EVALUATION OF MULTI-ELEMENT WING SAIL
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
5 th High Performance Yacht Design Conference<br />
Auckland, 10-12 March, 2015<br />
<strong>EVALUATION</strong> <strong>OF</strong> <strong>MULTI</strong>-<strong>ELEMENT</strong> <strong>WING</strong> <strong>SAIL</strong> AERODYNAMICS FROM<br />
TWO-DIMENSIONAL WIND TUNNEL INVESTIGATIONS<br />
Alexander W. Blakeley 1 , abla089@aucklanduni.ac.nz<br />
Richard G.J. Flay 2 , r.flay@auckland.ac.nz<br />
Hiroyuki Furukawa 3 , furukawa@meijo-u.ac.jp<br />
Peter J. Richards 4 , pj.richards@auckland.ac.nz<br />
Abstract. Following the 33rd America's Cup which featured a trimaran versus a catamaran, and the recent 34th America's Cup in<br />
2013 featuring AC72 catamarans with multi-element wing sail yachts sailing at unprecedented speeds, interest in wing sail<br />
technology has increased substantially. Unfortunately there is currently very little open peer-reviewed literature available with a<br />
focus on multi-element wing design for yachts. The limited available literature focuses primarily on the structures of wings and their<br />
control, rather than on the aerodynamic design. While there is substantial available literature on the aerodynamic properties of<br />
aircraft wings, the differences in the flow domains between aeroplanes and yachts is significant. A yacht sail will operate in a<br />
Reynolds number range of 0.2 to 8 million while aircraft operate regularly in excess of 10 million. Furthermore, yachts operate in the<br />
turbulent atmospheric boundary layer and require high maximum lift coefficients at many apparent wind angles, and minimising drag<br />
is not so critical. This paper reviews the literature on wing sail design for high performance yachts and discusses the results of wind<br />
tunnel testing at the Yacht Research Unit at the University of Auckland. Two wings with different symmetrical profiles have been<br />
tested at low Reynolds numbers with surface pressure measurements to measure the effect of gap geometry, angle of attack and<br />
camber on a wing sail’s performance characteristic. It has been found that for the two element wing studied, the gap size and pivot<br />
point of the rear element have only a weak influence on the lift and drag coefficients. Reynolds number has a strong effect on<br />
separation for highly cambered foils.<br />
NOMENCLATURE<br />
AWA<br />
c<br />
C l<br />
C d<br />
D<br />
g<br />
L<br />
R<br />
S<br />
TWA<br />
V A<br />
V mg<br />
V S<br />
V T<br />
<br />
<br />
<br />
<br />
A<br />
H<br />
<br />
Apparent wind angle<br />
Chord<br />
Lift coefficient<br />
Drag coefficient<br />
Drag on wing and any sails<br />
Gap<br />
Lift on wing and any sails<br />
Resistance of hull and appendages<br />
Side force on hull and appendages<br />
True wind angle<br />
Apparent wind velocity<br />
Velocity made good<br />
Yacht velocity<br />
True wind velocity<br />
Angle of attack of front element<br />
Angle between rear and front elements<br />
Apparent wind angle<br />
True wind angle<br />
Aerodynamic drag angle<br />
Hydrodynamic drag angle<br />
Leeway angle<br />
1. INTRODUCTION<br />
Following the 33rd America's Cup in 2010 which<br />
featured a trimaran versus a catamaran, and the recent<br />
34th America's Cup in 2013 featuring AC72 catamarans<br />
with multi-element wing sail yachts sailing at<br />
unprecedented speeds, interest in wing sail technology<br />
has increased substantially. Unfortunately there is<br />
currently very little open peer-reviewed literature<br />
available with a focus on multi-element wing design for<br />
yachts. Hence a wind tunnel study of multi-element wing<br />
sails was carried out by the University of Auckland.<br />
Some preliminary results of this research are available in<br />
[1].<br />
Solid wing sails offer several advantages over flexible<br />
sails for high speed sailing. One advantage is that the<br />
wing has an internal structure which is used to give the<br />
wing its shape. The shape is not dependent on the tension<br />
in the lines at the sail corners. The shape of a flexible sail<br />
cloth mainsail is highly dependent on the tension in the<br />
mainsheet, particular the vertical component which is<br />
required to keep the leech from twisting excessively. In a<br />
wing sail, this vertical component is eliminated<br />
completely from the mainsheet, which is used simply to<br />
alter the angle of attack using a horizontal force. The<br />
twist in the solid sail is controlled by internal control<br />
lines which are under considerably less tension than a<br />
conventional mainsheet. Thus the angle of attack can be<br />
changed relatively easily and quickly, and the power of<br />
the sail can be controlled much more easily than in a<br />
