in</strong>g efforts. There are no real benefits for you.<br />
Put the source code <strong>in</strong>to the public doma<strong>in</strong>. To unlock all advantages of source code access to everybody,<br />
you could explicitly give up on the copyright, and put your work <strong>in</strong> the public doma<strong>in</strong>. This<br />
has the potential of greatly <strong>in</strong>creas<strong>in</strong>g the use of your programm<strong>in</strong>g effort, and elim<strong>in</strong>ates the motivation<br />
for any parallel effort. Some people do this, simply because they like to see a wide adoption<br />
of their contribution. There may be some secondary economic advantages (Raymond, 1999), but<br />
the economic disadvantage can be far greater. Anybody or any corporation may legally take your<br />
code and build a proprietary computer programme on it, or <strong>in</strong>tegrate parts of it <strong>in</strong>to exist<strong>in</strong>g proprietary<br />
software. By giv<strong>in</strong>g up on the copyright completely, you allow others to make direct profit<br />
off your effort for no compensation, and you might not even know.<br />
Release the source code under relaxed copyright. Most commercial software licenses impose further<br />
restrictions, for which an agreement is necessary between licensee and licenser. But it is equally<br />
possible to formulate a license that states additional freedoms, relax<strong>in</strong>g the protection of the copyright,<br />
<strong>in</strong> exchange of specific obligations by the licensee. The license can state itself to be void <strong>in</strong><br />
case any of the obligations are breached, which effectively turns the act <strong>in</strong>to a violation of copyright<br />
law. For someone that does not accept the license and the obligations, the work is automatically<br />
protected by ord<strong>in</strong>ary copyright. For people who do accept the license, no signed agreement needs<br />
to exist for the obligations to be b<strong>in</strong>d<strong>in</strong>g.<br />
This mechanism can be used to give others access to all advantages mentioned <strong>in</strong> section 1.1, and<br />
at the same time to prohibit corporate “hijack<strong>in</strong>g” of your code. This can be achieved by requir<strong>in</strong>g<br />
any derivative work or modification of your code to be released under the same license as you<br />
did. Suddenly, for you as the copyright holder, this implies a great advantage: any improvement<br />
that anybody makes to your work, you may <strong>in</strong>corporate. This opens an enormous resource that<br />
has the potential to <strong>in</strong>crease the quality of your software to production level and evolve it <strong>in</strong>to a<br />
more powerful tool almost for free. This is the mechanism that makes Free/Libre and Open Source<br />
Software successful.<br />
2 Def<strong>in</strong>ition of Terms<br />
The English language has difficulties to put a name on software that is licenced under relaxed copyright.<br />
The term “Free Software” has been used for many years now, and is often associated with a man called<br />
401
Richard Stallman. In 1984, Stallman launched the GNU project (www.gnu.org) to develop a free operat<strong>in</strong>g<br />
system. GNU is still an important part of the now widely distributed free operat<strong>in</strong>g system that is<br />
<strong>in</strong>formally called L<strong>in</strong>ux 2 . In 1985, Stallman founded the Free Software Foundation (FSF, www.fsf.org),<br />
to promote computer users’ rights to use, study, copy, modify, and redistribute computer programs. Both<br />
Stallman and his projects still play an important role <strong>in</strong> the Free Software movement today.<br />
However, there is an ambiguity <strong>in</strong> the word “free” because it has different mean<strong>in</strong>gs when used as <strong>in</strong> “free<br />
beer” and as <strong>in</strong> “free speech”. In practice, both mean<strong>in</strong>gs often apply to Free Software, but the price of<br />
Free Software is not really important. You may sell Free Software if you can. The latter mean<strong>in</strong>g of the<br />
word “free” is about freedom and liberty, which really is the essence of Free Software.<br />
Unfortunately, most people th<strong>in</strong>k of economics when they th<strong>in</strong>k of someth<strong>in</strong>g free, and you always need<br />
to disambiguate the term when you speak of Free Software. It is possible that therefore “Free Software”<br />
does not appeal to bus<strong>in</strong>ess people. This has been considered to be a problem, which has motivated an<br />
other movement, the Open Source Initiative (www.opensource.org), <strong>in</strong> 1998. The Open Source Initiative<br />
(OSI) is a market<strong>in</strong>g program for Free Software, and aims to advocate it on solid pragmatic grounds.<br />
It has registered the term “Open Source” as a trade mark, and popularised it as a synonym for “Free<br />
Software”. OSI certifies software as Open Source, and carries a list of approved licenses (amount<strong>in</strong>g to<br />
almost 60 licenses at the time of this writ<strong>in</strong>g).<br />
Although the term “Open Source” is probably the most widely used term today, it is not without its<br />
problems either. Some people feel that it does not cover the essence, i.e., freedom, well enough (Stallman,<br />
2002). Indeed, sometimes software is claimed to be Open Source, whereas it really only is published<br />
source code, without the copyright hav<strong>in</strong>g been relaxed. As a matter of fact, OSI clarifies that “Open<br />
Source doesn’t just mean access to the source code” and gives a detailed def<strong>in</strong>ition of the term, consist<strong>in</strong>g<br />
of ten criteria that software must comply with before it may be called Open Source. These criteria differ<br />
slightly from the ones that the FSF practices; some of the license restrictions that the OSI accepts for<br />
Open Source are too restrictive <strong>in</strong> the eyes of the FSF (FSF, 2001).<br />
Dur<strong>in</strong>g a translation of the English text of the most important Free Software license, the GNU General<br />
Public License (GPL), <strong>in</strong>to Spanish, it appeared that Roman languages do not suffer the ambiguity<br />
problem (González-Barahona, 2004). The word “free” with respect to price is translated <strong>in</strong>to Spanish<br />
as “gratis”, and “free” with respect to liberty is translated as “libre”. Because of the problems with the<br />
terms “Free Software” and “Open Source”, some people have started to use the term “Libre Software”,<br />
even though this is a mix of languages. This term is ga<strong>in</strong><strong>in</strong>g popularity, notably <strong>in</strong> official circles <strong>in</strong> Europe.<br />
For example, <strong>in</strong> 1999, the Information Society Directorate General of the European Commission<br />
<strong>in</strong>itiated the European Work<strong>in</strong>g Group on Libre Software (eu.conecta.it).<br />
In 2002, the acronym FLOSS has been <strong>in</strong>troduced, <strong>in</strong> a survey and study commissioned by the European<br />
Commission (Ghosh and Glott, 2002), to cover all of Free/Libre and Open Source Software at once.<br />
It is important to note that both Free Software and Open Source do not differentiate between the last two<br />
options of section 1.3. That is, both Free Software and Open Source software recognise source code that<br />
has been put <strong>in</strong> the public doma<strong>in</strong>, because everyone is free to use such software <strong>in</strong> any way they like.<br />
Software that is guaranteed to be FLOSS <strong>in</strong> all its derivative variants, by means of a clause that makes<br />
derivative works <strong>in</strong>herit the licence of the orig<strong>in</strong>al and the requirement that source code always must be<br />
made available, is often identified with the term “Copyleft” (as opposed to “Copyright”). In general, the<br />
advantages of FLOSS are considered to be highest under a Copyleft license.<br />
2.1 FLOSS Licences<br />
To give an exhaustive list of officially recognised Free/Libre and Open Source Software licenses is<br />
far beyond the scope of this paper. For this the reader is referred to the sites of the FSF and OSI,<br />
who ma<strong>in</strong>ta<strong>in</strong> extensive lists of approved licenses (www.gnu.org/licenses/license-list.html and<br />
2 To be precise, L<strong>in</strong>ux is actually only the kernel <strong>in</strong> this operat<strong>in</strong>g system.<br />
402
www.opensource.org/licenses/). However, the important question is not whether any particular license<br />
is approved by either organisation. The important question is whether a piece of software that you<br />
are <strong>in</strong>terested <strong>in</strong> is licensed under terms that you can accept, and for this you should always consult<br />
the actual license text that follows with the software. The other important question is which license to<br />
choose or construct when you consider to write FLOSS or consider to release exist<strong>in</strong>g work as FLOSS.<br />
To help answer the latter question, both organisations have put together guides and “HOWTOs”, see<br />
(www.fsf.org/licens<strong>in</strong>g) and (Raymond and Raymond, 2002).<br />
For completeness however, we briefly expla<strong>in</strong> the licenses that are used for relevant FLOSS as listed <strong>in</strong><br />
the appendix. Both the MIT and BSD licences allow almost everyth<strong>in</strong>g, except removal of the license<br />
statement and copyright notice. They also disclaim warranty and liability, and the BSD licence adds a<br />
non-endorsement clause. The follow<strong>in</strong>g licenses are all Copyleft. The most widely used license is the<br />
GNU GPL license, or General Public Licence. Proprietary software can not <strong>in</strong>clude or even l<strong>in</strong>k with<br />
GPLed code. The QPL, or Qt Public License, is rarely used, and is comparable but <strong>in</strong>compatible with<br />
the GPL. The GNU LGPL, or Lesser General Public Licence, is similar to the GPL but allows LGPLed<br />
libraries to be l<strong>in</strong>ked to by proprietary software. The wxW<strong>in</strong>dows Library Licence is identical to the<br />
LGPL, however with an exception that states additional freedoms for b<strong>in</strong>ary distribution.<br />
3 The Case for FLOSS<br />
Due to space limitations, this paper is not go<strong>in</strong>g to do enough justice to the case for Free/Libre and<br />
Open Source Software. Many essays, articles and even books have been written on the subject by<br />
far better analysts and FLOSS advocates. For example, “The Cathedral and the Bazaar” by Raymond<br />
(1999) is very much worth a read, especially the onl<strong>in</strong>e version that conta<strong>in</strong>s references to comments<br />
and replies by others. The <strong>in</strong>formation on the OSI website is a very good place to start, because it<br />
concisely discusses the case for FLOSS from the viewpo<strong>in</strong>ts of bus<strong>in</strong>ess management, users and of<br />
developers <strong>in</strong>dividually (www.opensource.org/advocacy/). The phenomenon of FLOSS is actively<br />
be<strong>in</strong>g studied, and MIT carries an impressive list of onl<strong>in</strong>e articles from the FLOSS research community<br />
(opensource.mit.edu/onl<strong>in</strong>e_papers.php). Large scale surveys of the FLOSS developer community<br />
have been performed by Gosh and Glott (2002) and David et al. (2003), <strong>in</strong> which respectively 2784<br />
and 1588 developers participated.<br />
Especially the OSI helps you weigh the advantages and disadvantages between an open development<br />
model and a closed development model. Here it is often assumed that you are able to extract direct<br />
revenue out of your proprietary software, or <strong>in</strong> Raymond’s (1999) words “collect rent from your secret<br />
bits”. The OSI expla<strong>in</strong>s that even if this is the case, the pay-off of convert<strong>in</strong>g to FLOSS can be higher<br />
than the pay-off of rema<strong>in</strong><strong>in</strong>g <strong>in</strong> proprietary mode. However, that consideration is outside the scope of<br />
this paper, and we will proceed with the simpler assumption that you are the only user of your <strong>in</strong>-house<br />
developed software.<br />
There are three ma<strong>in</strong> advantages of writ<strong>in</strong>g FLOSS as opposed to writ<strong>in</strong>g software <strong>in</strong>-house. Firstly, you<br />
get valuable peer review. Secondly, you get to use free build<strong>in</strong>g blocks. And lastly, you work <strong>in</strong> an organisational<br />
mode that is arguably the most effective for the management of complex systems (Raymond<br />
1999).<br />
3.1 Peer Review<br />
The peer review helps improv<strong>in</strong>g the quality and feature-richness of your software. Because the software<br />
is available freely, a community of “beta testers” (your users) will connect with you and help you f<strong>in</strong>d<br />
bugs. And because they have access to the code and are allowed to play with it, they will send you bug<br />
fixes if they can, <strong>in</strong> exchange for just gratification and recognition. For the same reasons, you may get<br />
sent patches that add new features to your software.<br />
403
3.2 Free Build<strong>in</strong>g Blocks<br />
Because you are writ<strong>in</strong>g FLOSS, you are free to build on exist<strong>in</strong>g FLOSS. You can do that by l<strong>in</strong>k<strong>in</strong>g<br />
with one or more of the many excellent libraries that exist for different purposes, and thereby use the<br />
peer reviewed work of others <strong>in</strong>stead of re<strong>in</strong>vent<strong>in</strong>g the wheel yourself. Also, <strong>in</strong>stead of writ<strong>in</strong>g a new<br />
computer program from the ground up to help do your work, you may be able to f<strong>in</strong>d an exist<strong>in</strong>g project,<br />
either fully functional or <strong>in</strong> early stages of development, that you can extend to make it satisfy your<br />
requirements. In these ways you can have a fly<strong>in</strong>g start and reach your goal much sooner than when<br />
do<strong>in</strong>g it all alone.<br />
The risk of depend<strong>in</strong>g on third party code that exists <strong>in</strong> a proprietary sett<strong>in</strong>g, namely that the software<br />
may disappear because of a bus<strong>in</strong>ess situation with the vendor, is non-existent with FLOSS. If a piece<br />
of FLOSS gets dropped by its orig<strong>in</strong>ators, you are free to pick it up and ma<strong>in</strong>ta<strong>in</strong> and evolve it yourself.<br />
Therefore, a FLOSS project will thrive as long as there are people <strong>in</strong>terested <strong>in</strong> it. The same advantage<br />
applies to the accessibility of data files. The <strong>in</strong>formation <strong>in</strong> old data files, written <strong>in</strong> a proprietary format<br />
with a proprietary program that is not available anymore, is <strong>in</strong>accessible and effectively lost. But if you<br />
have the source code of the application that wrote the data file, you can understand the format and extract<br />
the <strong>in</strong>formation.<br />
3.3 Your Competitive Position<br />
If you are still reluctant to publish work from which your competitor may benefit, you may consider<br />
the follow<strong>in</strong>g. Although maritime <strong>in</strong>dustries <strong>in</strong>creas<strong>in</strong>gly play on a global market, most of your bus<strong>in</strong>ess<br />
likely still relates to a certa<strong>in</strong> geographic area. However, FLOSS development takes place on the Internet,<br />
where location is irrelevant. Whilst you compete more or less locally, you will collaborate globally <strong>in</strong><br />
FLOSS development. Also, even when your nearest competitor starts us<strong>in</strong>g your software <strong>in</strong>tensively, he<br />
will make improvements and he will have to contribute them back to you and make your software better.<br />
Therefore, it is important to be the first to embark <strong>in</strong> FLOSS mode, because it gives you more control of<br />
the situation. You still own the software, and it is you who decides where to take it. Raymond (1999) also<br />
says about this:<br />
[T]here’s a serious opportunity risk <strong>in</strong> wait<strong>in</strong>g too long: you could get scooped by a competitor<br />
go<strong>in</strong>g open-source <strong>in</strong> the same market niche.<br />
The reason this is a serious issue is that both the pool of users and the pool of talent available<br />
to be recruited <strong>in</strong>to open-source cooperation for any given product category is limited, and<br />
recruitment tends to stick. If two producers are the first and second to open-source compet<strong>in</strong>g<br />
code of roughly equal function, the first is likely to attract the most users and the most and<br />
best-motivated co-developers; the second will have to take leav<strong>in</strong>gs.<br />
So yes, your competitor may benefit from your actions, but you may benefit more.<br />
3.4 Invitation to Take the Lead<br />
If you are a scientist, commercial competition may not be so much of an issue, and with that the choice<br />
for FLOSS should be easy. FLOSS poses great opportunities for learn<strong>in</strong>g, and universities are encouraged<br />
to take the lead <strong>in</strong> develop<strong>in</strong>g FLOSS, also for maritime applications.<br />
3.5 Play the Game Well<br />
Before jump<strong>in</strong>g on the FLOSS bandwagon <strong>in</strong> all enthusiasm, be sure to be prepared for the rules of the<br />
game. FLOSS development is a voluntary community effort, and people skills are essential when lead<strong>in</strong>g<br />
one. This is especially the case because the Internet is a poor medium for social communication. In order<br />
to make FLOSS development work for you, you need to understand the process very well, or have a<br />
natural feel<strong>in</strong>g for it. Aga<strong>in</strong>, the writ<strong>in</strong>gs of Raymond (1999) can be a source of <strong>in</strong>sight.<br />
404
4 Conclusion<br />
This article has discussed how Free/Libre and Open Source Software can be an attractive alternative to<br />
<strong>in</strong>-house developed software for ship design. The freedom to use exist<strong>in</strong>g FLOSS build<strong>in</strong>g blocks is an<br />
important motivator to get <strong>in</strong>volved <strong>in</strong> FLOSS development. Therefore, this article has an appendix with<br />
references to selected FLOSS projects of various categories that are relevant to ship design. The use of<br />
one or more of these build<strong>in</strong>g blocks makes a fly<strong>in</strong>g start possible <strong>in</strong> the <strong>in</strong>itiation of FLOSS for ship<br />
design. The appendix also shows that it is already possible to use FLOSS for some aspects of ship design<br />
today, and that you will not have to pioneer the concept. If you decide to get actively <strong>in</strong>volved, you will<br />
not be alone.<br />
References<br />
DAVID, P.A.; WATERMAN, A.; ARORA, s. (2003), FLOSS-US — the free/libre/open source software<br />
survey for 2003, Stanford Institute for Economic Policy Research. www.stanford.edu/group/<br />
floss-us/<br />
FSF (2001), Categories of Free and Non-Free Software, Free Software Foundation. www.gnu.org/<br />
philosophy/categories.html<br />
GHOSH, R.A.; GLOTT, R. (2002), Free/libre and open source software: survey and study, International<br />
Institute of Infonomics, University of Maastricht, The Netherlands; and Berlecon Research GmbH,<br />
Berl<strong>in</strong>, Germany. www.<strong>in</strong>fonomics.nl/FLOSS/<br />
GONZÁLEZ-BARAHONA, J.M. (2004), Quo vadis, libre software?, Universidad Rey Juan Carlos,<br />
Spa<strong>in</strong>. s<strong>in</strong>etgy.org/jgb/articulos/libre-software-orig<strong>in</strong>/<br />
MUTU, V.; IONAS, O.; GAVRILESCU, I. (2003), SURF — an <strong>in</strong>-house development for a ship hull<br />
design software, 2 nd Int. Conf. Computer and IT Applic. Mar. Ind., <strong>COMPIT</strong>, <strong>Hamburg</strong>, pp.387–399<br />
RAYMOND, E.S.; RAYMOND, C.O. (2002), Licens<strong>in</strong>g HOWTO, draft Open Source Initiative work<strong>in</strong>g<br />
paper. www.catb.org/˜esr/Licens<strong>in</strong>g-HOWTO.html<br />
RAYMOND, E.S. (1999), The Cathedral & the Bazaar, O’Reilly, ISBN 1-56592-724-9, www.catb.<br />
org/esr/writ<strong>in</strong>gs/cathedral-bazaar/<br />
STALLMAN, R.M. (2002) Why “Free Software” is Better Than “Open Source”, <strong>in</strong> Free<br />
Software, Free Society, GNU Press, ISBN 1-882114-98-1. www.gnu.org/philosophy/<br />
free-software-for-freedom.html<br />
APPENDIX<br />
A<br />
FLOSS Software<br />
Examples of exist<strong>in</strong>g Free/Libre and Open Source Software are given <strong>in</strong> six categories: geometric modell<strong>in</strong>g,<br />
visualisation, numerics, f<strong>in</strong>ite element method, computational fluid dynamics and graphical user<br />
<strong>in</strong>terfaces. The <strong>in</strong>formation about each project also <strong>in</strong>cludes the latest version and release date at the time<br />
of this writ<strong>in</strong>g, which is an <strong>in</strong>dication of the degree of maturity and the developer activity respectively.<br />
Note that the list is by no means exhaustive.<br />
A.1 Geometric Modell<strong>in</strong>g<br />
A.1.1<br />
Blender<br />
Blender is a 3D graphics creation suite with a focus on the enterta<strong>in</strong>ment <strong>in</strong>dustry. Its modell<strong>in</strong>g capabilities<br />
<strong>in</strong>clude polygon meshes, NURBS surfaces, Bézier and B-spl<strong>in</strong>e curves, meta-balls, and “Smooth<br />
405
proxy” style Catmull-Clark subdivision surfaces with optimal iso-l<strong>in</strong>es display and sharpness edit<strong>in</strong>g. It<br />
supports Python script<strong>in</strong>g for the creation of custom tools.<br />
Homepage: www.blender.org<br />
Latest version: 2.36, 22 Dec 2004<br />
License: GPL 2<br />
Language: C, C++, Python<br />
Platforms: W<strong>in</strong>dows, Mac OS X, L<strong>in</strong>ux,<br />
Solaris, Irix, FreeBSD<br />
Fig.1: The Blender modeller<br />
A.1.2<br />
W<strong>in</strong>gs 3D<br />
W<strong>in</strong>gs 3D is a polygon mesh modeller with a user <strong>in</strong>terface that is easy to use for both beg<strong>in</strong>ners and<br />
advanced users. W<strong>in</strong>gs 3D uses subdivision surfaces rather than NURBS surfaces.<br />
Fig.2: The W<strong>in</strong>gs3D modeller<br />
Homepage: www.w<strong>in</strong>gs3d.com<br />
Latest version: 0.98.26b, 16 Dec 2004<br />
License: BSD License<br />
Language: Erlang<br />
Platforms: W<strong>in</strong>dows, Mac OS X, L<strong>in</strong>ux and<br />
other Unices.<br />
(Image source:<br />
www.naval-architecture.co.uk/phpbb/<br />
viewtopic.php?t=12)<br />
A.1.3<br />
Co<strong>in</strong>3d<br />
Co<strong>in</strong>3D is a set of libraries used for creat<strong>in</strong>g 3D graphics applications. Co<strong>in</strong>3D is fully compatible with<br />
SGI Open Inventor 2.1, the de facto standard for 3D visualisation and visual simulation software <strong>in</strong><br />
the scientific and eng<strong>in</strong>eer<strong>in</strong>g community. Additional features <strong>in</strong> Co<strong>in</strong>3D <strong>in</strong>clude VRML97 support, 3D<br />
sound, 3D textures, and parallel render<strong>in</strong>g on multiple processors. See Fig.4.<br />
A.1.4<br />
Open CASCADE<br />
Open CASCADE is an <strong>in</strong>dustrial Open Source alternative to proprietary 3D modell<strong>in</strong>g kernels, a library<br />
that <strong>in</strong>cludes components for 3D surface and solid modell<strong>in</strong>g, visualisation, data exchange and rapid<br />
application development. Open CASCADE Technology can be best applied <strong>in</strong> development of numerical<br />
simulation software <strong>in</strong>clud<strong>in</strong>g CAD/CAM/CAE, AEC and GIS, as well as PDM applications. The<br />
Technology exists from the mid 1990-s and has already been used by numerous commercial clients. See<br />
Fig.5.<br />
406
A.1.5<br />
FreeCAD<br />
FreeCAD will be a general purpose 3D CAD system based on OpenCasCade. FreeCAD will aim directly<br />
at mechanical eng<strong>in</strong>eer<strong>in</strong>g, product design and related features (like CatiaV4 and V5, and SolidWorks).<br />
It will be a Feature-Based parametric modeller. Script<strong>in</strong>g is supported through Python.<br />
Homepage:<br />
Latest version:<br />
License:<br />
Language:<br />
Platforms:<br />
free-cad.sourceforge.<br />
net<br />
0.1 B107<br />
GPL, LGPL<br />
C++, Python<br />
L<strong>in</strong>ux, W<strong>in</strong>dows<br />
Fig.3: The FreeCAD system<br />
Homepage: http://www.co<strong>in</strong>3d.<br />
org/<br />
Latest version: 2.3.0, 22 Jun 2004<br />
License: GPL (commercial license<br />
available)<br />
Language:<br />
Platforms:<br />
C++<br />
any UNIX / L<strong>in</strong>ux / *BSD<br />
platform, all Microsoft W<strong>in</strong>dows<br />
operat<strong>in</strong>g systems, and Mac OS X<br />
Fig.4: The Co<strong>in</strong>3D graphics library<br />
Homepage: www.opencascade.org<br />
Latest version: 5.2, July 2004<br />
License: LGPL-like with certa<strong>in</strong> differences<br />
Language: C++<br />
Platforms: L<strong>in</strong>ux Intel, W<strong>in</strong>dows Intel and<br />
Sun Sparc<br />
(Image: RINA F<strong>in</strong>ite-element pre- and post-processor<br />
for ship certification)<br />
Fig.5: The OpenCASCADE surface and solid modell<strong>in</strong>g kernel<br />
A.1.6<br />
BRL-CAD<br />
A powerful constructive solid geometry (CSG) solid modell<strong>in</strong>g system that <strong>in</strong>cludes an <strong>in</strong>teractive geometry<br />
editor, ray trac<strong>in</strong>g support for render<strong>in</strong>g and geometric analysis, network distributed frame-buffer<br />
support, and image and signal-process<strong>in</strong>g tools.<br />
407
Homepage: brlcad.sourceforge.<br />
net<br />
Latest version: 7.0.2, 6 Jan <strong>2005</strong><br />
License: BSD License, GPL, LGPL<br />
Language: C, C++, Java, PHP, Tcl,<br />
Unix Shell<br />
Platforms: W<strong>in</strong>dows and all BSD and POSIX<br />
platforms, <strong>in</strong>clud<strong>in</strong>g L<strong>in</strong>ux,<br />
Mac OS X, Solaris, Irix etc.<br />
Fig.6: The BRL-CAD solid modell<strong>in</strong>g system<br />
A.1.7<br />
Varkon<br />
VARKON can be used as a traditional CAD-system with draft<strong>in</strong>g, modell<strong>in</strong>g and visualisation if you<br />
want to, but the real power of VARKON is <strong>in</strong> parametric modell<strong>in</strong>g and CAD applications development.<br />
VARKON <strong>in</strong>cludes <strong>in</strong>teractive parametric modell<strong>in</strong>g <strong>in</strong> 2D or 3D but also the unique MBS programm<strong>in</strong>g<br />
language <strong>in</strong>tegrated <strong>in</strong> the graphical environment. VARKON is designed to be modified and extended<br />
with knowledge and functionality specific to a certa<strong>in</strong> product or problem. It is not a true solid modeller.<br />
Homepage: www.tech.oru.se/cad/<br />
varkon/<br />
Latest version: 1.18A, 18 Oct 2004<br />
License: LGPL 2<br />
Language:<br />
Platforms:<br />
C<br />
HP-UX, (Sun), AIX, Irix, VMS,<br />
SCO/UNIX, FreeBSD,<br />
GNU/L<strong>in</strong>ux, Cray, (W<strong>in</strong>dows at a<br />
fee)<br />
Fig.7: The Varkon CAD system<br />
A.1.8<br />
Sailcut CAD<br />
Sailcut CAD is a sail design and plott<strong>in</strong>g software. It allows you to design and visualise your own sail<br />
and compute the accurate development of all panels <strong>in</strong> flat sheets.<br />
Homepage: sailcut.sourceforge.<br />
net<br />
Latest version: 1.0, 20 Dec 2004<br />
License: GPL<br />
Language:<br />
Platforms:<br />
C++<br />
GNU/L<strong>in</strong>ux, Mac OS X and<br />
W<strong>in</strong>dows<br />
Fig.8: The Sailcut CAD sail design and plott<strong>in</strong>g software<br />
408
A.1.9<br />
SketchBoard<br />
SketchBoard is a sketch oriented CAD software. It is a tool for designers to develop their design by<br />
sketch<strong>in</strong>g and modell<strong>in</strong>g <strong>in</strong> 3D. It has a system that helps convert<strong>in</strong>g the sketch <strong>in</strong>to a polygon.<br />
Homepage: sketchboard.<br />
sourceforge.net<br />
Latest version: 1.22, 22 Apr 2004<br />
License: GPL<br />
Language: C++<br />
Platforms: W<strong>in</strong>dows<br />
Fig.9: The SketchBoard CAD software<br />
A.2 Visualisation<br />
A.2.1<br />
ADMesh<br />
Processor for repair<strong>in</strong>g, transform<strong>in</strong>g and merg<strong>in</strong>g triangulated solid meshes. Reads STL file format<br />
(used for rapid prototyp<strong>in</strong>g applications) and writes STL, VRML, OFF, and DXF files.<br />
(console application)<br />
Homepage: www.varlog.com/<br />
products/admesh/<br />
Latest version: 0.95, 3 Sep 2003<br />
License: GPL 2<br />
Language: C<br />
Platforms: L<strong>in</strong>ux, SunOS 4.1.3, IRIX 5.2,<br />
HP-UX, (W<strong>in</strong>dows)<br />
Fig.10: The ADMesh processor<br />
A.2.2<br />
Pixie<br />
Pixie is a photo-realistic renderer that uses a RenderMan-like <strong>in</strong>terface. Features <strong>in</strong>clude programmable<br />
shad<strong>in</strong>g, motion blur, depth of field, ray-trac<strong>in</strong>g, scan-l<strong>in</strong>e render<strong>in</strong>g, occlusion cull<strong>in</strong>g, global illum<strong>in</strong>ation,<br />
caustics, etc.<br />
Homepage: pixie.sourceforge.net<br />
Latest version: 1.3.26, 29 Dec 2004<br />
License: GPL<br />
Language: C, C++<br />
Platforms: W<strong>in</strong>dows, All POSIX<br />
(L<strong>in</strong>ux/BSD/UNIX-like OSes)<br />
Fig.11: The Pixie renderer<br />
409
A.2.3<br />
YafRay<br />
YafRay is a powerful ray-tracer. It enables you to create fantastic images and animations of a photo<br />
realistic quality. Thanks to its API (Application Programm<strong>in</strong>g Interface) and its modular structure, it is<br />
possible to develop render<strong>in</strong>g plug-<strong>in</strong>s, mak<strong>in</strong>g it possible to use YafRay from any program or 3D suite.<br />
At the time of this writ<strong>in</strong>g, suites as Blender, W<strong>in</strong>g3D or Aztec take advantage of this feature.<br />
Homepage: www.yafray.org<br />
Latest version: 0.0.7, 5 Aug 2004<br />
License: LGPL 2.1<br />
Language: C++<br />
Platforms: GNU/L<strong>in</strong>ux, W<strong>in</strong>dows 9x/XP/2K,<br />
Mac OS X and Irix<br />
Fig.12: The YafRay ray-tracer<br />
A.2.4<br />
GeomView<br />
GeomView is an <strong>in</strong>teractive 3D view<strong>in</strong>g program for Unix. It can be used as a stand-alone viewer for<br />
static objects or as a display eng<strong>in</strong>e for other programs which produce dynamically chang<strong>in</strong>g geometry.<br />
It can display objects described <strong>in</strong> a variety of file formats.<br />
Homepage: www.geomview.org<br />
Latest version: 1.8.1, 25 Mar 2001<br />
License: LGPL 2.1<br />
Language: C, C++<br />
Platforms: Unix-like OSes <strong>in</strong>clud<strong>in</strong>g but not<br />
limited to GNU/L<strong>in</strong>ux, Solaris,<br />
IRIX, and AIX<br />
Fig.13: The GeomView viewer<br />
A.2.5<br />
MayaVi<br />
MayaVi is a free, easy to use scientific data visualiser. See Fig.15.<br />
A.3 Numerics<br />
Beyond the software referenced below, there is a host of other powerful numerics libraries. Many of them<br />
are listed at www.oonumerics.org/oon.<br />
410
A.3.1<br />
Octave<br />
GNU Octave is a high-level language, primarily <strong>in</strong>tended for numerical computations. It provides a<br />
convenient command l<strong>in</strong>e <strong>in</strong>terface for solv<strong>in</strong>g l<strong>in</strong>ear and nonl<strong>in</strong>ear problems numerically, and for perform<strong>in</strong>g<br />
other numerical experiments us<strong>in</strong>g a language that is mostly compatible with Matlab. It may<br />
also be used as a batch-oriented language.<br />
Octave has extensive tools for solv<strong>in</strong>g common numerical l<strong>in</strong>ear algebra problems, f<strong>in</strong>d<strong>in</strong>g the roots of<br />
