21.09.2014 Views

Chapter 6. Engineering design optimization examples

Chapter 6. Engineering design optimization examples

Chapter 6. Engineering design optimization examples

SHOW MORE
SHOW LESS

Transform your PDFs into Flipbooks and boost your revenue!

Leverage SEO-optimized Flipbooks, powerful backlinks, and multimedia content to professionally showcase your products and significantly increase your reach.

<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Author: Tzu-Chi Liu (2006-08-04); recommended: Yeh-Liang Hsu (2006-08-12).<br />

Note: This article is <strong>Chapter</strong> 6 of Tzu-Chi Liu’s PhD thesis “Developing a Fuzzy<br />

Proportional-Derivative Controller Optimization Engine for <strong>Engineering</strong> Optimization<br />

Problems.”<br />

<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

In this <strong>Chapter</strong>, the fuzzy PD controller <strong>optimization</strong> engine is applied to 6<br />

engineering <strong>design</strong> <strong>optimization</strong> problems commonly seen in literature to demonstrate the<br />

practicality of the fuzzy PD controller <strong>optimization</strong> engine.<br />

<strong>6.</strong>1 Three-bar truss <strong>design</strong> <strong>optimization</strong> problem<br />

<strong>6.</strong>1.1 Formulating the <strong>optimization</strong> model<br />

Figure <strong>6.</strong>1 shows a <strong>design</strong> <strong>optimization</strong> problem of the three-bar truss with the<br />

cross-section areas of members A 1 (and A 3 ) and A 2 as <strong>design</strong> variables [Rao, 1987]. The<br />

objective of this problem is to find the optimal values of the cross-sectional areas of the<br />

bars A 1 , A 2 , A 3 , which minimize the weight of the structure, subject to maximize stress<br />

constraints on the three bars.<br />

Figure <strong>6.</strong>1 Three bar truss <strong>design</strong> problem<br />

1<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Equation (<strong>6.</strong>1) shows the <strong>optimization</strong> model of this problem. Note that the <strong>design</strong><br />

variables have been changed to x 1 and x 2 . It is assumed that A 1 = A 3 , so there are only two<br />

<strong>design</strong> variables in Equation (<strong>6.</strong>1).<br />

min.<br />

f = 2 2x<br />

+ x<br />

1<br />

2<br />

( u )<br />

( x , x ) σ<br />

s.t. g1 = σ<br />

1 1 2<br />

−1<br />

≤ 0 ,<br />

g<br />

g<br />

( u )<br />

( x x ) σ<br />

1,<br />

2<br />

−1<br />

2<br />

= σ<br />

2<br />

≤<br />

0,<br />

() l<br />

σ ( x ) −1<br />

0,<br />

= σ<br />

3 1, x2<br />

3<br />

≤<br />

() l<br />

g<br />

4<br />

= x1<br />

x1<br />

−1<br />

≤ 0,<br />

( u )<br />

g<br />

5<br />

= x1<br />

x1<br />

−1<br />

≤ 0,<br />

() l<br />

g<br />

6<br />

= x2<br />

x2<br />

−1<br />

≤ 0,<br />

( u )<br />

g = x x −1<br />

0,<br />

(<strong>6.</strong>1)<br />

7 2 2<br />

≤<br />

where σ 1 , σ 2 and σ 3 are the stress induced in members 1, 2 and 3 respectively, σ (u) is the<br />

maximum permissible stress in tension, σ (l) is the maximum permissible stress in<br />

compression, x (l) 1 and x (l) 2 are the lower bound on x 1 and x 2 respectively, and x (u) (u)<br />

1 and x 2<br />

are the upper bound on x 1 and x 2 respectively. The stresses are given by:<br />

σ<br />

1<br />

( x , x )<br />

1<br />

2<br />

= P<br />

2x<br />

2x<br />

1<br />

2<br />

1<br />

+ x<br />

2<br />

+ 2x<br />

x<br />

1<br />

2<br />

σ<br />

2<br />

( x , x )<br />

1<br />

2<br />

( x , x )<br />

= P<br />

x<br />

= P<br />

1<br />

+<br />

1<br />

2x<br />

x<br />

2<br />

2<br />

σ<br />

3 1 2<br />

(<strong>6.</strong>2)<br />

2<br />

2x1<br />

+ 2x1x2<br />

In this example, the parameters are:<br />

P = 20000lb, σ (u) = 20000psi, σ (l) = -15000psi,<br />

x (l) i = 0.1in 2 (i = 1, 2), x (u) i = 5in 2 (i = 1, 2) (<strong>6.</strong>3)<br />

2<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

In the fuzzy PD controller <strong>optimization</strong> engine, the objective function and all<br />

constraints can be implicit functions, which cannot be written explicitly in terms of <strong>design</strong><br />

variables. Only the monotonicities of the <strong>design</strong> variables with respect to the objective and<br />

constraint function are required. Therefore, the <strong>optimization</strong> model required by the fuzzy<br />