conventional soft sail. This makes sailing a high<br />
performance catamaran much easier with a wing sail than<br />
a soft flexible sail.<br />
1 PhD Student, Yacht Research Unit, Department of Mechanical Engineering, University of Auckland, NZ<br />
2 Professor, Director Yacht Research Unit, Department of Mechanical Engineering, University of Auckland, NZ<br />
3 Associate Professor, Department of Mechanical Engineering, Meijo University, Japan<br />
4 Associate Professor, Department of Mechanical Engineering, University of Auckland, NZ<br />
37
Multi-hull yachts have considerably less hydrodynamic<br />
drag than mono-hulls since they have much lower<br />
displacements than monohulls of similar length. This is<br />
because they derive their righting moment from their<br />
wide beam, and not from having a heavy keel.<br />
Furthermore they heel only a small amount. A dangerous<br />
feature of multi-hulls is that once the overturning<br />
moment exceeds their maximum righting moment, they<br />
will overturn unless the overturning moment is reduced<br />
quickly. When coupled with a high performing solid sail<br />
or wing, this means that they are able to sail with small<br />
apparent wind angles both upwind and downwind as<br />
illustrated in Error! Reference source not found.. It<br />
was shown by Lanchester [2] that the minimum angle<br />
( A ) at which a boat can shape its course relative to the<br />
wind is the sum of the under and above water drag<br />
angles, namely H and A . A wing sail operating at low<br />
angles of attack has a low aerodynamic drag angle, and<br />
so the result of this is that high performance multi-hulls<br />
with wing sails operate at low apparent wind angles in<br />
both upwind and downwind sailing. These angles are<br />
illustrated in Error! Reference source not found.. This<br />
means that it is only necessary to have aerodynamic data<br />
for wing sails for low angles of attack.<br />
V A<br />
A<br />
V S<br />
Upwind<br />
V T<br />
T<br />
V mg<br />
V mg<br />
Figure 1 Velocity triangles for up and downwind sailing<br />
showing the low apparent wind angle A which occurs for<br />
fast multi-hulls with wing sails.<br />
Understanding the performance of multi-hulls can be<br />
enhanced by analysing a free-body diagram of the yacht<br />
and considering the forces in the horizontal plane. This is<br />
close to reality, since they do not heel much. The<br />
aerodynamic and hydrodynamic forces on a catamaran<br />
are shown in Error! Reference source not found.. Sail<br />
forces are lift (L) and (D) being normal and parallel to<br />
the apparent wind direction, and hull/appendage forces<br />
are sideforce (S) and resistance (R) which are normal and<br />
parallel to the yacht’s course through the water. There<br />
may be a small leeway angle (). Note that for<br />
equilibrium, the hydrodynamic and aerodynamic forces<br />
in the plane of the water must be equal and opposite, so<br />
. Furthermore, as noted above,<br />
. Reducing D and R lead to reductions in<br />
and and thus to a reduction in AWA, which will lead<br />
V S<br />
V A<br />
A<br />
Downwind<br />
V T<br />
T<br />
to increased velocity made good (V mg ) to the top or<br />
bottom mark.<br />
S<br />
H<br />
Error! Reference source not found.Figure 2 Sketch of<br />
horizontal forces acting on a catamaran showing the<br />
important directions.<br />
Ignoring the structural aspects and considering primarily<br />
the aerodynamic aspects, high speed multi-hull yachts<br />
with wing sails need wings that are symmetrical so that<br />
they can tack and gybe at will. They need to be able to<br />
develop high lift and have low drag. They need to be able<br />
to twist so that the height of the thrust can be lowered<br />
when required to reduce the overturning moment. From<br />
observations of existing vessels, it is apparent that severe<br />
twist is important so that the direction of the lift force can<br />
be reversed at the top of the wing to provide a righting<br />
moment. Generally it is apparent from the 2013<br />
America’s Cup yachts that for a two-element wing, the<br />
chords of the main wing and flap are usually of similar<br />
length, and the gap between them is relatively small. The<br />
first element is thick so that it can take the required<br />
structural loads, while the flap is thin. Typical thickness<br />
ratios are 25% and 9% for the first and second elements<br />
respectively.<br />
2. SOME THEORETICAL CONSIDERATIONS<br />
By applying the sine rule to the velocity triangles in<br />
Figure 1 it is evident that<br />
The velocity made good is given by<br />
R<br />
, so (1)<br />
(2)<br />
(3)<br />
Thus to improve performance (i.e. increase the ratios<br />
and ) at a fixed heading (i.e. fixed ) it is necessary to<br />