nonl<strong>in</strong>ear equations, <strong>in</strong>tegrat<strong>in</strong>g ord<strong>in</strong>ary functions, manipulat<strong>in</strong>g polynomials, and <strong>in</strong>tegrat<strong>in</strong>g ord<strong>in</strong>ary<br />
differential and differential-algebraic equations. It is easily extensible and customisable via user-def<strong>in</strong>ed<br />
functions written <strong>in</strong> Octave’s own language, or us<strong>in</strong>g dynamically loaded modules written <strong>in</strong> C++, C,<br />
Fortran, or other languages.<br />
Homepage: www.octave.org,<br />
octave.sourceforge.<br />
net<br />
Latest version: 2.9.0, 15 Mar <strong>2005</strong><br />
License: GPL 2<br />
Language: C, C++, Fortran<br />
Platforms: L<strong>in</strong>ux, Mac OS X, W<strong>in</strong>dows<br />
Fig.14: The Octave language (mostly Matlab compatible)<br />
Homepage: mayavi.sourceforge.<br />
net<br />
Latest version: 1.3, 18 Nov 2003<br />
License: BSD license<br />
Language:<br />
Platforms:<br />
Python<br />
L<strong>in</strong>ux, Mac OS X and other<br />
Unices, W<strong>in</strong>dows<br />
Fig.15: The MayaVi data visualiser<br />
A.3.2<br />
FreeMat<br />
FreeMat is a free environment for rapid eng<strong>in</strong>eer<strong>in</strong>g and scientific prototyp<strong>in</strong>g and data process<strong>in</strong>g. It<br />
is similar to proprietary systems such as MATLAB from Mathworks, and IDL from Research Systems.<br />
FreeMat <strong>in</strong>cludes several novel features such as a code-less <strong>in</strong>terface to external C/C++/FORTRAN<br />
code, parallel/distributed algorithm development (via MPI), and plott<strong>in</strong>g and visualisation capabilities.<br />
See Fig.17.<br />
A.3.3<br />
Maxima<br />
Maxima is a fairly complete computer algebra system written <strong>in</strong> lisp with an emphasis on symbolic<br />
computation. Its abilities <strong>in</strong>clude symbolic <strong>in</strong>tegration, 3D plott<strong>in</strong>g, and an ODE solver. See Fig.18.<br />
A.3.4<br />
ATLAS<br />
The ATLAS (Automatically Tuned L<strong>in</strong>ear Algebra Software) project is an ongo<strong>in</strong>g research effort focus<strong>in</strong>g<br />
on apply<strong>in</strong>g empirical techniques <strong>in</strong> order to provide portable performance. At present, it provides<br />
411
C and Fortran77 <strong>in</strong>terfaces to a portably efficient BLAS implementation, as well as a few rout<strong>in</strong>es from<br />
LAPACK.<br />
Homepage: math-atlas.<br />
sourceforge.net<br />
Latest version: 3.7.8, 23 Jul 2004<br />
License: BSD License<br />
Language: C, Fortran77<br />
Platforms: L<strong>in</strong>ux and other Unices, W<strong>in</strong>dows<br />
Fig.16: The Automatically Tuned L<strong>in</strong>ear Algebra Software<br />
Homepage: freemat.sourceforge.<br />
net<br />
Latest version: 1.09, 27 Oct 2004<br />
License: MIT-type<br />
Language: C, C++, FORTRAN<br />
Platforms: L<strong>in</strong>ux, W<strong>in</strong>dows and Mac OS X<br />
Fig.17: FreeMat, an other alternative to Matlab<br />
Homepage: maxima.sourceforge.<br />
net<br />
Latest version: 5.9.1, 22 Sept 2004<br />
License: GPL<br />
Language:<br />
Platforms:<br />
C, Lisp<br />
All POSIX<br />
(L<strong>in</strong>ux/BSD/UNIX-like OSes)<br />
Fig.18: The Maxima computer algebra system<br />
A.4 FEM<br />
A.4.1<br />
Gmsh (FEM Mesh Generator)<br />
Gmsh is an automatic 3D f<strong>in</strong>ite element grid generator (primarily Delaunay) with a build-<strong>in</strong> CAD eng<strong>in</strong>e<br />
and post-processor. Its design goal is to provide a simple mesh<strong>in</strong>g tool for academic problems with<br />
parametric <strong>in</strong>put and advanced visualisation capabilities.<br />
412
Homepage: www.geuz.org/gmsh/<br />
Latest version: 1.58, 1 Jan <strong>2005</strong><br />
License: GPL 2<br />
Language: C++<br />
Platforms: W<strong>in</strong>dows, L<strong>in</strong>ux, Mac OS X<br />
Fig.19: The Gmsh mesh generator<br />
A.4.2<br />
NETGEN — Automatic Mesh Generator<br />
NETGEN is an automatic 3D tetrahedral mesh generator. It accepts <strong>in</strong>put from constructive solid geometry<br />
(CSG) or boundary representation (BRep) from STL file format. The connection to a geometry<br />
kernel allows the handl<strong>in</strong>g of IGES and STEP files. NETGEN conta<strong>in</strong>s modules for mesh optimisation<br />
and hierarchical mesh ref<strong>in</strong>ement. Works together with NGSolve (see Section A.4.3).<br />
Homepage: www.hpfem.jku.at/<br />
netgen/<br />
Latest version: 4.4, 17 Nov 2004<br />
License: LGPL 2.1<br />
Language: C++<br />
Platforms: FreeBSD, L<strong>in</strong>ux<br />
Fig.20: The NETGEN mesh generator<br />
A.4.3<br />
NGSolve<br />
NGSolve is a general purpose 3D f<strong>in</strong>ite element solver, support<strong>in</strong>g boundary value problems, <strong>in</strong>itialboundary<br />
value problems and Eigenvalue problems for the available types of equations, namely scalar<br />
(heat flow), elasticity, and magnetic field. NGSolve performs adaptive mesh ref<strong>in</strong>ement, the matrix equations<br />
are solved by optimal order multi-grid methods. Works together with NETGEN (see Section A.4.2).<br />
See also Fig.22.<br />
A.4.4<br />
CalculiX<br />
A Three-Dimensional Structural F<strong>in</strong>ite Element Program designed to solve field problems. With CalculiX<br />
F<strong>in</strong>ite Element Models can be build, calculated and post-processed. The pre- and post-processor is<br />
an <strong>in</strong>teractive 3D-tool us<strong>in</strong>g the OpenGL API. The solver is able to do l<strong>in</strong>ear and non-l<strong>in</strong>ear calculations.<br />
Static, dynamic and thermal solutions are available. Both programs can be used <strong>in</strong>dependently. Because<br />
the solver makes use of the Abaqus <strong>in</strong>put format, it is possible to use proprietary pre-processors as well.<br />
413
In turn, the pre-processor is able to write mesh related data for Nastran, Abaqus and Ansys and for the<br />
free cfd code duns. A VDA CAD <strong>in</strong>terface is available.<br />
Homepage: www.calculix.de<br />
Latest version: 1.2, 25 Jul 2004<br />
License: GPL 2<br />
Language: Fortran77, C<br />
Platforms: Unix, <strong>in</strong>clud<strong>in</strong>g L<strong>in</strong>ux and Irix<br />
Fig.21: The CalculiX solver<br />
Homepage: www.hpfem.jku.at/<br />
ngsolve/<strong>in</strong>dex.html<br />
Latest version: 4.4, 17 Nov 2004<br />
License: LGPL 2.1<br />
Language: C++<br />
Platforms: FreeBSD, L<strong>in</strong>ux<br />
Fig.22: The NGSolve solver<br />
A.4.5<br />
SLFFEA<br />
SLFFEA stands for San Le’s Free F<strong>in</strong>ite Element Analysis. It is a package of scientific software and<br />
graphical user <strong>in</strong>terfaces for use <strong>in</strong> (non-l<strong>in</strong>ear) f<strong>in</strong>ite element analysis.<br />
Homepage: slffea.sourceforge.<br />
net<br />
Latest version: 1.3, 14 Nov 2003<br />
License: LGPL 2<br />
Language: C<br />
Platforms: L<strong>in</strong>ux, (W<strong>in</strong>dows)<br />
Fig.23: San Le’s Free F<strong>in</strong>ite Element Analysis<br />
A.4.6<br />
TOCHNOG<br />
TOCHNOG is a feature-rich f<strong>in</strong>ite element program. Among the FE models supported are: differential<br />
equations (materials), convection-diffusion equations, Stokes and Navier-Stokes (fluids), elasticity<br />
(isotropy and transverse isotropy), plasticity (Von-Mises, Mohr-Coulomb, etc.; plastic surfaces can be<br />
414
arbitrarily comb<strong>in</strong>ed). Residues <strong>in</strong> equations and error estimates for all data can be pr<strong>in</strong>ted or plotted<br />
us<strong>in</strong>g gnuplot/plotmtv, CalculiX or gmsh.<br />
Homepage: tochnog.sourceforge.<br />
net<br />
Latest version: 26 Nov 2001<br />
License: GPL 2<br />
Language:<br />
Platforms:<br />
C++<br />
POSIX (such as L<strong>in</strong>ux, BSD,<br />
UNIX-like OSes) and W<strong>in</strong>dows<br />
Fig.24: The TOCHNOG f<strong>in</strong>ite element programme<br />
A.4.7<br />
OFELI<br />
OFELI (Object F<strong>in</strong>ite Element LIbrary) is a library of f<strong>in</strong>ite element C++ classes for multipurpose development<br />
of f<strong>in</strong>ite element software. It is <strong>in</strong>tended for teach<strong>in</strong>g, research and <strong>in</strong>dustrial developments as<br />
well.<br />
Homepage: www.ofeli.net<br />
Latest version: 1.3.0-3, 5 Dec 2004<br />
License: GPL<br />
Language: C++<br />
Platforms: W<strong>in</strong>dows, All POSIX<br />
(L<strong>in</strong>ux/BSD/UNIX-like OSes)<br />
Fig.25: The Object F<strong>in</strong>ite Element LIbrary<br />
A.5 CFD<br />
A.5.1<br />
FEATFLOW<br />
FEATFLOW is both a user oriented as well as a general purpose subrout<strong>in</strong>e system for the numerical<br />
solution of the <strong>in</strong>compressible Navier-Stokes equations <strong>in</strong> two and three space dimensions. It supports<br />
stationary and non-stationary problems.<br />
Homepage: www.featflow.de<br />
Latest version: 1.2d, 12 Apr 2000<br />
License: “Open Source”<br />
Language: Fortran77, C<br />
Platforms: Unix and L<strong>in</strong>ux, W<strong>in</strong>dows and<br />
DOS<br />
Fig.26: The FEATFLOW Navier-Stokes solver program and library<br />
415
A.5.2<br />
DUNS<br />
The DUNS (Diagonalised Upw<strong>in</strong>d Navier-Stokes) code is a 2D/3D, structured, multi-block, multispecies,<br />
react<strong>in</strong>g, steady/unsteady, Navier Stokes fluid dynamics code with q-omega turbulence model.<br />
Homepage: duns.sourceforge.net<br />
Latest version: 2.7.1, 25 Sept 2003<br />
License: GPL 2<br />
Language: C, Fortran77<br />
Platforms: L<strong>in</strong>ux and other POSIX<br />
compliant platforms<br />
Fig.27: The DUNS code<br />
A.5.3<br />
OpenFlower<br />
OpenFlower is an open source CFD software (FLOW solvER, literally) written <strong>in</strong> C++. It is ma<strong>in</strong>ly<br />
devoted to the resolution of the turbulent unsteady <strong>in</strong>compressible Navier-Stokes equations with a LES<br />
approach. It can deal with arbitrary complex 3D geometries with its f<strong>in</strong>ite volume approach.<br />
Homepage: openflower.<br />
sourceforge.net<br />
Latest version: 0.3, 11 Oct 2004<br />
License: GPL<br />
Language: C++<br />
Platforms: POSIX, <strong>in</strong>clud<strong>in</strong>g L<strong>in</strong>ux<br />
Fig.28: The OpenFlower CFD software<br />
A.5.4<br />
Gerris Flow Solver<br />
Gerris is a tool for generic numerical simulations of flows, possibly <strong>in</strong> geometrically complex geometries<br />
and <strong>in</strong>clud<strong>in</strong>g adaptive, multi phase and <strong>in</strong>terfacial flows capabilities.<br />
Homepage: gfs.sourceforge.net<br />
Latest version: 0.6-patch-1, 21 Oct 2004<br />
License: GPL<br />
Language: C<br />
Platforms: All POSIX<br />
(L<strong>in</strong>ux/BSD/UNIX-like OSes)<br />
Fig.29: The Gerris Flow Solver<br />
416
A.6 GUI<br />
A.6.1<br />
Qt<br />
One of the key design goals beh<strong>in</strong>d Qt is to make cross-platform application programm<strong>in</strong>g <strong>in</strong>tuitive, easy<br />
and fun. Qt achieves this goal by abstract<strong>in</strong>g low-level <strong>in</strong>frastructure functionality <strong>in</strong> the underly<strong>in</strong>g w<strong>in</strong>dow<br />
and operat<strong>in</strong>g systems, provid<strong>in</strong>g a coherent and logical <strong>in</strong>terface that makes sense to programmers.<br />
The Qt API and tools are consistent across all supported platforms, enabl<strong>in</strong>g platform <strong>in</strong>dependent application<br />
development and deployment. Qt applications run natively, compiled from the same source code,<br />
on all supported platforms. Qt is the basis of the “K” Desktop Environment for L<strong>in</strong>ux (KDE).<br />
Homepage: www.trolltech.com/<br />
Latest version: 3.3.3, 11 Aug 2004<br />
License: dual GPL and QPL, and<br />
commercial (<strong>in</strong>clud<strong>in</strong>g support<br />
for W<strong>in</strong>dows)<br />
Language: C++<br />
Platforms: L<strong>in</strong>ux, Solaris, HP-UX, IRIX,<br />
AIX, many other Unix variants,<br />
Mac OS X, embedded L<strong>in</strong>ux<br />
Fig.30: The Qt application framework<br />
A.6.2<br />
GTK+<br />
GTK+ is a multi-platform toolkit for creat<strong>in</strong>g graphical user <strong>in</strong>terfaces. Offer<strong>in</strong>g a complete set of widgets,<br />
GTK+ is suitable for projects rang<strong>in</strong>g from small one-off projects to complete application suites.<br />
GTK+ has been designed from the ground up to support a range of languages, not only C/C++. Us<strong>in</strong>g<br />
GTK+ from languages such as Perl and Python (especially <strong>in</strong> comb<strong>in</strong>ation with the Glade GUI builder)<br />
provides an effective method of rapid application development. GTK+ is the basis of the “Gnome” desktop<br />
environment on L<strong>in</strong>ux.<br />
Homepage: www.gtk.org<br />
Latest version: 2.6.1, 8 Jan <strong>2005</strong><br />
License: LGPL<br />
Language: C++<br />
Platforms: GNU/L<strong>in</strong>ux, BSD, Solaris, IRIX,<br />
HP-UX, AIX. There is a<br />
W<strong>in</strong>dows port.<br />
Fig.31: The GTK+ GUI library<br />
A.6.3<br />
wxWidgets<br />
wxWidgets gives you a s<strong>in</strong>gle, easy-to-use API for writ<strong>in</strong>g GUI applications on multiple platforms, with<br />
native look and feel. On top of great GUI functionality, wxWidgets gives you: onl<strong>in</strong>e help, network<br />
programm<strong>in</strong>g, streams, clipboard and drag and drop, multi-thread<strong>in</strong>g, image load<strong>in</strong>g and sav<strong>in</strong>g <strong>in</strong> a<br />
variety of popular formats, database support, HTML view<strong>in</strong>g and pr<strong>in</strong>t<strong>in</strong>g etc.<br />
417
Homepage: www.wxwidgets.org<br />
Latest version: 2.5.3, 10 Oct 2004<br />
License: wxW<strong>in</strong>dows Licence<br />
(OSI certified)<br />
Language: C++<br />
Platforms: W<strong>in</strong>dows, Unix, Mac OS X, MGL,<br />
OS/2<br />
Fig.32: The wxWidgets GUI library<br />
A.6.4<br />
Leonardo<br />
The Leonardo Library (LL) is a cross-platform, open-source C toolkit for program development. The<br />
library is light, well-documented, easy to learn, and covers a large number of functionalities, <strong>in</strong>clud<strong>in</strong>g<br />
graphic user <strong>in</strong>terface, thread, I/O and memory management. Differently from other programm<strong>in</strong>g<br />
toolkits, the LL also <strong>in</strong>cludes components with fundamental algorithms and data structures. Graphic user<br />
<strong>in</strong>terface components def<strong>in</strong>ed by the library preserve the look and feel of the target platform.<br />
Homepage: www.leonardo-vm.org<br />
Latest version: Qt-1.0.0b, 30 Nov 2003;<br />
W<strong>in</strong>32-1.1.0b, 13 Feb 2004<br />
License: LGPL 2.1<br />
Language: C<br />
Platforms: L<strong>in</strong>ux, W<strong>in</strong>dows<br />
Fig.33: The Leonardo Library<br />
A.6.5<br />
FOX Toolkit<br />
FOX is a C++ toolkit for easy and effective development of Graphical User Interfaces. It offers a wide<br />
collection of controls, and provides state of the art facilities such as drag and drop, selection, as well<br />
as OpenGL widgets for 3D graphical manipulation. FOX also implements icons, images, and userconvenience<br />
features such as status l<strong>in</strong>e help, and tool-tips. Tool-tips may even be used for 3D objects.<br />
Homepage: www.fox-toolkit.org<br />
Latest version: 1.3.22, 26 Dec 2004<br />
License: Relaxed LGPL 2.1<br />
Language: C++<br />
Platforms: L<strong>in</strong>ux, FreeBSD, SGI IRIX, HP-UX,<br />
IBM AIX, SUN Solaris,<br />
DEC/Compaq Tru64 UNIX,<br />
W<strong>in</strong>dows<br />
(Image CFD Research Corporation)<br />
Fig.34: The FOX GUI toolkit<br />
418
Optimization of Surface Utilization Us<strong>in</strong>g Heuristic Approaches<br />
Yves Langer, Maud Bay, Yves Crama, Frédéric Bair, Jean-David Caprace,<br />
Philippe Rigo, University of Liège, Liege/Belgium<br />
{Yves.Langer,Maud.Bay,Y.Crama,F.Bair,JD.Caprace,Ph.Rigo}@ulg.ac.be<br />
Abstract<br />
A schedul<strong>in</strong>g problem aris<strong>in</strong>g <strong>in</strong> factories produc<strong>in</strong>g large build<strong>in</strong>g blocks is modelled apply<strong>in</strong>g<br />
optimization techniques. The application is a shipyard workshop produc<strong>in</strong>g prefabricated keel<br />
elements. The objective is to maximize the number of build<strong>in</strong>g blocks produced <strong>in</strong> the factory<br />
dur<strong>in</strong>g a certa<strong>in</strong> time w<strong>in</strong>dow. The solution comb<strong>in</strong>es a Guided Local Search heuristic with Fast<br />
Local Search techniques. A f<strong>in</strong>al discussion expla<strong>in</strong>s the additional real-life issues aris<strong>in</strong>g <strong>in</strong> the<br />
<strong>in</strong>dustrial application and how firm-specific constra<strong>in</strong>ts can be conveniently considered by the<br />
model.<br />
1. Introduction<br />
This paper presents a new method to solve a schedul<strong>in</strong>g problem that arises <strong>in</strong> factories produc<strong>in</strong>g<br />
large build<strong>in</strong>g blocks (<strong>in</strong> our case, a shipyard workshop produc<strong>in</strong>g prefabricated keel elements).<br />
The factory is divided <strong>in</strong> equal size areas. The blocks produced <strong>in</strong> the factory are very large.<br />
Once a build<strong>in</strong>g block is placed <strong>in</strong>to the factory, it cannot be moved until all processes on the<br />
build<strong>in</strong>g block are f<strong>in</strong>ished. The blocks cannot overlap. The objective is to maximize the number<br />
of build<strong>in</strong>g blocks produced <strong>in</strong> the factory dur<strong>in</strong>g a certa<strong>in</strong> time w<strong>in</strong>dow.<br />
More precisely, we are given a set of n rectangular-shaped blocks. Each block is characterized by<br />
its geometric dimensions (width w j , length l j and height h j ) but also by process<strong>in</strong>g <strong>in</strong>formation<br />
such as its process<strong>in</strong>g time t j , its ready time r j and its due date d j (j ∈ {1, ..., n}). We are also<br />
given a number A of identical two-dimensional areas, hav<strong>in</strong>g width W and length L. Time is<br />
considered as a third dimension. The areas are fully dedicated to the production of the blocks.<br />
The problem consists of orthogonally order<strong>in</strong>g the blocks <strong>in</strong>to the areas, while respect<strong>in</strong>g the<br />
time constra<strong>in</strong>ts, and with the objective to produce the largest number of build<strong>in</strong>g blocks. In<br />
practical terms, we have to assign six variables for each block j:<br />
• p j = {0, 1} <strong>in</strong>dicat<strong>in</strong>g whether the block j is produced or not<br />
• the name a j = {1, . . . , A} of the area where block j is to be produced<br />
• x j and y j coord<strong>in</strong>ates, represent<strong>in</strong>g the position of the upper left corner of the block j <strong>in</strong><br />
the area<br />
• an orientation o j = {0, 1} (either horizontally or vertically) for block j<br />
• a start<strong>in</strong>g date s j<br />
A solution will be considered as feasible if the <strong>in</strong>dividual and the collective constra<strong>in</strong>ts are met.<br />
We call <strong>in</strong>dividual constra<strong>in</strong>ts those which focus on one block only, regardless of the other blocks.<br />
Major <strong>in</strong>dividual constra<strong>in</strong>ts are:<br />
• blocks must fit with<strong>in</strong> the width of an area (x j ≥ 0 and x j + [o j w j + (1 − o j )l j ] ≤ W )<br />
• blocks must fit with<strong>in</strong> the length of an area (y j ≥ 0 and y j + [o j l j + (1 − w j )l j ] ≤ L)<br />
• blocks must fit <strong>in</strong> their time w<strong>in</strong>dows (s j ≥ r j and s j + t j ≤ d j )<br />
419
Collective constra<strong>in</strong>ts focus on the <strong>in</strong>teraction between the positions of different blocks. In a<br />
first step, the only collective constra<strong>in</strong>t considered is that we need to prevent the blocks from<br />
overlapp<strong>in</strong>g.<br />
We assume <strong>in</strong>itially that there exists at least one feasible solution for the set of blocks <strong>in</strong>itially<br />
given. I.e. all the constra<strong>in</strong>ts can be satisfied when p j = 1 for all j ∈ {1, ..., n}.<br />
2. Analogy to the 3D-BPP<br />
In the three-dimensional b<strong>in</strong> pack<strong>in</strong>g problem (3D-BPP), we are given a set of n rectangularshaped<br />
items, each characterized by width w j , height h j , and depth d j (j ∈ {1, . . . , n}) and<br />
an unlimited number of identical three-dimensional conta<strong>in</strong>ers (b<strong>in</strong>s) hav<strong>in</strong>g width W, height<br />
H, and depth D. The 3D-BPP consists of orthogonally pack<strong>in</strong>g all the items <strong>in</strong>to the m<strong>in</strong>imum<br />
number of b<strong>in</strong>s.<br />
The major difference between 3D-BPP and our <strong>in</strong>itial problem is that, <strong>in</strong> the former,<br />
items/blocks must fit <strong>in</strong>to the conta<strong>in</strong>er height (z j ≥ 0 and z j + h j ≤ H), whereas they must<br />
fit <strong>in</strong>to their time w<strong>in</strong>dow <strong>in</strong> the latter (s j ≥ r j and s j + t j ≤ d j ). (One can compare a b<strong>in</strong><br />
<strong>in</strong> the 3D-BPP to a timel<strong>in</strong>e of the two-dimensional representation of an area, which gives us<br />
a three-dimensional representation of the problem.) The differences between m<strong>in</strong>imiz<strong>in</strong>g the<br />
number of b<strong>in</strong>s and maximiz<strong>in</strong>g the number of blocks will not complicate the formulation, s<strong>in</strong>ce<br />
we will assume (<strong>in</strong>itially) that there exists at least one feasible solution for a fixed number of<br />
block and of areas/b<strong>in</strong>s.<br />
The 3D-BPP is strongly NP-hard. (See Garey and Johnson (1979) for more <strong>in</strong>formation about<br />
the complexity of comb<strong>in</strong>atorial optimization problems. Indeed, it is a generalization of the<br />
one-dimensional b<strong>in</strong> pack<strong>in</strong>g problem (1D-BPP), <strong>in</strong> which a set of n positive values w j has to be<br />
partitioned <strong>in</strong>to the m<strong>in</strong>imum number of subsets so that the total value <strong>in</strong> each subset does not<br />
exceed a given b<strong>in</strong> capacity W. 1D-BPP is the special case of 3D-BPP for h j = H and d j = D<br />
for all j ∈ {1, ..., n} and it has been proven that the 1D-BPP is NP-Hard, Coffman et al. (1997).<br />
For such difficult problems, one way to conta<strong>in</strong> comb<strong>in</strong>atorial explosion is to allow algorithms<br />
to reach fairly good solutions, without guarantee<strong>in</strong>g that the best possible solution is reached.<br />
Local search heuristics use this strategy.<br />
Faroe et al. (2003) proposed a new heuristic for 3D-BPP. Their method is very flexible allow<strong>in</strong>g<br />
to adapt it to various additional constra<strong>in</strong>ts. Therefore it fits perfectly to our problem, as to<br />
many other real-life problems.<br />
3. F<strong>in</strong>d<strong>in</strong>g feasible solutions<br />
3.1. General approach<br />
The local search heuristic proposed to f<strong>in</strong>d a feasible schedule strictly enforces the <strong>in</strong>dividual<br />
constra<strong>in</strong>ts only. Then, penalties l<strong>in</strong>ked to the collective constra<strong>in</strong>ts are summed up <strong>in</strong> an<br />
objective function that is m<strong>in</strong>imized. With no additional real-life collective constra<strong>in</strong>ts, the<br />
objective function value of a given solution is the total pairwise overlap between the blocks.<br />
Therefore, with a randomly generated unfeasible solution where blocks can overlap, search<strong>in</strong>g<br />
for a feasible solution is equivalent to m<strong>in</strong>imiz<strong>in</strong>g the objective function, s<strong>in</strong>ce an objective value<br />
of zero <strong>in</strong>dicates that also the collective constra<strong>in</strong>ts are met. For any solution X, let overlap ij (X)<br />
be the overlap (<strong>in</strong> square meters days) between blocks i and j. The objective function can now<br />
be formulated as<br />
f(X) = ∑ overlap ij (X) (1)<br />
i
of a block around one of its four corners.’ A neighbor of X is therefore constructed by assign<strong>in</strong>g<br />
a new value to one of the variables x j , y j , s j , a j , o j . This def<strong>in</strong>ition of a solution space <strong>in</strong>cludes<br />
all feasible schedules and that there is a path of moves between every pair of solution.<br />
A typical local search procedure proceeds by mov<strong>in</strong>g from the current solution X p to a neighbor<strong>in</strong>g<br />
solution X p+1 ∈ ν(X p ) whenever this move improves the value of the objective function.<br />
This may lead to two types of difficulties. First, the solution may settle <strong>in</strong> a local m<strong>in</strong>imum.<br />
Several standard methods, such as the Simulated Anneal<strong>in</strong>g, Aarts and Korst (1989) or the Tabu<br />
Search, Glover (1990), exist to avoid this shortcom<strong>in</strong>g of local search procedures. Secondly, the<br />
neighborhood of any given solution may be quite large (even if cont<strong>in</strong>uous, variables like x j , y j<br />
or s j can be discretized for practical purposes). Therefore, explor<strong>in</strong>g the neighborhood to f<strong>in</strong>d<br />
an improv<strong>in</strong>g move can be very costly <strong>in</strong> comput<strong>in</strong>g time. To deal with these issues, we present<br />
<strong>in</strong> this paper an application of the Guided Local Search (GLS) heuristic, and its accompany<strong>in</strong>g<br />
neighborhood reduction scheme called Fast Local Search (FLS).<br />
3.2. Guided local search<br />
The Guided Local Search Heuristic (GLS) has its root <strong>in</strong> a Neural Network architecture named<br />
GENET, Wang and Tsang (1991), which is applicable to a class of problems known as Constra<strong>in</strong>t<br />
Satisfaction Problems. The actual GLS version with its accompany<strong>in</strong>g FLS has been first shown<br />
by Voudouris (1997), Voudouris and Tsang (1997,1999), and applied to the 3D-BPP by Faroe<br />
et al. (2003).<br />
Basically, GLS augments the objective function of a problem to <strong>in</strong>clude a set of penalty terms<br />
and considers this function, <strong>in</strong>stead of the orig<strong>in</strong>al one, for m<strong>in</strong>imization by the local search<br />
procedure. Local search is conf<strong>in</strong>ed by the penalty terms and focuses attention on promis<strong>in</strong>g<br />
regions of the search space, Voudouris and Tsang (1999). Iterative calls are made to a local<br />
search procedure, denoted as LocalOpt(X). Each time LocalOpt(X) gets caught <strong>in</strong> a local<br />
m<strong>in</strong>imum, the penalties are modified and local search is called aga<strong>in</strong> to m<strong>in</strong>imize the modified<br />
objective function. In a certa<strong>in</strong> measure, the heuristic may be classified as a tabu search heuristic;<br />
it uses memory to control the search <strong>in</strong> a manner similar to Tabu Search.<br />
GLS is based on the concept of ’features’, a set of attributes that characterizes a solution to the<br />
problem <strong>in</strong> a natural way. In our adaptation of the model, features are the overlaps between<br />
the blocks, and the <strong>in</strong>dicator I ij (X) = {0, 1} denotes whether blocks i and j overlap or not. In<br />
a particular solution, a feature with a high overlap is not attractive and may be penalized. As<br />
a result, the value of overlap ij (X) can measure the impact of a feature on a solution X (’cost<br />
function’ <strong>in</strong> Faroe et al. (2003)).<br />
p ij denotes the number of times a feature has been penalized. p ij is <strong>in</strong>itially zero. We want to<br />
penalize the features with the maximum overlap, that have not been penalized too often <strong>in</strong> the<br />
past. The source of <strong>in</strong>formation that determ<strong>in</strong>es which features will be penalized should thus<br />
be the overlap and the amount of previous penalties assigned to the features. For this purpose,<br />
we def<strong>in</strong>e a utility function µ(X) = overlap ij (X)/(1 + p ij ). After each LocalOpt(X) iteration,<br />
the procedure adds one to the penalty of the pairs with maximum utility.<br />
After <strong>in</strong>crement<strong>in</strong>g the penalties of the selected features, they are <strong>in</strong>corporated <strong>in</strong> the search<br />
with an augmented objective function<br />
h(X) = f(X) + λ ∑ i,j<br />
p ij · I ij (X) = ∑ overlap ij (X) + λ ∑<br />
i
3.3. Fast local search<br />
The ’fast local search’ (FLS) procedure, Voudouris and Tsang (1997), Faroe et al. (2003) is used<br />
to transform a current solution X cur <strong>in</strong>to a local m<strong>in</strong>imum X ∗ = LocalOpt(X cur ). It allows to<br />
reduce the size of the neighborhood with a selection of the moves that are likely to reduce the<br />
maximum utility overlaps.<br />
We def<strong>in</strong>e the sets ν m (X) as subsets of the neighborhood ν(X) where all solutions <strong>in</strong> ν m (X)<br />
only differ from X by the value of the variable m (m = {x j , y j , t j , a j , o j } with j ∈ 1...n). (In the<br />
case of m = {o 1 , ..., o n }, ν m also <strong>in</strong>cludes the particular change <strong>in</strong> x j and <strong>in</strong> y j that considers a<br />
rotation around the four corners of a block. To simplify the explanation, this technical issue is not<br />
detailed.) The neighborhood ν(X) is thus divided <strong>in</strong>to a number of smaller sub-neighborhoods<br />
that can be either ’active’ or ’<strong>in</strong>active’. Initially, only some sub-neighborhoods are active.<br />
FLS now cont<strong>in</strong>uously visits the active sub-neighborhoods <strong>in</strong> a random order. If there exists<br />
a solution X m with<strong>in</strong> the sub-neighborhood ν m (X cur ) such that f(X m ) < f(X cur ), then X cur<br />
becomes X m ; otherwise we suppose that the selected sub-neighborhood will provide no more<br />
significant improvements at this step, and thus it becomes <strong>in</strong>active. When there is no active subneighborhoods<br />
left, the FLS procedure is stopped and X cur , the best solution found, is returned<br />
to GLS. From a less formal po<strong>in</strong>t of view, FLS selects at random a variable m with<strong>in</strong> a list of<br />
active variables, as long as this list is not empty. Then, it searches with<strong>in</strong> the doma<strong>in</strong> of m any<br />
improvement of the objective function. If it does not exist the variable m becomes <strong>in</strong>active and<br />
is removed from the list. By do<strong>in</strong>g so, we focus specially on variables open for improvement.<br />
The size of the sub-neighborhoods related to the a j and the o j variables is relatively small:<br />
A <strong>in</strong> the first case, 5 <strong>in</strong> the second (<strong>in</strong>itial + four corners). Therefore FLS is set to test all<br />
the neighbors of these sets. But, on the other hand, us<strong>in</strong>g an enumerative method for the<br />
translations along the x, y and t axis would become very expensive <strong>in</strong> terms of comput<strong>in</strong>g time,<br />
if areas and/or time w<strong>in</strong>dows are large. However, only certa<strong>in</strong> coord<strong>in</strong>ates of such neighborhoods<br />
need to be <strong>in</strong>vestigated. If m represents x j , changes <strong>in</strong> the overlap function only depend on x j<br />
(h(X) = h(x j )). Most of the terms of this function are constant, thus, s<strong>in</strong>ce we want to compare<br />
values, only the few terms dependent on x j should be computed. Furthermore, an overlap is<br />
the product of four partial overlaps (three for the overlaps on each of the x, y and z axis, and<br />