PD controller <strong>optimization</strong> engine is:<br />

min.<br />

f<br />

+ +<br />

( x x )<br />

1<br />

,<br />

2<br />

− −<br />

( x , x ) 0<br />

s.t. g<br />

1 1 2<br />

≤ ,<br />

− −<br />

( x x ) 0<br />

g ,<br />

2 1<br />

,<br />

2<br />

≤<br />

+ −<br />

( x x ) 0<br />

g ,<br />

g<br />

g<br />

g<br />

3 1<br />

,<br />

2<br />

≤<br />

−<br />

( x ) 0<br />

4 1<br />

≤<br />

+<br />

( x ) 0<br />

5 2<br />

≤<br />

−<br />

( x ) 0<br />

6 2<br />

≤<br />

+<br />

( x ) 0<br />

g (<strong>6.</strong>4)<br />

7 2<br />

≤<br />

<strong>6.</strong>1.2 Monotonicity Analysis<br />

Table <strong>6.</strong>1 is the corresponding monotonicity table for Equation (<strong>6.</strong>4). Figure <strong>6.</strong>2<br />

shows the complete output from MONO, including rigorous Monotonicity Analysis steps<br />

and global knowledge about the <strong>optimization</strong> model. Note that this conclusion is obtained<br />

using only the monotonicity signs of the <strong>design</strong> variables with respect to the objective<br />

function and constraints (the monotonicity table in Table <strong>6.</strong>1). Explicit formulations of the<br />

objective function and the constraints are not required.<br />

3<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>1 Monotonicity table of the three-bar truss problem<br />

x 1 x 2<br />

f + +<br />

g 1 - -<br />

g 2 - -<br />

g 3 + -<br />

g 4 - .<br />

g 5 + .<br />

g 6 . -<br />

g 7 . +<br />

----------------- PC MONO V3.2 : C:\MONO\threebar.out ------------------<br />

TABLE 0 .<br />

x1 x2<br />

F + +<br />

G1 - -<br />

G2 - -<br />

G3 + -<br />

G4 - .<br />

G5 + .<br />

G6 . -<br />

G7 . +<br />

******************************<br />

* GLOBAL FACTS *<br />

4<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

******************************<br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

None<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

* Conditionally Critical Sets :<br />

One of ( G1 G2 G4 ) must be critical for x1.<br />

One of ( G1 G2 G3 G6 ) must be critical for x2.<br />

Figure <strong>6.</strong>2 The output of MONO program of the three-bar truss example<br />

As shown in Figure <strong>6.</strong>2, two conditionally critical sets exist in the three-bar truss<br />

<strong>design</strong> <strong>optimization</strong> problem. One of the constraints in the conditionally critical set (g 1 , g 2 ,<br />

g 4 ) must be critical for <strong>design</strong> variable x 1 , and one of the constraints in the other<br />

conditionally critical set (g 1 , g 2 , g 3 , g 6 ) must be critical for <strong>design</strong> variable x 2 .<br />

<strong>6.</strong>1.3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the constraint functions (g 1 , g 2 , g 3 , g 4 and g 6 ). The<br />

quantitative definitions for the error change inputs (Δe) are the change of the values of the<br />

constraint functions (Δg 1 , Δg 2 , Δg 3 , Δg 4 and Δg 6 ).<br />

5<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

x 1 = 3 in 2 , x 2 = 3 in 2 (<strong>6.</strong>5)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

initial<br />

g 1<br />

initial<br />

g 4<br />

initial<br />

initial<br />

= -0.7643, = -0.8620, = -1.0732,<br />

g 2<br />

initial<br />

=-0.9667, = -0.9667, (<strong>6.</strong>6)<br />

g 6<br />

g 3<br />

The <strong>design</strong>er can define the “move limits” according to the characteristics and<br />

physical properties of the <strong>design</strong> variables. In this example, the values of the move limits<br />

are:<br />

(Δx 1 ) max = 0.1, (Δx 2 ) max = 0.11. (<strong>6.</strong>7)<br />

Note that there is a slight difference in move limits between <strong>design</strong> variables x 1 and x 2<br />

to avoid premature convergence in this case. The values of the quantization level of error<br />

change inputs are evaluated by Equation (5.10) as follows:<br />

Δg 1 (x 1 , x 2 ) max = 0.0082, Δg 2 (x 1 , x 2 ) max = 0.0051 Δg 3 (x 2 ) max = 0.0044,<br />

Δg 4 (x 1 ) max = 0.0012, Δg 6 (x 2 ) max = 0.0013 (<strong>6.</strong>8)<br />

<strong>6.</strong>1.4 The <strong>optimization</strong> results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 75 iterations, when the change in<br />

objective function value in consecutive iterations is less than 0.01%, and all constraints are<br />

satisfied with a tolerance of 0.01% of the initial values of the constraints. Table <strong>6.</strong>2<br />

compares the results obtained by the fuzzy PD controller <strong>optimization</strong> engine and that from<br />

the literature. Note that only constraint g 1 is active at the optimum <strong>design</strong> point, though<br />

there are two <strong>design</strong> variables. Figure <strong>6.</strong>3 shows the iteration histories of the objective<br />

function.<br />

Note that in this process, the function value of each constraint is evaluated 77 times,<br />

and no sensitivity information required. The <strong>design</strong>ers only define the initial values and the<br />

move limits of the <strong>design</strong> variables in the process of the fuzzy PD controller <strong>optimization</strong><br />

engine.<br />

6<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>2 Comparison of the result for the three-bar truss <strong>design</strong> problem<br />