reduce . Figure 3 shows curves of the ratio for<br />
constant and for various , which are evidently<br />
circular arcs. Figure 3 shows clearly how smaller<br />
<br />
L<br />
D<br />
A<br />
38
gives more speed. Maximum yacht speed occurs when<br />
and it is the diameter of the circule. The<br />
maximum velocity made good is determined from the<br />
horizontal tangents to the circle, and occurs for the<br />
angles and for upwind and<br />
downwind respectively.<br />
Thus the importance of being able to determine and<br />
minimise the aerodynamic drag angle has been<br />
demonstrated. Wing performance between candidate<br />
options can therefore be compared on the basis of the<br />
aerodynamic drag angle as it influences directly.<br />
1<br />
0.5<br />
0<br />
‐0.5<br />
V T<br />
‐1<br />
‐1.5<br />
‐2<br />
T<br />
V S<br />
A<br />
V A<br />
bA = 25 deg<br />
bA = 45 deg<br />
bA = 90 deg<br />
bA = 150 deg<br />
0 0.5 1 1.5 2 2.5<br />
Figure 3 Velocity triangles (speed polars) for fixed values of<br />
A for a range of T .<br />
3. LITERATURE REVIEW<br />
The limited available literature on wing sails focuses<br />
primarily on their structure and control, rather than on<br />
their aerodynamic design. While there is substantial<br />
available literature on the aerodynamic properties of<br />
aircraft wings, the differences in the flow domains<br />
between aeroplanes and yachts is significant. A yacht sail<br />
will operate in a Reynolds number range of 0.2 to 8<br />
million while aircraft operate regularly in excess of 10<br />
million. Furthermore, yachts operate in the turbulent<br />
atmospheric boundary layer and require high maximum<br />
lift coefficients at many apparent wind angles. A brief<br />
review of some of the literature on wing sails is<br />
presented below.<br />
Elkaim [3] at the University of California undertook a<br />
large amount of work on wing sails while involved in the<br />
Atlantis Project, which focused on designing an<br />
Autonomous Marine Surface Vehicle for a wide range of<br />
applications. A rigid wing sail was used on this vehicle<br />
because it was regarded as more aerodynamically<br />
efficient than a soft sail, by pivoting it near the centre of<br />
pressure it required less force to actuate than a regular<br />
soft sail, and it could be designed to be self-trimming.<br />
Marchaj [4] reviewed some research on wing sails,<br />
pointing out their advantages in some situations. In<br />
discussing the 1976 C-Class race between Aquarius V<br />
with a soft sail, and Miss Nylex with an advanced wing<br />
sail, he pointed out the advantage of the soft sail with the<br />
lighter rig in light conditions, compared to the more<br />
aerodynamically efficient but heavier wing sail rig.<br />
Some wing sails have been made of double-surface soft<br />
sails, and are a good idea for cruising as they can be<br />
reefed and dropped when wind conditions are<br />
unfavourable, e.g. the Wally and Omer wing sail design<br />
[5]. They do not have all the benefits of rigid wing sails,<br />
e.g. the leech tension needs to be maintained, requiring a<br />
sophisticated boom and powerful mainsheet.<br />
C-Class catamaran racing has often produced very<br />
innovative ideas regarding wing sails. Examples are the<br />
slot gap on the yacht Cogito, described by MacLane [6],<br />
which could be changed onshore, but was fixed once<br />
racing. Killing and Clarke [7] describe further<br />
enhancements of C-Class wings for the yachts Alpha and<br />
Rocker, which had more advanced cross-sectional shapes<br />
than Cogito which were developed using software<br />
developed by Drela at MIT [8,9].<br />
There are maximum heeling and pitching moments that<br />
can be applied to catamarans before they tip over. Thus<br />
in strong wind conditions it is advantageous to lower the<br />
height of the line of action of the side and thrust forces.<br />
Thus it can be shown that the optimum lift distribution<br />
for a catamaran is such that it changes sign at the top of<br />
the wing, requiring large amounts of wing twist [10].<br />
Wing sails can be designed with mechanisms for<br />
producing such large amounts of twist, which is<br />
impossible for soft sails, and so this is another significant<br />
difference between wing and soft sails.<br />
One of the commonest areas of wing sail use is for highspeed<br />
land yacht sailing, more specifically ice yachts.<br />
These yachts skate on ice and are sometimes powered by<br />
large wing sails. They are capable of very high speeds.<br />
The Greenbird project is arguably one of the most<br />
successful versions of wind powered land/ice yachts. On<br />
land, the Greenbird holds the land speed record of 126.2<br />
mph, while on ice, they hope to beat the current record of<br />