the fourth equals one if a i = a j ; zero otherwise). S<strong>in</strong>ce we know that only the partial overlaps<br />
for the x axis depend on x j , comput<strong>in</strong>g efforts can be reduce to their smallest size. Also,<br />
overlap ij (X) = overlap ji (X), so that the comput<strong>in</strong>g time of one solution is l<strong>in</strong>ear (n) <strong>in</strong>stead<br />
of quadratic (n 2 ). Additionally, all functions overlap ij (x j ) are piecewise l<strong>in</strong>ear functions, and<br />
therefore the functions will atta<strong>in</strong> their m<strong>in</strong>imum <strong>in</strong> one of their breakpo<strong>in</strong>ts (or at the limits<br />
of their doma<strong>in</strong>s). As a result, FLS only needs to compute the values of f(x j ) with x j at<br />
breakpo<strong>in</strong>ts or at extreme values. In fact, there are at most four breakpo<strong>in</strong>ts for each function,<br />
and only the first and the last one are evaluated. Indeed, <strong>in</strong> regard to the analogy with the<br />
3D-BPP, a good pack<strong>in</strong>g <strong>in</strong>tuitively supposes that the boxes touch each other.<br />
FLS represents a relatively fast procedure that leads to a local m<strong>in</strong>imum if the amount of active<br />
sub-neighborhoods is relatively small. Remember that LocalOpt(X) is called iteratively by GLS,<br />
and that penalties are changed with an objective of escap<strong>in</strong>g local m<strong>in</strong>ima. Activation of subneighborhoods<br />
should therefore allow moves on penalized features. The follow<strong>in</strong>g reactivation<br />
scheme is used, Faroe et al. (2003): (1) moves on the two blocks i and j, correspond<strong>in</strong>g to the<br />
penalized features, are reactivated. (2) We reactivate the moves on all blocks that overlap with<br />
blocks i and j. The latter reactivation is added to allow FLS to pay attention not only to the<br />
two overlapp<strong>in</strong>g blocks but also to the whole area around the penalized feature.<br />
4. Select<strong>in</strong>g the blocks<br />
In the previous chapter, we described a method that m<strong>in</strong>imizes the collective constra<strong>in</strong>ts under<br />
restriction of the <strong>in</strong>dividual constra<strong>in</strong>ts and we supposed that there exists at least one feasible<br />
422
solution for the set of blocks <strong>in</strong>itially given. Let us denote this procedure by GlobalOpt(X).<br />
If GlobalOpt(X) is efficient, it should f<strong>in</strong>d a solution with an objective function of zero after<br />
a certa<strong>in</strong> time and this solution would be one of the feasible solutions. However, <strong>in</strong> the <strong>in</strong>itial<br />
formulation of the problem, we do not know whether a set of blocks is feasible or not. The<br />
comb<strong>in</strong>ation of GLS and FLS can be used anyway if we rely on the follow<strong>in</strong>g heuristic assumption:<br />
there exists no feasible solution if none is found with<strong>in</strong> a certa<strong>in</strong> amount of comput<strong>in</strong>g time T .<br />
Consequently, the search heuristic GlobalOpt(X, T ) is utilized as a test of feasibility and gives<br />
the correspondant schedules if a feasible solution is identified with<strong>in</strong> T .<br />
Several methods have been tested us<strong>in</strong>g this concept. The objective was to rema<strong>in</strong> as close as<br />
possible to the work<strong>in</strong>g methods and habits used <strong>in</strong> the factory under study. From this po<strong>in</strong>t<br />
of view, an efficient approach for the <strong>in</strong>dustrial application is to start GLS with a randomly<br />
generated solution X 0 that <strong>in</strong>cludes the entire set of blocks (p j = 1 for all j = {1, ..., n}). After<br />
a search of T seconds, the algorithm is stopped and returns X 1 = GlobalOpt(X 0 , T ), the best<br />
solution found (<strong>in</strong> terms of overlap). One of the blocks with the highest overlap is removed from<br />
the set (X 1 → X 1 ′ ) and the heuristic GlobalOpt(X 1 ′ , T ) is restarted. The entire procedure ends<br />
if a solution X n with zero overlap is found.<br />
A variant procedure is to start with an empty set X 0 (p j = 0 for all j = {1, ..., n}). At<br />
each iteration, if the solution X n+1 = GlobalOpt(X n , T ) is feasible, then an additional block<br />
is <strong>in</strong>serted <strong>in</strong> the set; otherwise an overlapp<strong>in</strong>g block is removed. This procedure is stopped<br />
after a certa<strong>in</strong> comput<strong>in</strong>g time, or by any more sophisticated stopp<strong>in</strong>g criterion, and returns<br />
the solution with the largest collection of blocks. Fig.1 shows the iterative processus of this<br />
procedure.<br />
Fig.1: Test <strong>in</strong>stance (T = 1 s, Time limit = 600 s)<br />
Both approaches suffers from one major default: they are likely to have aversion for the largest<br />
blocks. Indeed, we do not have an appropriate weight<strong>in</strong>g scheme to evaluate the preferences<br />
between blocks, and, s<strong>in</strong>ce small blocks generally provide smaller overlaps, they are preferred<br />
to larger ones. In the real-life situation, when the entire set of block cannot be produced, the<br />
person <strong>in</strong> charge of schedul<strong>in</strong>g can either subcontract specific blocks <strong>in</strong> other factories, or change<br />
some temporal parameters (e.g. <strong>in</strong>tensify the workforce to reduce process<strong>in</strong>g times or postpone<br />
due dates). No formal <strong>in</strong>formation can describe all the aspects of these choices. For this reason,<br />
the operator should be able to change manually the collection of blocks to be produced. Start<strong>in</strong>g<br />
from our ”fairly good“ feasible solution X n , iterative X n+1 = GlobalOpt(X ′ n) calls are ordered<br />
423
manually after deliberate changes (X n → X n ′ ) <strong>in</strong> the assignment. In addition, a last procedure<br />
provides a list with each block that is not assigned even though a feasible solution that <strong>in</strong>cludes<br />
the block can be found.<br />
By not regenerat<strong>in</strong>g solutions on a random basis, some of the <strong>in</strong>formation from previous solutions<br />
is preserved. A drawback to this approach is that the structure of a previous solution can<br />
conf<strong>in</strong>e GLS to an area of the solution space that can be difficult to escape. (Faroe et al. (2003)<br />
suggest a similar problem <strong>in</strong> their approach for 3D-BPP.) We may therefore not reach the very<br />
best solution. However, the modus operandi described <strong>in</strong> this section is developed for a daily<br />
<strong>in</strong>dustrial use. In that sett<strong>in</strong>g, the above drawback may actually be viewed as an advantage.<br />
Indeed, it may be very costly for the company to mix up the schedules over and over aga<strong>in</strong>.<br />
Traditionally, methods for problems of similar classes utilize a construction algorithm dur<strong>in</strong>g<br />
the search. A slight improvement may disturb the whole solution; with GLS, non-problematic<br />
regions are not perturbed. Murata et al. (1996) developed a tricky construction technique based<br />
on partial-orders cod<strong>in</strong>g scheme for the 2D-BPP. Imahori et al. (<strong>2005</strong>) adapted this approach<br />
for a problem very close to our’s.<br />
5. Additional real-life issues<br />
Additional constra<strong>in</strong>ts may occur <strong>in</strong> any firm-specific situation. The tool proposed here is easily<br />
customizable to most of them. For example, we may need to restrict or force the position of a<br />
block (e.g. a tool is only available <strong>in</strong> one area or the block is already <strong>in</strong> process). Intgrat<strong>in</strong>g those<br />
constra<strong>in</strong>ts is trivial: restricted positions are not generated and unfeasible neighbors simply do<br />
not exist. As a result, the end-user may fix the value of any variable (<strong>in</strong>clud<strong>in</strong>g p j ) or reduce its<br />
doma<strong>in</strong>.<br />
Specific collective constra<strong>in</strong>ts may also appear <strong>in</strong> the word<strong>in</strong>g of a problem. In our case, the<br />
areas of the factory have one s<strong>in</strong>gle door, and the crane bridge can only carry blocks up to a<br />
certa<strong>in</strong> height. As a result, a large block may obstruct a door, and some blocks might not be<br />
deliverable <strong>in</strong> time because there is no route to transport them out. We dealt with this issue <strong>in</strong><br />
the same way as for overlaps. For each generated solution X, we add to the objective function<br />
h(X) a new term account<strong>in</strong>g for exit difficulties:<br />
g(X) = h(X) + ExitP roblems(X) = ∑ overlap ij (X) + λ · ∑<br />
p ij · I ij (X) + ExitP roblems(X)<br />
i
References<br />
AARTS, E.; KORST, J. (1989), Simulated Anneal<strong>in</strong>g and Boltzmann Mach<strong>in</strong>es – A stochastic<br />
approach to comb<strong>in</strong>atorial optimisation and neural comput<strong>in</strong>g, Wiley<br />
COFFMAN, E.G.; GAREY, M.R.; JOHNSON, D.S. (1997), Approximation algorithms for b<strong>in</strong><br />
pack<strong>in</strong>g: A survey, D.S. Hochbaum (Ed.), Approximation Algorithms for NP-Hard Problems,<br />
PWS Publ. Company<br />
FAROE, O.; PISINGER, D.; ZACHARIASEN, M. (2003), Guided Local Search for the threedimensional<br />
b<strong>in</strong>-pack<strong>in</strong>g problem, INFORMS J. Comput<strong>in</strong>g 15/3, pp.267-283<br />
GAREY, M.R.; JOHNSON, D.S. (1979), Computers and Intractability: A Guide to the Theory<br />
of NP-Completeness, W.H. Freeman<br />
GLOVER, F. (1990), Tabu Search: A Tutorial, Interfaces 20/44, pp.74-94<br />
IMAHORI, S.; YAGIURA, M.; IBARAKI, T. (<strong>2005</strong>), Improved local search algorithms for the<br />
rectangle pack<strong>in</strong>g problem with general spatial costs, European J. Operational Research 167,<br />
pp.48-67<br />
MARTELLO, S.; PISINGER, D.; VIGO, D. (2000), The three dimensional b<strong>in</strong> pack<strong>in</strong>g problem,<br />
INFORMS Operations Research 48/2, pp.256-267<br />
MURATA, H.; FUJIYOSHI, K.; NAKATAKE, S.; KAJITANI, Y. (1996), VLSI module placement<br />
based on rectangle-pack<strong>in</strong>g by the sequence-pair, IEEE Trans. Computer-Aided Design of<br />
Integrated Circuits and Systems 15/12, pp.1518-1524<br />
SCHOLL, A.; KLEIN, R.; JÜRGENS, C. (1997), BISON: A fast hybrid procedure for exactly<br />
solv<strong>in</strong>g the one-dimensional b<strong>in</strong> pack<strong>in</strong>g problem, Computers & Operations Research 24, pp.627-<br />
645<br />
VOUDOURIS, C. (1997), Guided local search for comb<strong>in</strong>atorial optimization problems, Ph.D.<br />
Thesis, Dept. Computer Science, Univ. of Essex, Colchester/UK<br />
VOUDOURIS, C.; TSANG, E. (1997), Fast local search and guided local search and their application<br />
to British Telecom’s workforce schedul<strong>in</strong>g problem, Operations Research Letters 20,<br />
pp.119-127<br />
VOUDOURIS, C.; TSANG, E. (1999), Guided local search and its application to the travel<strong>in</strong>g<br />
salesman problem, European J. Operational Research 113, pp.469-499<br />
WANG, C.J.; TSANG, E. (1991), Solv<strong>in</strong>g constra<strong>in</strong>t satisfaction problems us<strong>in</strong>g neural-networks,<br />
IEE 2 nd Int. Conf. Artificial Neural Network, pp.295-299<br />
425
! <br />
" <br />
<br />
<br />
<br />
<br />
!"#<br />
<br />
<br />
!<br />
"!# "" $ <br />
"! % <br />
& ! <br />
& ' ! <br />
<br />
<br />
!!! !<br />
! <br />
!(!! )! ! !! <br />
$ !*! <br />
" ! <br />
+& ! <br />
&<br />
<br />
<br />
# & #,-./0<br />
. ,1./00 !2 3/ 4+,# <br />
& <br />
# & / 2 5 6 7 8 <br />
2 #$%&&'(<br />
<br />
! <br />
! <br />
& <br />
<br />
9 & <br />
& :&
$" <br />
<br />
&< <br />
/! : <br />
& &! )$*++,(-<br />
" = ><br />
"=& ><br />
"= ! <br />
7 ! )$*++,(-?2$ <br />
&' <br />
@<br />
<br />
%" <br />
<br />
! &<br />
# $ ! ! <br />
9 & <br />
! & &=<br />
" ><br />
" !><br />
" ><br />
" <br />
" <br />
< & <br />
'& & !<br />
& = & &<br />
<br />
&"' <br />
<br />
<br />
!#$%&&'(!9 <br />
# <br />
<br />
!9=
6 & <br />
<br />
/ <br />
2 <br />
7<<br />
7<<br />
6 & <br />
/<br />
<<br />
9<br />
4 &<br />
8' C <br />
: <br />
7<<br />
7<<br />
7<<br />
7<<br />
7<<br />
7<<br />
7<<br />
&=%" 7" <br />
,= (&)><br />
E , E= @ @ $ ( <br />
& $ '=<br />
@ ! $ @)> F <br />
@><br />
, = @!! @ !<br />
>F @><br />
1 , 1= @ @ <br />
()! (3:G3: )><br />
F@><br />
,= @!!@ "! <br />
"!(H3:GH 3: )& <br />
@62: @62:<br />
<br />
)"' <br />
<br />
# * ! <br />
2 !<br />
I <br />
! !<br />
<br />
! ! <br />
<br />
# &#$%&&.%&&/%&&'(-<br />
. <br />
& <br />
<br />
& <br />
! J. 2 <br />
! & 9 <br />
' <br />
E <br />
&! <br />
6 2 : (62:) # 62:<br />
&= * <br />
$ & = * <br />
/! ! & <br />
! & <br />
! & <br />
D
*" <br />
# * ! <br />
# $%&&'( 2 ! <br />
I <br />
! ! <br />
< & ! = <br />
(62:) !01 $*++2%&&%( 6 2 : # =<br />
* <br />
: & ! & &<br />
&! <br />
' & !"#$%&&'(-<br />
" "& ! ! ><br />
" "!! !!"! " ! <br />
!<br />
6 & =<br />
" ! !& <br />
<br />
" & ! <br />
<br />
<br />
& =!" "& "<br />
& ! & ! <br />
& "<br />
! * & <br />
<br />
<br />
! # =<br />
" ><br />
" ><br />
" ><br />
" & <br />
& #$%&&'(-<br />
" $ ><br />
" ><br />
" !#><br />
" ><br />
" = * * <br />
( !)><br />
" (!! !!* )><br />
" ><br />
" ( )><br />
" "! ><br />
" ><br />
" ><br />
" ><br />
" <br />
<br />
6.<br />
<br />
<br />
<br />
<br />
<br />
-
2<br />
3: /<br />
<br />
27!#<br />
<br />
/+ / <br />
<br />
< % 3$ <br />
<br />
% /L/ <br />
2L8 % <br />
% : <br />
33 <br />
% < <br />
,<br />
3/ <br />
3!<br />
% 2<br />
8 <br />
M<br />
5*9 L<br />
5*: <br />
2 9 <br />
5 <br />
% : <br />
3: <br />
6.=A <br />
#$%&&.%&&/%&&'(<br />
<br />
B <br />
8 ! <br />
& 6 <br />
! & $ ! ! & <br />
:A:3 (: A& : 3 ! !) 6", $ <br />
! ' !01 $*++2%&&%( <br />
$ & <br />
$ 9
!!"! : ! <br />
$ !" "& ! 3<br />
$%&&&( 9 "& ! <br />
& <br />
"! * $ <br />
& & ! <br />
$ ! & ! <br />
! <br />
3 $%&&&( # $%&&'( ! !<br />
& *<br />
* <br />
* <br />
<br />
$ !$ !<br />
!" <br />
<br />
# &#$%&&'(-<br />
" @ @ ><br />
" ? @ & ><br />
" ><br />
" ><br />
" ! <br />
! !! ! ! *<br />
9
" @!!77 "&@><br />
" @!!/* // @& 9
. * ><br />
! ><br />
E & =<br />
" ><br />
" <br />
<br />
1"./ <br />
<br />
: !! <br />
& ?: 2 ! 9' :@ <br />
& =<br />
<br />
:≥3 (.)<br />
&= :" ! ' :O ><br />
G!! ><br />
"!! ><br />
3"$ ! '<br />
+ !9
! ' $ ! '<br />
<br />
6 &$ ! : <br />
5:3%83 8 <br />
& & ? @ ! ! <br />
& !01<br />
$%&&%%&&/(5 $*++'(5 $*++'(64$%&&*(#$%&&'(-<br />
" (.)><br />
" ()><br />
" (E)><br />
" ()<br />
% ! '!99 & <br />
> &> ! > & > <br />
> !! ! <br />
!!& =<br />
<br />
O . Q Q E Q (E)<br />
<br />
& "!! ><br />
. E " !!& <br />
<br />
!99=: '' <br />
<br />
2. 2 2E 2<br />
& R R R R<br />
& R R R R<br />
! R R<br />
/& R R R<br />
R R <br />
"&! R R<br />
<br />
!2"./ 3 4/56<br />
<br />
6 & ! 6 & <br />
! <br />
! <br />
: 2 ! 9': <br />
9' : <br />
! ! ! <br />
<br />
!!"./ <br />
<br />
&<br />
& !<br />
& 5 $*++7(-<br />
<br />
∆: # O( )( ) ()<br />
<br />
∆: # & ! <br />
∆: # !$ ' !<br />
61&∆: # <br />
?+.@ 6<br />
<br />
<br />
<br />
E
% @+.@= <br />
wskaznik A<br />
1<br />
0,8<br />
0,6<br />
0,4<br />
0,2<br />
0<br />
1 2 3 4 5<br />
iteracje<br />
w ersja B1<br />
% @+@=<br />
<br />
wskaznik A<br />
1<br />
0,8<br />
0,6<br />
0,4<br />
0,2<br />
0<br />
1 2 3 4 5<br />
iteracje<br />
% @+E@=<br />
w ersja B2<br />
<br />
wskaznik A<br />
1<br />
0,8<br />
0,6<br />
0,4<br />
0,2<br />
0<br />
1 2 3 4 5<br />
iteracje<br />
w ersja B3<br />
<br />
6=9 "<br />
<br />
Local safety <strong>in</strong>dices<br />
Local safety <strong>in</strong>dices<br />
1st iteration<br />
2nd iteration<br />
deltaA values<br />
1<br />
0,5<br />
0<br />
0,8363 0,914<br />
0,4953<br />
0,3522 0,4311<br />
0,608<br />
0,3521<br />
1 2 3 4 5 6 7<br />
deltaA values<br />
1<br />
0,5<br />
0<br />
0,8363<br />
0,49530,3539<br />
0,62390,5478 0,627 0,9286<br />
1 2 3 4 5 6 7<br />
1st iteration 0,8363 0,4953 0,3522 0,4311 0,3521 0,608 0,914<br />
compartments<br />
2nd iteration 0,836 0,495 0,354 0,624 0,548 0,627 0,929<br />
compartments<br />
<br />
Local safety <strong>in</strong>dices<br />
Local safety <strong>in</strong>dices<br />
3rd iteration<br />
4th iteration<br />
deltaA values<br />
1<br />
0,5<br />
0<br />
0,89790,8137 0,6104<br />
0,74570,67240,6453 0,9286<br />
1 2 3 4 5 6 7<br />
deltaA values<br />
1<br />
0,5<br />
0<br />
0,89790,9096 0,6228 0,74150,67260,6453 0,9286<br />
1 2 3 4 5 6 7<br />
3rd iteration 0,898 0,814 0,61 0,746 0,672 0,645 0,929<br />
4th iteration 0,898 0,91 0,623 0,742 0,673 0,645 0,929<br />
compartments<br />
compartments<br />
<br />
61=7 ∆: # @:@ 6<br />
<br />
E1
: ! !! <br />
& 9 & <br />
<br />
$ <br />
<br />
!$"./ <br />
<br />
!& <br />
(52/) $ && !!"! <br />
! 7 !& & <br />
<br />
"! <br />
<br />
2 ! <br />
& : ' ! !! * <br />
! ! <br />
! !&<br />
<br />
!!* ! !! &=<br />
<br />
/9 O.G 29 (1)<br />
/% O.G 2% ()<br />
&= /9 29 "!!* ! ! ><br />
/% 2% "!!* !! <br />
<br />
!!!! ! ! <br />
$*++2(#$%&&.%&&/(-<br />
<br />
29 O(((5: 9 SO . 3: <br />
3: <br />
: (')"!! : <br />
<br />
! ( ) ! <br />
! ! +&<br />
E
!! &&&<br />
!!! <br />
<br />
: " G ' '<br />
U' ! <br />
RU ' '<br />
&=<br />
<br />
<br />
K 3J O(R . R V R ) (..)<br />
<br />
<br />
<br />
W KOXY O. Y #O. (ZZR )(ZZR # )(R R # )[ (.)<br />
<br />
<br />
&= K 3J G K><br />
W K G K<br />
<br />
! ! '=<br />
<br />
<br />
39OK 3J W K (.E)<br />
<br />
&39( )K<br />
<br />
9K ! !!* / &=<br />
<br />
/ O.G\(39) (.)<br />
<br />
&\ ! <br />
<br />
' !!& K <br />
! 6 " ! ! <br />
!! ' <br />
<br />
: ' ! ! &<br />
]O.1000 -0Q^6 !<br />
&<br />
<br />
!999=!!: . : <br />
3 ! < X [ / %! <br />
: . D0D0 0. ,<br />
: .-;.0 0. ,<br />
<br />
!9J=* !! !<br />
<<br />
3 <br />
% ! * <br />
6" " 39O.E0.;0; / O00-10<br />
: " 39O.E0.;0; / O00-10<br />
( )<br />
<br />
!%" <br />
<br />
% 8 <br />
! # &!<br />
# & &= ! <br />
<br />
E;
(! ) ' (& & ) <br />
<br />
' 5:3%83<br />
!8 & 27A:2/99".:+ +". <br />
& & ! !01$%&&%(<br />
<br />
& ! !9<br />
" & ><br />
" 9 <br />
<br />
!&" <br />
<br />
: <br />
9! <br />
<br />
" "! "<br />
!! "!<br />
!! <br />
2/ 6 2: <br />
& * <br />
! <br />
& && ! <br />
9 ! 9 <<br />
7 (9
7<br />
<br />
::4>:MM++:3/: 3KM+7B 249 (00.) <br />
8 # . <br />
2 2?22@ <br />
<br />
8394 < (00E) <br />
D 9/ 2!2 7 J2:+U00E<<br />
<br />
8394 < (00E!) <br />
<
Optimisation of Vessel Resistance us<strong>in</strong>g Genetic Algorithms<br />
and Artificial Neural Networks<br />
Andrew Mason, Patrick Couser, Formation Design Systems, Fremantle/Australia,<br />
AndyM@formsys.com<br />
Garth Mason, Cameron R. Smith, Brian R. von Konsky, Curt<strong>in</strong> University of Technology<br />
Abstract<br />
It has been found that an artificial neural network is able to produce results of sufficient accuracy to<br />
be useful for prelim<strong>in</strong>ary prediction of vessel resistance, with the major benefits of: be<strong>in</strong>g relatively<br />
simple to set up; be<strong>in</strong>g easily retra<strong>in</strong>ed with new data; and that Froude number may be easily<br />
<strong>in</strong>cluded as an <strong>in</strong>dependent variable. In this work the Neural Network is fitted directly to the orig<strong>in</strong>al<br />
tank test data, rather than to a set of smoothed curves vary<strong>in</strong>g <strong>in</strong> the Froude number axis. In addition,<br />
different network architectures have been <strong>in</strong>vestigated, with networks conta<strong>in</strong><strong>in</strong>g two hidden layers<br />
be<strong>in</strong>g seen to perform well. A procedure for the optimisation of design parameters us<strong>in</strong>g a Genetic<br />
Algorithm has also been evaluated. The Genetic Algorithm uses the approximation provided by the<br />
Neural Network response surfaces for its objective function. This has been found to be effective and of<br />
acceptable performance.<br />
Nomenclature<br />
B<br />
C F<br />
C R<br />
C T<br />
C W<br />
F n<br />
g<br />
L<br />
L/∇ 1/3<br />
R e<br />
S<br />
T<br />
υ<br />
WSA<br />
∇<br />
ν<br />
ρ<br />
Demi-hull maximum beam of waterl<strong>in</strong>e<br />
Coefficient of frictional resistance [ITTC-57 Correlation l<strong>in</strong>e]<br />
Coefficient of residuary resistance<br />
Coefficient of total resistance<br />
Coefficient of wave resistance<br />
Froude number [υ/ √(gL)]<br />
Acceleration due to gravity<br />
Vessel length between perpendiculars<br />
Slenderness ratio<br />
Reynolds number [υL/ν]<br />
Catamaran demihull centerl<strong>in</strong>e separation<br />
Demi-hull draft<br />
Vessel velocity<br />
Wetted surface area<br />
Demi-hull volume of displacement<br />
Fluid k<strong>in</strong>ematic viscosity<br />
Fluid density<br />
1. Introduction<br />
Artificial Neural Networks (ANNs) have been used for classification and prediction across many<br />
discipl<strong>in</strong>es <strong>in</strong>clud<strong>in</strong>g medic<strong>in</strong>e, eng<strong>in</strong>eer<strong>in</strong>g, and computer science, primarily because of their ability<br />
to model non-l<strong>in</strong>ear functions quickly and accurately.<br />
In the field of naval architecture, <strong>in</strong>terpolation and prediction of hull resistance from model<br />
experiments and tank test<strong>in</strong>g have traditionally been made us<strong>in</strong>g statistical regression equations.<br />
However, this is a problem that is also well suited to neural networks. Specifically a neural network<br />
may be a more favourable option than statistical methods s<strong>in</strong>ce they have been shown to provide<br />
greater flexibility and can more quickly create accurate models of complex systems, Ja<strong>in</strong> et al.<br />
(1996).<br />
440
This paper describes an <strong>in</strong>vestigation <strong>in</strong>to the accuracy of ANNs as prediction tools for hull resistance<br />
and follows on from the work presented <strong>in</strong> Couser et al. (2004). The goal of the orig<strong>in</strong>al <strong>in</strong>vestigation<br />
was to determ<strong>in</strong>e a predictive model for residuary resistance (C R ), based on <strong>in</strong>put values of Froude<br />
number (F n ), Separation-Length ratio (S/L), Breadth-Draught ratio (B/T) and Slenderness ratio<br />
(L/∇ 1/3 ). The data used for the <strong>in</strong>vestigation orig<strong>in</strong>ated from a series of tank tests described <strong>in</strong><br />
Molland et al. (1994,1995), Insel and Molland (1992).<br />
In the previous work, the authors had the choice of us<strong>in</strong>g the raw tank test data, or us<strong>in</strong>g a larger<br />
dataset derived from a least squares fit to the F n axis of the raw data. At that time it was felt that the<br />
larger dataset would give superior results, however <strong>in</strong> retrospect it was felt that the additional fitt<strong>in</strong>g<br />
step may have <strong>in</strong>troduced errors. As a result it was decided to attempt fitt<strong>in</strong>g of the orig<strong>in</strong>al<br />
experimental data directly, despite the data be<strong>in</strong>g highly non-l<strong>in</strong>ear <strong>in</strong> the F n axis, there be<strong>in</strong>g a<br />
relatively small number of po<strong>in</strong>ts and the presence of significant noise <strong>in</strong> the data.<br />
The non-l<strong>in</strong>earities <strong>in</strong> the solution surface fitted to the resistance data by the neural network raise an<br />
additional problem when it comes to us<strong>in</strong>g the output of the tra<strong>in</strong>ed ANN with<strong>in</strong> an optimisation<br />
procedure. Because the five-dimensional solution surface is multimodal, it is not feasible to use a<br />
gradient-based optimisation method to reliably f<strong>in</strong>d the optimum set of parameters.<br />
To address this problem the authors have implemented a simple Genetic Algorithm (GA) based<br />
optimisation framework to <strong>in</strong>vestigate the application of GAs to catamaran design optimisation. The<br />
framework was designed to be as flexible as possible <strong>in</strong> order to experiment with the GA parameters<br />
to determ<strong>in</strong>e the best comb<strong>in</strong>ation of these sett<strong>in</strong>gs.<br />
2. Neural Networks<br />
Accord<strong>in</strong>g to Sarle (1994)<br />
Neural Networks are general purpose, flexible, non-l<strong>in</strong>ear models that, given enough hidden<br />
neurons and enough data, can approximate virtually any function to any desired degree of<br />
accuracy. In other words, Neural Networks are universal approximators. Neural Networks can be<br />
used when there is little knowledge about the form of the relationship between the <strong>in</strong>dependent<br />
and dependent variables.<br />
Most ANNs have some sort of tra<strong>in</strong><strong>in</strong>g rule whereby the weights of connections are adjusted on the<br />
basis of tra<strong>in</strong><strong>in</strong>g data. In other words, ANNs learn by example and exhibit some capability for<br />
generalization beyond the tra<strong>in</strong><strong>in</strong>g data. There are many different types and topologies of ANNs. The<br />
form that has found widest application is the feed-forward multi-layer perceptron or MLP.<br />
Fig.1: MLP topology<br />
441
MLPs have an <strong>in</strong>put layer with a series of <strong>in</strong>puts, one or more hidden layers and an output layer. The<br />
number of <strong>in</strong>put elements is equal to the number of variables <strong>in</strong> the <strong>in</strong>put dataset and the number of<br />
outputs equal to the number of result values required. The number of hidden layers and the number of<br />
elements <strong>in</strong> them can vary, and there are several techniques used to determ<strong>in</strong>e the optimum structure.<br />
There is considerable overlap between the fields of ANNs and statistics. In ANN term<strong>in</strong>ology,<br />
statistical <strong>in</strong>ference means learn<strong>in</strong>g to generalize from noisy data. Most ANNs that can learn to<br />
generalize effectively from noisy data are analogous to statistical methods. For example:<br />
• MLPs are a subset of the class of non-l<strong>in</strong>ear regression models.<br />
• MLPs with no hidden layer are basically generalized l<strong>in</strong>ear models.<br />
• MLPs with one hidden layer are closely related to projection pursuit regression.<br />
• MLPs with multiple <strong>in</strong>puts, multiple outputs and one or more hidden layers are analogous to<br />
multivariate, multiple non-l<strong>in</strong>ear regression.<br />
Despite the similarities between ANNs and statistical methods, ANNs have some specific advantages-<br />
• Power. ANNs are capable of modell<strong>in</strong>g extremely complex functions. In particular, ANNs are<br />
non-l<strong>in</strong>ear. For many years l<strong>in</strong>ear modell<strong>in</strong>g has been the commonly used technique <strong>in</strong> most<br />
modell<strong>in</strong>g doma<strong>in</strong>s, s<strong>in</strong>ce l<strong>in</strong>ear models had well-known optimisation strategies. Where the<br />
l<strong>in</strong>ear approximation was not valid (which was frequently the case) the models suffered<br />
accord<strong>in</strong>gly. ANNs also keep <strong>in</strong> check the “curse of dimensionality” problem, which bedevils<br />
attempts to model non-l<strong>in</strong>ear functions with large numbers of variables.<br />
• Ease of use. ANNs learn by example. The ANN user gathers representative data, and then<br />
<strong>in</strong>vokes tra<strong>in</strong><strong>in</strong>g algorithms which enable the ANN to learn the structure of the data. Although<br />
the user requires some knowledge of how to select and prepare the data, the level of user<br />
knowledge needed to successfully apply ANNs is much lower than that required for more<br />
traditional statistical methods.<br />
2.3. Tra<strong>in</strong><strong>in</strong>g and Test<strong>in</strong>g Overview<br />
The ANNs discussed <strong>in</strong> this paper were developed and tested us<strong>in</strong>g software called NeuroIntelligence,<br />
Alyuda (<strong>2005</strong>). NeuroIntelligence is a commercial ANN program, which specialises <strong>in</strong> the tra<strong>in</strong><strong>in</strong>g of<br />
multi-layer perceptrons. NeuroIntelligence conta<strong>in</strong>s a variety of tra<strong>in</strong><strong>in</strong>g algorithms, <strong>in</strong>clud<strong>in</strong>g back<br />
propagation, conjugate gradient descent, quasi-Newton and Levenberg-Marquardt methods.<br />
2.4. Tra<strong>in</strong><strong>in</strong>g and Test<strong>in</strong>g Data<br />
The data set used to tra<strong>in</strong> and test the ANN is a series of geometrically similar catamaran models<br />
based on the hull shape shown together with the parameters described <strong>in</strong> Table I.<br />
Model B/T L/∇ 1/3 Demihull<br />
WSA m 3<br />
3b 2.0 6.27 0.434<br />
4a 1.5 7.40 0.348<br />
4b 2.0 7.41 0.338<br />
4c 2.5 7.39 0.34<br />
5a 1.5 8.51 0.282<br />
5b 2.0 8.50 0.276<br />
5c 2.5 8.49 0.277<br />
6a 1.5 9.50 0.24<br />
6b 2.0 9.50 0.233<br />
6c 2.5 9.50 0.234<br />
Table I<br />
442
Both C T and C W data from a series of tank test experiments on scale models were available from<br />
Molland (1994), <strong>in</strong> four catamaran configurations (S/L=0.2, 0.3, 0.4 and 0.5) for each model. In this<br />
work the ANN was fitted directly to the C T data, rather than deriv<strong>in</strong>g C W for each model. Over the<br />
past decade there has been considerable discussion over the breakdown of C T <strong>in</strong>to its F n and R e<br />
dependent components for these types of catamaran vessels. The decision to fit the model C T data was<br />
made as it allows the reader to select their own scal<strong>in</strong>g method.<br />
It is essential when tra<strong>in</strong><strong>in</strong>g an ANN that an unbiased estimate of the generalisation error be available.<br />
In order to achieve this, it is necessary to partition the data <strong>in</strong>to mutually exclusive tra<strong>in</strong><strong>in</strong>g and<br />
validation sets. Data from the validation set are held out from the tra<strong>in</strong><strong>in</strong>g process and are used to<br />
estimate the error between the fitted neural network and the orig<strong>in</strong>al function. The network<br />
architecture and early stopp<strong>in</strong>g of the tra<strong>in</strong><strong>in</strong>g process can be determ<strong>in</strong>ed us<strong>in</strong>g the validation set<br />
error. However this can conta<strong>in</strong> some bias toward the chosen network architecture and as such a<br />
separate, <strong>in</strong>dependent test set should also be kept aside for a truly unbiased estimate of the network<br />
error.<br />
The resistance data available for neural network tra<strong>in</strong><strong>in</strong>g was randomly partitioned <strong>in</strong>to mutually<br />
exclusive tra<strong>in</strong><strong>in</strong>g (85%), validation (13%) and test (2%) sets. Because of the small number of data<br />
po<strong>in</strong>ts <strong>in</strong> the dataset it was not possible to allocate a large number of po<strong>in</strong>ts to the validation and test<br />
sets, and these were set to reasonable m<strong>in</strong>imum values based on recommendations from Guyon (1997)<br />
2.5. ANN Architecture<br />
The <strong>in</strong>itial test network architectures were constra<strong>in</strong>ed to three layers, one for <strong>in</strong>put, hidden and output<br />
layers respectively. This architecture was chosen as previous research <strong>in</strong>dicated that multiple hidden<br />
layers are rarely effective <strong>in</strong> terms of both accuracy and tra<strong>in</strong><strong>in</strong>g speed, Neocleous and Schizas (1995).<br />
A search was performed us<strong>in</strong>g the architecture search function <strong>in</strong> NeuroIntelligence. This allowed<br />
multiple networks to be tra<strong>in</strong>ed with different numbers of hidden layer neurons and the results<br />
collated to display the optimum network. Tra<strong>in</strong><strong>in</strong>g runs used the Quasi-Newton method for 50,000<br />