Fuzzy PD Rao [1987]<br />

Objective Function 2.6507 2.6335<br />

Iteration 75 -<br />

x 1 0.7511 0.7871<br />

x 2 0.5262 0.4074<br />

g 1 0.0000 0.0415<br />

g 2 -0.3312 -5.3280<br />

g 3 -1.2485 -20.3695<br />

g 4 -0.8669 -0.6871<br />

g 5 -0.8498 -4.2129<br />

g 6 -0.8100 -0.3074<br />

g 7 -0.8948 -4.5926<br />

Figure <strong>6.</strong>3 The iteration histories of the objective function of the three-bar truss example<br />

7<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

<strong>6.</strong>2 A tension/compression spring <strong>design</strong> <strong>optimization</strong> problem<br />

<strong>6.</strong>2.1 Formulating the <strong>optimization</strong> model<br />

This problem is from Arora [1989] and Belegundu [1982], which minimizes the<br />

weight of a tension/compression spring as shown Figure <strong>6.</strong>4. The tension/compression<br />

spring <strong>design</strong> <strong>optimization</strong> problem subject to constraints on minimum deflection (g 1 ),<br />

shear (g 2 ), surge frequency (g 3 ), limits on outside diameter (g 4 ) and on <strong>design</strong> variables.<br />

The <strong>design</strong> variables are the mean coil diameter D, the wire diameter d and the number of<br />

active coils N. The mathematical formulation of this <strong>design</strong> problem can be described as<br />

Equation (<strong>6.</strong>9).<br />

N<br />

D<br />

d<br />

Figure <strong>6.</strong>4The tension/compression spring <strong>design</strong> <strong>optimization</strong> problem<br />

min.<br />

F<br />

=<br />

2<br />

( N + 2) Dd<br />

4<br />

71785xd<br />

s.t. g<br />

1<br />

= −1<br />

≤ 0 ,<br />

3<br />

D N<br />

2<br />

4D<br />

− dD 1<br />

g<br />

2<br />

=<br />

+ −1<br />

≤ 0 ,<br />

2<br />

12566<br />

5108d<br />

3 4<br />

( Dd − d )<br />

2<br />

D N<br />

g<br />

3<br />

= −1<br />

≤ 0 ,<br />

140.45d<br />

d + D<br />

g<br />

4<br />

= −1<br />

≤ 0 . (<strong>6.</strong>9)<br />

1.5<br />

8<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

The <strong>optimization</strong> model required by the fuzzy PD controller <strong>optimization</strong> engine is:<br />

min. F<br />

+ + +<br />

( d , D , N )<br />

+ − −<br />

( d , D N )<br />

s.t. g<br />

1<br />

,<br />

,<br />

− +<br />

( d D )<br />

g ,<br />

2 ,<br />

− + +<br />

( d , D N )<br />

g ,<br />

3<br />

,<br />

+ +<br />

( d D )<br />

g<br />

4<br />

,<br />

. (<strong>6.</strong>10)<br />

<strong>6.</strong>2.2 Monotonicity Analysis<br />

Table <strong>6.</strong>3 is the corresponding monotonicity table for Equation (<strong>6.</strong>10). Figure <strong>6.</strong>5<br />

shows the complete output from MONO. Design variable D has only one critical constraint<br />

g 1. One of the constraints in the conditionally critical set (g 2 and g 3 ) must be critical for<br />

<strong>design</strong> variable d. The monotonicity sign of <strong>design</strong> variable N is “indeterminate” in the<br />

objective function of the final monotonicity table in Figure <strong>6.</strong>5.<br />

Table <strong>6.</strong>3 Monotonicity table of the spring <strong>design</strong> example<br />

d D N<br />

f + + +<br />

g 1 + - -<br />

g 2 - + .<br />

g 3 - + +<br />

g 4 + + .<br />

----------------- PC MONO V3.2 : C:\MONO\spring02.out ------------------<br />

TABLE 0 .<br />

d D N<br />

F + + +<br />

9<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

G1 + - -<br />

G2 - + .<br />

G3 - + +<br />

G4 + + .<br />

1. G1 is critical for D by MP1.<br />

2. D and G1 are eliminated.<br />

3. Monotonicity of G2 wrt d becomes 'i'.<br />

4. Monotonicity of G3 wrt d becomes 'i'.<br />

5. Monotonicity of F wrt N becomes 'i'.<br />

<strong>6.</strong> Monotonicity of G2 wrt N becomes '-'.<br />

7. Monotonicity of G3 wrt N becomes 'i'.<br />

8. Monotonicity of G4 wrt N becomes '-'.<br />

TABLE 1 .<br />

d<br />

N<br />

F + i<br />

G2 i -<br />

G3 i<br />

i<br />

G4 + -<br />

******************************<br />

* GLOBAL FACTS *<br />

******************************<br />

10<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

D eliminated by MP1, G1 critical below.<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