84 mph by setting a speed equal or higher than the<br />
126.2mph that they set on land [11].<br />
Recently relatively small yachts have been<br />
experimenting with wing sails. The X-Wing sail project<br />
[12] is a 2-element wing designed for small sailing<br />
dinghies, specifically the “Sunfish” class. The designer<br />
was inspired by the 2010 America’s Cup and has<br />
designed a low-cost wing for those wanting to<br />
experiment with the technology.<br />
39
The literature cited above shows that work on wing sail<br />
development has taken place for a variety of reasons, but<br />
mainly to obtain high speeds or to improve control, and<br />
that developments have occurred in both small as well as<br />
larger craft. However, it is evident that there is a paucity<br />
of multi-element wing aerofoil data obtained specifically<br />
for yacht sail design. This paper discusses the results of<br />
wind tunnel testing at the Yacht Research Unit at the<br />
University of Auckland to help provide this information.<br />
4. CFD OPTIONS<br />
To analyse the flow around upwind sails, the potential<br />
flow assumption and the panel method are basic<br />
methods, but they cannot model viscous effects which<br />
are dominant in locations of flow separation. The Navier-<br />
Stokes equations can be solved directly (DNS) in cases<br />
with flow separation phenomena. When the Reynolds<br />
number is high and the flow becomes turbulent, it is<br />
difficult to use DNS because it needs an enormous<br />
number of grid points. The turbulence must be modelled<br />
with turbulence or sub-grid models. Reynolds Average<br />
Navier Stokes (RANS) turbulence models obtain a<br />
stationary solution that is representative of the mean<br />
flow. They are robust and easy to use, but cannot be<br />
adopted to unsteady flow. Large-eddy simulation (LES)<br />
is used to model the larger, three-dimensional unsteady<br />
turbulent motions. In LES, the smaller scale motions are<br />
filtered and subsequently modelled. LES can be expected<br />
to be more accurate and reliable than RANS models for<br />
flows where large-scale unsteadiness is significant.<br />
The program MSES [8] is a coupled viscid/inviscid Euler<br />
method. It solves the Euler equations on a discrete two<br />
dimensional grid, coupled with an integral boundary<br />
layer formation. MSES was created to analyse the flow<br />
around an aerofoil more accurately than inviscid codes<br />
by incorporating viscous effects and a boundary layer<br />
formation into its solutions. Therefore it theoretically<br />
offers more realistic results than a vortex lattice method<br />
(VLM), although it possibly requires more computational<br />
power. A derivation of MSES is the MISES [9] code,<br />
which has been developed to determine the flow around<br />
multi-element aerofoils.<br />
While CFD analysis of wing sails is highly advanced and<br />
capable of accurate predictions, it is not easy to predict<br />
sail performance if there is flow separation. As<br />
mentioned in the present paper, it is important for<br />
designers of wing sails to know where flow separation<br />
occurs, and physical testing such as wind tunnel testing<br />
can help in this regard. Wind tunnel testing can also give<br />
good information for checking the output from CFD<br />
codes, although they may be unable to give results for the<br />
very high Reynolds numbers appropriate for large yachts<br />
sailing very fast, e.g. America’s Cup AC72 yachts.<br />
However, the Reynolds numbers from wind tunnels<br />
would be commensurate to those for smaller C-class<br />
yacht wing sails.<br />
CFD analysis of wing sails is not considered further in<br />
the present paper.<br />
5. EXPERIMENTAL SETUP<br />
5.1 Wind Tunnel<br />
The experiments were performed in the University of<br />
Auckland open-return wind tunnel. The wind tunnel has<br />
been specifically designed for testing yacht sails and has<br />
a standard operating cross section of 7 m wide by 3.5 m<br />
high. The flow is produced by two 3-m diameter 4-<br />
bladed fans and then driven through a 1 m thick<br />
honeycomb screen and two tight mesh screens to remove<br />
swirl and to give a uniform velocity profile and relatively<br />
low turbulence flow. Permanent pitot-static probes set up<br />
well upstream of the wing recorded the dynamic and<br />
static pressures whilst another probe measured the<br />
atmospheric pressure outside the working section.<br />
The wind tunnel walls were brought inwards to give a<br />
2.5m wide by 3.5m high test section and the wing models<br />
were located near the outlet of the nozzle as shown in<br />
Figure 4.<br />
Figure 4 Photograph of wing model in wind tunnel showing<br />
base of support struts and endplate marking for angle<br />