iterations, and 10 retra<strong>in</strong>s were performed for each network topology to m<strong>in</strong>imise error. The range of<br />
architectures searched was from 4-4-1 (i.e. 4 <strong>in</strong>puts, 4 hidden layer neurons and 1 output), to 4-17-1<br />
(i.e. 4 <strong>in</strong>puts, 17 hidden layer neurons and 1 output).<br />
Fig.2 displays the results of the search process with both the Average Absolute Error (AAE) and the<br />
standard deviation of the results reduc<strong>in</strong>g as the number of hidden layer neurons <strong>in</strong>creased.<br />
Improvement <strong>in</strong> error values tended to level off at about 10 hidden layer neurons, with improvements<br />
from that po<strong>in</strong>t onward be<strong>in</strong>g small.<br />
Average Absolute Error<br />
0.33<br />
0.31<br />
0.29<br />
0.27<br />
0.25<br />
0.23<br />
0.21<br />
Validation error<br />
Mean Validation Error<br />
0.19<br />
0.17<br />
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17<br />
Hidden Layer Neurons<br />
Fig.2: Validation set error for 10 retra<strong>in</strong>s of each network architecture<br />
443
It is important that both tra<strong>in</strong><strong>in</strong>g and test set error be <strong>in</strong> broad agreement with the validation set error,<br />
and it can be seen from Fig.3 that the three sets of values correlate well as the number of hidden layer<br />
neurons <strong>in</strong>creases.<br />
M<strong>in</strong>. Average Absolute Error<br />
0.28<br />
0.26<br />
0.24<br />
0.22<br />
0.2<br />
0.18<br />
0.16<br />
0.14<br />
0.12<br />
Tra<strong>in</strong><strong>in</strong>g Set<br />
Validation Set<br />
Test Set<br />
0.1<br />
4 5 6 7 8 9 10 11 12 13 14 15 16 17<br />
Hidden Layer Neurons<br />
Fig.3: M<strong>in</strong>imum error values<br />
Neural networks tra<strong>in</strong>ed with large numbers of hidden layer neurons tend to suffer from overfitt<strong>in</strong>g.<br />
This overfitt<strong>in</strong>g cannot always be determ<strong>in</strong>ed from the m<strong>in</strong>imum validation set error if the dataset is<br />
small. One way of m<strong>in</strong>imis<strong>in</strong>g the risk of overfitt<strong>in</strong>g is to select the network architecture based on a<br />
criterion that balances m<strong>in</strong>imum error aga<strong>in</strong>st network complexity. One such measure is Akaike’s<br />
Information Criterion, Akaike (1974), which has been widely used for model selection for both<br />
conventional statistical models and neural networks.<br />
-7050<br />
-7100<br />
-7150<br />
-7200<br />
AIC<br />
-7250<br />
-7300<br />
-7350<br />
-7400<br />
-7450<br />
-7500<br />
4 5 6 7 8 9 10 11 12 13 14 15 16 17<br />
Hidden Layer Neurons<br />
Fig.4: Akaike’s Information Criterion<br />
A graph of Akaike’s criterion for the tra<strong>in</strong>ed networks, Fig.4, shows a m<strong>in</strong>imum value at 14 hidden<br />
layer neurons and it can be concluded that this network architecture is a reasonable compromise<br />
between error values and network complexity. An additional consideration is that the standard<br />
deviation of the Average Absolute Error values for the validation set is at a m<strong>in</strong>imum at 14 hidden<br />
layer neurons, which gives some confidence that these results will be repeatable <strong>in</strong> subsequent<br />
tra<strong>in</strong><strong>in</strong>g.<br />
444
Couser et al. (2004) found that a hidden layer conta<strong>in</strong><strong>in</strong>g approximately 15 neurons was optimal for<br />
this dataset, and it is encourag<strong>in</strong>g that despite the different treatment of the data, the new analysis is <strong>in</strong><br />
broad agreement with this conclusion. In this work architectures conta<strong>in</strong><strong>in</strong>g two hidden layers were<br />
<strong>in</strong>vestigated but were not found to be advantageous, except for an <strong>in</strong>dication that architectures with<br />
two hidden layers may be effective if the number of neurons <strong>in</strong> the first hidden layer was small.<br />
As a result, an <strong>in</strong>vestigation was performed to determ<strong>in</strong>e whether network architectures with the same<br />
or less total number of degrees of freedom could be found that would outperform the s<strong>in</strong>gle layer<br />
architecture. It was felt that if these degrees of freedom could be redistributed <strong>in</strong>to two hidden layers<br />
the ability of the ANN to model a highly non-l<strong>in</strong>ear dataset might improve.<br />
As the total number of two layer architectures was high it was not feasible to search all possible<br />
configurations. Instead architectures were <strong>in</strong>vestigated that had either the same total number of hidden<br />
layer nodes as the previously determ<strong>in</strong>ed optimum (i.e. 14), one less total hidden layer nodes than the<br />
optimum (i.e. 13), two less total hidden layer nodes than the optimum (i.e. 12) or three less total<br />
hidden layer nodes than the optimum (i.e. 11). For each of these total numbers of hidden layer<br />
neurons different architectures were evaluated conta<strong>in</strong><strong>in</strong>g between 1 and 8 neurons <strong>in</strong> the second<br />
hidden layer.<br />
0.2<br />
M<strong>in</strong>. Average Absolute Error<br />
0.195<br />
0.19<br />
0.185<br />
0.18<br />
0.175<br />
Total 11 HL neurons<br />
Total 12 HL neurons<br />
Total 13 HL neurons<br />
Total 14 HL neurons<br />
0.17<br />
0.165<br />
0 1 2 3 4 5 6 7 8<br />
Neurons <strong>in</strong> 2nd Hidden Layer<br />
Fig.5: M<strong>in</strong>imum error values for two hidden layers<br />
Fig.5 clearly shows a trend towards lower validation set errors when more than one neuron is placed<br />
on a second hidden layer. The networks conta<strong>in</strong><strong>in</strong>g a total of 11 and 12 hidden layer neurons have<br />
m<strong>in</strong>ima at 4 and 5 second layer neurons respectively, while the networks with a total of 13 and 14<br />
hidden layer neurons cont<strong>in</strong>ued to improve up to the po<strong>in</strong>t where there were 8 neurons <strong>in</strong> the second<br />
hidden layer, the maximum tested.<br />
On the basis of this test<strong>in</strong>g, the network selected used a 4-7-5-1 architecture. This network had<br />
tra<strong>in</strong><strong>in</strong>g and validation set errors only slightly higher than the 4-14-1 architecture, but scored<br />
significantly better on Akaike’s criterion due to it hav<strong>in</strong>g two fewer nodes <strong>in</strong> the hidden layers.<br />
Appendix 1 displays sample C T curves derived from this ANN architecture for models 4b, 6a and 6b.<br />
These models have been selected for illustration as they have proven to be the configurations most<br />
difficult to fit <strong>in</strong> previous work, due to the comb<strong>in</strong>ation of significant noise and nonl<strong>in</strong>earities <strong>in</strong> the<br />
F n axis.<br />
445
3. Optimisation us<strong>in</strong>g Genetic Algorithms<br />
An ANN derived from the catamaran resistance data permits C T to be estimated when a set of specific<br />
<strong>in</strong>put parameters is def<strong>in</strong>ed. However it is often desirable to fix the value of one or more parameters<br />
and then determ<strong>in</strong>e the optimum values of the rema<strong>in</strong><strong>in</strong>g parameters.<br />
To achieve this, a search or optimisation method is required. This process is complicated <strong>in</strong> the case<br />
of catamaran resistance due to the <strong>in</strong>teraction of wave patterns between the two demi-hulls, which<br />
may cause significant non-l<strong>in</strong>earities <strong>in</strong> the vessel’s resistance and an associated likelihood of a multimodal<br />
solution space, Molland et al. (1994). In this case it is necessary to use a method capable of<br />
global optimisation to f<strong>in</strong>d the m<strong>in</strong>imum resistance.<br />
One such method is the Genetic Algorithm (GA). GAs are a search and optimisation method based on<br />
the process of biological evolution. GAs differ from traditional optimisation techniques <strong>in</strong> that they<br />
<strong>in</strong>volve a search from a population of solutions, not from a s<strong>in</strong>gle po<strong>in</strong>t.<br />
Each iteration of a GA <strong>in</strong>volves a competitive selection that penalises poor solutions. The solutions<br />
with high fitness are recomb<strong>in</strong>ed with others to produce members of the next generation, which<br />
<strong>in</strong>herit properties from their parents. Solutions are also mutated by mak<strong>in</strong>g small, random changes to<br />
one or more free parameters. Recomb<strong>in</strong>ation and mutation are used to generate new solutions, which<br />
are biased towards regions of the space for which good solutions have already been seen.<br />
The steps <strong>in</strong>volved <strong>in</strong> a genetic algorithm are as follows:<br />
Initialise the population<br />
Evaluate <strong>in</strong>itial population<br />
REPEAT<br />
Perform competitive selection<br />
Apply genetic operators to generate new solutions<br />
Evaluate solutions <strong>in</strong> the population<br />
UNTIL some convergence criteria is satisfied<br />
One of the strengths of GAs is that they perform well on noisy solution spaces where there may be<br />
multiple local optima. GAs tend not to get stuck on a local m<strong>in</strong>ima and can often f<strong>in</strong>d the global<br />
optimum.<br />
GAs have been used by several researchers to optimise catamaran designs, Doctors and Day (2000),<br />
Tuck and Lazauskas (1996), Hearn and Wright (1998). However these have calculated resistance<br />
directly us<strong>in</strong>g first pr<strong>in</strong>ciples analysis. One weakness of GAs is that they are slow due to the high<br />
number of evaluations of the objective function required. This means that the performance penalty<br />
implicit <strong>in</strong> first pr<strong>in</strong>ciple analysis of resistance is a significant impediment to effective optimisation<br />
us<strong>in</strong>g GAs.<br />
The use of an ANN based objective function allows the fitness to be evaluated <strong>in</strong> a small fraction of<br />
the time taken to perform first pr<strong>in</strong>ciples analysis and permits the GA optimisation to complete <strong>in</strong> a<br />
reasonably small amount of time. It also allows a researcher to use exist<strong>in</strong>g experimental data that<br />
may have been derived from tank test<strong>in</strong>g which cannot be reproduced easily.<br />
4. GA parameter sett<strong>in</strong>gs<br />
4.1. Exploration versus exploitation<br />
A GA based search can be viewed as a trade off between the ma<strong>in</strong>tenance of sufficient diversity <strong>in</strong> the<br />
population to permit the coverage of the entire solution space (exploration), and the need to converge<br />
on the optimum solution with<strong>in</strong> an acceptable time (exploitation), Eiben and Schippers (1998).<br />
446
There are several variables or algorithms that can be applied to manipulate the degree of exploration<br />
or exploitation that the GA exhibits, Goldberg (1989). These <strong>in</strong>clude:<br />
• Selection method<br />
• Elitism<br />
• Population nich<strong>in</strong>g - fitness shar<strong>in</strong>g<br />
4.2. Selection<br />
Much research has been done <strong>in</strong>to different selection methods with two of the most widely used<br />
methods be<strong>in</strong>g roulette wheel selection and tournament selection. Roulette wheel selection works by<br />
choos<strong>in</strong>g two <strong>in</strong>dividuals from the population and compar<strong>in</strong>g their fitness scores. The <strong>in</strong>dividual with<br />
the highest fitness score is selected to be a parent for an <strong>in</strong>dividual <strong>in</strong> the subsequent generations.<br />
Roulette wheel selection assigns a likelihood of representation <strong>in</strong> the follow<strong>in</strong>g generation<br />
proportional to the magnitude of an <strong>in</strong>dividual’s fitness score. The difficulty <strong>in</strong>herent <strong>in</strong> the use of<br />
roulette wheel selection is that the likelihood of representation <strong>in</strong> the follow<strong>in</strong>g generation is<br />
dependent on the scal<strong>in</strong>g factor used to scale the results of the objective function to create fitness<br />
scores. Chang<strong>in</strong>g this scal<strong>in</strong>g factor can dramatically change the results of the selection process, and it<br />
is difficult to determ<strong>in</strong>e the correct scal<strong>in</strong>g factor ahead of time.<br />
On the other hand, tournament selection is a simple comparison of fitness values and is <strong>in</strong>dependent<br />
of their scal<strong>in</strong>g. As a result, tournament selection is both easier to implement and more robust <strong>in</strong> use<br />
than roulette wheel selection.<br />
4.3. Elitism<br />
Elitism ensures that the next generation’s best <strong>in</strong>dividual will be at least as good as any from the<br />
previous generation by automatically <strong>in</strong>clud<strong>in</strong>g the best <strong>in</strong>dividual from the previous population,<br />
Andris and Frollo (2002).<br />
4.4 Fitness shar<strong>in</strong>g<br />
Fitness shar<strong>in</strong>g works by scal<strong>in</strong>g the fitness of <strong>in</strong>dividuals based on how similar they are to all other<br />
<strong>in</strong>dividuals <strong>in</strong> the population. Individuals that are very similar to others <strong>in</strong> the population are given a<br />
slight penalty, while <strong>in</strong>dividuals that show novel features are rewarded by <strong>in</strong>creas<strong>in</strong>g their fitness<br />
relative to the population. This is <strong>in</strong>tended to ensure that the population ma<strong>in</strong>ta<strong>in</strong>s sufficient diversity<br />
to avoid premature convergence on false optima, Goldberg (1987).<br />
5. GA based optimiser prototype<br />
A prototype GA based optimisation framework, HullGA, has been developed by the authors <strong>in</strong> order<br />
to evaluate the feasibility of us<strong>in</strong>g an ANN model for the objective function of an optimisation<br />
process. The GA code used for the prototype developed by the authors was developed and tested <strong>in</strong><br />
MATLAB 5.3 (The MathWorks 2003). The objective function used is the predicted total resistance<br />
coefficient (C T ) for the catamaran configuration at a selected F n .<br />
A design was adopted that allowed some experimentation with different parameters which may affect<br />
the evolution process. Parameters which can be adjusted via the user <strong>in</strong>terface are –<br />
• Population size<br />
• Limit on number of generations evaluated<br />
• Selection method – roulette wheel or tournament selection<br />
• Mutation rate<br />
• Elitism - on or off<br />
• Fitness shar<strong>in</strong>g - on or off<br />
447
The prototype also codifies specific exploration and exploitation phases, each of which used their own<br />
sett<strong>in</strong>gs for the above parameters.<br />
A prelim<strong>in</strong>ary evaluation of the effects of the <strong>in</strong>terplay of the different evolution parameters has been<br />
conducted, but no comprehensive study has been done due to the large amount of time required.<br />
Initial conclusions are as follows –<br />
5.1. Population size<br />
The goal of test<strong>in</strong>g population size is to ensure that the population is large enough to ma<strong>in</strong>ta<strong>in</strong><br />
sufficient diversity to avoid premature convergence.<br />
Tests were run for population sizes rang<strong>in</strong>g between 10 and 150. Negligible differences <strong>in</strong> rate of<br />
convergence or the fitness level of the best solution were observed for populations greater than 20<br />
<strong>in</strong>dividuals. From these results, it appears that there is no advantage to be ga<strong>in</strong>ed from a population<br />
size greater than 20 for this particular dataset.<br />
5.2. Selection method<br />
Much research has been done <strong>in</strong>to different selection methods and there is support <strong>in</strong> the literature for<br />
both roulette wheel selection and tournament selection.<br />
In the case of the HullGA prototype it was found that tournament selection gave the best results,<br />
primarily due to the difficulty of choos<strong>in</strong>g suitable fitness scal<strong>in</strong>g parameters to avoid premature<br />
convergence when us<strong>in</strong>g roulette wheel selection.<br />
5.3. Elitism<br />
The use of elitism dur<strong>in</strong>g the exploration phase may lead to premature convergence <strong>in</strong> a multi-modal<br />
solution space. Initial experimentation with the HullGA optimiser <strong>in</strong>dicated that it may be best to use<br />
elitism only dur<strong>in</strong>g the exploitation phase of the optimisation process.<br />
5.4. Fitness shar<strong>in</strong>g<br />
Fitness shar<strong>in</strong>g has been shown to be effective <strong>in</strong> ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g population diversity dur<strong>in</strong>g the<br />
exploration phase, which helps to prevent premature convergence on local rather than global optima.<br />
Dur<strong>in</strong>g the exploitation phase, however, this diversity may <strong>in</strong>hibit convergence to an optimum<br />
solution and thus fitness shar<strong>in</strong>g may not be useful dur<strong>in</strong>g this phase.<br />
Although roulette wheel selection appears to benefit from fitness shar<strong>in</strong>g, previous research has found<br />
that fitness shar<strong>in</strong>g can display chaotic <strong>in</strong>teractions with tournament selection, with unexpected<br />
results, Oei et al. (1991). This <strong>in</strong>teraction between evolution parameters is an area for future<br />
<strong>in</strong>vestigation.<br />
6. Example catamaran optimisation<br />
Based on the test results, a GA with tournament selection and fitness shar<strong>in</strong>g <strong>in</strong> the exploration phase,<br />
and tournament selection and elitism <strong>in</strong> the exploitation phase, was used to optimise catamaran hull<br />
forms for m<strong>in</strong>imum resistance The exploration and exploitation phase were evenly divided over the<br />
total number of generations.<br />
Fig.6 demonstrates an example at Fn 0.3. In this case resistance was optimised based on range<br />
constra<strong>in</strong>ts for S/L Ratio (0.15 to 0.5), B/T ratio (1.5 to 2.5), and L/∇ 1/3 ratio (6.27 to 9.51) as<br />
described <strong>in</strong> Molland et al. (1994).<br />
448
Fig.6: The HullGA user <strong>in</strong>terface<br />
Due to the multi-dimensional solution space, <strong>in</strong> this case four free parameters, it is not feasible to<br />
visualise the complete surface. The 3D surface plot <strong>in</strong> Fig.6 captures C T values (vertical axis) aga<strong>in</strong>st<br />
B/T ratio and L/∇ 1/3 ratio, at a fixed S/L ratio, equal to the average S/L ratio of the population.<br />
The graph <strong>in</strong> the lower left corner of Fig.6 illustrates the best, worst and average fitness values over<br />
the generations evaluated. The highest fitness values <strong>in</strong> the population are <strong>in</strong>dicated by the dashed<br />
l<strong>in</strong>e (upper), the average fitness values by the solid l<strong>in</strong>e (middle), and the lowest fitness values by the<br />
dash-dot l<strong>in</strong>e (lower).<br />
The <strong>in</strong>itial oscillations <strong>in</strong> the fitness graph illustrate the exploration of the solution space us<strong>in</strong>g fitness<br />
shar<strong>in</strong>g. As the three fitness plots meet it can be seen that the GA is converg<strong>in</strong>g to an optimal part of<br />
the solution space.<br />
7. Future work<br />
The work described here has used a GA to f<strong>in</strong>d the optimal solution <strong>in</strong> a problem space with four free<br />
parameters. More useful results could be achieved if additional catamaran hull parameters were<br />
<strong>in</strong>cluded for optimisation, such as longitud<strong>in</strong>al centre of buoyancy position (LCB), prismatic or block<br />
coefficients (C P or C B ), or midship area coefficient (C M )<br />
Another natural extension to the work described here is to extend the GA to perform multi-objective<br />
optimisation. Example objectives might be the simultaneous optimisation of both catamaran<br />
resistance and seakeep<strong>in</strong>g qualities.<br />
449
8. Conclusions<br />
The work detailed <strong>in</strong> this paper has demonstrated that a sparse dataset with a relatively high degree of<br />
noise can be fitted effectively with a feed-forward neural network. In addition we have demonstrated<br />
that there may be benefits result<strong>in</strong>g from the <strong>in</strong>vestigation of networks with two hidden layers rather<br />
than one. In particular it may be possible to f<strong>in</strong>d network architectures that have fewer degrees of<br />
freedom comb<strong>in</strong>ed with lower error levels.<br />
The work detailed <strong>in</strong> this paper has also demonstrated that a comb<strong>in</strong>ation of GAs and ANNs can be<br />
successfully used as an optimisation tool for catamaran design parameters. In this problem doma<strong>in</strong> it<br />
appears that us<strong>in</strong>g tournament selection alone dur<strong>in</strong>g the exploration phase, and tournament selection<br />
with elitism <strong>in</strong> the exploitation phase gives the best results.<br />
This research has also shown that the success of a GA is closely related to the selection methods and<br />
other parameters used. However, a particular comb<strong>in</strong>ation of parameters may not produce a<br />
successful optimiser <strong>in</strong> a different problem doma<strong>in</strong>. In particular, if the solution space becomes more<br />
multi-modal <strong>in</strong> nature, the importance of select<strong>in</strong>g the correct GA parameters <strong>in</strong>creases.<br />
References<br />
AKAIKE, H. (1974), A New Look At The Statistical Model Identification. IEEE Trans. Auto. Control.,<br />
19, 723<br />
ALYUDA (<strong>2005</strong>), Neuro<strong>in</strong>telligence Manual, http://www.alyuda.com<br />
ANDRIS, P.; FROLLO, I. (2002), Optimisation of NMR Coils by Genetic Algorithm, Measurement<br />
Science Review 2/2<br />
COUSER, P.R.; MASON, A.P.; MASON, G.; SMITH, C.R.; VON KONSKY, B.R. (2004), Artificial<br />
Neural Networks for Hull Resistance Prediction, 3 rd <strong>COMPIT</strong>, Siguenza<br />
DARPA (1998), Neural Network Study, AFCEA International Press<br />
DOCTORS, L.J.; DAY, A.H. (2000), The Survival of the Fittest – Evolutionary Tools for<br />
Hydrodynamic Design of Ship Hull Form, Royal Inst. Naval Arch.<br />
EIBEN, A.E.; SCHIPPERS, C.A. (1998), On evolutionary exploration and exploitation, Fundamenta<br />
Informaticae 35<br />
GOLDBERG, D.E. (1989), Genetic Algorithms <strong>in</strong> Search, Optimisation and Mach<strong>in</strong>e Learn<strong>in</strong>g,<br />
Addison-Wesley Inc.<br />
GOLDBERG, D.E.; RICHARDSON, J. (1987), Genetic algorithms with shar<strong>in</strong>g for multimodal<br />
function optimization, 2 nd Int. Conf. Genetic Algorithm<br />
GUYON, I. (1997), A scal<strong>in</strong>g law for the validation_set tra<strong>in</strong><strong>in</strong>g_set size ratio, Bell Laboratories<br />
HEARN, G.E.; WRIGHT, P.N.H. (1998), Michell wave resistance characteristics result<strong>in</strong>g from the<br />
optimisation of a catamaran hullform via genetic algorithms, EMAC’98<br />
INSEL, M.; MOLLAND, A. (1992), An <strong>in</strong>vestigation <strong>in</strong>to the resistance components of high speed<br />
displacement catamarans, Trans. RINA 134(A)<br />
450
JAIN, A.K.; MAO, J.C.; MOHIUDDIN, K.M. (1996), Artificial neural networks: A tutorial,<br />
Computer 29(3), pp.31-44<br />
KOUSHAN, K. (2003), Artificial neural networks for prediction of ship resistance and wetted surface<br />
area, Mar<strong>in</strong>tek<br />
MATLAB version 5.3: The MathWorks [Onl<strong>in</strong>e] http://www.themathworks.com.<br />
MOLLAND, A.F.; WELLICOME, J.F.; COUSER, P.R. (1994), Resistance experiments on a<br />
systematic series of high speed displacement catamaran forms: variation of length-displacement ratio<br />
and breadth-draught ratio, Ship Science Report 71, Dept. Ship Science, Univ. of Southampton<br />
NEOCLEOUS, C.; SCHIZAS, C. (1995), Artificial neural networks <strong>in</strong> mar<strong>in</strong>e propeller design,<br />
ICNN’95, Int. Conf. on Neural Networks, Vol. 2, pp.1098-1102, New York, IEEE Computer Society<br />
Press<br />
OEI, C.K.; GOLDBERG, D.E.; CHANG, S.-J. (1991), Tournament selection, nich<strong>in</strong>g, and the<br />
preservation of diversity. IlliGAL Techn. Report 91011, Univ. of Ill<strong>in</strong>ois at Urbana-Champaign<br />
SARLE, W.S. (1994), Neural networks and statistical models, 19 th Annual SAS Users Group Int.<br />
Conf.<br />
SCHIZAS, C.N., NEOCLEOUS, C.C. (1995), Artificial Neural Networks <strong>in</strong> mar<strong>in</strong>e propeller design,<br />
Int. Conf. Neural Networks<br />
TUCK, E.O.; LAZAUSKAS, L. (1996), Unconstra<strong>in</strong>ed ships of m<strong>in</strong>imum drag, Dept. Applied<br />
Mathematics Techn. Report, The University of Adelaide<br />
451
Appendix 1<br />
4b - S/L = 0.2<br />
4b - S/L = 0.3<br />
10<br />
10<br />
9.5<br />
Experiment<br />
ANN<br />
9.5<br />
Experiment<br />
ANN<br />
9<br />
9<br />
8.5<br />
8.5<br />
8<br />
8<br />
CT (x 1000)<br />
7.5<br />
7<br />
CT (x 1000)<br />
7.5<br />
7<br />
6.5<br />
6.5<br />
6<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9<br />
Fn<br />
Fn<br />
4b - S/L = 0.4<br />
4b - S/L = 0.5<br />
9<br />
9<br />
Experiment<br />
Experiment<br />
8.5<br />
ANN<br />
8.5<br />
ANN<br />
8<br />
8<br />
7.5<br />
7.5<br />
CT (x 1000)<br />
7<br />
6.5<br />
CT (x 1000)<br />
7<br />
6.5<br />
6<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
Fn<br />
Fig.7: C T curves for model 4b<br />
452
6a - S/L = 0.2<br />
6a - S/L = 0.3<br />
8<br />
8<br />
Experiment<br />
Experiment<br />
7.5<br />
ANN<br />
7.5<br />
ANN<br />
7<br />
7<br />
CT (x 1000)<br />
6.5<br />
6<br />
CT (x 1000)<br />
6.5<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
6a - S/L = 0.4<br />
6a - S/L = 0.5<br />
8<br />
8<br />
Experiment<br />
Experiment<br />
7.5<br />
ANN<br />
7.5<br />
ANN<br />
7<br />
7<br />
CT (x 1000)<br />
6.5<br />
6<br />
CT (x 1000)<br />
6.5<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
Fig.8: C T curves for model 6a<br />
453
6b - S/L = 0.2<br />
6b - S/L = 0.3<br />
8<br />
8<br />
Experiment<br />
Experiment<br />
7.5<br />
ANN<br />
7.5<br />
ANN<br />
7<br />
7<br />
CT (x 1000)<br />
6.5<br />
6<br />
CT (x 1000)<br />
6.5<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
6b - S/L = 0.4<br />
6b - S/L = 0.5<br />
8<br />
Experiment<br />
8<br />
Experiment<br />
7.5<br />
ANN<br />
7.5<br />
ANN<br />
7<br />
7<br />
CT (x 1000)<br />
6.5<br />
6<br />
CT (x 1000)<br />
6.5<br />
6<br />
5.5<br />
5.5<br />
5<br />
5<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
4.5<br />
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />
Fn<br />
Fig.9: C T curves for model 6b<br />
454
! <br />
"! # <br />
$ #%<br />
<br />
#%<br />
&#<br />
<br />
! "<br />
<br />
! ! ! <br />
!" !#$ <br />
!! %<br />
&' !"! (" ( <br />
<br />
)! !<br />
# (( ( *! <br />
! ! <br />
<br />
+#( "(! ! <br />
<br />
<br />
, ' !! ! ! * ! <br />
! $! " <br />
! !"$&" ! ! <br />
( <br />
" (" ( ( ! <br />
# " ! " <br />
" ! ( ( " <br />
( # - ! <br />
! !. <br />
!! " ! <br />
!! <br />
!! ( / & * <br />
" ( * 0- 1 <br />
/&*" <br />
!! (<br />
<br />
#! $% <br />
<br />
2 3'())* <br />
" " "(<br />
# *"# "(<br />
(" ("!("3 <br />
3 , !(" <br />
( !' ! (" !
' ! !( (" !# ("3<br />
# * (! <br />
(" + ,-())(#<br />
<br />
. !' ! +<br />
' 5 ! " <br />
! <br />
! !! " (<br />
67* ! <br />
1 " ! <br />
!* (" <br />
" ' ! ( /"<br />
!!!<br />
! (<br />
( # !( ( /&*<br />
! 8<br />
!!( %<br />
• / <br />
• !'#((" !<br />
(" !!( <br />
• ("3(" !3 <br />
• ("#"<br />
• * (" <br />
• & (<br />
! # ! <br />
/&*# * # " (!<br />
! <br />
<br />
&! <br />
<br />
,9(!/& * <br />
: ! <br />
(" " ! !"<br />
(!<br />
• ; 3<br />
• + <br />
• 8 <br />
• . < ! <br />
• *"<br />
"! ! !# (<br />
'<br />
<br />
/&* ((!3 =<br />
! ! <br />
, ! ! (<br />
" ( <br />
( ! $ ' <br />
/ ! (1 <br />
(!("!6 <br />
7 ( "( ! 0 3 <br />
!!" ! " $<br />
<br />
4
,? !/&*+&+/<br />
! 1 ?/&* ( ,?<br />
!( <br />
<br />
<br />
<br />
(%(<br />
<br />
<br />
+ <br />
/<br />
@ <br />
8 <br />
&<br />
. <br />
) <br />
*+<br />
%<br />
*+<br />
. "<br />
*"<br />
<br />
<br />
<br />
<br />
,9%, !/&* ,?%@ !/&*<br />
&#())*<br />
<br />
<br />
<br />
&!! %<br />
<br />
! ' ! !! 3<br />
! !! (" !<br />
!!(" 3 (";<br />
!("# A;0 <br />
) +(" "1! !' <br />
! " !(" !!( <br />
!! 3 8*5) <br />
! .! /00(1!((!<br />
! !!"<br />
" 1/00/#' !(! <br />
'!A;!! ( '!" !<br />
(" (" <br />
/&*<br />
<br />
&!#! '<br />
<br />
/&* !! !<br />
)# ("" ! !<br />
" +! ! !<br />
+<br />
!!!! ! ! <br />
<br />
<br />
&!&! <br />
<br />
!" !8/ /&*"<br />
! !$! <br />
! " !" !$#1<br />
! !! 0 " <br />
@ !" 6( ()*37 (!"<br />
! ( <br />
<br />
<br />
<br />
>
&!,! ) <br />
<br />
!! !(" ! <br />
• ( ! ("<br />
/&* ("! <br />
• -"6( (7 !" <br />
<br />
• ! " !<br />
!! /&* ! <br />
"( ( <br />
<br />
&!-! .<br />
<br />
: ! ! (<br />
! (" " " 3 <br />
: " ! !<br />
! ! !! ! : <br />
/&* !#&+/ !<br />
" <br />
!!! ! 3<br />
"" ,A,( #!3<br />
C1!6C17<br />
<br />
<br />
,A% + ( <br />
6"7 ! 6(7 <br />
3"<br />
,%;(( ! (<br />
3 <br />
<br />
<br />
,! $" <br />
<br />
@ ( ( <br />
?DD? ?DDA.&())2#"( /&*<br />
!88 # 8 # "<br />
!/&* ! ! <br />
! ! /&* ( ! @3 4ADD<br />
)C88 # 2/ 5(E 3 ! !<br />
: ?DD? @3 BDD )C 8 # 25( E )#3 ( !<br />
'(/&* !/?DDA<br />
<br />
<br />
B
,%G/ 5(E G<br />
<br />
<br />
,4%G5(E )#G<br />
1%8 <br />
+3 ! +3 !-4<br />
" ADABA ?FD?<br />
( ?F?DD ?BAAD<br />
067 DDD A??D<br />
;67 ??D ?9BD<br />
; !67 9?DD 9?DD<br />
/ !67 9DD 9A<br />
/! 49B AB<br />
; ! 9D>>4? B4F4B<br />
/" ?4 ? <br />
<br />
/&* !( %+. "#<br />
!5()))!( 3= &<br />
* 3:!(! <br />
+ ;/8:/( +@ H <br />
<br />
<br />
<br />
<br />
,>%& !#( !!' D<br />
$+3 !"<br />
<br />
" ( ' ( '( <br />
! " ( : <br />
# " ! ! ! <br />
!! 8*5).! /00(#<br />
!3 ( <br />
: ( !<br />
<br />
F
!! !#( " :<br />
!! ( ("/<br />
! 1 ,><br />
!!!" " (" # 3<br />
,! #?D!' <br />
""3 !" <br />
<br />
,!! / <br />
<br />
"(" <br />
"(" ! ' '! ,<br />
" (" <br />
!( ! $ ! "!(<br />
! ! *( ! !<br />
# (! " $ <br />
! ! ! !!! !<br />
<br />
,!#! 0/ $ <br />
<br />
"/&* "! ( ( <br />
" (<br />
! <br />
!/&* !! " (<br />
(!<br />
8 # ,B<br />
<br />
<br />
,B%/ 25(E )#3<br />
<br />
! ! " <br />
" ( (" ! !<br />
( !( !<br />
( <br />
! ! FDD<br />
$/ ( ! <br />
( " ( ! <br />
# " " ( 9 !<br />
! <br />
" /" ! " <br />
!! ( # "#" ": <br />
" ""( <br />
31 " <br />
"(! <br />
<br />
4D
-! 1 %<br />
<br />
" ! /&* 3 ! <br />
!3"! !"<br />
( .&())2#)#! <br />
! " # " ! <br />
+#" !A> " ( <br />
I4J#" %<br />
<br />
<br />
,46<br />
<br />
% ?<br />
= ? ⋅ <br />
D<br />
<br />
<br />
? π D<br />
<br />
( !( D " ? <br />
"!2/ 5(E 3!!"!!<br />
06?DD?99?FK?DD?9?D?7( ,<br />
"!25(E )#3!! /!<br />
(?? ( " <br />
<br />
-!! ( <br />
<br />
" ( "" <br />
/&* ' ! * 2/ 5(<br />
E 3"6 L?F?7( (" ADD<br />
9A ,F(("((,9D(<br />
(" (" ( <br />
!(" ! ("". "# <br />
("!,99( (" <br />
<br />
,9? 9 ( " ! 88 # " ( # " <br />
!" / ,949F!8 #"<br />
" ( ,9? 9B<br />
! " " !#""<br />
!( "0" /&*<br />
(". ("! (" <br />
,9?<br />
<br />
! (" 6/07 68/7<br />
( ,9A 9!2/ 5(E 3 94 9>!25(E )#3<br />
@ ! ! <br />
! !" $ %<br />
, , 8<br />
σ = 68<br />
− 8D7<br />
± 6<br />
− D7<br />
<br />
7<br />
7<br />
<br />
! 7 7 88 ! ! <br />
#8 ) ) ( !"3# ,)1$<br />
/&* ! "!"!<br />
2 "! 5"<br />
!( !(<br />
<br />
"!!2/ 5(E 3 <br />
( (" (( !+ ("<br />
(<br />
' "<br />
<br />
<br />
88<br />
<br />
49
2/ 5(E 3<br />
Hs (m)<br />
6.0<br />
5.0<br />
4.0<br />
3.0<br />
2.0<br />
1.0<br />
Significant wave height<br />
Hs (WAVEX)<br />
Forecast 2002-11-29<br />
Forecast 2002-11-30<br />
Forecast 2002-12-01<br />
0.0<br />
2002-11-29 00:00 2002-11-30 00:00 2002-12-01 00:00 2002-12-02 00:00 2002-12-03 00:00<br />
,F%& ! ("<br />
<br />
Zero-upcross<strong>in</strong>g wave period<br />
14.0<br />
12.0<br />
WAVEX<br />