* Conditionally Critical Sets :<br />

One of ( G2 G3 ) must be critical for d.<br />

Figure <strong>6.</strong>5 The output of MONO program of the spring <strong>design</strong> example<br />

<strong>6.</strong>2.3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the all constraint functions. The quantitative definitions<br />

for the error change inputs (Δe) are the change of the values of the all constraint functions.<br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

d = 0.1, D = 0.25, N = 13. (<strong>6.</strong>11)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

11<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

initial<br />

g 1<br />

initial<br />

g 3<br />

initial<br />

= 34.34, = -0.86,<br />

g 2<br />

initial<br />

g 4<br />

= -0.94, = -0.77, (<strong>6.</strong>12)<br />

In this example, the values of the move limits are:<br />

(Δd) max = 0.001, (ΔD) max = 0.01, (ΔN) max = 0.1 (<strong>6.</strong>13)<br />

Therefore, the values of the quantization level of error change inputs are evaluated by<br />

Equation (5.10) as follows:<br />

Δg 1 (x 0 ) max = <strong>6.</strong>5482, Δg 2 (x 0 ) max = 0.0089,<br />

Δg 3 (x 0 ) max = 0.0058, Δg 4 (x 0 ) max = 0.0073. (<strong>6.</strong>14)<br />

<strong>6.</strong>2.4 Optimization Results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 225 iterations, when the change in<br />

objective function value in consecutive iterations is less than 0.01%, and all constraints are<br />

satisfied with a tolerance of 0.01% of the initial values of the constraints. Table <strong>6.</strong>4 shows<br />

the comparison of the results for the speed reducer <strong>design</strong> <strong>optimization</strong> problem. Note that<br />

only constraint g 1 , g 2 and g 4 are active at the optimum <strong>design</strong> point, though there are three<br />

<strong>design</strong> variables. Figure <strong>6.</strong>6 shows the iteration histories of the objective function.<br />

Note that in this process, the function value of each constraint is evaluated 227 times,<br />

and no sensitivity information required. The <strong>design</strong>ers only define the initial values and the<br />

move limits of the <strong>design</strong> variables in the process of the fuzzy PD controller <strong>optimization</strong><br />

engine.<br />

12<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>4 Comparison of the results for the spring <strong>design</strong> example<br />

Fuzzy PD<br />

Coello and<br />

Monts (2002)<br />

Arora (1989)<br />

Belegundu<br />

(1982)<br />

Objective<br />

function value<br />

0.0126503 0.0126810 0.0127303 0.0128334<br />

Iteration 225 - - -<br />

d 0.052362 0.051989 0.053396 0.050000<br />

D 0.373153 0.363965 0.399180 0.315900<br />

N 10.36486 10.89052 9.185400 14.25000<br />

g 1 0.001987 -0.000013 0.000019 -0.000014<br />

g 2 0.000084 -0.000021 -0.000018 -0.003782<br />

g 3 -0.803753 -4.061338 -4.123832 -3.938302<br />

g 4 -0.716237 -0.722698 -0.698283 -0.756067<br />

Figure <strong>6.</strong>6 The iteration history of objective function of the spring <strong>design</strong> example<br />

13<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

<strong>6.</strong>3 Tubular column <strong>design</strong> <strong>optimization</strong> problem<br />

<strong>6.</strong>3.1 Formulating the <strong>optimization</strong> model<br />

Figure <strong>6.</strong>7 shows an example for <strong>design</strong>ing a uniform column of tubular section to<br />

carry a compressive load P = 2500 kgf for minimum cost [Rao, 1996]. The column is made<br />

up of a material that has a yield stress (σ y ) of 500 kgf/cm 2 , modulus of elasticity (E) of<br />

0.85×106 kgf/cm 2 , and density (ρ) of 0.0025 kgf/cm 3 . The length of the column (L) is 250<br />

cm. The stress included in the column should be less than the bucking stress as well as the<br />

yield stress. The mean diameter of the column is restricted between 2 and 14 cm, and<br />

columns with thickness outside the range 0.2 to 0.8 cm are not available in the market. The<br />

cost of the column includes material and construction costs and can be taken as 9.82dt+2d,<br />

where d is the mean diameter of the column in centimeters, and t is tube thickness.<br />

Figure <strong>6.</strong>7 Tubular column under compression <strong>design</strong> problem<br />

Equation (<strong>6.</strong>9) is the <strong>optimization</strong> model of this problem. The <strong>design</strong> variables are the<br />

mean diameter (d) and tube thickness (t). The objective is to minimize the cost. Constraint<br />

g 1 and g 2 indicates that the stress induced is lower than the yield stress (g 1 ) and buckling<br />

stress (g 2 ). Constraints g 3 to g 6 are side constraints. This <strong>design</strong> <strong>optimization</strong> model can be<br />

summarized as in Equation (<strong>6.</strong>15).<br />

14<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

( d,<br />

t) = 9.82dt<br />

d<br />

min . f + 2<br />

P<br />

s. t.<br />

g1 = −1<br />

≤ 0,<br />

π d tσ<br />

y<br />

2<br />

8PL<br />

g<br />

2<br />

=<br />

−1<br />

≤ 0,<br />

3<br />

π Edt<br />

2 2<br />

( d + t )<br />

g<br />

3<br />

= 2.0 d −1<br />

≤ 0,<br />

g<br />

4<br />

= d 14 −1<br />

≤ 0,<br />

g<br />

5<br />

= 0.2 t −1<br />

≤ 0,<br />

g = t 0.8 −1<br />

0,<br />

(<strong>6.</strong>15)<br />

6<br />

≤<br />

In the fuzzy PD controller <strong>optimization</strong> engine, only the monotonicities of the <strong>design</strong><br />

variables with respect to the objective and constraint function are required. Therefore, the<br />