fixing.<br />
5.2 Multi-element Wing Construction<br />
The multi-element wing used in the wind tunnel<br />
comprised a NACA 0025 main element and a NACA<br />
0009 aft element with a 50:50 chord ratio. The wing<br />
spans 2.5 m across the width of the wind tunnel and has a<br />
chord length of 1 m (with zero aft element deflection and<br />
no gap). 20 mm thick MDF boards at each end of the<br />
model contained the placement locations for the elements<br />
and tested configurations. Once attached to the wind<br />
tunnel walls the model can be rotated and then securely<br />
fixed to achieve various angles of attack. By constructing<br />
the model in this fashion the trailing vortices as<br />
experienced by a finite spanned wing are virtually<br />
eliminated.<br />
40
Each element has 64 independent pressure taps which<br />
measure pressure differentials through 1 mm diameter<br />
holes placed around the centreline of the windward and<br />
leeward surface of each wing, sufficiently far away from<br />
the wind tunnel side-wall boundary layers. They are<br />
dispersed from the leading to the trailing edge with a<br />
higher concentration around the leading edges of the<br />
elements. The pressure taps can be seen in Figure 5.<br />
They are slightly off-set from the stream-wise direction<br />
to avoid interference from wakes of upstream holes<br />
interfering with downstream taps. Tubes connecting to<br />
each individual pressure tap feed out an end of each wing<br />
to a box containing pressure transducers. The pressures<br />
were logged at a rate of 200 Hz over a 30-s period. The<br />
pressure transducer box was located downstream of the<br />
wing out of the free-stream.<br />
5.3 Experimentation Variables<br />
The variables altered during the wind tunnel testing were<br />
as follows:<br />
• Front wing model angle of attack, relative to<br />
main element chord<br />
• Aft element chord-line deflection, , relative to<br />
main element chord<br />
• Gap between main and aft element, g, defined at<br />
zero aft deflection angle , as a % of the main<br />
wing chord<br />
• Aft element pivot point position, as a % of the<br />
main wing chord.<br />
• Free stream flow velocity, V.<br />
Figure 6 shows these variables with respect to the model.<br />
It also defines the pivot location for the aft element and<br />
defines the gap between the foils.<br />
Figure 5 Close up photograph of leading edge of a wing<br />
model showing the 1 mm diameter pressure tap holes.<br />
The pressure measurements, along with the locations of<br />
the pressure taps can be used to obtain a pressure<br />
distribution across the surface of each element.<br />
Aerodynamic forces result from these pressure<br />
distributions acting over the surface of the wings. By<br />
integrating the pressure distributions in the vertical and<br />
horizontal directions, the vertical and horizontal forces<br />
can be computed, respectively. Since pressure acts<br />
normal to a surface, the surface curvature of the aerofoil<br />
between two points should not be neglected. However,<br />
the pressure taps were distributed closely together around<br />
regions of high curvature on the wing elements and<br />
therefore a straight line approximation of the surface<br />
between any two pressure taps was regarded as<br />
sufficiently accurate for the present work. The integrated<br />
x and y forces were firstly resolved into directions<br />
relative to the chord of the aerofoil. Therefore their<br />
resultants were then decomposed appropriately to find<br />
the lift and drag relative to the free-stream velocity.<br />
Finally non-dimensional representations of the<br />
coefficients of pressure (Cp), lift (Cl), and drag (Cd)<br />
were computed.<br />
Figure 6 Wing variables altered in experiment.<br />
Reynolds number Re is defined by the reference length<br />
2c (=1.0 m), the reference velocity V and the kinematic<br />
viscosity for air.<br />
41
6. RESULTS AND DISCUSSION<br />
6.2 The influence of angle of attack<br />
6.1 Pressure coefficients<br />
Figure 7 Example of pressure coefficient distribution. =<br />
10°, g = 1%, = 2°, Re = 800,000.<br />
The shapes of the main and aft elements are NACA0025<br />
and NACA0009, respectively [13]. The chord of the<br />
main element c1 is 500 mm, and it is same as the chord<br />
of the aft element c2. The total wing chord c = c1 + c2 =<br />
1 m.<br />
Figure 7 shows an example of the pressure coefficient<br />
variation over the foils. In Figure 7 the deflection angle<br />
of the aft element is 10°, the gap between main and aft<br />
elements is 1% referenced to the main element chord c1,<br />
the angle of attack, is 2° and Reynolds number based<br />
on the total wing chord c as the reference length is<br />