10.0<br />
8.0<br />
6.0<br />
4.0<br />
2.0<br />
0.0<br />
2002-11-29 00:00 2002-11-30 00:00 2002-12-01 00:00 2002-12-02 00:00 2002-12-03 00:00<br />
Tz (s)<br />
,9D%&$ <br />
<br />
Relative wave peak direction<br />
250<br />
WAVEX<br />
200<br />
150<br />
100<br />
50<br />
0<br />
2002-11-29 00:00 2002-11-30 00:00 2002-12-01 00:00 2002-12-02 00:00 2002-12-03 00:00<br />
Dir (deg)<br />
<br />
Acceleration (m/s2)<br />
1.0<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
Vertical acceleration at bow<br />
(SDEV)<br />
Measurements<br />
Octopus us<strong>in</strong>g WAVEX<br />
Octopus us<strong>in</strong>g forecast<br />
,99%& !("<br />
"6DM!("7<br />
Stress (N/mm2)<br />
14.0<br />
12.0<br />
10.0<br />
8.0<br />
6.0<br />
4.0<br />
2.0<br />
Stress SB<br />
(SDEV)<br />
Measurements<br />
Octopus us<strong>in</strong>g WAVEX<br />
Octopus us<strong>in</strong>g forecast<br />
0.0<br />
2002-11-29 2002-11-30 2002-12-01 2002-12-02 2002-12-03<br />
<br />
0.0<br />
2002-11-29 2002-11-30 2002-12-01 2002-12-02 2002-12-03<br />
<br />
,9?% + /&* "<br />
(6 " 7<br />
<br />
,9A% + /&* ,9% + /&* <br />
6
70<br />
2.5<br />
Stress [N/mm²]<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
Measurement<br />
Octopus<br />
ACCv [m/s²]<br />
2<br />
1.5<br />
1<br />
0.5<br />
Measurement<br />
Octopus<br />
0<br />
04-02-15<br />
00:00<br />
04-02-15<br />
06:00<br />
04-02-15<br />
12:00<br />
04-02-15<br />
18:00<br />
04-02-16<br />
00:00<br />
04-02-16<br />
06:00<br />
04-02-16<br />
12:00<br />
<br />
0<br />
04-02-15<br />
00:00<br />
04-02-15<br />
06:00<br />
04-02-15<br />
12:00<br />
04-02-15<br />
18:00<br />
04-02-16<br />
00:00<br />
04-02-16<br />
06:00<br />
04-02-16<br />
12:00<br />
<br />
,9B% + /&* #<br />
" (<br />
,9>%+ /&*#<br />
<br />
<br />
,25(E )#3 ( /&* <br />
!!( " ! !1 <br />
! "!(" 1<br />
" (" !( (" ,9<br />
(" ! "<br />
<br />
2! 34<br />
<br />
! !!" <br />
( " "(" #/009#<br />
! #!<br />
!!" !!! !!#<br />
! !! ( " ' ! <br />
<br />
25<br />
20<br />
Measurement<br />
Octopus<br />
Roll [°]<br />
15<br />
10<br />
5<br />
0<br />
04-02-15<br />
00:00<br />
04-02-15<br />
06:00<br />
04-02-15<br />
12:00<br />
04-02-15<br />
18:00<br />
04-02-16<br />
00:00<br />
04-02-16<br />
06:00<br />
04-02-16<br />
12:00<br />
<br />
,9F%+ /&*# 5(E )#<br />
<br />
,9F( ! !+3 !-4<br />
! !! !"(<br />
!! ! <br />
: !" #<br />
<br />
. " !! <br />
!(( !- <br />
" ( ,?D, <br />
! !!" !<br />
5 !!(" ! ", <br />
( ! /0::'/000<br />
( ' (<br />
<br />
4A
! ! <br />
!! 5 !! !<br />
"!& /00;&#()))#<br />
<br />
3000<br />
2500<br />
2000<br />
Vertical bend<strong>in</strong>g moment [MNm]<br />
1500<br />
1000<br />
500<br />
0<br />
-500<br />
-1000<br />
Hogg<strong>in</strong>g(l<strong>in</strong>ear)<br />
Hogg<strong>in</strong>g<br />
Still-water<br />
(Hogg<strong>in</strong>g+Sagg<strong>in</strong>g)/2<br />
-1500<br />
Sagg<strong>in</strong>g(l<strong>in</strong>ear)<br />
-2000<br />
Sagg<strong>in</strong>g<br />
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0<br />
Wave amplitude [m]<br />
<br />
<br />
,?D% @" (" " ! ?DD )C <br />
("&#()))<br />
<br />
, !(! " <br />
/&* (" !! ( <br />
&#()))#)" (" ( (<br />
*!/&* "<br />
""(" ( " <br />
"(+""(" ! <br />
(( " <br />
,?D<br />
<br />
5! 01<br />
<br />
1'"! ! !!! /&*<br />
" ( ! <br />
!( " !/&* $( !!<br />
0 "" " ! ! <br />
!! ! (<br />
# " 23 *! <br />
! ( / " !<br />
!! " ! !! <br />
<br />
4
1! ( !<br />
# 1/&*"( <br />
" ( (!<br />
! !#" 2* <br />
1! (<br />
" <br />
<br />
,# ( " <br />
!!' ("!!( "'# <br />
<br />
6! $%$<br />
<br />
/&*"! "3 "<br />
$ "! <br />
# ( "(<br />
<br />
<br />
!!/&* 2/ 5(E 3 25(E <br />
)#3!( $ !!<br />
!! ! <br />
!! <br />
! ! (#<br />
" ! 1" C1 <br />
<br />
<br />
( : /&*<br />
" " ! "<br />
(/&*( !! <br />
<br />
* /&*( ! <br />
" ! <br />
1 (! ( !!<br />
<br />
0 ((! ( "<br />
! " <br />
<br />
<br />
N ! "$ !!!! <br />
( (<br />
! !!" !. " # <br />
! # ( ! <br />
<br />
<br />
/&*" ! 6 8/("<br />
7 ( !"!#!<br />
!"! , !/&*"<br />
" ! 3" <br />
1( !! <br />
<br />
7! <br />
<br />
( ( (<br />
! !" ( ( <br />
" "" ! <br />
<br />
4
(" (!!!("( <br />
1 "( " (/& *<br />
/&*! !! <br />
# <br />
<br />
0 !("( = <br />
!! ! " /&*(" ( <br />
8 #("88 # (<br />
" '( 3 <br />
<br />
"! !" ( ! ! !<br />
"!#", <br />
("(" " <br />
!( ! !/&*<br />
<br />
! 3! ( <br />
!(!2/ 5(E 3 25(E )#3 !,<br />
!(" ( !! <br />
! /&*<br />
( ! (" "<br />
<br />
. "#(" '= <br />
"(" ! ( <br />
(* (" = <br />
!( ("! ! <br />
(" / " <br />
" <br />
<br />
) ! (!(( ( <br />
!/&*" <br />
<br />
<br />
8(%<br />
<br />
" !#/&*( !!"! <br />
" ! * 0- + */ @ / ! <br />
(!2/ 5(E 3 25(E )#<br />
3!" " <br />
!/&*<br />
<br />
<br />
<br />
<br />
01&*5*6?DDA7 0(O <br />
<br />
0C+) 8 69FFF7 , <br />
&/ABFF/-*<br />
<br />
*5)/P*& &< &Q51) Q 6?DDD7 <br />
@ <br />
<br />
*@+*55 ; 69FF97 < " / & AB<br />
9999A?<br />
<br />
44
15. +
! " ! "#$%&'("# <br />
#$% ! "#$%&')#*!+ <br />
& ! "#$%&', <br />
<br />
'<br />
<br />
<br />
<br />
<br />
<br />
<br />
!"!<br />
# <br />
<br />
$# <br />
<br />
%<br />
$!&%$!'<br />
( # <br />
# <br />
### <br />
# <br />
## # <br />
##<br />
<br />
)"<br />
<br />
()* <br />
<br />
#'*-**-.###-#''-#<br />
&#'*''# #%*'# #--#'**'#<br />
/ * '-# '& - & - #'<br />
' +& * '' &#' ' - &+ <br />
#-#&<br />
<br />
0--'--*''#'-#'# <br />
** ' # &+ -' * ' ' - --# <br />
--#1'2*'#3124#'-#'* '<br />
*12'#-#'***&-'-#'<br />
''* - &+- *<br />
12**&#'-#'*#--*<br />
-.<br />
<br />
0*-#'-# #+-'-.<br />
#--#*#'##-#-<br />
# - ' - -- * & * -' /#<br />
'*#''#'
+) ,- & % <br />
+)( <br />
<br />
&+*#'##*-*-'<br />
-*&#'#'&#*#<br />
-#'#+'+#'#'# #+'#'-'<br />
# ' &+ # --# **' -' #' #+ <br />
#+ +'&-##''#**'<br />
<br />
#*#--#*#' -&'#<br />
-.#*'#'2/*-#*<br />
%# '# -' #+ - '# # * '#<br />
''-'--%*''*%#'#-<br />
6#''#.-7&'-# '''#<br />
''-' #+<br />
89:2*'*!#66#6<br />
+)+ , - .<br />
<br />
#-'-#- '-#-<br />
-'-.#&''& /#'-' ''<br />
<br />
% #' ;# - ' # # *#*#<br />
#'
8?:6#6' **#--#*%#6#'#-'<br />
*&'-'#<br />
<br />
/) ,0<br />
<br />
/)( , - .<br />
<br />
'#'-&# *#--# +**<br />
'# ' - --# ;'' '<br />
%#6#'#'--#--*#'127<br />
# + #' -*' '# ' &+ ' & -- ''<br />
'''--' --*'**'-#-*&*<br />
'# @' -' & * &+ & '' ' &<br />
*'-<br />
<br />
8A:6#6' **#--#*%#6&'#'<br />
'-''&1'2*'#<br />
<br />
=>
)+0<br />
<br />
%# &*12'*&+*#'2<br />
&'-#&+&*## . '#'##<br />
.&'*''#-'#'*<br />
'##*&-' #+#&# <br />
--#''#<br />
<br />
)2- <br />
<br />
%* &;-*"#+%'-*%# &'<br />
1' & #' '*#*'&&''#<br />
& '** *'&--**'*<br />
1' &<br />
<br />
, # -# &+ & ; '# '# <br />
%# ' 1' & - '# '# <br />
%# '# - '# '7 & 6# '# #<br />
'#'**-'-#''*"#+%<br />
12 ' - --# -# * '# <br />
&&-* '#<br />
<br />
*# -# 3 '+4 %# '# & "#+ %<br />
# '* -' -' -# ' # & #<br />
* #+ -# + ' ** # ' &# <br />
'#"#+% &&# '*''#-'<br />
-#&-'-#21' &-'-#<br />
# #*-'#'-'#''7#+#**<br />
'''#-#+ -#'#<br />
'*# '<br />
<br />
# ' ** * *# '+ -# &#<br />
# -''#'&*-'-*'+2*<br />
'#''' * ##-'-#&#&-<br />
#-'*#'+ #'<br />
<br />
)-'&*&+&##&*:<br />
<br />
• *'#'-*#--#<br />
•
- # 3 C%04 &# --' *#<br />
<br />
<br />
<br />
<br />
8D:-'#'C*%*'*-#<br />
<br />
#'&'*#*##+'+'<br />
. #+'<br />
<br />
<br />
8:8#*#'#<br />
<br />
%**12'*#''#<br />
**#''''-&&+1' &<br />
;---*'#1' & --#*<br />
*"#+6 ' ''**"#+%<br />
''# *# -# -* # * ''# C%0 #+ # 2<br />
#' '"#+%##' '''''#'#<br />
'-*--*1' & '#<br />
*#-' #+<br />
<br />
<br />
=A
0<br />
1 0<br />
<br />
<br />
8=:1-*1' & '*"#+%2'<br />
C*%*'*"#+%<br />
<br />
*--*12*1' & &%# &<br />
'1' & #'''#&# # <br />
<br />
B%# &#+''#'+&# --#+<br />
1' &*#"#+#<br />
<br />
0<br />
1 0<br />
<br />
<br />
8:2-*"#+%%#'1' &*'#<br />
<br />
B#*"#+%*##*-'<br />
-'-#1' &<br />
<br />
0<br />
1 0<br />
<br />
<br />
85:1' &*"#+%'#<br />
1'1#<br />
<br />
=
)-''#-*'1' &-#''*''#'#<br />
''**-'#'*%1' &#<br />
'#' * # ' * %# & # -<br />
*##<br />
<br />
1 0<br />
0<br />
<br />
<br />
89>:%##'''1' &<br />
<br />
-- * &+ -' '' -- # *<br />
'-' '-# *# -# '
899:*-'-#<br />
<br />
0 & 12 --# ' ' --' * ' %# ' 1'<br />
&'&+-'&# '' <br />
<br />
- <br />
<br />
'' -# * 1' 2* '# ' - ' --# <br />
- #*&'-#-*&*'#-''#<br />
'#''#<br />
<br />
-##&*&+*#'''#<br />
' - * /# '-' & - *<br />
-#<br />
<br />
--' &+ ' # ' '' * # '# * ' ' &#<br />
* # - &# -# * ' #' -'<br />
#/#*1'2*'#<br />
<br />
=
Global Optimization for Safety and Comfort<br />
Daniele Peri, Emilio Campana, INSEAN, Rome/Italy, {e.campana,d.peri}@<strong>in</strong>sean.it<br />
Abstract<br />
Design eng<strong>in</strong>eers community seems to be still Lipschitzian on the usefulness of numerical<br />
optimization. Criticisms are often focused on the simplicity of the optimization problem solved<br />
(e.g. resistance m<strong>in</strong>imization) when compared to the complexity of a real-life design problem,<br />
<strong>in</strong>volv<strong>in</strong>g tens of different criteria. Indeed, much more challeng<strong>in</strong>g requirements are on the<br />
desk of the design team and solutions tend to become more complex by the hour, particularly<br />
when new features, missions, or capabilities are to be considered for the f<strong>in</strong>al design. In order<br />
to address complex requirements, the design team typically explores different solutions by means<br />
of systematic variations of the hull form, while <strong>in</strong> a second phase the results are analyzed <strong>in</strong><br />
order to reach the objectives. A much more efficient way is to adopt numerical optimization<br />
techniques. The ma<strong>in</strong> features requested to the new design (objectives and constra<strong>in</strong>ts) are<br />
translated <strong>in</strong> a mathematical form and then passed to a numerical optimization framework, able<br />
to manage all the tools necessary for the solution of the problem. Hence, new problems can<br />
be easily faced, discover<strong>in</strong>g the right direction for the design development. In this paper some<br />
orig<strong>in</strong>al Global Optimization (GO) algorithms are illustrated: a Multistart Gradient Method<br />
(MsGM), an evolutionary algorithm (Particle Swarm Optimization, PSO) and a Lipschitzian<br />
method (Diagonal Rectangular Algorithm for Global Optimization, DRAGO) are <strong>in</strong>troduced and<br />
applied to the shape optimization of a cruise ship for improv<strong>in</strong>g safety and comfort, compar<strong>in</strong>g<br />
the numerical results.<br />
1. Introduction<br />
Efficiency, economy and great environmental compatibility of shipp<strong>in</strong>g with respect to land and<br />
air transport seems to support the growth of the fleets, as witnessed by the fact that about 90%<br />
of the world trade is be<strong>in</strong>g transported by ship, Payer (2004). In this scenario, the need for<br />
new designs with <strong>in</strong>creased safety, efficiency, flexibility of use, speed and comfort, is becom<strong>in</strong>g<br />
evident.<br />
Far from be<strong>in</strong>g the only factor of <strong>in</strong>novation, safety is however a major element of public concern,<br />
especially when pollution of the environment is <strong>in</strong>volved, and the quest for designs capable of<br />
reduc<strong>in</strong>g the risks of shipp<strong>in</strong>g is challeng<strong>in</strong>g design eng<strong>in</strong>eers. Progress <strong>in</strong> ship design is also<br />
fundamental to improve the comfort of both crew and passengers.<br />
The quest for new design might give impulse for more numerical optimization <strong>in</strong> ship design.<br />
The <strong>in</strong>creas<strong>in</strong>g attention on this topic is partially due to the appeal<strong>in</strong>g opportunities given by<br />
these techniques. In fact, the (naive) dream of each shipowner would be a computer program<br />
able to design the perfect ship for a given task. On the other hand, the design community is<br />
suspicious about the real ability of an optimization framework to manage all the aspects of the<br />
design activity with a sufficient degree of reliability.<br />
For this reason, the researchers and code developers have to move ma<strong>in</strong>ly along three directions:<br />
(1) Increase problem complexity: <strong>in</strong>clude as much as possible different discipl<strong>in</strong>es <strong>in</strong> the<br />
optimization problem, enlarg<strong>in</strong>g the number of different issues the optimizer is able to treat simultaneously;<br />
(2) Ga<strong>in</strong> reliability by apply<strong>in</strong>g high-fidelity solvers, with a solid ground of physics<br />
<strong>in</strong>side, and <strong>in</strong>creas<strong>in</strong>g the grid density of the adopted meshes. Examples are Duvigneau et al.<br />
(2003), Jacqu<strong>in</strong> et al. (2004), Tahara et al. (<strong>2005</strong>); (3) Explore the algorithmic side, look<strong>in</strong>g<br />
for techniques able to <strong>in</strong>crease the chance to f<strong>in</strong>d the maximum improvement with<strong>in</strong> affordable<br />
overall time.<br />
This last po<strong>in</strong>t is the focus of this paper. Algorithms underly<strong>in</strong>g different optimization philos-<br />
477
ophy are described and applied to solve an optimization problem of the redesign of a cruis<strong>in</strong>g<br />
ship, eventually compar<strong>in</strong>g the numerical results.<br />
2. The quest for a global m<strong>in</strong>imum<br />
The easiest (and maybe naive) way to perform optimization is to start from an assigned design,<br />
evaluate its performances, explore the performances <strong>in</strong> a neighborhood of the <strong>in</strong>itial design: once<br />
a direction of improvement is identified <strong>in</strong> the design space, move toward this direction until<br />
an improvement of the design is still possible and keep iterat<strong>in</strong>g for a few steps. The identified<br />
m<strong>in</strong>imum is clearly a local one, which may be enough for our purposes.<br />
If more improvements are necessary or appeal<strong>in</strong>g, global optimization might be the proper<br />
choice. Once a local m<strong>in</strong>imum is found <strong>in</strong>deed, we are not sure about the existence of better<br />
local m<strong>in</strong>ima, located <strong>in</strong> different attraction bas<strong>in</strong>s, with better performances than the present<br />
one.<br />
The m<strong>in</strong>imum value among all the local m<strong>in</strong>ima is the global m<strong>in</strong>imum, def<strong>in</strong>ed as the po<strong>in</strong>t<br />
such that:<br />
F : X → R , X ⊆ R N<br />
x, x ′ ∈ X<br />
x = argm<strong>in</strong>{F(x ′ )} ⇐⇒ F(x) ≤ F(x ′ ) ∀ x ′ ∈ X<br />
where X is the subset of R N where the objective function is def<strong>in</strong>ed and x, x’ are design vectors.<br />
A few comments on certify<strong>in</strong>g the global m<strong>in</strong>imizer. Follow<strong>in</strong>g the def<strong>in</strong>ition of global m<strong>in</strong>imum,<br />
we can state that the global m<strong>in</strong>imum has been found only if sufficient <strong>in</strong>formation about all the<br />
existent local m<strong>in</strong>ima are available. This represents an h<strong>in</strong>t about the complexity of the problem<br />
we want to solve, s<strong>in</strong>ce a detailed search on the Feasible Solution Set (FSS) becomes necessary,<br />
enhanc<strong>in</strong>g the price of the solution due to the larger number of objective function evaluations<br />
requested. Furthermore, def<strong>in</strong><strong>in</strong>g stopp<strong>in</strong>g criteria for this algorithm is also more complex, s<strong>in</strong>ce<br />
the usual condition of null gradient for the objective function is poor. As a consequence, the<br />
extension of the search <strong>in</strong> the design space is sometime adopted as an <strong>in</strong>direct <strong>in</strong>dication of the<br />
detection of a global m<strong>in</strong>imizer.<br />
Beside <strong>in</strong>formation about the objective function and on the FSS, to guarantee the completeness of<br />
the FSS <strong>in</strong>vestigation the algorithms too have to possess some special features. Non-determ<strong>in</strong>istic<br />
algorithms, like Genetic Algorithms, are not suited to this particular aim. They are able to<br />
improve substantially the objective function, but the randomness of the search guarantees their<br />
completeness only if the number of attempts is huge and deal<strong>in</strong>g with real-life problems (those<br />
for which the price of the computation of the objective function is not negligible), this feature<br />
makes the GA less attractive than determ<strong>in</strong>istic algorithms.<br />
3. GO Algorithms<br />
In this section, some GO algorithms are def<strong>in</strong>ed. All of them are substantially determ<strong>in</strong>istic algorithms.<br />
S<strong>in</strong>ce no assumption is possible about the objective function (cont<strong>in</strong>uity, boundedness<br />
etc.), no mention is given about the theoretical properties of the here described algorithms.<br />
3.1. Multistart Gradient Method (MsGM)<br />
The simplest way to <strong>in</strong>vestigate more than a s<strong>in</strong>gle bas<strong>in</strong> of attraction is to perform different<br />
local optimizations start<strong>in</strong>g from different po<strong>in</strong>ts <strong>in</strong> the FSS. The here adopted MsGM is based<br />
on the Conjugate-Gradient algorithm described <strong>in</strong> Peri et al. (2004). Two different choices<br />
are possible: local searches can be performed sequentially or <strong>in</strong> parallel. Here we adopt the<br />
second choice, and hence, at each step, M gradients of the objective function must be computed,<br />
be<strong>in</strong>g M the number of different start<strong>in</strong>g po<strong>in</strong>ts. If NDV is the number of design variables,<br />
each gradient evaluation requires 2*NDV evaluations of the objective function when a centered<br />
difference scheme is adopted. Moreover, at least one more evaluation for each mov<strong>in</strong>g po<strong>in</strong>t<br />
478
is requested after the descent direction for that po<strong>in</strong>t has been def<strong>in</strong>ed. Summariz<strong>in</strong>g, the<br />
price of the algorithm is M*(2*NDV+1) solutions for each iteration. In order to reduce the<br />
computational costs, the algorithm has been modified as follows:<br />
1. Distribute M po<strong>in</strong>ts over the FSS.<br />
2. Evaluate the M po<strong>in</strong>ts <strong>in</strong> their actual position.<br />
3. Build up a meta-model of the objective function adopt<strong>in</strong>g all the available known po<strong>in</strong>ts.<br />
4. Evaluate the objective function gradient <strong>in</strong> all the M po<strong>in</strong>ts by us<strong>in</strong>g the meta-model (no<br />
further evaluation of the objective function).<br />
5. Compute the M descent directions.<br />
6. Move the M po<strong>in</strong>ts along the descent directions.<br />
7. go to 2<br />
In this way, the price of each iteration becomes M <strong>in</strong>stead of M*(2*NDV+1). More explicitly,<br />
if we have 6 design variables and 6 different start<strong>in</strong>g po<strong>in</strong>t, the price of each iteration passes<br />
from 78 to 6. In the algorithm description, the concept of meta-model has been <strong>in</strong>troduced.<br />
A meta-model is an approximation of the objective function obta<strong>in</strong>ed from sampled values.<br />
Details are reported <strong>in</strong> Peri and Campana (<strong>2005</strong>). A last comment related to the strategy for<br />
the start<strong>in</strong>g po<strong>in</strong>t selection is necessary. In fact, the uniformity of the distribution of these<br />
po<strong>in</strong>ts is of paramount importance, s<strong>in</strong>ce the probability to fall <strong>in</strong>side as many attraction bas<strong>in</strong>s<br />
as possible is connected with the sparsity of the <strong>in</strong>itial distribution. For this reason, an LP τ<br />
sequence, Statnikov and Matusov (1995) is adopted <strong>in</strong> this phase: its eas<strong>in</strong>ess <strong>in</strong> be<strong>in</strong>g generated<br />
and uniformity of the distribution are two key features of this k<strong>in</strong>d of sequence.<br />
3.2. Diagonal Rectangular Algorithm for Global Optimization (DRAGO)<br />
Another algorithm is <strong>in</strong>spired by the family of the <strong>in</strong>terval search algorithms Schubert (1972),<br />
Jones et al. (1993). Here some bound values for the design variables are assumed, and the so<br />
def<strong>in</strong>ed hyper-rectangle is considered as a s<strong>in</strong>gle <strong>in</strong>terval. This orig<strong>in</strong>al <strong>in</strong>terval will be subdivided<br />
<strong>in</strong>to smaller <strong>in</strong>tervals across the promis<strong>in</strong>g value of the objective function. The estimate of the<br />
value of the objective function y i <strong>in</strong>side a potential new <strong>in</strong>terval i is made by assum<strong>in</strong>g the<br />
Lipschitz constant value, and then comput<strong>in</strong>g the expected objective function value y i as<br />
y i = (f u i + f l i )/2. − K ∗ (u ij − l ij ) (1)<br />
where i denotes the <strong>in</strong>terval <strong>in</strong>dex, j the design variable, u i and l i the upper corner of the i th<br />
hyper-rectangle, f l and f u the objective function value <strong>in</strong> l i and u i , K the Lipschitz constant and<br />
y i the estimated <strong>in</strong>fimum of the objective function <strong>in</strong>side the i th hyper-rectangle. This estimate<br />
is based on the assunption that if we fix the rate of improvement as the Lipschitz constant, the<br />
objective function <strong>in</strong>side the i th <strong>in</strong>terval cannot be improved more than y i . It is clear that the<br />
orig<strong>in</strong>al Schubert (1972) algorithm has been modified, and each hyper-rectangle is characterized<br />
by two po<strong>in</strong>ts only, that is, the extreme corners.<br />
The <strong>in</strong>itialization is made by comput<strong>in</strong>g the objective function <strong>in</strong> these two vertex of the orig<strong>in</strong>al<br />
hyper-rectangle only, regardless the dimension of the problem. At each iteration, the hyperrectangle<br />
to be divided is detected, and it is partitioned sequentially along all the directions.<br />
For each subdivision, two new hyper-rectangles are obta<strong>in</strong>ed, and only two new values of the<br />
objective function are needed, Fig.1. The Lipschitz constant is estimated as the maximum value<br />
of the ratio of change of the objective function along each <strong>in</strong>terval augmented by a certa<strong>in</strong> factor.<br />
479
This algorithm is characterized by the coexistence of two different souls, global and local. In<br />
fact, large unexplored <strong>in</strong>tervals become promis<strong>in</strong>g due to the term −K ∗ (u ij − l ij ), <strong>in</strong> which the<br />
dimensions of the <strong>in</strong>terval dom<strong>in</strong>ates, while the absolute value of the objective function at the<br />
extremes plays <strong>in</strong> the term (fi<br />
u + fi l )/2. A global phase is active when some huge <strong>in</strong>tervals are<br />
still unexplored, and a local phase occurs when the objective function values at the boundaries<br />
of an <strong>in</strong>terval are extremely low.<br />
3<br />
ξ2<br />
1<br />
f<br />
1<br />
ξ2 f<br />
1 2<br />
f<br />
2<br />
1<br />
3<br />
ξ1<br />
ξ1<br />
ξ1<br />
ξ1<br />
ξ2<br />
1 2<br />
f<br />
1<br />
2<br />
ξ2 f<br />
1 2<br />
3 4<br />
ξ1<br />
ξ2<br />
ξ1<br />
Fig.1: Course of the DRAGO algorithm. Sequence of partitions of the design variable space<br />
(ξ 1 , ξ 2 ), obta<strong>in</strong>ed us<strong>in</strong>g <strong>in</strong>formations com<strong>in</strong>g from the objective function space (ξ 1 , f and ξ 2 .f)<br />
Asymptotically, this algorithms guarantees the exploration of the whole FSS. In order to avoid<br />
an extensive but unessential exploration of the FSS, some weights on the local and global phase<br />
are necessary.<br />
3.3. Particle Swarm Optimization (PSO)<br />
In its first implementation, Kennedy and Eberhart (1995), Particle Swarm Optimization (PSO)<br />
was a pure stochastic method. The swarm simulates the social behavior of some <strong>in</strong>dividuals,<br />
shar<strong>in</strong>g <strong>in</strong>formation among them while explor<strong>in</strong>g the design variables space. Each particle has<br />
its own (<strong>in</strong>dividual) memory to remember the places visited dur<strong>in</strong>g the exploration, while the<br />
swarm has its own (collective) memory, to memorize the best locations visited by the particles.<br />
The particles have an adaptable velocity and <strong>in</strong>vestigate the design space analyz<strong>in</strong>g their own<br />
fly<strong>in</strong>g experience, and the one of all the particles of the swarm. Each particle is a potential<br />
solution of the optimization problem under consideration, as happen <strong>in</strong> the classical Evolutionary<br />
Computation (EC) techniques. However, PSO method has not a direct “recomb<strong>in</strong>ation operator”<br />
and it does not use the “survival of the fittest” concept.<br />
The basic PSO formulation def<strong>in</strong>es each particle as a potential solution of a problem <strong>in</strong><br />
a N-dimensional space, where the i th particle of the swarm can be represented by a N-<br />
dimensional vector, X i = (x i1 , x i2 , ..., x iN ) T . The velocity of this particle is given by the vector<br />
V i = (v i1 , v i2 , ..., v iN ) T . The best previously visited position of the i th particle is denoted as<br />
P i = (p i1 , p i2 , ..., p iN ) T . Def<strong>in</strong><strong>in</strong>g b as the <strong>in</strong>dex of the best visited location by the whole swarm,<br />
and lett<strong>in</strong>g the superscript denote the iteration number, the swarm is mov<strong>in</strong>g accord<strong>in</strong>g to the<br />
follow<strong>in</strong>g two equations given <strong>in</strong> Shi and Eberhart (1998,2001).<br />
{ v<br />
n+1<br />
id<br />
= c 1 r1 n (pn id(<br />
− xn id ) + c 2r<br />
) 2 n (pn bd − xn id )<br />
x n+1<br />
id<br />
= x n id + χ vid<br />
n+1 + ωvid<br />
n (2)<br />
where d = 1, 2, ..., N; i = 1, 2, ..., n sw , and n sw is the size of the swarm; ω is called <strong>in</strong>ertia<br />
weight; c 1 , c 2 are two positive constant, called cognitive and social parameter respectively; χ is a<br />
constriction function, which is used to limit the velocity; r 1 , r 2 are random numbers, uniformly<br />
distributed <strong>in</strong> [0,1]; and n=1,2.., determ<strong>in</strong>es the iteration number.<br />
The role of the <strong>in</strong>ertia weight ω, <strong>in</strong> equation (2), is considered critical for the PSO’s convergence<br />
behavior. The <strong>in</strong>ertia weight is employed to control the impact of the previous history of<br />
480
velocities on the current one. Accord<strong>in</strong>gly, the parameter ω regulates the trade-off between the<br />
global (wide-rang<strong>in</strong>g) and local (nearby) exploration abilities of the swarm. A large <strong>in</strong>ertia weight<br />
facilitates global exploration (search<strong>in</strong>g for new areas), while a small one tends to facilitate local<br />
exploration, i.e. f<strong>in</strong>e-tun<strong>in</strong>g of the current search area. A suitable value for the <strong>in</strong>ertia weight ω<br />
usually provides balance between global and local exploration abilities and consequently results<br />
<strong>in</strong> a reduction of the number of iterations required to locate the optimum solution. Experimental<br />
results <strong>in</strong>dicated that it is better to <strong>in</strong>itially set the <strong>in</strong>ertia to a large value, <strong>in</strong> order to promote<br />
global exploration of the search space, and gradually decrease it to get more ref<strong>in</strong>ed solutions,<br />
Shi and Eberhart (1998,2001). An <strong>in</strong>itial value for ω was set and the gradual decrease was<br />
calculated as follows:<br />
(<br />
ω = ω 0 1 − n − 1 )<br />
n max<br />
The parameters c 1 and c 2 , are not critical for the PSO’s convergence. However, proper f<strong>in</strong>etun<strong>in</strong>g<br />
may result <strong>in</strong> faster convergence and alleviation of local m<strong>in</strong>ima. Random parameters<br />
have been excluded from this implementation, elim<strong>in</strong>at<strong>in</strong>g every non-determ<strong>in</strong>istic behavior of<br />
the algorithm, Campana et al. (<strong>2005</strong>).<br />
3.4. Genetic Algorithm<br />
As a further reference po<strong>in</strong>t, a Genetic Algorithm (GA) code provided by the author and described<br />
<strong>in</strong> Yang et al. (1998) has been applied. The algorithm has been simply applied with the<br />
default sett<strong>in</strong>gs.<br />
4. Hull form parameterizations<br />
In the past years, a number of different parameterizations styles have been proposed <strong>in</strong> literature,<br />
follow<strong>in</strong>g three ma<strong>in</strong> guidel<strong>in</strong>es:<br />
• CAD-based: a commercial CAD system is <strong>in</strong>cluded <strong>in</strong>to the optimization cycle. The ship<br />
hull is modified us<strong>in</strong>g the features of the selected CAD. This option requires the license for a<br />
specific CAD system, that could be expensive, and new releases of the product may require<br />
modifications <strong>in</strong> the <strong>in</strong>terface with the optimizer. Moreover, the mesh<strong>in</strong>g operation, that<br />
is, the production of a computational mesh on the modified design must to be performed<br />
each time the hull is modified, that is time-consum<strong>in</strong>g and also <strong>in</strong>secure, because the<br />
mesh quality could be not ensured if strong modifications are applied. Moreover, <strong>in</strong> some<br />
parametric CAD system is uneasy to obta<strong>in</strong> an exist<strong>in</strong>g design produced with a different<br />
product, and the modification of some details of an exist<strong>in</strong>g hull cold require a strong effort<br />
<strong>in</strong> produc<strong>in</strong>g the orig<strong>in</strong>al design, without ensur<strong>in</strong>g the precision of the orig<strong>in</strong>al geometry.<br />
• CAD-emulation mode: some features of a CAD system are emulated and the new hull<br />
is obta<strong>in</strong>ed <strong>in</strong> a similar way, but without the use of a specific product. An example is<br />
reported <strong>in</strong> Tahara et al. (<strong>2005</strong>): the IGES file of the modify<strong>in</strong>g geometry conta<strong>in</strong>s all the<br />
<strong>in</strong>formations about the control net of the hull surfaces, and the control po<strong>in</strong>ts of the net are<br />
moved as <strong>in</strong>to the CAD system. An alternative approach is to produce a net around the<br />
hull surface, than controll<strong>in</strong>g the net po<strong>in</strong>ts. An approach of this family, called Free-Form<br />
Deformation (FFD) is described and applied <strong>in</strong> this paper. The computational grid is also<br />
deformed, and it is not required to be generated at each step, ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g the quality of<br />
the orig<strong>in</strong>al one.<br />
• Perturbation approach: an analytical patch is applied to the computational mesh, <strong>in</strong> an<br />
additive or multiplicative manner. These approach have been explored <strong>in</strong> Peri et al.<br />
(2001), Peri and Mandolesi (<strong>2005</strong>). The easy way <strong>in</strong> which the new geometry is produced<br />
has, as a s<strong>in</strong>gle drawback, the difficulty to ma<strong>in</strong>ta<strong>in</strong> everytime the fairness of the orig<strong>in</strong>al<br />
surface, while flexibility is one of the major features of this approach.<br />
(3)<br />
481
In this paper, we are adopt<strong>in</strong>g the FFD approach Sederberg and Parry (1986). The portion of<br />
the hull to be modified is embedded <strong>in</strong>to a rectangular control region. The control region is<br />
partitioned with regular subdivisions along all the three coord<strong>in</strong>ate directions, and the <strong>in</strong>tersections<br />
between the subdivisions represent the control po<strong>in</strong>ts of the doma<strong>in</strong>. Once the control<br />
po<strong>in</strong>ts are shifted, an set of hexaedrical shapes is obta<strong>in</strong>ed, and the <strong>in</strong>cluded volume is modified<br />