<strong>optimization</strong> model required by the fuzzy PD controller <strong>optimization</strong> engine is:<br />

min.<br />

( d<br />

+ t + )<br />

f ,<br />

( d<br />

− , t − ) 0<br />

s.t.<br />

g1 ≤<br />

g<br />

g<br />

g<br />

g<br />

( d<br />

− , t − ) 0<br />

2<br />

≤<br />

−<br />

( d ) 0<br />

3<br />

≤<br />

+<br />

( d ) 0<br />

4<br />

≤<br />

−<br />

( t ) 0<br />

5<br />

≤<br />

+<br />

( t ) 0<br />

g (<strong>6.</strong>16)<br />

6<br />

≤<br />

<strong>6.</strong>3.2 Monotonicity Analysis<br />

Table <strong>6.</strong>5 is the corresponding monotonicity table for Equation (<strong>6.</strong>16). Figure <strong>6.</strong>8<br />

shows the complete output from MONO. There are two conditionally critical sets. One of<br />

the constraints in the conditionally critical set (g 1 , g 2 , g 3 ) must be critical for <strong>design</strong><br />

15<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

variable d, and one of the constraints in the other conditionally critical set (g 1 , g 2 , g 5 ) must<br />

be critical for <strong>design</strong> variable t.<br />

Table <strong>6.</strong>5 Monotonicity table of the tubular column problem<br />

d<br />

t<br />

f + +<br />

g 1 - -<br />

g 2 - -<br />

g 3 - .<br />

g 4 + .<br />

g 5 . -<br />

g 6 . +<br />

----------------- PC MONO V3.2 : C:\MONO\tubular.out ------------------<br />

TABLE 0 .<br />

x1 x2<br />

F + +<br />

G1 - -<br />

G2 - -<br />

G3 - .<br />

G4 + .<br />

G5 . -<br />

G6 . +<br />

******************************<br />

* GLOBAL FACTS *<br />

16<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

******************************<br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

None<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

* Conditionally Critical Sets :<br />

One of ( G1 G2 G3 ) must be critical for x1.<br />

One of ( G1 G2 G5 ) must be critical for x2.<br />

Figure <strong>6.</strong>8 The output of MONO program of the tubular column example<br />

<strong>6.</strong>3.3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the constraint functions (g 1 , g 2 , g 3 and g 5 ). The<br />

quantitative definitions for the error change inputs (Δe) are the change of the values of the<br />

constraint functions (Δg 1 , Δg 2 , Δg 3 and Δg 5 ).<br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

17<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

d = 14.0 cm, t = 0.8 cm (<strong>6.</strong>17)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

initial<br />

g 1<br />

initial<br />

g 3<br />

initial<br />

= -0.8579, = -0.9785,<br />

g 2<br />

initial<br />

= -0.8571, = -0.7500, (<strong>6.</strong>18)<br />

g 5<br />

In this example, the values of the move limits are:<br />

(Δd) max = 1.0, (Δt) max = 0.1. (<strong>6.</strong>19)<br />

Therefore, the values of the quantization level of error change inputs are evaluated by<br />

Equation (5.10) as follows:<br />

Δg 1 (d, t) max = 0.0328, Δg 2 (d, t) max = 0.0092,<br />

Δg 3 (d) max = 0.0110, Δg 5 (t) max = 0.0357. (<strong>6.</strong>20)<br />

<strong>6.</strong>3.4 The <strong>optimization</strong> results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 53 iterations, when the change in<br />

objective function value in consecutive iterations is less than 0.01%, and all constraints are<br />

satisfied with a tolerance of 0.01% of the initial values of the constraints. Table <strong>6.</strong>6<br />

compares the results obtained from the fuzzy PD controller <strong>optimization</strong> engine and that<br />

from the literature. Note that two stress constraints g 1 and g 2 are active constraints, and<br />

side constraint g 5 is close to active too. Figure <strong>6.</strong>9 shows the iteration histories of the<br />

objective function.<br />

Note that in this process, the function value of each constraint is evaluated 55 times,<br />

and no sensitivity information required. The <strong>design</strong>ers only define the initial values and the<br />

move limits of the <strong>design</strong> variables in the process of the fuzzy PD controller <strong>optimization</strong><br />

engine.<br />

18<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>6 Comparison of the result for the tubular column <strong>design</strong> problem<br />