800,000.<br />
Figure 8 Example of stall. = 10°, g = 1%, Re = 700,000.<br />
Figure 8 shows the distributions of the lift coefficient C l<br />
and the drag coefficient C d versus angles of attack C l<br />
increases almost linearly with increase in up to 10° and<br />
C d also increases a little. C l reaches a maximum value at<br />
around = 10°. At larger angles of attack the wing stalls<br />
and C l decreases sharply and C d increases rapidly, as<br />
expected.<br />
6.3 The influence of aft element deflection<br />
The horizontal axis of the graph indicates the<br />
dimensionless distance along the chord direction. The<br />
vertical axis shows the pressure coefficient. The position<br />
x/c = 0 corresponds to the leading edge of the main<br />
element, and x/c = 0.5 indicates the trailing edge of the<br />
main element and the leading edge of the aft element.<br />
The vertical axis has been inverted with negative values<br />
at the top. Thus the red line indicates the pressure<br />
coefficient at the top of the elements, and those along the<br />
bottom are shown by the blue line. Like ordinary wings,<br />
the pressure at the top is negative, and lift occurs due to<br />
the pressure difference between the top and bottom of<br />
elements. The distance between the lines is indicative of<br />
the normal force acting on each element.<br />
= 0<br />
= 0<br />
= 15<br />
= 15<br />
= 30<br />
= 30<br />
Figure 9 C l and C d versus for varying aft element<br />
deflection. g = 2%, Re = 700,000.<br />
42
= 0<br />
= 15<br />
= 30<br />
= 25<br />
Figure 10 Pressure coefficient distribution for varying .<br />
g = 2%, Re = 700,000, = -1°.<br />
Figure 9 shows C l and C d versus for varying aft<br />
element deflections. Deflecting the aft element<br />
downwards increases C l substantially, although the<br />
values of C l for = 15° and 30° are almost the same. The<br />
gradients for each are nearly identical. The stall angle<br />
for at high aft element deflections (15°) is lower<br />
than that at = 0°. The drag coefficients are substantially<br />
higher when the aft element is deflected at angles of 15°<br />
and 30° compared to = 0°. It can be seen that<br />
increasing the flap angle from = 15° to = 30°<br />
produces almost no increase in lift and a significant<br />
increase in the drag.<br />
To show why the lift is so low for = 30°, Figure 10<br />
compares the pressure distributions for three different<br />
flap deflection angles for = -1°. At = 30°, the<br />
pressure coefficients are smaller than for = 0° and 15°,<br />
and the difference between the values at the upper and<br />
lower surface is small. In particular, it can be seen that<br />
the large aft element deflection of = 30° produces<br />
virtually no pressure difference across it, as well as<br />
virtually no pressure difference across the front element<br />
as well.<br />
Figure 11 C l and C d versus for varying aft element<br />
deflection. g = 2%, Re = 300,000, pivot = 80%.<br />
Figure 11 shows another example when the cambers are<br />
varied. C l at ° is larger than that at ° over<br />
whole rangebefore the flow stalls. C l at ° is<br />
larger than that at ° when is small, but the slope<br />
of C l -a curve for ° is lower than the other curves,<br />
and C l at ° becomes larger beyond °This<br />
means C l has a local maximum at a certain camber.<br />
6.4 The influence of gap size<br />
Figure 12 shows a comparison of wing performance for<br />
different gap sizes. As can be seen in the figure, the<br />
distributions of C l and C d for each gap are almost the<br />
same and the only small difference is that g = 1% gives a<br />
slightly higher C l at = 10°. In other tests in this study it<br />
has been observed that the influence of gap size is very<br />
small.<br />
43
Figure 14 C l and C d versus for varying gap size for<br />
negative gaps, = 15°, pivot = 90%, Re = 700,000.<br />
Figure 12 C l and C d versus for varying gap size for = 10°<br />
and Re = 700,000.<br />
Figure 14 shows lift and drag coefficient result when the<br />
gap is negative, i.e. the main and aft elements are<br />
partially overlapping. Note that this is only physically<br />
possible for non-zero values of . While C d at g = 0 and<br />
at g = -2% are amost identical, C l at g = 0 is larger than<br />
that for g = -2% near the angle where the flow stalls.<br />
Thus negative gaps do not seem to be beneficial.<br />
6.5 The influence of aft element pivot point position<br />
Figure 13 C l and C d versus for varying gap size for =<br />
10°, Re = 300,000.<br />
Figure 13 shows the effect of varying gap size at a<br />
smaller Reynolds number. While the flow stalls at<br />
° in Figure 12, C l begins to decrease at a smaller <br />
in Figure 13. Thus a smaller Reynolds number causes<br />