via Bernste<strong>in</strong> polynomials def<strong>in</strong>ed by the position of the control po<strong>in</strong>ts. Different control region<br />
could be sequentially applied to the same hull portion. In order to reduce the number of design<br />
variable at the price of limit<strong>in</strong>g the variety of obta<strong>in</strong>able shapes, a subset of the control po<strong>in</strong>ts<br />
can be moved together. If necessary, the displacement of one control po<strong>in</strong>t can be connected<br />
with the displacement of a different po<strong>in</strong>t, <strong>in</strong> the same direction or not, with the same or a<br />
different (prescribed) law. This few examples may give an idea about the flexibility of the tool.<br />
The global movement of the embedded volume makes the modification of the hull shape very<br />
smooth, prevent<strong>in</strong>g unfaired regions.<br />
5. Problem description: m<strong>in</strong>imiz<strong>in</strong>g the whipp<strong>in</strong>g of a cruis<strong>in</strong>g ship<br />
Comfort and safety of a cruise ship are of paramount importance, s<strong>in</strong>ce the achievement of these<br />
tasks is the core of the success of a ship of this class. In fact, customer satisfaction is the primary<br />
goal of a cruise company. Dur<strong>in</strong>g the years, more and more comfortable ships, with improved<br />
seago<strong>in</strong>g features have been produced, <strong>in</strong>creas<strong>in</strong>g the active control features as well as the passive<br />
security systems. In this framework, one of the most <strong>in</strong>terest<strong>in</strong>g problems to be solved for a<br />
cruise ship is the whipp<strong>in</strong>g problem. With this word we are <strong>in</strong>dicat<strong>in</strong>g the resonance effect on<br />
the structures <strong>in</strong>duced by small but cont<strong>in</strong>uous impacts of the stern of the ship on the sea water,<br />
particularly when the ship is at rest, just outside the harbor, for example dur<strong>in</strong>g the wait<strong>in</strong>g<br />
time for the free dock availability. In this situation, the ship is nearly <strong>in</strong> open seas at rest and<br />
the active control systems are <strong>in</strong>efficient and only the hull shape might help. By mov<strong>in</strong>g up the<br />
stern of the ship the waterl<strong>in</strong>e is reduced and this might reduce the performances of the ship,<br />
while by lower<strong>in</strong>g the stern location, the transom is also moved toward the propellers, probably<br />
<strong>in</strong>creas<strong>in</strong>g the <strong>in</strong>duced vibrations. F<strong>in</strong>ally, the occurrence of whipp<strong>in</strong>g is also connected with<br />
the slope of the stern and of the characteristics of the <strong>in</strong>com<strong>in</strong>g wave. Clearly, there is not a<br />
simple solution to this problem.<br />
Bear<strong>in</strong>g <strong>in</strong> m<strong>in</strong>d the focus of the paper, that is to compare different GO algorithms <strong>in</strong> the redesign<br />
problem of an exist<strong>in</strong>g ship, a case study for m<strong>in</strong>imiz<strong>in</strong>g the number of expected stern<br />
slamm<strong>in</strong>g events and their probability has been set up.<br />
Dur<strong>in</strong>g a prelim<strong>in</strong>ary analysis of the behavior of the orig<strong>in</strong>al hull with <strong>in</strong>com<strong>in</strong>g waves of 3.5<br />
meters of amplitude, it was found that the worst condition occurs for <strong>in</strong>com<strong>in</strong>g waves of a period<br />
of 9 seconds and for a direction of 75 degrees. These conditions have been adopted for the test.<br />
The objective function F is<br />
f = 0.5 S o<br />
S i<br />
+ 0.5 P o<br />
P i<br />
(4)<br />
where S is the number of slamm<strong>in</strong>g events per hour, P is the probability of the slamm<strong>in</strong>g event,<br />
and the underscores are o for optimized and i for <strong>in</strong>itial.<br />
The adopted constra<strong>in</strong>ts <strong>in</strong>volve the displacement of the ship, that cannot be varied more than<br />
1% of the orig<strong>in</strong>al value, and the beam, that is fixed. As to the geometry manipulation, three<br />
different boxes for the FFD approach have been used <strong>in</strong> order to modify bulb, stern and bow<br />
regions, Fig.2. Ten design variables have been used to move the hull shape and these are the<br />
variables of the optimization problem too.<br />
482
X<br />
Z<br />
Y<br />
Z<br />
X<br />
Y<br />
Z<br />
Z<br />
Y<br />
X<br />
X<br />
Y<br />
Fig.2: Details on the parameterization scheme for the cruise ship: three FFD blocks. Black<br />
dots <strong>in</strong>dicate the mov<strong>in</strong>g vertexes.<br />
6. Numerical solution: Native form algorithms<br />
In this section, all the orig<strong>in</strong>al algorithms have been applied <strong>in</strong> their native formulation, while<br />
<strong>in</strong> the next section improved versions of the algorithms will be presented.<br />
The same <strong>in</strong>itial distribution of po<strong>in</strong>ts has been applied for PSO and MsGM. For GA, a random<br />
population is enforced. The number of <strong>in</strong>itial po<strong>in</strong>ts is equal for all the algorithms that requires<br />
an <strong>in</strong>itial population, and it has been set to 25. DRAGO is the only one that do not require<br />
an <strong>in</strong>itial population. A limit on the number of objective function evaluations has been set<br />
(100*NDV = 1000).<br />
All the previously described algorithms have been applied, and different results have been produced<br />
by each algorithm. Results are reported <strong>in</strong> Table I.<br />
Table I: Optimal values of the objective function (OF) given by the different algorithms. The<br />
iteration number (It.#) at which the m<strong>in</strong>imum value is found are also reported, together with<br />
the absolute value, the percentage ga<strong>in</strong> (Ga<strong>in</strong>%) and the unit cost (Cost) of the percentage<br />
improvement (It.#/Ga<strong>in</strong>%).<br />
Algorithm It.# OF Ga<strong>in</strong>% Cost<br />
GA 306 0.6548 34.52 8.86<br />
PSO 546 0.6905 30.95 17.64<br />
DRAGO 665 0.6905 30.95 21.48<br />
MsGM 24 0.8095 19.05 1.26<br />
GA gives the best performances, obta<strong>in</strong><strong>in</strong>g the lowest value of the objective function <strong>in</strong> the<br />
smaller number of iterations. One the other hand, MsGM algorithm do not improve at all<br />
the objective function, s<strong>in</strong>ce the best value is obta<strong>in</strong>ed dur<strong>in</strong>g the evaluation of the start<strong>in</strong>g<br />
population. PSO seems to perform better than DRAGO, reach<strong>in</strong>g the same level of the objective<br />
function <strong>in</strong> a smaller number of iterations. These results are partly <strong>in</strong> disagreement with previous<br />
483
esults obta<strong>in</strong>ed on a different test case, as reported <strong>in</strong> Campana et al. (<strong>2005</strong>) and <strong>in</strong> Table II for<br />
completeness. Here the objective function was the peak of the heave’s RAO for a fast monohull.<br />
The only strong difference between these two applications is related to some features of the<br />
objective function. Here the objective function is nearly discrete, s<strong>in</strong>ce the number of slamm<strong>in</strong>g<br />
per hour S is an <strong>in</strong>teger number. Consequently, the objective function (4) is a compound of the<br />
step function S and a smooth function (the probability of the slam event) P , and the comb<strong>in</strong>ation<br />
results <strong>in</strong> a function with many plateau. On the other hand, the peak of the RAO’s heave used<br />
<strong>in</strong> the other application is a cont<strong>in</strong>uous function. As a consequence, as well expla<strong>in</strong>ed by the<br />
result of MsGM, the gradient of the objective function is nearly zero on a large portion of the<br />
FSS, and the multistart gradient method meets strong difficulties <strong>in</strong> f<strong>in</strong>d<strong>in</strong>g a descent direction.<br />
The same feature seems to <strong>in</strong>fluence also DRAGO method, s<strong>in</strong>ce efficiency of the method is<br />
related to the estimation of the Lipschitz constant, whose value is connected with the values at<br />
the bounds of each <strong>in</strong>terval. For the PSO algorithm, a reason of the discussed results is still<br />
under analysis.<br />
Table II: Optimal values of the objective function (OF) given by the different algorithms for the<br />
test case depicted <strong>in</strong> Campana et al. (<strong>2005</strong>). The iteration number (It.#) at which the m<strong>in</strong>imum<br />
value is found are also reported, together with the absolute value, the percentage ga<strong>in</strong> (Ga<strong>in</strong>%)<br />
and the unit cost (Cost) of the percentage improvement (It.#/Ga<strong>in</strong>%).<br />
Algorithm It.# OF Ga<strong>in</strong>% Cost<br />
PSO 286 0.9872 1.34 213.43<br />
DRAGO 277 0.9884 1.23 225.20<br />
GA 309 0.9904 1.02 302.94<br />
Simplex 130 0.9999 0.08 1625.00<br />
7. Numerical solution: Improved algorithms<br />
Different variants of PSO exists <strong>in</strong> literature and they have been tested recently, Campana et al.<br />
(<strong>2005</strong>). The ma<strong>in</strong> difference between the native form and the improved one is the total absence<br />
of random selection, <strong>in</strong> the <strong>in</strong>itial swarm distribution and <strong>in</strong> the evolution coefficient. The <strong>in</strong>itial<br />
swarm has been selected by us<strong>in</strong>g an LP τ sequence, Statnikov and Matusov (1995), of feasible<br />
solutions. Moreover, the computation of the velocity of the particle, once a non admissible region<br />
has been reached, follows the <strong>in</strong>dications of Venter and Sobieszcanski-Sobieski (2004), and the<br />
selection of the coefficients has been carried out <strong>in</strong> order to stabilize the search, as performed <strong>in</strong><br />
Clerk and Kennedy (2002).<br />
For the MsGM algorithm, different metamodels have been applied dur<strong>in</strong>g the course of the<br />
optimization. In fact, the use of a s<strong>in</strong>gle metamodel might be not be the best solution, because<br />
at the beg<strong>in</strong>n<strong>in</strong>g of the search very few po<strong>in</strong>ts are known, and a highly precise metamodel could<br />
trap the local search too close to the <strong>in</strong>itial po<strong>in</strong>t. On the contrary, a simple polynomial fit<br />
do not locate precisely the global optimum, but is able to give smooth tendencies about the<br />
region the global optimum is located <strong>in</strong>. As a consequence, simple metamodels (polynomials<br />
and polynomials plus ANOVA, Giunta et al. (1997)) are adopted <strong>in</strong> the early stage of the<br />
search, while much more flexible metamodels, like Krig<strong>in</strong>g or neural networks, are adopted once<br />
a sufficient number of solutions are available. Moreover, a limit on the descent step has been<br />
imposed, <strong>in</strong> order to ma<strong>in</strong>ta<strong>in</strong> diversity among the search<strong>in</strong>g elements.<br />
For the DRAGO algorithm, the local Lipschitz constant has been estimated bas<strong>in</strong>g on the<br />
<strong>in</strong>formations given by a metamodel analysis. Furthermore, at each iteration, two <strong>in</strong>tervals are<br />
partitioned, one selected among the smaller ones and the second one from the bigger ones. In<br />
order to establish the dimension of an <strong>in</strong>terval, the mean value of the size of the actual <strong>in</strong>tervals<br />
and the variance of the dimensions are adopted.<br />
The results from these new implementations of the different algorithms are reported <strong>in</strong> Table I.<br />
484
Here the GA results are the same as <strong>in</strong> Table I. The improvements obta<strong>in</strong>ed for the modified<br />
algorithms are evident. All the orig<strong>in</strong>al algorithms have an higher efficiency than the GA.<br />
DRAGO algorithm still f<strong>in</strong>d a value better than the GA: the improvements are comparable but<br />
the number of iterations needed is strongly reduced.<br />
Table III: Optimal values of the objective function (OF) given by the improved algorithms. The<br />
iteration number (It.#) at which the m<strong>in</strong>imum value is found are also reported, together with<br />
the absolute value, the percentage ga<strong>in</strong> (Ga<strong>in</strong>%) and the unit cost (Cost) of the percentage<br />
improvement (It.#/Ga<strong>in</strong>%).<br />
8. Conclusions<br />
Algorithm It.# OF Ga<strong>in</strong>% Cost<br />
MsGM 182 0.3810 61.90 2.94<br />
PSO 189 0.3810 61.90 3.05<br />
GA 306 0.6548 34.52 8.86<br />
DRAGO 152 0.6905 30.95 4.91<br />
Four different optimization algorithms have been applied to the design of a cruise ship for the<br />
optimization of the comfort qualities. Usefulness of the application of global optimization algorithms<br />
has been clearly shown, <strong>in</strong> terms of efficiency and obta<strong>in</strong>ed improvements. Indications<br />
about the future directions for the development of more efficient global optimization algorithms<br />
can be traced by this paper.<br />
Acknowledgments<br />
This work has been partially supported by M<strong>in</strong>istero delle Infrastrutture e dei Trasporti <strong>in</strong> the<br />
framework of the research plan “Programma di Ricerca sulla Sicurezza”, Decreto 17/04/2003<br />
G.U. n. 123 del 29/05/2003.<br />
References<br />
CAMPANA, E.F.; LIUZZI, G.; LUCIDI, S.; PERI, D.; PICCIALLI, V.; PINTO A. (<strong>2005</strong>), New<br />
global optimization methods for ship design problems, Techn. Rep. INSEAN, <strong>2005</strong> – 1, submitted<br />
to Optimization and Eng<strong>in</strong>eer<strong>in</strong>g.<br />
CLERK, M.; KENNEDY, J. (2002), The Particle Swarm-explosion, stability and convergence<br />
<strong>in</strong> multidimensional complex space, IEEE Trans. Evolutionary Computation 6(1)<br />
DUVIGNEAU, R.; VISONNEAU, M.; DENG, G. (2003). On the role played by turbulence<br />
closures for hull shape optimization at model and full scale, J. Mar<strong>in</strong>e Science and Technology<br />
153/8<br />
GIUNTA, A.A.; BALABANOV, V.; KAUFMAN, M.; BURGEE, S.; GROSSMAN, B.;<br />
HAFTKA, R.T.; MASON, W.H.; WATSON, L.T. (1997), Variable-complexity response surface<br />
design of an HSCT configuration, Multidiscipl<strong>in</strong>ary Design Optimization, SIAM, Philadelphia<br />
JACQUIN, E.; DERBANNE Q.; BELLEVRE, D.; CORDIER, S.; ALESSANDRINI, B.; ROUX,<br />
Y. (2004), Hull form optimization us<strong>in</strong>g a free-surface RANSE solver, Symp. Naval Hydrodynamics,<br />
St. John’s<br />
JONES, D.R.; PERTTUNEN, C.D.; STUCKMAN, B.E. (1993). Lipschitzian Optimization<br />
Without the Lipschitz Constant, J. Optimization Theory and Application, Vol. 79, No. 1,<br />
October 1993, pp.157-181.<br />
KENNEDY, J.; EBERHART, R. (1995), Particle swarm optimization, IEEE Int. Conf. Neural<br />
Networks, Perth, IV:pp.1942-1948.<br />
485
MEYERS W.G.; APPLEBEE T.R.; BAITIS, A.E. (1981), User’s manual for the standard ship<br />
motion program SMP, David Taylor <strong>in</strong>ternal report DTNSRDC/SPD-0936-01<br />
MEYERS W.G.; BAITIS A.E. (1985), SMP84: improvements to capability and prediction<br />
accuracy of the standard ship motion program SMP84, David Taylor <strong>in</strong>ternal report<br />
DTNSRDC/SPD-0936-04<br />
NELDER J.A.; MEAD R. (1965), A simplex method for function m<strong>in</strong>imization, Computer J. 7,<br />
pp.308-313<br />
PAYER H.G. (2004), Challenges <strong>in</strong> ship design, fabrication and <strong>in</strong>spection - A class view,<br />
PRADS Symp., Lubeck-Travemunde<br />
PERI, D.; CAMPANA, E.F. (2003), Multidiscipl<strong>in</strong>ary design optimization of a naval surface<br />
combatant, J. Ship Research 47/1, pp.1-12<br />
PERI, D.; CAMPANA, E.F. (<strong>2005</strong>), High fidelity models and multiobjective global optimization<br />
algorithms <strong>in</strong> simulation based design, J. Ship Research 49<br />
PERI, D.; CAMPANA, E.F.; STERN, F.; TAHARA, Y. (<strong>2005</strong>), A comparison of global optimization<br />
methods with application to ship design, 5th Osaka Colloquium, Osaka<br />
PERI, D.; MANDOLESI, F. (<strong>2005</strong>), Multiobjective optimization of an IACC sail<strong>in</strong>g yacht by<br />
means of CFD high-fidelity solvers, 17 th Chesapeake Sail<strong>in</strong>g Yacht Symp., Annapolis<br />
PERI, D.; ROSSETTI, M.; CAMPANA, E.F. (2001), Design optimization of ship hulls via CFD<br />
techniques, J. Ship Research 45/2, pp.140-149<br />
PINTO, A.; PERI, D.; CAMPANA, E.F. (2004), Global optimization algorithms <strong>in</strong> naval hydrodynamics,<br />
Ship Technology Research 51/3, pp.123-133<br />
SEDERBERG, T.W.; PARRY, S.R. (1986), Free-form deformation of solid geometric models,<br />
SIGGRAPH ’86, 20/4, pp.151-161<br />
SHI, Y.H.; EBERHART, R.C. (1998), Parameter selection <strong>in</strong> particle swarm optimization, 7th<br />
Annual Conf. Evolutionary Programm<strong>in</strong>g, San Diego<br />
SHI, Y.H., EBERHART, R.C. (2001), Particle swarm optimization: developments, applications<br />
and resources, IEEE Congress on Evolutionary Computation, pp.27-30<br />
SCHUBERT, B. (1972), A sequential method seek<strong>in</strong>g the global maximum of a function, SIAM<br />
J. Numerical Analysis 9, pp.379-388<br />
STATNIKOV, R.B.; MATUSOV, J.B. (1995), Multicriteria optimization and eng<strong>in</strong>eer<strong>in</strong>g, Chapman<br />
& Hall<br />
TAHARA, Y.; PERI, D.; CAMPANA, E.F.; STERN, F. (<strong>2005</strong>), CFD-based multiobjective optimization<br />
of a surface combatant, 5th Osaka Colloquium, Osaka<br />
TÖRN, A.A.; ŽILINSKAS A. (1989), Global optimization, Spr<strong>in</strong>ger<br />
VENTER, G.; SOBIESZCZANSKI-SOBIESKI, J. (2004), Multidiscipl<strong>in</strong>ary optimization of a<br />
transport aircraft us<strong>in</strong>g particle swarm optimization, Structural and Multidiscipl<strong>in</strong>ary Optimization<br />
26/1-2, pp.121-131<br />
YANG, G.; REINSTEIN, L.E.; PAI, S.; XU, Z.; CARROLL, D.L. (1998), A new genetic algorithm<br />
technique <strong>in</strong> optimization of permanent 125-I prostate implants, Medical Physics 25/12,<br />
pp.2308-2315<br />
486
Abstract<br />
Maneuver<strong>in</strong>g Simulations for Ships and Sail<strong>in</strong>g Yachts us<strong>in</strong>g<br />
FRIENDSHIP-Equilibrium as an Open Modular Workbench<br />
Tanja Richardt, Stefan Harries, Karsten Hochkirch, FRIENDSHIP-Systems GmbH,<br />
Richardt@FRIENDSHIP-Systems.com<br />
An open modular workbench called FRIENDSHIP-Equilibrium has been developed for the analysis<br />
of stationary and <strong>in</strong>stationary modes of motion. FRIENDSHIP-Equilibrium offers an <strong>in</strong>tegration<br />
environment and suitable algorithms for determ<strong>in</strong><strong>in</strong>g the equilibrium of forces and moments act<strong>in</strong>g on<br />
ships and yachts, follow<strong>in</strong>g a force model approach. In the stationary mode the steady state properties<br />
of the vessel are determ<strong>in</strong>ed (hydrostatics, VPP) while <strong>in</strong> the <strong>in</strong>stationary mode the accelerated rigid<br />
body motions are computed (maneuver<strong>in</strong>g motions). The equations of motions may be solved by means<br />
of various time <strong>in</strong>tegration methods one of which be<strong>in</strong>g an advanced time-step adapt<strong>in</strong>g Runge-Kutta<br />
algorithm. The functionality comprises the computation of added mass and damp<strong>in</strong>g as needed for time<br />
simulations. Special consideration is given to the aero- and hydrodynamics of sail<strong>in</strong>g. The program<br />
therefore also constitutes a ref<strong>in</strong>ed Velocity Prediction Program. With<strong>in</strong> the paper various maneuver<strong>in</strong>g<br />
situations are presented for both a conventional cargo ship and a contemporary IMS yacht. The<br />
examples serve to show the applicability of the approach and to illustrate the usage of the tool. Attention<br />
will also be given to IT <strong>in</strong>tegration.<br />
1. Introduction<br />
The importance of the performance prediction of ships and sail<strong>in</strong>g yachts at an early design stage <strong>in</strong>creases<br />
significantly these days. Besides the prediction of the maximum velocity <strong>in</strong> a balanced condition<br />
the analysis of the ship motions <strong>in</strong> unsteady conditions receives grow<strong>in</strong>g attention. Of major concern<br />
are the evaluation of the maneuverability for ships and the speed loss while chang<strong>in</strong>g courses for rac<strong>in</strong>g<br />
yachts.<br />
FRIENDSHIP-Equilibrium has been developed for both velocity predictions and maneuver<strong>in</strong>g simulations.<br />
The forces act<strong>in</strong>g on a ship or sail<strong>in</strong>g yacht are calculated via specific force modules of the<br />
software system. FRIENDSHIP-Equilibrium represents a workbench <strong>in</strong> which each act<strong>in</strong>g force is taken<br />
<strong>in</strong>to account by an <strong>in</strong>dividual module. Us<strong>in</strong>g the stationary mode steady state calculations are performed<br />
by balanc<strong>in</strong>g the hydrodynamic, aerodynamic, buoyant and gravitational forces. In the maneuver<strong>in</strong>g<br />
mode the equations of motions are <strong>in</strong>tegrated to record <strong>in</strong>stationary behavior. The state model comprises<br />
l<strong>in</strong>earized equations of motions <strong>in</strong> six degrees of freedom. The dynamic forces, dependent on the<br />
velocity and the acceleration, are considered via the <strong>in</strong>stationary terms <strong>in</strong> the equations of motion. The<br />
<strong>in</strong>stationary mode therefore enables a simulation of motions, especially maneuver<strong>in</strong>g, <strong>in</strong> which the<br />
different state variables can be regarded as functions of time. Hence, with the maneuver<strong>in</strong>g simulations<br />
the behavior of the vessel <strong>in</strong> unsteady conditions can be assessed and improved before the first sea trials<br />
commence. Naturally, the quality of the simulation depends on the validity and accuracy of the available<br />
force modules.<br />
2. Software system – FRIENDSHIP-Equilibrium<br />
FRIENDSHIP-Equilibrium is an advanced workbench for the analysis of stationary and <strong>in</strong>stationary<br />
modes of motion of both ships and sail<strong>in</strong>g yachts. The external forces act<strong>in</strong>g on a vessel for a given<br />
state are calculated via various force modules. Each force type like buoyant forces, gravitational forces,<br />
rudder forces, keel forces, hull resistance, aerodynamic forces, added resistance <strong>in</strong> waves, w<strong>in</strong>dage etc.<br />
is calculated <strong>in</strong> specific modules. All forces are added up by the program to determ<strong>in</strong>e the result<strong>in</strong>g<br />
forces on the vessel.<br />
These modules can be added to and taken out from the simulations <strong>in</strong>dividually depend<strong>in</strong>g on the design<br />
487
Conditions States Excitation<br />
Input files<br />
Boat Name FShip<br />
Water {<br />
Density 1.<br />
Temperature 18<br />
}<br />
DegreesOfFreedom Speed |<br />
Leeway | S<strong>in</strong>k | Heel<br />
Module Displacement<br />
Module Mass<br />
Module Rudder<br />
Module Keel<br />
Module GenericRig<br />
above upper limit<br />
axis symmetry<br />
@<strong>in</strong>clude hydrostatic.feq 4<br />
7 extra<br />
Labels Heel Angle, CoB<br />
Data
Fig. 2: Modules used <strong>in</strong> FRIENDSHIP-Equilibrium<br />
For all calculations the desired degree of freedom can be chosen by the user. Up to all six degrees of<br />
freedom may be considered <strong>in</strong> the simulations. The l<strong>in</strong>k between the degrees of freedom and the free<br />
variables is summarized <strong>in</strong> Table I. The applications of FRIENDSHIP-Equilibrium for typical scenarios<br />
are shown <strong>in</strong> Table II. The free variables used with<strong>in</strong> these operations are marked with ’X’. The required<br />
variables for hydrostatic calculations are <strong>in</strong>dicated as ’<strong>in</strong>put’.<br />
Fig. 3 illustrates the structure of FRIENDSHIP-Equilibrium. The used modules and the required parameters<br />
are def<strong>in</strong>ed <strong>in</strong> one or several <strong>in</strong>put files, see also Fig. 1. The act<strong>in</strong>g forces are determ<strong>in</strong>ed for<br />
specified environmental conditions. Depend<strong>in</strong>g on the force modules these conditions are part of the<br />
required <strong>in</strong>put. For <strong>in</strong>stance the force calculation on sail<strong>in</strong>g yachts requires the w<strong>in</strong>d speed and the w<strong>in</strong>d<br />
direction as <strong>in</strong>put.<br />
Table II: Typical scenarios <strong>in</strong> FRIENDSHIP-Equilibrium (free variables are marked with ’X’)<br />
Scenario Mode V S λ dz ϕ θ δ r δ t<br />
Hydrostatics hydrostatic X <strong>in</strong>put X<br />
Curves of form hydrostatic <strong>in</strong>put <strong>in</strong>put <strong>in</strong>put<br />
IMS type standard sail<strong>in</strong>g yacht VPP stationary X (X) X<br />
4 DOF sail<strong>in</strong>g yacht VPP stationary X X X X<br />
Advanced sail<strong>in</strong>g yacht VPP stationary X X X X X X X<br />
VPP self propulsion stationary X X X<br />
Maneuver<strong>in</strong>g <strong>in</strong>stationary X X X X X X X<br />
489
INPUT Simulation MODES OUTPUT<br />
Ship / yacht<br />
Conditions<br />
Stationary mode<br />
Velocities<br />
Module file(s):<br />
- Hull data<br />
- Keel data<br />
- Rig data<br />
- Mass<br />
- Rudder<br />
- Propeller<br />
- User module<br />
- ...<br />
- True w<strong>in</strong>d speed vs<br />
- True w<strong>in</strong>d angle b<br />
States<br />
T<br />
Equilibrium<br />
mode via<br />
Newton Raphson<br />
Hydrostatic mode<br />
- Ship speed vs<br />
- Heel angle <br />
- Leeway angle <br />
- Rudder angle <br />
- Trim angle <br />
- S<strong>in</strong>k z<br />
Hydrostatic<br />
Sail<strong>in</strong>g trim<br />
- Heel angle <br />
- Heel angle <br />
- Trim angle <br />
- S<strong>in</strong>k z<br />
Equilibriuum<br />
<strong>in</strong> Fz, My<br />
Force Module<br />
Displacment<br />
Float<strong>in</strong>g position<br />
- Trim<br />
- S<strong>in</strong>k<br />
Curves of form<br />
Excitation<br />
- Maneuver<br />
- Rudder angle<br />
- Tab angle<br />
- Course<br />
(Autopilot)<br />
Instationary mode<br />
Time-Step-Mode<br />
with<br />
Integration-Method<br />
(e.g. Cash -Karp<br />
Runge-Kutta)<br />
Motion<br />
- Path: longitude<br />
latitude<br />
- State: x,y,z, <br />
. . . . . .<br />
x,y,z, <br />
.. .. .. .. .. ..<br />
x,y,z, <br />
Fig. 3: Modes with<strong>in</strong> FRIENDSHIP-Equilibrium<br />
Three simulation modes are offered by the program for different applications:<br />
– stationary mode<br />
– hydrostatic mode<br />
– <strong>in</strong>stationary mode<br />
2.1. Stationary mode<br />
In the stationary mode the steady state of the ship will be determ<strong>in</strong>ed for specified environmental conditions.<br />
The program will resolve the equilibrium <strong>in</strong> which the sum of all forces add up to zero <strong>in</strong> the<br />
def<strong>in</strong>ed degrees of freedom by a nonl<strong>in</strong>ear equation solver. The balance will be computed by means of a<br />
Newton-Raphson algorithm. Velocity predictions <strong>in</strong> steady conditions may also be calculated with<strong>in</strong> this<br />
mode. Depend<strong>in</strong>g on the considered degrees of freedom, the output conta<strong>in</strong>s the state variables displayed<br />
<strong>in</strong> Table I.<br />
In addition, an arbitrarily set of trim parameters may be def<strong>in</strong>ed <strong>in</strong> the force modules. These parameters<br />
are optimized to achieve the maximum speed. Various optimizations rout<strong>in</strong>es are available with<strong>in</strong> the<br />
program and may be selected by the user. Consequently, desired parameters must be implemented <strong>in</strong> the<br />
modules. The usage of trim parameters is shown <strong>in</strong> the second example, see section .<br />
2.2. Hydrostatic mode<br />
The float<strong>in</strong>g position will be calculated depend<strong>in</strong>g on the heel angle <strong>in</strong> the hydrostatic mode. In this<br />
mode the equilibria for vertical, roll and pitch movements are resolved. Furthermore, curves of form are<br />
determ<strong>in</strong>ed.<br />
2.3. Instationary mode<br />
For the analysis of motions the program offers an <strong>in</strong>stationary mode. For <strong>in</strong>stationary analysis the excitation<br />
forces have to become part of the <strong>in</strong>put. Maneuver<strong>in</strong>g simulations for <strong>in</strong>stance operate with<br />
chang<strong>in</strong>g rudder angles. Fixed rudder angles can be set. The rudder angles may also be controlled by<br />
490
Instationary mode<br />
Parameters Initial values Output parameters<br />
TWS, TWA<br />
Rudder<br />
S 0<br />
INPUT<br />
Ship / yacht<br />
Maneuver<br />
Time<br />
Force modules<br />
State function Integration rout<strong>in</strong>e Output function<br />
Controller (autopilot)<br />
F( S )<br />
OUTPUT<br />
t i+1= t i + tstep Integration method Longitude,<br />
State vector<br />
(4th order, 5th order, latitude,<br />
S<br />
variable time . step,..) head<strong>in</strong>g<br />
S<br />
next<br />
= f ( s, s,t)<br />
solve F = m a<br />
S<br />
Fig. 4: Maneuver<strong>in</strong>g simulation with<strong>in</strong> FRIENDSHIP-Equilibrium<br />
the use of predef<strong>in</strong>ed maneuvers. A PID-controller (Proportional Integral Derivative) is implemented as<br />
autopilot to keep a desired course. An additional manual maneuver<strong>in</strong>g module offers the possibility to<br />
steer the boat <strong>in</strong>teractively with a joystick. Either way, steer<strong>in</strong>g changes are accounted for while runn<strong>in</strong>g<br />
the process.<br />
The motions are described via the equations of motions <strong>in</strong>clud<strong>in</strong>g all external forces, calculated by their<br />
respective force modules. Additional added mass and damp<strong>in</strong>g forces are determ<strong>in</strong>ed via a further force<br />
module and are then considered as part of the excitation force. A module for l<strong>in</strong>ear coefficients of added<br />
mass and damp<strong>in</strong>g is readily available. Alternatively, an additional module which describes these forces<br />
can be def<strong>in</strong>ed by the user. Nonl<strong>in</strong>ear coefficients as used by Masayuma et al. (1995) for <strong>in</strong>stance, may<br />
be implemented <strong>in</strong> more advanced modules. The accelerations are calculated by solv<strong>in</strong>g the equation of<br />
motions via a time-step method with a chosen <strong>in</strong>tegration rout<strong>in</strong>e. FRIENDSHIP-Equilibrium currently<br />
makes available a fourth-order Runge-Kutta scheme, a fifth-order Runge-Kutta-Feldberg scheme as well<br />
as a fifth-order Cash-Karp Runge-Kutta variable time step scheme. In the variable time step <strong>in</strong>tegration<br />
the time step is adjusted so that the difference of the result us<strong>in</strong>g a fourth order Runge-Kutta scheme and<br />
the result us<strong>in</strong>g a fifth order scheme is less than a selected tolerance for each of the state variables. A<br />
time scale can be set such that the maneuver is executed <strong>in</strong> real time or a specified fraction of that. The<br />
process flow of the time-stepp<strong>in</strong>g procedure is sketched <strong>in</strong> Fig. 4.<br />
The velocity and acceleration are comb<strong>in</strong>ed <strong>in</strong>to one state vector:<br />
s = (x,y,z,φ,θ,ψ,ẋ,ẏ,ż, ˙φ, ˙θ, ˙ψ,ẍ,ÿ, ¨z, ¨φ, ¨θ, ¨ψ) T . (1)<br />