Fuzzy PD Rao [1996]<br />

Objective Function 25.5316 2<strong>6.</strong>5323<br />

Iteration 53<br />

d 5.4507 5.440<br />

t 0.2920 0.293<br />

g 1 -7.8×10 -5 -0.8579<br />

g 2 -7.8×10 -5 -0.9785<br />

g 3 -0.6331 -0.8571<br />

g 4 -0.6107 0.0000<br />

g 5 -0.3151 -0.7500<br />

g 6 -0.6350 0.0000<br />

Figure <strong>6.</strong>9 The iteration histories of the objective function of the tubular column example<br />

19<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

<strong>6.</strong>4 Welded beam <strong>design</strong> <strong>optimization</strong> problem<br />

<strong>6.</strong>4.1 Formulating the <strong>optimization</strong> model<br />

Figure <strong>6.</strong>10 shows the welded beam structure which consists of beam A and the weld<br />

required to hold the beam to member B [Ragsdell and Phillips, 1976]. The objective is to<br />

find a feasible set of dimensions h, l, t and b to carry a certain load (P) while maintaining<br />

the minimum total fabrication cost. The constraints are on shear stress (τ), bending stress in<br />

the beam (σ), bucking load on the bar (P c ), and end deflection on the beam (δ). The <strong>design</strong><br />

<strong>optimization</strong> model can be summarized in Equation (<strong>6.</strong>21).<br />

Figure <strong>6.</strong>10 Welded Beam <strong>design</strong> problem<br />

min.<br />

s.t.<br />

( L l)<br />

2<br />

f = 1.10471h<br />

l + 0. 04811tb<br />

+<br />

( )/<br />

−1<br />

0<br />

g = τ x τ ,<br />

1 max<br />

≤<br />

( ) −1<br />

0<br />

g = σ x σ ,<br />

g<br />

2<br />

/<br />

max<br />

≤<br />

= h / b −1<br />

3<br />

≤<br />

0,<br />

( L + ) −1<br />

0<br />

2<br />

g<br />

4<br />

= 0.02094h<br />

+ 0.00962tb<br />

l ≤ ,<br />

g = 0.125/ h −1<br />

0,<br />

5<br />

≤<br />

( ) −1<br />

0<br />

g = δ x δ ,<br />

6<br />

/<br />

max<br />

≤<br />

20<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


( ) −1<br />

0<br />

<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

g<br />

7<br />

= P / x ≤ , (<strong>6.</strong>21)<br />

P c<br />

where x is a vector of the <strong>design</strong> variables (h, l, t, b), L is the overhang length if the beam<br />

(14 in), τ max is the allowable <strong>design</strong> shear stress of weld (136,000 psi), τ(x) is the weld<br />

shear stress, σ max is the allowable <strong>design</strong> yield stress for the bar material (36,600 psi), σ(x)<br />

is the bending stress, P c (x) is the bar bucking load, P is the loading condition (6,000<br />

lb), δ max is the allowable bar end deflection, and δ(x) is the bar end deflection. The terms<br />

τ(x), σ(x), P c (x) and δ(x) are expressed as follows:<br />

l<br />

τ ( x ) = ( τ ′) + 2τ<br />

′ τ ′′ + ( τ ′<br />

) 2<br />

,<br />

R<br />

2 2<br />

6PL<br />

σ ( x)<br />

= ,<br />

2<br />

t b<br />

3<br />

4PL<br />

δ ( x)<br />

= ,<br />

3<br />

Et b<br />

2 6<br />

4.013E<br />

t b 36 ⎛ t E<br />

( )<br />

⎟ ⎞<br />

P = c<br />

x ⎜<br />

1<br />

2 −<br />

L ⎝ 2L<br />

4 G , (<strong>6.</strong>22)<br />

⎠<br />

where<br />

P<br />

τ ′ = ,<br />

2hl<br />

MR ⎛ l ⎞<br />

τ ′′ = , M = P⎜<br />

L + ⎟ ,<br />

J ⎝ 2 ⎠<br />

2<br />

l ⎛ h + t ⎞<br />

R = + ⎜ ⎟ ,<br />

4 ⎝ 2 ⎠<br />

2<br />

⎪⎧<br />

⎡<br />

2<br />

2<br />

l ⎛ h + t ⎞ ⎤⎪⎫<br />

J = 2⎨<br />

2hl⎢<br />

+ ⎜ ⎟ ⎥⎬<br />

. (<strong>6.</strong>23)<br />

⎪⎩ ⎢⎣<br />

12 ⎝ 2 ⎠ ⎥⎦<br />

⎪⎭<br />

The <strong>optimization</strong> model required by the fuzzy PD controller <strong>optimization</strong> engine is:<br />

min.<br />

+ + + +<br />

( h , l , t b )<br />

f ,<br />

− − −<br />

( h , l t )<br />

s.t. g<br />

1<br />

,<br />

,<br />

( t<br />

− b − )<br />

g ,<br />

2 ,<br />

( h<br />

+ b − )<br />

g ,<br />

3<br />

,<br />

21<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

+ + + +<br />

( h , l , t b )<br />

g ,<br />

4 ,<br />

−<br />

( h )<br />

g 5<br />

,<br />

( t<br />

− b − )<br />

g ,<br />

6<br />

,<br />

( t<br />

− b − )<br />

g<br />

7<br />

,<br />

. (<strong>6.</strong>24)<br />

<strong>6.</strong>4.2 Monotonicity Analysis<br />

Table <strong>6.</strong>7 is the corresponding monotonicity table for Equation (<strong>6.</strong>24). Figure <strong>6.</strong>11<br />

shows the complete output from MONO. Constraint g 1 is critical for <strong>design</strong> variable l. One<br />

of the constraints in the conditionally critical set (g 2 , g 3, g 6 , g 7 ) must be critical for <strong>design</strong><br />

variable b. The monotonicity sign of <strong>design</strong> variables h and t are “indeterminate” in the<br />

objective function of the final monotonicity table in Figure <strong>6.</strong>11.<br />