stall to occur at a smaller <br />
Figure 15 C l and C d versus for varying pivot points for =<br />
10°, g = 0%, Re = 700,000.<br />
The pivot point is the location about which the aft<br />
element rotates as shown in Figure 6. It is defined by the<br />
distance aft of the leading edge of the main wing in terms<br />
of % main element chord. Thus for example, when the<br />
pivot point is 0%, it means that the pivot point is at the<br />
leading edge of the main element, and 100% means that<br />
44
pivot point is at the trailing edge of the main element.<br />
Figure 15 shows results for various pivot points. g and<br />
Re are all constant. It can be seen that C l at pivot = 40%<br />
is slightly larger than other results, but otherwise there is<br />
little obvious effect, even for the very large changes of<br />
position from 40% to 90%.<br />
Figure 17 shows a comparison of the C p variation when<br />
Re is varied. For Re = 800,000, the pressure coefficient<br />
curve is smooth. On the other hand, when Re is 200,000,<br />
the pressure coefficient has a step-like distribution at<br />
x/c = 0.18. The authors speculate that this step-like<br />
distribution at this low Reynolds number may be a<br />
laminar separation bubble. Furthermore, there may also<br />
be a laminar separation bubble on the pressure surface at<br />
x/c = 0.4. When the Reynolds number is larger, flow<br />
separation does not occur in these results and the curves<br />
are smooth. The pressure coefficient distributions on the<br />
aft elements are almost the same for both Reynolds<br />
numbers.<br />
Figure 16 C l and C d versus for varying pivot points for =<br />
10°, g = 2%, Re = 300,000.<br />
Figure 16 shows the results when Re is smaller<br />
(300,000). It can be seen that C l and C d are almost<br />
identical, but that the flow separates at smaller when<br />
the pivot point is 40% compared to 90%.<br />
Figure 18 C l and C d versus Re. = 10°, = 2°, g = 1%.<br />
6.6 The influence of Reynolds number<br />
Figure 19 C l and C d versus Re. = 10°, = 10°, g = 1%<br />
Figure 17 Pressure coefficient variation over the elements<br />
for two different Re, for = 10°, g = 1%, = 2°.<br />
Figure 18 and Figure 19 show C l and C d versus Reynolds<br />
number. In many cases in this study, C l and C d do not<br />
change significantly when the Reynolds number is varied<br />
as shown in Figure 18. On the other hand, when the aft<br />
45
element deflection and the angle of attack are large, C l<br />
and C d change with Re as indicated in Figure 19.<br />
Generally, C d increases with increase in C l as shown in<br />
Figure 11 and Figure 16.But, in Figure 19, C d becomes<br />
small when C l becomes large for this value of<br />
To analyse why this reduction in C d occurs, the<br />
pressure coefficients corresponding to Re = 400,000,<br />
500,000 and 600,000 in Figure 19 are shown in Figure<br />
20.<br />
has the same effect as roughening the surface of an<br />
aerofoil, and causes early transition, and thus the<br />
emulation of higher Reynolds number behaviour. Even<br />
the turbulence from a small upstream rod [14] has been<br />
found to be sufficient to substantially alter the separation<br />
behaviour of bluff bodies.<br />
For testing aerofoils like those in the present tests,<br />
turbulence in the onset flow will affect the location of<br />
transition, as will the presence of the pressure gradients<br />
on the top and bottom surfaces. It is expected that the<br />
tested wings will simulate the behaviour of C-class<br />
catamarans, which operate at similar Reynolds Numbers.<br />
It is not known with certainty in the present tests where<br />
transition was occurring, and whether or not bypass<br />
transition was present.<br />
The effect of the transition location on the behaviour of<br />
wings can be investigated by tripping the boundary layer<br />
artificially using a wire or similar, and it is planned to<br />
carry out such tests in the future.<br />
Re = 0.4 x 10 6<br />
Figure 20 Pressure coefficient variation over the elements<br />
for three different values of Re. = 10°, = 10°, g = 1%.<br />
The pressure coefficient over the top of the wing at Re =<br />
400,000 is almost constant downstream of the leading<br />
edge. This indicates that the flow has separated at the<br />
leading edge of the main element for this small Reynolds<br />
number. The wing is stalled over its entire surface, and<br />
thus C l becomes small and C d is large. When the<br />
Reynolds number is larger (Re = 600,000), no separation,<br />
or only a very small amount of flow separation occurs,<br />
and C d is again small. The pressure coefficient near the<br />
leading edge of the main element is significantly large,<br />
and C l is large. When the angle of attack and the aft<br />