All state variables are computed as functions of time. The trajectory is displayed <strong>in</strong> global coord<strong>in</strong>ates.<br />
Moreover, any desired state variable can be plotted dur<strong>in</strong>g the operation.<br />
3. Mathematical model<br />
3.1. Coord<strong>in</strong>ate systems<br />
The coord<strong>in</strong>ate systems used are displayed <strong>in</strong> Fig. 5. All systems are right handed and orthogonal. The<br />
forces are calculated <strong>in</strong> absolute coord<strong>in</strong>ates A which is a partially body fixed mov<strong>in</strong>g coord<strong>in</strong>ate system.<br />
The x-axis is directed along the vessel’s centerl<strong>in</strong>e while the z-axis is perpendicular to the undisturbed<br />
free surface. The heel and trim angles are determ<strong>in</strong>ed via the rotation between the body fixed coord<strong>in</strong>ate<br />
system B and the absolute coord<strong>in</strong>ates A. The hydrodynamic coord<strong>in</strong>ate system H is a mov<strong>in</strong>g coord<strong>in</strong>ate<br />
system directed along the track to determ<strong>in</strong>e the leeway angle. The boat’s track is shown <strong>in</strong> the global<br />
coord<strong>in</strong>ate system which represents the earth-bound coord<strong>in</strong>ates.<br />
491
Latitude<br />
Apparent W<strong>in</strong>d CS<br />
[+]<br />
Leeway<br />
VS<br />
X A<br />
X B<br />
X H<br />
View from above<br />
Y<br />
Y B A<br />
Y H<br />
[+]<br />
Hydrodynamic Tab<br />
CS<br />
[+]<br />
Delta<br />
A absolute coord<strong>in</strong>ate system<br />
Longitude<br />
x along the centerl<strong>in</strong>e forward<br />
x-y plane <strong>in</strong> undisturbed free surface<br />
y directed to port side<br />
B body fixed coord<strong>in</strong>ate system<br />
[-]<br />
Heel<br />
x along the centerl<strong>in</strong>e forward<br />
x-y plane <strong>in</strong> designed water plane<br />
y perpendicular to the mast<br />
H hydrodynamic coord<strong>in</strong>ate system<br />
Z H<br />
Z ZA<br />
x system along the track horizontally<br />
B<br />
to the water plane<br />
x-y plane <strong>in</strong> water plane<br />
G global coord<strong>in</strong>ate system<br />
Y<br />
Y A<br />
earth coord<strong>in</strong>ates<br />
H<br />
x-y ebene <strong>in</strong> water plane<br />
Y B<br />
Body Fixed<br />
Absolute CS<br />
Global Coord<strong>in</strong>ate System (CS)<br />
Absolute CS<br />
Hydrodynamic<br />
CS<br />
[+]<br />
AWA<br />
[+]<br />
TWA<br />
Body Fixed<br />
View from beh<strong>in</strong>d<br />
X AW<br />
Y AW<br />
X MW<br />
TWS<br />
Mean W<strong>in</strong>d CS<br />
Y MW<br />
[180 - 360]<br />
YAW<br />
MW mean w<strong>in</strong>d coord<strong>in</strong>ate system<br />
x along the true w<strong>in</strong>d direction<br />
AW apparent w<strong>in</strong>d coord<strong>in</strong>ate system<br />
x along the apparent w<strong>in</strong>d direction<br />
Fig. 5: Coord<strong>in</strong>ate systems used <strong>in</strong> FRIENDSHIP-Equilibrium<br />
492
Fig. 6: Def<strong>in</strong>ition of ship coord<strong>in</strong>ate systems <strong>in</strong> six degrees of freedom<br />
The location of the orig<strong>in</strong> must be body fixed but can be arbitrarily selected, e.g. the center of gravity.<br />
The transformation of the forces from a body fixed system B to the absolute coord<strong>in</strong>ate system A is<br />
accomplished by left multiplication with the transformation matrix<br />
⎡<br />
T BA<br />
= ⎣<br />
cosθ s<strong>in</strong>ϕs<strong>in</strong>θ cosϕs<strong>in</strong>θ<br />
0 cosϕ −s<strong>in</strong>ϕ<br />
−s<strong>in</strong>θ s<strong>in</strong>ϕcosθ cosϕcosθ<br />
for arbitrary heel and pitch angles ϕ and θ, respectively.<br />
⎤<br />
⎦ (2)<br />
3.2. Specification of ship motions<br />
The ship motions comprises <strong>in</strong> six degrees of freedom – three translations and three rotations. These<br />
motions are described <strong>in</strong> the absolute coord<strong>in</strong>ate system illustrated <strong>in</strong> Fig. 6.<br />
Table III: Generalized displacements of a vessel<br />
x Translation <strong>in</strong> longitud<strong>in</strong>al direction surge<br />
y Translation <strong>in</strong> transversal direction sway<br />
z Translation <strong>in</strong> vertical direction heave<br />
φ Rotation about the longitud<strong>in</strong>al axis roll<br />
θ Rotation about the transversal axis pitch<br />
ψ Rotation about the vertical axis yaw<br />
The state variables are expressed <strong>in</strong> vector form:<br />
x = (x,y,z,φ,θ,ψ) T<br />
ẋ = ( ẋ,ẏ,ż, ˙φ, ˙θ, ˙ψ ) T<br />
ẍ = ( ẍ,ÿ, ¨z, ¨φ, ¨θ, ¨ψ ) T<br />
3.3. Equations of motion<br />
Accord<strong>in</strong>g to Newton’s second law the motion of a rigid body <strong>in</strong> the earth-bound coord<strong>in</strong>ate system can<br />
be expressed as follows:<br />
F external = M · ẍ (3)<br />
493
The forces and moments are comb<strong>in</strong>ed <strong>in</strong> a generalized force vector<br />
F = (F x ,F y ,F z ,M x ,M y ,M z ) T<br />
while the accelerations are given <strong>in</strong> the state vector<br />
ẍ = ( ẍ,ÿ, ¨z, ¨φ, ¨θ, ¨ψ ) T<br />
.<br />
The mass matrix M then conta<strong>in</strong>s the follow<strong>in</strong>g components:<br />
⎛<br />
M =<br />
⎜<br />
⎝<br />
⎞<br />
m 0 0 0 m · z CG −m · y CG<br />
0 m 0 −m · z CG 0 m · x CG<br />
0 0 m m · y CG −m · x CG 0<br />
0 −m · z CG m · y CG I xx −I xy −I xz<br />
⎟<br />
m · z CG 0 −m · x CG −I yx I yy −I yz<br />
⎠<br />
−m · y CG m · x CG 0 −I zx −I zy I zz<br />
with m = ρ · ∇ for steady conditions.<br />
A l<strong>in</strong>ear force system is assumed to specify the act<strong>in</strong>g forces on the vessel. The hydrodynamic forces are<br />
divided <strong>in</strong>to their translation, velocity and acceleration components.<br />
F Hydrodynamic = −A · ẍ − B · ẋ −C · x (4)<br />
The matrix A <strong>in</strong>cludes the added mass coefficients and B refers to the damp<strong>in</strong>g coefficients. Hydrostatic<br />
forces are <strong>in</strong>cluded <strong>in</strong> the third term. Velocity and acceleration depend on the reference system. Therefore,<br />
the centripetal acceleration a cp must be considered <strong>in</strong> the accelerated absolute coord<strong>in</strong>ate system,<br />
see e.g. Gummert ; Reckl<strong>in</strong>g (1985). Forces result<strong>in</strong>g from w<strong>in</strong>d and waves and all other additional<br />
external forces are comb<strong>in</strong>ed <strong>in</strong> so-called further environmental forces F Environmental to be determ<strong>in</strong>ed by<br />
the appropriate modules of the program.<br />
The follow<strong>in</strong>g equations are solved by the <strong>in</strong>tegration rout<strong>in</strong>e:<br />
(M + A) · (ẍ − a cp ) = −B · ẋ −C · x + F Environmental (5)<br />
3.4. Coefficients<br />
In order to solve the equations of motion the matrices of the added mass and damp<strong>in</strong>g coefficients must<br />
be determ<strong>in</strong>ed. In general, several methods are available to identify these coefficients:<br />
• Estimation of the coefficients from simple shapes<br />
• Formulas developed by Clarke<br />
• Lewis transformation apply<strong>in</strong>g conformal mapp<strong>in</strong>g and strip theory<br />
• Inverse Fourier transformation of frequency dependent coefficients<br />
• Viscous flow simulations<br />
• Model experiments<br />
• (Full scale measurements)<br />
494
Latitude [m]<br />
-200.0-100.0 0.0 100.0 200.0 300.0 400.0 500.0 600.0 700.0 800.0<br />
-900.0-800.0-700.0-600.0-500.0-400.0-300.0-200.0-100.0 0.0 100.0<br />
Longituide [m]<br />
Fig. 7: Turn<strong>in</strong>g test for cargo ship<br />
turn<strong>in</strong>g rate [rad/s]<br />
-0.02 -0.01 -0.01 0.0 0.0<br />
0 200 400 600 800 1000<br />
time [s]<br />
Fig. 8: Turn<strong>in</strong>g rate ψ dur<strong>in</strong>g turn<strong>in</strong>g test<br />
The order <strong>in</strong> this list roughly corresponds the associated effort, the quality of the outcome be<strong>in</strong>g proportional.<br />
For a first estimation of the added mass the formulas for simple shapes like cyl<strong>in</strong>ders and<br />
ellipsoids as published by Saunders (1957) can be utilized. Clarke, see Lewis (1988), developed formulas<br />
depend<strong>in</strong>g on the ship’s geometry to evaluate l<strong>in</strong>ear coefficients for added mass and damp<strong>in</strong>g. More<br />
accurate values of the coefficients can be derived from conformal mapp<strong>in</strong>g of each section by apply<strong>in</strong>g<br />
Lewis transformation and utiliz<strong>in</strong>g strip theory. Here the three-dimensional flow problem is solved via<br />
the <strong>in</strong>tegration of two-dimensional sub-doma<strong>in</strong>s along the hull. This method is by def<strong>in</strong>ition only valid<br />
for slender bodies, but reasonable values can be obta<strong>in</strong>ed. Bohlmann (1990) for <strong>in</strong>stance used this theory<br />
to identify the coefficients for submar<strong>in</strong>es and provided formulas to calculate the damp<strong>in</strong>g coefficients.<br />
The hydrodynamic mass <strong>in</strong> x-direction can be calculated from the the added mass for an ellipsoid given<br />
<strong>in</strong> Bertram (2000). However, this yields only rough estimates for typical sail<strong>in</strong>g yacht bow and stern<br />
shapes.<br />
Frequency dependent coefficients can be obta<strong>in</strong>ed from seakeep<strong>in</strong>g calculations. A rout<strong>in</strong>e to transform<br />
these coefficients with an <strong>in</strong>verse Fourier transformation from frequency-doma<strong>in</strong> to time-doma<strong>in</strong> <strong>in</strong><br />
order to identify the coefficients <strong>in</strong> the state space (Schmiechen, 1973) has been studied by Richardt<br />
(2004) for implementation <strong>in</strong>to FRIENDSHIP-Equilibrium. F<strong>in</strong>ally, the determ<strong>in</strong>ation of the coefficients<br />
by model test<strong>in</strong>g as carried out by Wolff (1981) for example is recommended, of course, but expensive<br />
with regard to both time and money.<br />
4. Simulations<br />
Maneuver<strong>in</strong>g simulations were processed for a representative cargo ship and a typical sail<strong>in</strong>g yacht. For<br />
the former a Mar<strong>in</strong>er ship was selected which is a common test case. For the latter the full scale research<br />
yacht DYNA was chosen. DYNA resembles a 10m IMS sail<strong>in</strong>g yacht and was developed at the Technical<br />
University Berl<strong>in</strong>, Hochkirch (2000). Full scale experiment data are available for this yacht which can<br />
be used for validation purposes.<br />
4.1. Maneuver simulations for a cargo ship<br />
For the force calculations of the Mar<strong>in</strong>er ship a simple module which <strong>in</strong>cludes the estimation of the<br />
propeller thrust, the hull resistance and the rudder forces was <strong>in</strong>tegrated. The ma<strong>in</strong> parameters were<br />
taken from Wolff (1981). The coefficients for added mass and damp<strong>in</strong>g were determ<strong>in</strong>ed by the formulas<br />
of Clarke as published <strong>in</strong> Lewis (1988). For the maneuverability analysis various common test maneuvers<br />
are implemented. The standard maneuvers like turn<strong>in</strong>g and zig zag tests are readily available as<br />
495
Latitude [m]<br />
-600.0 -400.0 -200.0 0.0 200.0 400.0<br />
0 1000 2000 3000 4000 5000<br />
Longitude [m]<br />
Fig. 9: Zig zag test for cargo ship (scaled 2.5:1)<br />
maneuver<strong>in</strong>g modules. The rudder angle and the maximum course angle are free to be specified. The<br />
fourth-order Runge-Kutta <strong>in</strong>tegration method with a time step of 0.2 seconds was applied for these simulations.<br />
The calculated track for a turn<strong>in</strong>g test with a rudder angle of 20 ° is displayed <strong>in</strong> Fig. 7. Fig. 8<br />
shows the turn<strong>in</strong>g rate dur<strong>in</strong>g this simulation. After 1000 seconds a constant turn<strong>in</strong>g rate was reached.<br />
The tactical diameter of 670 meters as determ<strong>in</strong>ed from Fig. 7 is met after 10 m<strong>in</strong>utes.<br />
The zig zag test is typically executed to <strong>in</strong>vestigate turn<strong>in</strong>g capabilities. Different rudder angles and<br />
maximum course angles can be used. The track for a zig zag test with a rudder and a maximum course<br />
angle of 20 ° is displayed <strong>in</strong> Fig. 9. For lack of space, the longitude and latitude axis are scaled 2.5:1.<br />
4.2. Velocity prediction for a sail<strong>in</strong>g yacht<br />
In a second example the FRIENDSHIP-Equilibrium was used for the velocity prediction of a 10m IMS<br />
yacht. Various modules have been developed from research results for that yacht, see Hochkirch (2000).<br />
The forces act<strong>in</strong>g on keel and rudder were def<strong>in</strong>ed <strong>in</strong> the modules called ’Keel’ and ’Rudder’. The<br />
aerodynamic rig forces were calculated with the module ’GenerigRig’. Different modules were used for<br />
displacement, mass, crew and w<strong>in</strong>dage. So as to demonstrate the application of trim parameters a set of<br />
different trim parameters was <strong>in</strong>cluded <strong>in</strong> the ’GenericRig’ module. These parameters are expla<strong>in</strong>ed <strong>in</strong><br />
Table IV.<br />
Table IV: Example of trim parameters <strong>in</strong> the module ’GenericRig’<br />
σ reef V S −→ max reef<strong>in</strong>g of the sails to be adjusted so as to ga<strong>in</strong> optimum speed<br />
τ flat V S −→ max flaten<strong>in</strong>g of the sails to be adjusted so as to ga<strong>in</strong> optimum speed<br />
t twist V S −→ max reduction of the center of effort of the aerodynamic forces so as to improve<br />
the overall performance of the boat when sail<strong>in</strong>g upw<strong>in</strong>d, see Jackson<br />
(2001) for details<br />
The offsets of the hull and geometric data are given <strong>in</strong> the <strong>in</strong>put file. Parameters like water density,<br />
gravity etc. can be def<strong>in</strong>ed <strong>in</strong> either the <strong>in</strong>put file or <strong>in</strong> the program. Before calculat<strong>in</strong>g the equilibrium<br />
for sail<strong>in</strong>g conditions the model must be <strong>in</strong> hydrostatic balance. Therefore, the equilibrium for trim and<br />
s<strong>in</strong>kage <strong>in</strong> rest position is determ<strong>in</strong>ed first. The states to be calculated can be def<strong>in</strong>ed <strong>in</strong> the ’Cycle Range’<br />
w<strong>in</strong>dow as <strong>in</strong>dicated <strong>in</strong> Fig. 2. In the example the true w<strong>in</strong>d velocity VWT is considered from 4 m/s to<br />
9 m/s for all courses between an angle of 35 ° to 180 ° with an <strong>in</strong>crement of 5 °. In the output table the<br />
velocities for the equilibrium of all states is listed as pictured <strong>in</strong> Fig. 10. In this figure a polar diagram is<br />
also shown for the six w<strong>in</strong>d speeds.<br />
496
Fig. 10: VPP for a sail<strong>in</strong>g yacht<br />
4.3. Maneuver simulations for a sail<strong>in</strong>g yacht<br />
For the <strong>in</strong>stationary calculations an additional module was def<strong>in</strong>ed to compute the added mass and damp<strong>in</strong>g.<br />
The module was called ’AddedMass’ and the coefficients were determ<strong>in</strong>ed by means of Lewis transformation<br />
as expla<strong>in</strong>ed <strong>in</strong> section . The dynamic changes of the effective angles at the keel, rudder and<br />
the sails dur<strong>in</strong>g the motions are accounted for <strong>in</strong> the lift calculations. The conditions were chosen to meet<br />
the situation encountered dur<strong>in</strong>g full-scale measurements. The tack<strong>in</strong>g simulations were processed with<br />
a true w<strong>in</strong>d velocity of 9.59 m/s. The close-hauled courses were steered at 37 ° to the w<strong>in</strong>d. The rudder<br />
angle was automatically adjusted by the autopilot. The factors of the PID controller can be adapted to<br />
the problem and consequently <strong>in</strong>fluence the course keep<strong>in</strong>g capabilities. Tacks can also be simulated <strong>in</strong><br />
manual maneuvers which enables the users of the program to steer themselves by means of a joystick. A<br />
tack<strong>in</strong>g maneuver <strong>in</strong> which the rudder is l<strong>in</strong>ked to a maximum angle until a certa<strong>in</strong> course angle on the<br />
other tack is reached can be applied. The fifth-order Cash-Karp <strong>in</strong>tegration rout<strong>in</strong>e was used to solve the<br />
equations of motions.<br />
Fig. 11 depicts the FRIENDSHIP-Equilibrium runn<strong>in</strong>g <strong>in</strong> the maneuver<strong>in</strong>g mode. The sett<strong>in</strong>gs of the<br />
autopilot can be seen <strong>in</strong> the ’ManeuverEditor’ on the upper left. Several graphs are <strong>in</strong>tegrated <strong>in</strong> the<br />
’Maneuver<strong>in</strong>g’ w<strong>in</strong>dow. On the left the velocity as function on time is displayed. In the other graphs the<br />
track and the heel<strong>in</strong>g angle are plotted. Naturally, the quantities to be shown can be selected as needed.<br />
The results of the tack<strong>in</strong>g simulations are compared to data of the full scale measurements. The yaw<br />
angle is displayed <strong>in</strong> Fig. 12 while the associated heel angle is shown <strong>in</strong> Fig. 13. The oscillation of<br />
the full-scale data was caused by heavy seas and strong unsteady w<strong>in</strong>ds dur<strong>in</strong>g the measurements.<br />
Nevertheless, a rather good correlation between the simulated and the measured data can be observed.<br />
From the start<strong>in</strong>g course to head w<strong>in</strong>d the tack needs 7 seconds. A stable heel angle is achieved after<br />
497
Fig. 11: Maneuver<strong>in</strong>g simulation for a 10m sail<strong>in</strong>g yacht<br />
43 seconds and the new course is fixed after about one m<strong>in</strong>ute. In real life these results depend on the<br />
helmsman. In the simulations the value may be optimized by adjust<strong>in</strong>g the autopilot. The optimum<br />
speed is reached after about 82 seconds. The sail trim was idealized dur<strong>in</strong>g the simulated tack. Naturally,<br />
this typically is not the case <strong>in</strong> reality. Therefore, additional parameters will have to be <strong>in</strong>tegrated <strong>in</strong> the<br />
future so as to take <strong>in</strong>to account the time needed by the crew to trim the sails optimally.<br />
5. Conclusion<br />
FRIENDSHIP-Equilibrium is an open modular workbench which follows a force approach. All forces<br />
and moments are def<strong>in</strong>ed <strong>in</strong> separate modules. A wide range of force modules is readily available with<strong>in</strong><br />
the software system like hydrostatics, mass, hydrodynamics from model tests, added resistance <strong>in</strong> waves,<br />
Delft series for resistance of sail<strong>in</strong>g yachts and keels (Keun<strong>in</strong>g ; Sonnenberg, 1999), lift<strong>in</strong>g surfaces<br />
such as keels and rudder, IMS type rig models and many more. Additional modules to accommodate<br />
<strong>in</strong>dividual forces and scenarios can be easily <strong>in</strong>tegrated for specific simulations.<br />
FRIENDSHIP-Equilibrium supports stationary, hydrostatic and <strong>in</strong>stationary simulation modes. The accuracy<br />
of the results, naturally depends on the implementation of the underly<strong>in</strong>g mathematical model<br />
used <strong>in</strong> the force modules. For <strong>in</strong>stationary conditions reasonable estimations of added mass and damp<strong>in</strong>g<br />
terms are of importance s<strong>in</strong>ce the simulations are sensitive to these forces. An accurate identification<br />
of this terms is required for realistic results.<br />
Special focus was given on <strong>in</strong>stationary simulations for which examples were presented for two types<br />
of vessels – a cargo ship and a sail<strong>in</strong>g yacht – to show the applicability and the potential usage of the<br />
system. The sail<strong>in</strong>g yacht maneuver presented was the tack<strong>in</strong>g of a 10m IMS yacht. It comprised a<br />
498
Yaw angle [deg]<br />
-60 -40 -20 0 20 40 60<br />
simulation<br />
full-scale measurements<br />
0 50 100 150 200 250<br />
Time [s]<br />
Fig. 12: Yaw angle<br />
Heel angle [deg]<br />
-40 -30 -20 -10 0 10 20 30 40<br />
simulation<br />
full-scale measurements<br />
0 50 100 150 200 250<br />
Time [s]<br />
Fig. 13: Heel angle<br />
comparison between the simulations based on FRIENDSHIP-Equilibrium and experimental data from<br />
full-scale sail<strong>in</strong>g tests which showed very promis<strong>in</strong>g correlation.<br />
References<br />
BERTRAM, V. (2000), Practical Ship Hydrodynamics, Butterworth - He<strong>in</strong>emann<br />
BOHLMANN, H. J. (1990), Berechnung hydrodynamischer Koeffizienten von U-Booten zur Vorhersage<br />
des Bewegungsverhaltens, IfS Report 513, Technische Universität <strong>Hamburg</strong><br />
GUMMERT, P.; RECKLING, K.-A. (1985), Mechanik, F. Vieweg & Sohn<br />
HANSEN, H.; JACKSON, P. S.; HOCHKIRCH, K. (2003), Real-time velocity prediction program for<br />
w<strong>in</strong>d tunnel test<strong>in</strong>g of sail<strong>in</strong>g yachts, Int. Conference on the Modern Yacht, Southampton<br />
HOCHKIRCH, K. (2000), Entwicklung e<strong>in</strong>er Meßyacht zur Analyse der Segelleistung im Orig<strong>in</strong>almaßstab,<br />
Mensch & Buch Verlag, Berl<strong>in</strong>, ISBN 3-89820-119-8, PhD-Thesis, Technische Universität Berl<strong>in</strong><br />
JACKSON, P. (2001), An improved upw<strong>in</strong>d sail model for VPP’s, 15 th Chesapeake Sail<strong>in</strong>g Yacht Symposium,<br />
Annapolis<br />
KEUNING, J. A.; SONNENBERG, U. B. (1999), Approximation of the calm water resistance on a sail<strong>in</strong>g<br />
yacht based on the ’Delft Systematic Hull Series’, 14 th Chesapeake Sail<strong>in</strong>g Yacht Symp., Annapolis<br />
LEWIS, E. V. (Editor) (1988), Pr<strong>in</strong>ciples of Naval Architecture Vol.III, SNAME<br />
MASAYUMA, Y.; FUKASAWA, T.; SASAGAWA, H. (1995), Tack<strong>in</strong>g simulation of sail<strong>in</strong>g yachts -<br />
Numerical <strong>in</strong>tegration of equation of motion and application of neural network technique, 12 th Chesapeake<br />
Sail<strong>in</strong>g Yacht Symp., Annapolis<br />
RICHARDT, T. (2004), Simulation dynamischer Bewegungsvorgänge von Segelyachten, Diploma-thesis,<br />
ILS Technischen Universität Berl<strong>in</strong><br />
SAUNDERS, H. E. (1957), Hydrodynamics <strong>in</strong> ship design Vol.I, SNAME<br />
SCHMIECHEN, M. (1973), On state space models and their application to hydromechanic systems,<br />
NAUT Report 5002, Dept. of Naval Architecture, University of Tokyo<br />
WOLFF, K. (1981), Ermittlung der Manövriereigenschaften fünf repräsentativer Schiffstypen mit Hilfe<br />
von CPMC-Modellversuchen, IfS Report 412, Technischen Universität <strong>Hamburg</strong><br />
499
! <br />
<br />
<br />
"<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
#$%"<br />
<br />
!<br />
" " # <br />
" $ % " " <br />
# & '"<br />
"( ") <br />
*"+" # "<br />
,*&% +"" <br />
& " ! <br />
- <br />
" " <br />
<br />
<br />
&$ <br />
<br />
. " " +"<br />
""/ +" "<br />
" " " " " " " "<br />
( " ") %& " " ( ") +"<br />
"" " " "<br />
" 0 (" ) " % <br />
&"1 &" *<br />
" " " "<br />
" " " " "* " <br />
<br />
" " " "<br />
" " & " <br />
& &+" "# " # <br />
" " " <br />
" " +" " * " "<br />
" " " 0 <br />
+0 " "<br />
%( 0 ) "*"
" " " <br />
( ) +" "/<br />
("" ) &"" " 1 <br />
( " " " ) +" " & " <br />
" "!<br />
* "" " " <br />
# & " " " 0 "/ <br />
" ( ) ""(")<br />
<br />
-* " " <br />
" <br />
+"" " "<br />
+" & "# *%& "<br />
0 <br />
<br />
" " " # * 1<br />
%& & " " " <br />
&% " "! " <br />
" & " % <br />
" " " " +" <br />
% " & 1 "+"<br />
"" "# * 1%& , " <br />
" " +" <br />
# 3 " " " <br />
"*0 " " ""<br />
# * * 4 <br />
* " " % +" <br />
" & % " <br />
"" 5 & " " " * <br />
<br />
<br />
+"*" <br />
" " "<br />
" +" " * % * <br />
"(%& # ) (* <br />
*1%& ) "" "<br />
(/)" "<br />
<br />
<br />
'$ <br />
<br />
+" 1 % <br />
" % " "<br />
" " &""% <br />
+" " 6 "<br />
& " 0 (" 0 <br />
%) " "" " <br />
. "% "&"" % "<br />
. " % "%<br />
<br />
+" " " "& 1<br />
"+"" " <br />
" " " +" " "<br />
%& * "% <br />
"# # . " <br />
<br />
2
" <br />
" & <br />
<br />
+" &,<br />
• "(8) "(+8)<br />
• <br />
• "1<br />
• "1<br />
• 8<br />
+" 8/ "5 <br />
. " " "()<br />
" " (8 9 )+" "<br />
"(+8) +8 9 & " <br />
<br />
+"* " " % "8 9 +8 9 +"% "<br />
:1 ",<br />
<br />
7<br />
7<br />
= ( ∗<br />
) + ( ∗<br />
) (2)<br />
<br />
&",!6 (& ")<br />
<br />
+" &"" " <br />
& " " 6 " <br />
+" " ,<br />
<br />
Ρ = %( $ # " ) (7)<br />
&",<br />
$ 6 <br />
# 6 <br />
" 6 <br />
6 "/ <br />
+" " " %+"&%"<br />
" * 1%& & "<br />
" &% & " +" &% " <br />
" " +" <br />
" *"( #)," <br />
(" "" ) " <br />
" <br />
<br />
+" 6"16 " +"<br />
". " " ""1<br />
& " " " 0 & & " "<br />
* 0 . "&% " <br />
" "" "&"" " % " "<br />
% &" " . " " <br />
" "" "( 9 ) " <br />
"( 9 ) " "& "1<br />
"" " " "" <br />
+" " 0 " "1<br />
& " " " &" "<br />
+" " " " "1<br />
" 0 " & "<br />
7
"1+" % <br />
" " " " "1 "<br />
" <br />
<br />
'())*+ " * " "1 <br />
"1 " "" " ""1 "<br />
"0 +" " " "<br />
< (< ∈ 2) +" "1 % "<br />
" " +" " "<br />
"# 1%& <br />
+" <br />
&%" <br />
<br />
""1& " <br />
" +" " &, " <br />
" " " , & &"<br />
" "+" = ">& " ".<br />
&= "> " , ( ? , )+&" " " "<br />
& " +" "* <br />
, ,* " < "<br />
2<br />
<br />
. " & " "<br />
" " <br />
" 1 %& " 0 + " " & <br />
"<br />
<br />
($) "<br />
<br />
-*"& " " <br />
+" ""& 6"& "<br />
% +" " & " # +"<br />
& &,&<br />
" +" " & & " " 2 7 ; @A +"<br />
"& "(2% ) "2 & +" <br />
6 "1 +" &<br />
,<br />
"1(&"1B@°A)<br />
" " "<br />
&"1 " " " <br />
+" " 8 9 +8 9 5 " <br />
" " "" * <br />
" " " "<br />
<br />
+" & +" " . 2 & " <br />
+" & &,<br />
2 8+8<br />
7 "1<br />
; "<br />
C "1<br />
<br />
<br />
;
($#$*" " " "<br />
<br />
+" " #"& <br />
+" " <br />
" 8 9 +8 9 +" & "1 " 8 9 +8 9 <br />
&*D 2+" "& *<br />
8 9 +" "". "& " <br />
"* " ""& "<br />
<br />
[Nm]<br />
3.50<br />
3.00<br />
2.50<br />
2.00<br />
1.50<br />
1.00<br />
0.50<br />
0.00<br />
L=100 [m] L=200 [m] L=300 [m]<br />
CPA average<br />
CPA m<strong>in</strong><br />
CPA max<br />
<br />
D 2,8 9 "<br />
<br />
+" " . &" " <br />
%& " ( ) " <br />
+" +8 9 " <br />
.& " """ " <br />
<br />
[m<strong>in</strong>]<br />
30.00<br />
25.00<br />
20.00<br />
15.00<br />
10.00<br />
5.00<br />
0.00<br />
L=100 [m] L=200 [m] L=300 [m]<br />
TCPA average<br />
TCPA m<strong>in</strong><br />
TCPA max<br />
<br />
D 7+8 9 "<br />
<br />
($&$+ <br />
<br />
"1& " " +" 1* <br />
" "" "0 - '*../+<br />
C
$ " "" " 6 "" " " " <br />
% "* .& " "<br />
" " "& "D ;+"<br />
""& " " " &"( ) <br />
<br />
+" "( " ) " "<br />
D ;D C<br />
<br />
315°<br />
3.00<br />
[Nm]<br />
2.00<br />
1.00<br />
45°<br />
Ship's doma<strong>in</strong><br />
L=100[m]-L=300[m]<br />
Ship's doma<strong>in</strong><br />
L=100[m]-L=200[m]<br />
270°<br />
0.00<br />
90°<br />
Ship's doma<strong>in</strong><br />
L=100[m]-L=100[m]<br />
225°<br />
135°<br />
180°<br />
D ;,E2" & ""27; "<br />
<br />
315°<br />
[Nm]<br />
0°<br />
3.00<br />
2.00<br />
1.00<br />
45°<br />
Ship's doma<strong>in</strong><br />
L=300[m] L=300[m]<br />
Ship's doma<strong>in</strong><br />
L=100[m]-L=300[m]<br />
<br />
270°<br />
0.00<br />
90°<br />
225°<br />
135°<br />
180°<br />
<br />
D C,E"& " "2; & ";"<br />
<br />
+" +" <br />
&"1 " ()<br />
& " +" & " & " <br />
+" # &"" " 8 9 " <br />
& " 1"
. " "&" "*" "<br />
" "& " ""D <br />
<br />
270°<br />
315°<br />
0°<br />
3.00<br />
[Nm]<br />
2.00<br />
1.00<br />
0.00<br />
45°<br />
90°<br />
Average doma<strong>in</strong>:<br />
Ship L=100[m]<br />
Average doma<strong>in</strong>:<br />
Ship L=200[m]<br />
Average doma<strong>in</strong>:<br />
Ship L=300[m]<br />
225°<br />
135°<br />
180°<br />
<br />
D , "& " "27;@A<br />
<br />
. " <br />
+" "" "&"+"# &" * "<br />
% " ""& " &+"<br />
& "(2; ")&*+"" " &<br />
& " "D F<br />
<br />
[Nm]<br />
0°<br />
3.00<br />
Ship's doma<strong>in</strong><br />
L=100[m]-L=300[m]<br />
315°<br />
2.00<br />
45°<br />
Ship's doma<strong>in</strong><br />
L=300[m]-L=100[m]<br />
1.00<br />
270°<br />
0.00<br />
90°<br />
225°<br />
135°<br />
180°<br />
<br />
D F,E"2; <br />
<br />
($'$," "<br />
<br />
+" " & " <br />
+" " "*<br />
# "<br />
" "& " " "<br />
F
" " (8 9D ) " "<br />
"(+8 9D )D G+"&* H <br />
<br />
<br />
D G,8 "" 8 9D<br />
<br />
! " "8 9D & γ∈?2I(6<br />
2 6 ) +" <br />
* "# .& " &<br />
"1" "8 9D D J<br />
<br />
<br />
<br />
D J,8 9D ""1 "()K);K)7)2<br />
<br />
($($+ <br />
<br />
"1 & " L<br />
&%& " & "* %& # <br />
'*..0+ ! " " &<br />
+"" 1'())2++""1<br />
<br />
G
& &/ ,/ .<br />
&"& " 1*( <br />
# )! " "" "( ) <br />
"# 6. "&<br />
"" " " "<br />
+" -*<br />
" ; 2 "& D M & " " +"<br />
& " % " "<br />
<br />
<br />
<br />
)<br />
)<br />
D M,"1,)"; / & ""; "K)"2<br />
6 & ""2 "<br />
<br />
% " 8 9 " " (8 9D ) " <br />
% " ""(&"1 <br />
" " ) -* " <br />
" "(&"@°A)"&D 2<br />
<br />
<br />
<br />
<br />
)<br />
)<br />
D 2, D " ; ( ) & " " " "<br />
( C ),) "2J K) "7G )<br />
<br />
-$ <br />
<br />
+" " " " <br />
4 " " " "1 <br />
+" " """ <br />
J
" " +"* <br />
1%& "+""" <br />
<br />
+" " "<br />
" ""1 % "" <br />
" " +" <br />
" " & "<br />
0 '())/+<br />
<br />
<br />
)"<br />
<br />
.-+5NO:!$ :.N(72)P 3.D88<br />
8 8172-9 <br />
<br />
.-+5NO:!$ :. N(7;) 3 22 " $ <br />
8 . . L ..L<br />
<br />
.-+5NO:!$ :.N(7;)4 M " . 8<br />
" 5 5<br />
<br />
.-+5NO:!$ :. NK 5.N Q(7C) 5 ; 8!.+<br />
<br />
<br />
5+:!$ :EK.9.R:.K5+:!$ :.9(2MMG)6 ! <br />
$ L$ &/9<br />
<br />
NS!QK$ NK$ LHD(2MM;) QL CF<br />
<br />
M
! <br />
!"!<br />
<br />
<br />
<br />
<br />
<br />
<br />
! <br />
" # <br />
$ % <br />
& & &<br />
<br />
' ' ( ) <br />
* <br />
+ <br />
* <br />
<br />
<br />
"#<br />
<br />
# $ $ % $ & <br />
' (!!) *$ + ,!,- <br />
. /!)0%12$0 %<br />
1!1$! " $# $ <br />
3!<br />
<br />
4 -%. <br />
/ ,5 ! ) 661 3<br />
- / ,-../0 + 37 <br />
!-8 % % 9 <br />
+:+!!/<br />
<br />
" " % <br />
! 3 % <br />
$ " !<br />
<br />
<br />
(!;# $
% %% "!?3" " =% <br />
. % $ ! 3 " <br />
:$ " :.".%""<br />
! < % <br />
!<br />
<br />
$#%&<br />
<br />
# $ % ! $ % <br />
3)4 9 + !)6+$ + <br />
% !2$ , @! "4 3<br />
! 2*, % ! 7 <br />
% $% 606$ + !% <br />
-+:*!&!/% <br />
% 0 $% % !8% <br />
" $ !<br />
<br />
# $ % % " <br />
% % " % !4% <br />
% % % A 4 % ! 3 <br />
9 9 % " %<br />
% ! % " <br />
B)))! - C % <br />
4!&!/!7$ !!)% <br />
% " !0% 12,<br />
% 2 "4 !% %$%<br />
"B)B! %% "<br />
$ 2,6! % -2*/ 9 <br />
% % 3D <br />
2*,!<br />
<br />
$#"#&#<br />
<br />
E % "%3> <br />
3 ' ' ! 3 % " 0<br />
4 ! 4 % ! 9 " <br />
$%1,23330<br />
<br />
% 7"<br />
6@@ : F , 6@ % ! ) <br />
+%= $!<br />
<br />
7% !7<br />
. = 9 "" B< <br />
!<br />
<br />
4 # $ % % ! <br />
-/@> :3=GD<br />
-" $/! % 20 !
"0% $H0<br />
%% % ,!<br />
<br />
49I "% 0% $"%@<br />
*"*% % !9 " <br />
"%$ !4"<br />
% " $ % + %<br />
6@06,!<br />
<br />
(% % J220>> <br />
-66*/%)#0+-666+/ ?0+-66@/3 ""4 ,23350!(!+<br />
%)#0+""C666! -)#0+<br />
+4DK/"$$!4($0' 26<br />
> % % $"4 ,233503<br />
$"% %$ !<br />
<br />
8 % "".<br />
!" 9 "<br />
$%. %$ <br />
I" " !<br />
<br />
<br />
<br />
3;;%%%! "!$+!;!%"!!<br />
(!+!;3 )#0+"" $666C<br />
<br />
% % $ <br />
" !<br />
<br />
79,2+ B
N=OJ+6+; + 0$ % =!<br />
"GJ+;4"% !)% . B
% !4<br />
""$"%,0 !"<br />
% " !&<br />
! ) " %$ <br />
" %""0<br />
" ' ( $! 4 7>?3 9 <br />
!) ""$%$<br />
" % " $ " <br />
% " !<br />
<br />
%;<br />
• 3=;2929*<br />
• 3%;*$<br />
• ?;+$"%<br />
• F%;+9,' -/<br />
• ?%;" $' <br />
• &;+! 2<br />
•
( 0" 6,23370<br />
<br />
'#$#!./-<br />
<br />
0 " H ! ' <br />
" !<br />
<br />
F0%$" !)%<br />
" $ I = . <br />
! ? 9 " $ !<br />
4$ <br />
=9!' %<br />
$!<br />
<br />
< "!<br />
<br />
9 !) <br />
90 -/ " !<br />
<br />
9 %0$% <br />
% 9="9!<br />
0 - / " $ !<br />
4 " <br />
" !<br />
<br />
-. / <br />
!7@0* "% %!<br />
% % " : % <br />
9: !<br />
<br />
0!<br />
= "-0<br />
/" 9! <br />
!<br />
<br />
3% ! <br />
% ! ' " <br />
;<br />
• <br />
• <br />
<br />
3% %: : <br />
"$!4 <br />
!< =%% <br />
=% % =<br />
!<br />
<br />
3 % 0 ! ) <br />
% !<br />
4 $0: 0"% <br />
L$M!<br />
<br />
" " <br />
!