Table <strong>6.</strong>7 Monotonicity table of the welded beam problem<br />

h l t b<br />

f + + + +<br />

g 1 - - - .<br />

g 2 . . - -<br />

g 3 + . . -<br />

g 4 + + + +<br />

g 5 - . . .<br />

g 6 . . - -<br />

g 7 . . - -<br />

----------------- PC MONO V3.2 : C:\MONO\Welded.out ------------------<br />

TABLE 0 .<br />

h l t b<br />

F + + + +<br />

G1 - - - .<br />

22<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

G2 . . - -<br />

G3 + . . -<br />

G4 + + + +<br />

G5 - . . .<br />

G6 . . - -<br />

G7 . . - -<br />

1. G1 is critical for l by MP1.<br />

2. l and G1 are eliminated.<br />

3. Monotonicity of F wrt h becomes 'i'.<br />

4. Monotonicity of G4 wrt h becomes 'i'.<br />

5. Monotonicity of F wrt t becomes 'i'.<br />

<strong>6.</strong> Monotonicity of G4 wrt t becomes 'i'.<br />

TABLE 1 .<br />

h t b<br />

F i i +<br />

G2 . - -<br />

G3 + . -<br />

G4 i i +<br />

G5 - . .<br />

G6 . - -<br />

G7 . - -<br />

23<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

******************************<br />

* GLOBAL FACTS *<br />

******************************<br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

l eliminated by MP1, G1 critical below.<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

* Conditionally Critical Sets :<br />

One of ( G2 G3 G6 G7 ) must be critical for b.<br />

Figure <strong>6.</strong>11 The output of MONO program of the welded beam example<br />

<strong>6.</strong>4.3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the constraint functions (g 1 , g 2 , g 3 , g 4 , g 5 , g 6 and g 7 ). The<br />

quantitative definitions for the error change inputs (Δe) are the change of the values of the<br />

constraint functions (Δg 1 , Δg 2 , Δg 3 , Δg 4 , Δg 5 , Δg 6 and Δg 7 ).<br />

24<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

h = 2 in, l = 10 in, t = 10 in, b = 2 in (<strong>6.</strong>25)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

initial<br />

g 1<br />

initial<br />

g 5<br />

g 2<br />

initial<br />

initial<br />

initial<br />

initial<br />

= -0.9620, = -0.9160, = 0.000, = 3.7023,<br />

g 6<br />

g 3<br />

initial<br />

g 7<br />

= -0.9375, = -0.9956, = -0.9990. (<strong>6.</strong>26)<br />

g 4<br />

The <strong>design</strong>er can define the “move limits” according to the characteristics and<br />

physical properties of the <strong>design</strong> variables. In this example, the values of the move limits<br />

are:<br />

(Δh) max = 0.1, (Δl) max = 0.1. (Δt) max = 0.1, (Δb) max = 0.1. (<strong>6.</strong>27)<br />

Therefore, the values of the quantization level of error change inputs are evaluated by<br />

Equation (5.10) as follows:<br />

Δg 1 (h, l, t) max = -0.0962, Δg 2 (t, b) max = -0.0916, Δg 3 (h) max = 0.0000,<br />

Δg 4 (h, l, t, b) max = 0.3702, Δg 5 (h) max = -0.0938, Δg 6 (t, b) max = -0.0996,<br />

Δg 7 (t, b) max = -0.1000 (<strong>6.</strong>28)<br />

<strong>6.</strong>4.4 The <strong>optimization</strong> results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 139 iterations, when the change in<br />

objective function value in consecutive iterations is less than 0.01%, and all constraints are<br />

satisfied with a tolerance of 0.01% of the initial values of the constraints. Table <strong>6.</strong>8<br />

compares the results obtained from the fuzzy PD controller <strong>optimization</strong> engine and that<br />

from the literature. Note that constraint g 1 , g 2 , g 3 , g 5 and g 6 are active at the optimum<br />

<strong>design</strong> point. Figure <strong>6.</strong>12 shows the iteration histories of the objective function.<br />

25<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Figure <strong>6.</strong>12 The iteration history of the objective function of the welded beam example<br />

Note that this problem has many local minimums. Table <strong>6.</strong>8 also shows the<br />

<strong>optimization</strong> results starting from 2 other different starting points:<br />

Fuzzy PD 02: h = 1.0, l = 5, t = 10, b = 2; move limits (Δh) max = 0.01, (Δl) max = 0.01,<br />