element deflection are large, it is necessary to operate the<br />
wing at Reynolds numbers in excess of 600,000 in order<br />
to avoid separation and poor performance. Note that the<br />
pressure distribution at Re = 600,000 may show small<br />
laminar separation bubbles on the suction surfaces of<br />
both the fore and aft elements near their leading edges in<br />
the strongly adverse pressure gradient regions.<br />
The results presented in the paper were obtained with the<br />
two-element, 1-m chord wing mounted between the walls<br />
of the 2.5-m wide duct near the outlet of the open jet.<br />
The turbulence intensity of the flow was about 1%,<br />
which is much lower than that in the atmosphere, but is<br />
quite large compared to 0.1% in good quality<br />
aeronautical wind tunnels. Turbulence in the onset flow<br />
7. CONCLUSIONS<br />
In this study, the influence of geometric factors such as<br />
the aft element deflection, the gap and the aft element<br />
pivot point have been investigated in a wind tunnel<br />
study. From the study, the following conclusions can be<br />
drawn.<br />
C l and C d are significantly related to the aft element<br />
deflection. As the aft element deflection is increased, C l<br />
and C d also increase. It is the same aerodynamically as<br />
adding camber to a single element wing. The addition of<br />
aft element deflection increases C l , but too high a<br />
deflection results in an increase in drag.<br />
The influences of the gap size and the pivot point<br />
location are found to be relatively small in this study.<br />
The Reynolds number has a large influence on the<br />
sensitivity of the wing to separation. As expected, a<br />
smaller Reynolds number caused the flow to separate<br />
more easily, especially with high aft element deflections<br />
and high angles of attack.<br />
Acknowledgements<br />
We would like to recognise the following people who<br />
have contributed to this project. We thank Joseph Nihotte<br />
and James Turner for carrying out a large amount of the<br />
wind tunnel testing, and for analysing some of the<br />
results. We also thank David Le Pelley for his help and<br />
advice in setting up and overseeing the wind tunnel<br />
testing.<br />
References<br />
1. Blakeley, A.W., Flay, R.G.J. and Richards, P.J.,<br />
Design and optimisation of multi-element wing sails<br />
for multihull yachts, Proceedings of the 18th<br />
Re = 0.6 x 10 6 46
Australasian Fluid Mechanics Conference 3rd - 7th<br />
December 2012 Edited by P.A. Brandner and B.W.<br />
Pearce. Published by the Australasian Fluid<br />
Mechanics Society. ISBN: 978-0-646-58373-0.<br />
2. Lanchester, F.W., Aerodynamics, Vol. 1, p 431, A.<br />
Constable and Co., London, 1907.<br />
3. Elkaim, G.H., “Autonomous Surface Vehicle Free-<br />
Rotating Wingsail Section Design and Configuration<br />
Analysis”. Journal of Aircraft, 2008. 45(6): p. 1835-<br />
1852.<br />
4. Marchaj, C.A., Sail Performance: Techniques to<br />
Maximize Sail Power. 1996: McGraw-Hill.<br />
5. Gonen, I. Omer Wing-Sail Ltd. Available from:<br />
www.omerwingsail.com.<br />
6. MacLane, D., “The Cogito Project- Design and<br />
Develop of an International C-Class Catamaran and<br />
Her Successful Challenge to Regain the Little<br />
Americas Cup”. Marine Technology and SNAME<br />
News, 2000. 37(4): p. 163-184.<br />
7. Killing, S., “Alpha and Rocker - Two Design<br />
Approaches that Led to the Successful Challenge for<br />
the 2007 International C Class Catamaran<br />
Championship”. Proceedings of the 19th<br />
Chesapeake Sailing Yacht Symposium. 2009.<br />
Annapolis, Maryland.<br />
8. Drela, M. "Newton Solution of Coupled<br />
Viscous/Inviscid Multielement Airfoil Flows,",<br />
AIAA paper 90-1470, presented at the AIAA 21st<br />
Fluid Dynamics, Plasma Dynamics and Lasers<br />
Conference, Seattle Washington, June 1990.<br />
9. Drela, M. and Youngren H., "A User's Guide to<br />
MISES 2.53", MIT Computational Sciences<br />
Laboratory, December 1998.<br />
10. Peart, C.R.C., “Numerical Modeling of a wing-sail<br />
Rigged Class in a Wind Gradient in Both Upwind<br />
and Off the Wind Conditions”. MSc Thesis, School<br />
of Engineering Science, University of Southampton,<br />
UK, 2011.<br />
11. Greenbird. Greenbird supported by ecotricity.<br />
Available from: http://www.greenbird.co.uk/.<br />
12. Sails, X.-W. X-Wing Sails - Wing Sails for Sunfish<br />
and More. Available from:<br />
http://www.solidwingsails.com/.<br />
13. Abbott, I.H. and A.E. Von Doenhoff, Theory of<br />
Wing Sections: Including a Summary of Airfoil Data.<br />
1959: Dover Publications.<br />
14. Gartshore, I.S. “Some effects of upstream turbulence<br />
on the lift forces imposed on prismatic two<br />
dimensional bodies”, J. Fluids Eng. 106(4), 418-<br />
424, 1984.<br />
47