#&7 % %<br />
: %!<br />
<br />
% % !<br />
%! $% <br />
" $ I ! <br />
% :%<br />
!40,7 " 9 !<br />
<br />
' ! <<br />
9 " :9 <br />
" :%" % !<br />
<br />
=<br />
% ! " <br />
" ! % <br />
" ! ) $ <br />
!<br />
<br />
,7%"% <br />
" %I!
40 ." "%0 <br />
% !%<br />
%9 %0 %!"<br />
$ " 9 0<br />
%!3 0 % % <br />
9 ! .<br />
" % " <br />
" $!4 <br />
" % 6$;
F 4 % % % " % ! F " <br />
B< % %$ ! B< . <br />
00 ""! " <br />
3' ' % 9" <br />
!<br />
<br />
) % = ! C <br />
"!! =" <br />
"! 4<br />
" !A " <br />
"" % !4 %$ % B<<br />
! " $ $ % <br />
!
(!*!;4#4A
(!6!;E +<br />
<br />
9 4#4A< %$ 9<br />
!" <br />
! " % 9! E <br />
"% !<br />
<br />
0#$# <br />
<br />
F "% " B
C)33!H
! <br />
<br />
<br />
<br />
<br />
<br />
<br />
!" # <br />
!$ !$ <br />
%<br />
& %<br />
% % ' <br />
& <br />
" <br />
!#<br />
<br />
!"#<br />
<br />
!"! $<br />
<br />
"# <br />
$ %&#% '!<br />
### ### ())*++,"<br />
(<br />
$ <br />
$ <br />
$ <br />
$ ! )<br />
$ <br />
<br />
###! ! <br />
* & + , )! ,<br />
! - )&.#(/#<br />
<br />
### (*0'+ <br />
* '+ <br />
1 ( ' <br />
2 3 4 ''5 %<br />
! <br />
<br />
<br />
6 <br />
! !,<br />
, '!7 ! 384 <br />
92 %8'! '<br />
& % <br />
38'*&:/# )+ <br />
( $ )
!"% & '$<br />
<br />
< ! <br />
'$1 ) <br />
) !0 & <br />
5 *5+ ! -<br />
,..."+ 5 <br />
<br />
% 5 9 5*95+ )'5*'5+ <br />
% 5 <br />
<br />
! ! 7 <br />
< ! <br />
7 55 *&<br />
+ ! ! 5% <br />
!95 ! <br />
<br />
%" &$<br />
<br />
5 $ ) <br />
7 (<br />
: 9! = . 7 = & <br />
! !<br />
8 <br />
; >2 ? )! <br />
/ @ ,! <br />
<br />
<br />
%"! <br />
<br />
! ! <br />
7>!? 6& ! !! <br />
! !& A )!82<br />
' *82'+ ! <br />
1 ! <br />
) 6B! 7<br />
<br />
<br />
<br />
!%( 5 <br />
C DB<br />
<br />
0 ;<br />
:<br />
'<br />
:")<br />
:(;;<br />
<br />
%"% <br />
<br />
! 82'7 <br />
!(<br />
<br />
• 1 !8E6<br />
1 8 !5<br />
A *5A + 7 :)41 <br />
8>?)8 ! $<br />
! !<br />
• 2 ! !88 !5A <br />
88 <br />
<br />
5A ! <br />
!7 #41 ! : (") <br />
: !) <br />
!) <br />
<br />
<br />
<br />
<br />
<br />
!!7 <br />
0 ! <br />
')?
; >A ?
?> ? > ? <br />
') ! <br />
!)!! !! <br />
!!) !!) <br />
!<br />
)" $$<br />
)"! <br />
<br />
> ? * :" <br />
38'+ (<br />
• ! <br />
• ! ! *. +<br />
7 + ! <br />
*+<br />
• ! *<br />
!+<br />
• 6! *<br />
6 6 +<br />
• 6!! <br />
• 6 <br />
• 6 * + <br />
1 * ) $ 7 +<br />
• 3 ) <br />
<br />
! <br />
! & <br />
(<br />
• !) ( <br />
<br />
• 7( 7 ! <br />
6 6 <br />
• ' ( ) <br />
<br />
• 6( ) 8 <br />
<br />
• ( 7 <br />
)<br />
• 5 ( ! <br />
! 5 <br />
) )<br />
<br />
3& <br />
!!& <br />
3 ) *!+ ) <br />
& !') H <br />
4 ) <br />
; 6, = I<br />
<br />
)"% <br />
<br />
< 5 7 <br />
1 ! %% <br />
! <br />
<br />
< !! !<br />
!%%(8 !! 5<br />
8 <br />
& <br />
2 !<br />
<br />
& <br />
!<br />
<br />
< <br />
! <br />
2 ! <br />
<br />
<br />
<br />
)") -$ <br />
<br />
' )! <br />
F ! &<br />
& 95 <br />
'5<br />
<br />
' )!! < <br />
! !! <br />
! ! ! <br />
7 ! !<br />
> ? E **++*"+<br />
!5% ! ! !<br />
5% <br />
! % <br />
1! ! 7 <br />
<br />
<br />
<br />
7?E <br />
<br />
<br />
D
$ & ! < <br />
& ! ! !<br />
<br />
A ! !<br />
<br />
!*D#&:;#+)<br />
*##&;#+!% !
! " # <br />
$%&' ( <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
)*("!* " !<br />
+( + **+"<br />
! * ! ! + ! ! , " <br />
* " -.* #*+!<br />
+!!*""!!! "<br />
!! % ! ( !! ! ! + */<br />
(!!!!*+!+* ! <br />
<br />
0** ! * " !+ **! !! <br />
+!**!"" !!!!"<br />
"(+* ! ! !"###$0**!<br />
!!+("( ,!+*<br />
*!**!(#*(!"" "( 1<br />
*!,(#*!" "*!#<br />
"" +!!!!"" <br />
<br />
,1""((#*!+<br />
2 + ! ! * * " "" "! % * <br />
**!"" !!+""(3<br />
<br />
,* !*""!""! 4 2<br />
"" . * *"" !<br />
!"!!+"$%&!!"(<br />
% , 56%)6 "* * *!<br />
*"" * (*!"!!*(#*<br />
+ " , (#* + " * ( * #"<br />
(**!5,""!!" *<br />
5&! *<br />
<br />
!"#$<br />
<br />
.7( """+!# &!+/ !*<br />
"""" !""!
Governor<br />
Command<br />
<br />
<br />
<br />
Propulsion<br />
Control<br />
System<br />
θ<br />
<br />
Pitch<br />
Control<br />
System<br />
Disturbances<br />
*<br />
θ<br />
Propulsion<br />
mach<strong>in</strong>e<br />
<br />
$ )*<br />
+ -<br />
$ 5"<br />
Propeller<br />
Torque<br />
<br />
9<br />
Propeller<br />
Thrust<br />
<br />
. 5"<br />
+<br />
.)"<br />
Ship<br />
Resistance<br />
-<br />
<br />
)<br />
<br />
Shaft<br />
rotation<br />
dynamics<br />
7<br />
: π& <br />
<br />
<br />
<br />
9<br />
÷<br />
%<br />
<br />
1-w<br />
7<br />
<br />
<br />
)<br />
<br />
Disturbances<br />
.7-,"""+!#<br />
Ship<br />
translation<br />
dynamics<br />
<br />
0" "(+*!+ !+("" <br />
" ! ' ( *! * + ! *! ( <br />
!!!*" )-<br />
<br />
<br />
7<br />
% = ⋅ !- <br />
<br />
<br />
= % + ;<br />
<br />
<br />
<br />
0 ** "( (+ !+("" <br />
1&( "1( !!!<br />
!** )-<br />
<br />
<br />
' = ⋅<br />
<br />
& ⋅<br />
7<br />
: π<br />
<br />
!& " -<br />
<br />
= ' +<br />
;<br />
<br />
: π ⋅ &<br />
<br />
& "" 1 , "" 1 ! +<br />
!"*"(*"" !-<br />
: 8<br />
<br />
= ρ ⋅<br />
<br />
⋅ ⋅ )<br />
<br />
(<br />
θ<br />
<br />
<br />
: <br />
' = ρ ⋅ ⋅ ⋅ ) (<br />
θ<br />
<br />
<br />
<br />
<br />
,""".:/ ,/ *<br />
* (<br />
<br />
<br />
= " !1*+ '<br />
<br />
= ρ ⋅<br />
<br />
⋅ ⋅ )<br />
<br />
<br />
<br />
<br />
& "!! !/ ""+ &<br />
! 2*""!<<br />
$%,%= )# ! 2 ! + ! ( ( <br />
% 2+ ("! "*""<br />
**!+**1! <br />
<br />
!%&'(<br />
<br />
,!" "!* 2!!**!(*<br />
!"##+$0 !<br />
<br />
!" ""(.7 2 )!**!*<br />
! "/ """>! ?" * ,<br />
"" ! "! ! ! <br />
2 , *>!""?(.:+<br />
;<br />
<br />
;<br />
:<br />
<br />
8
2 ,+!#!(>!""?8"-*1<br />
$ * ! "" "!@!+!!. " (#*+!<br />
!(#*!7/( ,*("!+! ( (+!<br />
+ ! + * ! . !- $ * ! + ! ! +<br />
! *!*(!( ,"" "!@<br />
! ! +""!!(! ( <br />
<br />
$ )*<br />
Shaft<br />
rotation<br />
dynamics<br />
7<br />
: π& <br />
θ<br />
$ 5"<br />
Propeller<br />
Torque<br />
<br />
9<br />
+ -<br />
<br />
<br />
9<br />
÷<br />
Propeller<br />
Thrust<br />
<br />
%<br />
<br />
. 5"<br />
1-w<br />
+<br />
7<br />
<br />
.)"<br />
Ship<br />
Resistance<br />
-<br />
<br />
)<br />
<br />
Disturbances<br />
.:-,!""+!#<br />
Ship<br />
translation<br />
dynamics<br />
Disturbances<br />
<br />
"" +! ! ( ! ( * <br />
! +"$ * @!(!"/+<br />
' +! " ! " ! !" * <br />
+! ! * " =* ! *<br />
! 2" "!!!(*+!#! ,( <br />
+( * " ! + * <br />
**!+(! "" "!%!" <br />
* 2""!1!"*""+<br />
#-<br />
<br />
* !"##+$ ( *1!<br />
+*""" "A<br />
"*",+& <br />
&<br />
<br />
τ ≡ : π ⋅<br />
'<br />
;<br />
; <br />
; <br />
,+&-5"*1! ≡ ⋅ ≡ ⋅ <br />
+*>!""?<br />
)<br />
<br />
∂(<br />
)<br />
; θ = ,<br />
;<br />
∂(<br />
θ = ,<br />
,+!" ; !" * θ ∂)<br />
θ<br />
;<br />
;<br />
∂)<br />
<br />
1 +"! <br />
≡ ⋅ ≡ ⋅ <br />
)<br />
<br />
∂θ<br />
)<br />
;<br />
<br />
( = ,<br />
;<br />
∂θ<br />
( = ,<br />
,"B B !*! (* "<br />
" " 5 + " 1 * " (<br />
* "" ! & * A " " ( +<br />
(!"*1!+*"""**! )"<br />
**!**"*1$ * "" "!@!+!<br />
C (#* +! C( ( + " A "<br />
" =))"!**!! +* <br />
<br />
, **))"!+-<br />
(<br />
∂)<br />
;<br />
<br />
τ<br />
<br />
<br />
)<br />
<br />
≡<br />
%<br />
<br />
⋅ <br />
(<br />
;<br />
;<br />
<br />
∂)
-<br />
= /- + .<br />
<br />
= ,- + <br />
/ 4 . " 4 , " 4 **( !<br />
!" 4<br />
<br />
,>!""?"))"!*+- 7 <br />
D<br />
: − <br />
7 ∂α<br />
<br />
D <br />
− −<br />
D <br />
;<br />
−<br />
D <br />
∂<br />
<br />
<br />
τ τ ∂<br />
τ τ τ ∂'<br />
<br />
= <br />
⋅ + <br />
⋅<br />
D <br />
<br />
− −<br />
D<br />
<br />
D<br />
∂<br />
<br />
: <br />
: <br />
∂<br />
7 ∂ <br />
<br />
−<br />
<br />
− ; −<br />
<br />
<br />
<br />
<br />
τ<br />
D<br />
<br />
τ<br />
τ<br />
<br />
τ<br />
<br />
τ<br />
<br />
∂θ<br />
<br />
<br />
D<br />
∂α<br />
<br />
D<br />
D<br />
D <br />
∂<br />
7<br />
; ∂<br />
∂<br />
=<br />
+<br />
<br />
⋅ ; ; ; ;<br />
<br />
<br />
⋅ '<br />
<br />
D<br />
D<br />
∂<br />
∂<br />
<br />
D<br />
; 7 ; ; ; ; ∂<br />
<br />
<br />
<br />
<br />
D<br />
<br />
∂θ<br />
<br />
,""" " &**<br />
*!***"*!E "++"""(<br />
! ' ))"!(4"E +-<br />
<br />
D<br />
: − <br />
7 ∂α<br />
<br />
D <br />
− −<br />
D <br />
;<br />
−<br />
D <br />
∂<br />
<br />
<br />
τ τ ∂<br />
τ τ τ ∂'<br />
<br />
= <br />
⋅ + <br />
⋅<br />
D <br />
<br />
− −<br />
D<br />
<br />
D<br />
∂<br />
<br />
: <br />
: <br />
∂<br />
7 ∂ <br />
<br />
−<br />
<br />
− ; −<br />
<br />
<br />
<br />
<br />
τ<br />
<br />
D<br />
<br />
τ<br />
τ<br />
<br />
τ<br />
<br />
τ<br />
<br />
∂θ<br />
<br />
<br />
D<br />
D<br />
∂α<br />
<br />
∂<br />
<br />
7 ; ;<br />
; ; ; <br />
D<br />
D <br />
D ∂<br />
<br />
<br />
∂'<br />
<br />
= ⋅ +<br />
⋅ <br />
∂<br />
<br />
; 7<br />
D <br />
; ; ; ;<br />
<br />
∂<br />
<br />
<br />
D<br />
∂ <br />
<br />
D<br />
<br />
<br />
<br />
<br />
∂)<br />
<br />
<br />
− <br />
<br />
; ; − + <br />
<br />
<br />
D<br />
<br />
∂θ<br />
<br />
<br />
<br />
!!)#<br />
<br />
)! "" ! !" 1 + " * (! = " <br />
" ! !* ! !" " ; , "<br />
"B B +"1 ",+&&
.-64" *=" <br />
<br />
<br />
.+"* (-%; ;7-1 ; 8F<br />
("/78 ,*"$ * (<br />
7;G*1!; ;7" " (8 FG ,*<br />
" 78 ( "! * 1 % 7; *<br />
" (" *H <br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
,+&&-!+"!<br />
!"" ; <br />
&!<br />
"" <br />
<br />
*" ""<br />
" "<br />
τ = 8 A τ<br />
<br />
= HI <br />
= −: IF = −8 A7<br />
<br />
<br />
= 8 I = A A<br />
*<br />
<br />
% $%& ! 56%)6 + " (! !"+ * <br />
"" !(, !+ <br />
!"##+$, !"+*! !"!*(!"" <br />
)+1 **! !"" *"<br />
" + "" *! (! ! ( (<br />
!!! <br />
<br />
, * *! + " "" " (! <br />
( ,! !*+"*+ <br />
<br />
. **! *"<br />
*"<br />
* !**** 1 ,*<br />
+" *! <br />
<br />
)!! "! ! !*(#* "<br />
+** 1* + <br />
,!"!+"* (!-<br />
<br />
*$'<br />
<br />
, ! ((#<br />
<br />
<br />
*"+*(#* &!<br />
(#*<br />
"" !(#* <br />
!(#*<br />
-<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
= + + ) (!(#*+++(<br />
<br />
. * + ! ( ! ! *<br />
,<br />
+ ( " * ! + "? + **! (<br />
!!"+(#!*<br />
<br />
<br />
<br />
F
%" (( 4"! ,<br />
!( *"? ,(# !*+<br />
! ! <br />
<br />
<br />
% " ! ( ! <br />
&*!*!""" " !++<br />
*<br />
"(!*"!++- = − ,"<br />
!!<br />
+ ! (<br />
!""! ((#*<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
#!!( !!"!+<br />
= + + + + <br />
<br />
<br />
<br />
<br />
<br />
**! ( <br />
<br />
<br />
<br />
<br />
<br />
-<br />
<br />
<br />
<br />
<br />
<br />
<br />
+ % <br />
% #*! (#*!"<br />
<br />
<br />
!+!(<br />
<br />
*!<br />
<br />
.!!! !*!++"+!!<br />
! * + " + ( ! ! <br />
== """ !1 -<br />
∂Φ 7<br />
:<br />
= −ρ<br />
− ρ( ∇φ<br />
⋅∇φ<br />
−<br />
<br />
) − ρ<br />
+ <br />
<br />
∂<br />
:<br />
φ +!" ,+"!+""-<br />
<br />
7 ,!"""2((*(*!<br />
"!"" +- − ρ + J<br />
7<br />
:<br />
: , ! " − ρ( ∇φ<br />
⋅∇φ<br />
− )<br />
:<br />
<br />
!+ " " <br />
(!"" " J<br />
∂Φ<br />
," − ρ " * (* <br />
∂<br />
7<br />
:<br />
< ( ! − ρ( ∇φ<br />
⋅∇φ<br />
−<br />
<br />
) !" ! %"" 2<br />
:<br />
∂Φ<br />
!+!**! − ρ = ρζ<br />
<br />
∂<br />
,"* +- = + ρ −<br />
+ ζ $**<br />
***!*" *""!<br />
<br />
**''+<br />
<br />
,! >!""?+! !*(** 1!+<br />
! )!"*"* **!"<br />
*1+" ,"*(1" <br />
+-<br />
<br />
+ = % 7<br />
− <br />
− <br />
(<br />
<br />
)<br />
<br />
%<br />
<br />
<br />
,"?!+"( !*!<br />
,"" . "" +"" " ,"!*!<br />
. " " *!++*! "?" % <br />
<br />
H
"!!*!
=! ! !#" !(-"! +<br />
( $&$0 ! (! , * ! <br />
(->! **/"*!#?)&)0>! *E *!#?<br />
$&)0>! *E +"!?)&)0>*+!#! ?<br />
<br />
= ! " "! ! 5& ! &" <br />
! **! +( ! * + +<br />
! &!*! "*! *!#K <br />
' ! "!"*! ! "!,<br />
!*"!! !" ! ," *""!+<br />
!! "<br />
<br />
''.('$<br />
,! "" ""! +"*">! **/"<br />
*!#?. ,>! **/"*!#?/! # #-<br />
= −<br />
<br />
<br />
<br />
<br />
<br />
0 = )<br />
<br />
+ ) )<br />
+<br />
<br />
<br />
(E " E E ! !**! <br />
,1 -<br />
<br />
<br />
'<br />
<br />
= )<br />
<br />
⋅ 0 = )<br />
)<br />
<br />
+ ) + )<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
/-0' ('$<br />
<br />
/1023'$<br />
<br />
<br />
'4 ('$<br />
& ! ! ! !" * ! E L M >! !"?<br />
%4" (.A 0+!#!!!*<br />
*! ," !( ! (&(* !<br />
*( "(( !!<br />
<br />
&*E #" *+!#4 "(!+"" * +<br />
*+!#!!! +"" <br />
<br />
("!!E ( + <br />
!""" "!! " %$&)0""!<br />
) <br />
! "+ .F = !+ E ( N "<br />
"!+! ( 1!#* !E ! +!"*,<br />
! (45&!**! ,! -<br />
<br />
= − <br />
<br />
<br />
<br />
)<br />
<br />
<br />
)<br />
<br />
<br />
= )<br />
<br />
0 = )<br />
<br />
<br />
<br />
<br />
<br />
⋅ <br />
)<br />
<br />
<br />
<br />
− )<br />
⋅ <br />
<br />
)<br />
+ )<br />
<br />
<br />
<br />
+ )<br />
<br />
⋅<br />
<br />
)<br />
<br />
<br />
<br />
<br />
⋅<br />
<br />
<br />
)<br />
+ )<br />
<br />
+ )<br />
<br />
<br />
)<br />
<br />
⋅<br />
<br />
<br />
<br />
<br />
<br />
<br />
)<br />
<br />
⋅<br />
<br />
<br />
+ <br />
<br />
.F- *E N +*!#<br />
'4 (5<br />
' E ! "!""*+ %)&)0! <br />
(.H!+#""" <br />
<br />
<br />
.H- *E +"!<br />
<br />
.I-*+!#! <br />
<br />
<br />
87
,! # #-<br />
)<br />
= )<br />
<br />
−<br />
<br />
)<br />
<br />
<br />
<br />
θ = ) ⋅ + θ<br />
<br />
<br />
<br />
)<br />
<br />
<br />
@ "" "!* *"*"" "!# ,<br />
**! ! ( ! "! . * "!<br />
!! %"!!+ <br />
"!!<br />
<br />
6'$<br />
," **! ?*+!#! ?,!(!"!<br />
!*!#,"!1!)!"" 1<br />
( (+4"!*"( !!% <br />
*"( *!"+ .I(<br />
"*! <br />
<br />
,&<br />
<br />
, 2""+*""+<br />
*1! &!+*""+ *<br />
/ !" "! + * 2<br />
"" , + + # !" ( 2 ( <br />
&" " "!" <br />
<br />
, !" + !" ( ! "! * ( <br />
! !! * !+ * ( *1! + , !<br />
! ( .7; (! +( ( *1! <br />
!*1!!"" 7 ,""#"""4!*4<br />
""*"" <br />
<br />
.7;-' !*1! O 7 <br />
. *( * *! * (# * +! " . "" E * <br />
!"(.77 0 ! 2<br />
1 )-<br />
• , " **!"++( +! "" <br />
1$ "" **"( **!"" <br />
, 2 "(1!**! <br />
<br />
8:
• ,(!*. "" !" (##!!!<br />
! ,!+**<br />
!("!=" <br />
• , 56%)6 !" " ! ! <br />
(#,!>2?"!**!<br />
=" !!<br />
• ,*I; (*(4.77 <br />
$(#!!# *>2?"!*<br />
; 7H; <br />
&("+"!**!**"1"" "!<br />
! (#* +! ! " E . "!<br />
!!!+("+ !!!""( ,<br />
*"!**!("E (.7: ! **!<br />
!**! E ( *1! , **! * (#* +! E <br />
! ( ( ! *1! , ! + 4" * (- ( *1!<br />
+ (!"( *!*!"<br />
& **!"( (<br />
"!"**! %"*1!"<br />
*!(#* <br />
! *!#"+ <br />
<br />
<br />
D<br />
∂<br />
<br />
.77-,**!* 2/ *- <br />
D<br />
∂<br />
D<br />
∂% <br />
( **! * (# */" $ - ( **! * (#<br />
D<br />
∂<br />
<br />
8
∂<br />
. "" -<br />
∂<br />
D<br />
) <br />
D<br />
(**!E *(# <br />
N ! * !# ! + ! (#* +! ! +<br />
!"*+1*"+"" "! <br />
N1!*"!( *1!<br />
P; 7*!E !.7:
- '<br />
<br />
,* (*! "" -<br />
• *+!#! .IJ<br />
• ! **"+*!#J<br />
• ! *E N +*!#.FJ<br />
• ! *E +"!.H <br />
&!!"*!**! " +*<br />
E <br />
<br />
% ! * ! *1! " (! ! *1! <br />
"+ ! >! * ? = ! ( *1! +<br />
!+; 7H; !*1!(+ )""<br />
#"""4! #"!* <br />
<br />
- )<br />
<br />
K t<br />
variations<br />
normalised by K variations of n control<br />
t shaft<br />
1.6<br />
1.4<br />
1.2<br />
1<br />
0.8<br />
0.6<br />
0.4<br />
n shaft<br />
control by fuelrack<br />
no control<br />
K t<br />
& V s<br />
control by fuelrack<br />
0.2<br />
10 −2 10 −1 ω [rad/s]<br />
enc,(1−w)<br />
10 0<br />
.7-E (*! "<br />
,*(.7 64!"*; ;7"! **<br />
" * E > *+!# ! ? >E N ! ? ( <br />
!"+ + 6"! ; 7 *! ! * E / <br />
! . E L N / ! ( 4"! + " * *<br />
D<br />
∂)<br />
<br />
*! <br />
D<br />
∂'<br />
<br />
&*1! !E * %" 4" <br />
*! ! #" ! * ! *1! % <br />
*1! 1 **! + ,<br />
" !+"" 4"!<br />
**! <br />
<br />
% (*1!! *E N E *"! <br />
,"++!+!*N ! "E ! " % (<br />
!*1!"" " <br />
<br />
<br />
8
+"!(!" "!! !<br />
(#(!"!/"*7%"**"!"*E <br />
(!+&" !"!"!( +<br />
( + 4"! !<br />
*1!! *E +"!( *!!*E +<br />
D<br />
∂) **! <br />
D <br />
∂θ<br />
- !)<br />
<br />
, " ! ( ( .78 , * ( #<br />
+#!!-<br />
• !*1!+"/!"*7/( J<br />
• * ω<br />
7−<br />
<br />
,!!#"""<br />
""4!!; 7H; % !<br />
,+&&&J<br />
• ! * E + "! "+ ( # " & * (<br />
( E ! G & ( 7 E <br />
!F;G &"2"!!+ <br />
!(J<br />
• 4"! (! * " +*!#+"*<br />
!"*( <br />
<br />
.78-E (+(!*1!<br />
<br />
<br />
,+&&&-% 2" !!"" ; <br />
0" !<br />
)8;;; " P ; <br />
)A;;; " P ; <br />
)87H; " P ; <br />
ω<br />
7−<br />
<br />
<br />
; :8<br />
; :F<br />
7 7<br />
8A
1 <br />
<br />
& !! "" ! ! + **! * !#<br />
4" !**! "" + % <br />
*!!"!!!+*! " <br />
!"##+$ & * ( ! *1! " ! <br />
!**!+*"! *! +*!#<br />
! +"" "!! +! .*1!!<br />
!+(1("" "!! !**!<br />
! " <br />
<br />
, 2 "" " "! * *<br />
*! *1! , * *! ( * <br />
**!"+(! "!*!#,*<br />
2!*+*/ <br />
<br />
."!**!4"!*"!<br />
"" ! ( ! * "" # !! + *<br />
"!!*" , ! <br />
"" ! ( * !# + +! <br />
"*!! <br />
<br />
$<br />
<br />
,(#+"+""*"!5!<br />
>56%)6? (! % ( $%&<br />
!" , 56%)6 "! + "! * + *<br />
"" **/! ,""!* !#( <br />
<br />
)'<br />
<br />
),%56)$% :;;8 <br />
- E&$/5.)/:;;/7A8 & = <br />
<br />
<br />
N6E
!"#$$$<br />
%<br />
! " #$$$ !<br />
!! & ' ( <br />
) <br />
<br />
<br />
! "# <br />
<br />
!""# ""$%<br />
<br />
& "'! # (" "" " " "<br />
'" )! " *<br />
&'+""'*""'* !+<br />
#("""<br />
& "+ !'""<br />
"&++" '" <br />
'*<br />
<br />
!!! *""'+"!!'""++<br />
'"!!!'+!!<br />
" "* ! !!' ! "' <br />
' "'* * !","<br />
+"" ! -" ! "<br />
!"<br />
<br />
!!!'!'' !!** "<br />
)! "+ !!'"++!!"!<br />
'+"*)"!!!'!!!'<br />
!!'""+".+!*'*&/+<br />
"0*!/!'"01 ! !!!'<br />
"" !' ! * '+ , ""<br />
)"+<br />
<br />
"+ '+ !!'" "+ + ,! <br />
)+'!!' * ""+ )<br />
"""2 "'+' <br />
)""''+""'''"3'<br />
*+!4 !*'!.!!!'<br />
'* " "& " !" ) " 5*
!""+,+"""<br />
<br />
7 ! ! !!" " "!" ! ! <br />
"+")"' ''+<br />
!"!!!#"'+""<br />
! '+<br />
<br />
$%%<br />
<br />
7 " * 8998 8999 :4 ; .38;& " +<br />
"" 4'%<br />
<br />
o 1!!!'!+1!4<br />
o 5" !!!'+<br />
o 1 !'!"'*!!!'<br />
o 5 "< !<br />
<br />
1 " ' ! !!' + " !<br />
!' ! "' . ;! " :& <br />
"'" " !4"+ !<br />
!4+ "'*<br />
<br />
38; +"+!'+""'! "' "<br />
+"++" *!!!'"!'" +<br />
*'*! *"')* !' <br />
<br />
" ! " + " " * ""<br />
+""+!!<br />
<br />
&'()+ !" '+'<br />
'+38;' :" :<br />
!#'+')"*# !" + 4<br />
"!4:*#'+'."" +&<br />
5!!'+'"+!41!"<br />
#'+'+"+!!!:4 +'<br />
+ ! #'+ ' * ! #'+<br />
'"+:4!'"*" !!"!<br />
, " : " * * " '<br />
!*: ' *"!<br />
<br />
*+,) :4 + .& )" " ,!"" +<br />
'"*'- "38;:4<br />
"+(;2:4"+*
'+ "* ! '+ + ) "<br />
"" '+ 1 "" ! " ' <br />
#'+' ! !4!", '+<br />
"'""! )" !#'+'*<br />
<br />
)=+ 38;' +"<br />
". 1 1& "'+!!'<br />
+* #'+ '* :4 ) !<br />
.#'+$*:& " ! * 3 7 + <br />
" " ' + + '" ! +*<br />
! *!!'1="'+<br />
+ "1='"+ * " <br />
" * + "" )" " ! <br />
'+"!'* !'<br />
" !4 + +" * ' !4<br />
""! <br />
<br />
/* ! 00%<br />
<br />
/!!$111!0<br />
<br />
! < ' " !'" ' <br />
!! *!!'" "'*+<br />
' "" '* " '* .<br />
!'" 3 '* " ; * " " <br />
! '" !'"'&<br />
<br />
!!* * "!%""<br />
!! 4 ! ! " ,! " <br />
! '+ ! " ! " " " '*<br />
"!!'<br />
<br />
/$02 <br />
<br />
5 * ! * !! ! :"<br />
2".:2&'* " :!!'";<br />
! " 2*#"! ! <br />
*$#<br />
<br />
" + +* ! * " <br />
;B +
#!38;': 2;,++ *"<br />
'" ! * " ! !" <br />
! !4""*""!""<br />
-""'"<br />
<br />
4$,7-6<br />
<br />
""+""*!!!!' <br />
!'")*'!!!<br />
" ! + * + )*<br />
,'+F"!!) + !'""%<br />
<br />
-6 0 6 , ' !4 ' " <br />
!'""* '
o !4""+" "<br />
o !""+"""<br />
o $+'+'<br />
o #'!!'""+"<br />
o $+"!" <br />
<br />
'* "<br />
<br />
<br />
8,<br />
<br />
* """.38;38H3&!'" '*<br />
!"+!"* " , <br />
"'++<br />
""'+* '*"+"!4<br />
! "!!'""+<br />
<br />
7 ,! * . +* " & :4<br />
2"+-"*"'** !<br />
++" !*+' <br />
+'%<br />
o :<br />
o "".#:*!!*"&<br />
o ".'"'""""&<br />
o D<br />
o @'*+!!*<br />
<br />
)<br />
** !!"!"
+"!-"!!<br />
<br />
' + * ' " " <br />
++!!"' "!!'<br />
+;*! !4 #<br />
<br />
!!!!'" !3#<br />
";+L'+,'*M2@ ""<br />
! *"'!!+NN" "';<br />
' ' /0<br />
<br />
5"+'"3" +!!"<br />
' "*!(*"<br />
"#"' ""!!*<br />
!""""" "" !" <br />
"" ,<br />
<br />
$+ "+ "" < + 'O 1( !" " + <br />
@'@ ' "+ ' " " "+ <br />
""*" !@'@+!*"!"<br />
!!
"+5 !)*+"*"""/"0<br />
. !& " " " * $( ."<br />
'!& !!
;0 !!<br />
<br />
8999 ; !'"* !' +"<br />
"*"+' " !!<br />
" " " +" " <br />
""!" !'""'*+<br />
++"" '*"*"<br />
!!"! -<br />
<br />
!4'++ !"!! -<br />
!!'!'""++ *" <br />
!4"!!!'<br />
<br />
"'!"* !! 2<br />
!' ! ' !" + "<br />
# * " !!' ' " "* ' <br />
+'*"'"'<br />
<br />
<br />
<br />
R
!<br />
<br />
<br />
<br />
<br />
<br />
<br />
"#<br />
<br />
<br />
#$ <br />
%& <br />
% <br />
%&$ <br />
%<br />
<br />
% <br />
%#<br />
<br />
%<br />
<br />
'# <br />
(<br />
<br />
)<br />
<br />
* <br />
*++<br />
<br />
*$<br />
<br />
*#<br />
<br />
*<br />
<br />
*# <br />
*#<br />
<br />
*#<br />
<br />
*#<br />
<br />
,<br />
<br />
,<br />
<br />
, <br />
,<br />
<br />
,<br />
<br />
,$$<br />
<br />
- <br />
.<br />
<br />
.<br />
<br />
.<br />
<br />
.#<br />
<br />
.<br />
<br />
/ <br />
/#<br />
<br />
/<br />
<br />
/!+<br />
<br />
<br />
01<br />
<br />
01#<br />
<br />
0<br />
<br />
0 <br />
0 <br />
0! <br />
0!+<br />
<br />
23<br />
<br />
2#<br />
<br />
2## <br />
2$<br />
<br />
4<br />
<br />
4 <br />
45 <br />
4+<br />
<br />
4 <br />
61<br />
<br />
6$<br />
<br />
6 <br />
61<br />
<br />
6 <br />
6#15<br />
<br />
7<br />
<br />
7$ <br />
7$#<br />
<br />
7$# <br />
7<br />
<br />
7&<br />
<br />
7#<br />
<br />
7+$+<br />
<br />
8<br />
<br />
8<br />
<br />
95<br />
<br />
9+<br />
<br />
:1<br />
<br />
:<br />
<br />
:<br />
<br />
:<br />
<br />
:1<br />
<br />
; <br />
; <br />
; <br />
; # <br />
; <br />
<strong>COMPIT</strong>'06<br />
!"" #$$%<br />
<br />
&<br />
<br />
<br />
!" <br />
!#<br />
<br />
$% <br />
<br />
<br />
<br />
<br />
<br />
$&' $'<br />
<br />
&<br />
<br />
'& -.<br />
<br />
()$*+ ,*# <br />
/$<br />
01 + $2<br />
11 2)$)<br />
)30 %$*<br />
<br />
<br />
<br />
(&<br />
<br />
<br />
4#030 +" $*<br />
*# +3$*5<br />
+$1<br />
*# $<br />
3 ' $ 67 <br />
# 0%$890-8: ;.<br />
; ##-:.;<br />
www.oudpoelgeest.nl<br />
<br />
<br />
<br />
) & # #3 ;<br />
3 /##69<br />
<br />
& ?:;:>;8::9 @3A <br />
"#*$"*#$$% + ' <br />
69;:6;8::B <br />
:7;:?;8::B <br />
88;:?;8::B <br />
<br />
)& $$, -.<br />
<br />
&h.t.grimmelius@wbmt.tudelft.nl<br />
www.3ME.TUDelft.nl/<strong>COMPIT</strong>06 <br />
<br />
-&C