(Δt) max = 0.01, (Δb) max = 0.01.<br />

Fuzzy PD 03: h = 2, l = 5, t = 10, b = 2; move limits (Δh) max = 0.1, (Δl) max = 0.1,<br />

(Δt) max = 0.1, (Δb) max = 0.01.<br />

Table <strong>6.</strong>9 shows the <strong>optimization</strong> results using different values of the quantization<br />

level of inputs and output. The cases of Fuzzy PD 04 to 06 use the same initial values and<br />

move limits to the case of Fuzzy PD 01 to 03. It can be observed that there is no significant<br />

difference using different initial values.<br />

26<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>8 Comparison of the result for the welded beam <strong>design</strong> problem<br />

Fuzzy PD 01 Fuzzy PD 02 Fuzzy PD 03<br />

Ragsdell and<br />

Phillips(1976)<br />

Coello (2000)<br />

Objective<br />

function<br />

2.0182 1.8052 1.8362 2.386 1.748<br />

Iteration 139 796 1157 - 900,000<br />

h 0.1740 1.6887 0.2326 0.2455 0.2088<br />

l 4.8075 4.4207 3.1795 <strong>6.</strong>1960 3.4208<br />

t 8.1836 9.1658 8.4355 8.2730 8.9975<br />

b 0.2508 2.0509 2.3609 0.2455 0.2100<br />

g 1 3.9×10 -5 8.2×10 -6 1.1×10 -4 -0.4223 -9.5×10 -5<br />

g 2 5.6×10 -5 -0.0250 4.9×10 -5 -1.6×10 -4 -0.0118<br />

g 3 -0.3062 -0.1766 -0.0144 0.0000 -0.0057<br />

g 4 -0.6279 -0.6662 -0.6697 -0.6041 -0.6824<br />

g 5 -0.2817 -0.2598 -0.4627 -0.4908 -0.4013<br />

g 6 -0.9361 -0.9444 -0.9380 -0.9368 -0.9426<br />

g 7 -0.4099 1.1×10 -4 -0.3069 -0.3753 -0.0571<br />

Table <strong>6.</strong>9 Comparison of the result using different quantization factor<br />

Fuzzy PD 04 Fuzzy PD 05 Fuzzy PD 06<br />

The quantization level of inputs 4 4 4<br />

The quantization level of output 4 4 4<br />

Objective function 2.1498 1.8107 1.8571<br />

Iteration 160 958 876<br />

h 1.3866 1.6674 2.4131<br />

l <strong>6.</strong>5110 4.4963 3.0785<br />

t 8.2413 9.1642 8.3195<br />

b 2.4735 2.0510 2.4271<br />

g 1 1.0×10 -4 9.1×10 -5 9.7×10 -5<br />

g 2 3.4×10 -5 -0.02467 1.3×10 -4<br />

g 3 -0.4394 -0.1871 -0.0057<br />

g 4 -0.5973 -0.6649 -0.6670<br />

g 5 -0.0985 -0.2503 -0.4820<br />

g 6 -0.9366 -0.9444 -0.9372<br />

g 7 -0.3875 1.7×10 -6 -0.3559<br />

27<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

<strong>6.</strong>5 Speed reducer <strong>design</strong> <strong>optimization</strong> problem<br />

<strong>6.</strong>5.1 Formulating the <strong>optimization</strong> model<br />

The <strong>design</strong> of the speed reducer [Golinski, 1973], shown in Figure <strong>6.</strong>13, is considered<br />

with the face width (b), module of teeth (m), number of teeth on pinion (z), length of shaft<br />

1 between bearings (l 1 ), length of shaft 2 between bearings (l 2 ), diameter of shaft 1 (d 1 ),<br />

and diameter of shaft 2 (d 2 ). The objective is to minimize the total weight of the speed<br />

reducer. The constraints include limitations on the bending stress of gear teeth, surface<br />

stress, transverse deflections of shafts 1 and 2 due to transmitted force, and stresses in<br />

shafts 1 and 2. The <strong>design</strong> <strong>optimization</strong> model can be summarized in Equation (<strong>6.</strong>29).<br />

l 2<br />

z1<br />

l 1<br />

d 2<br />

d 1<br />

min.<br />

f<br />

−<br />

Figure <strong>6.</strong>13 Speed reducer <strong>design</strong> problem<br />

2<br />

2<br />

( b,<br />

m,<br />

z,<br />

l1,<br />

l2<br />

, d1,<br />

d<br />

2<br />

) = 0.7854bm<br />

( 3.3333z<br />

+ 14.9334z<br />

− 43.0934)<br />

2 2<br />

3 3<br />

2 2<br />

1.508b( d + d ) + 7.477( d + d ) + 0.7854( l d + l d )<br />

27<br />

s.t. g<br />

1<br />

= −1<br />

≤ 0,<br />

2<br />

bm z<br />

g<br />

g<br />

1<br />

397.5<br />

= −1<br />

2 2<br />

bm z<br />

2<br />

≤<br />

1.93<br />

=<br />

3<br />

mzl d<br />

2<br />

−1<br />

3<br />

≤<br />

4<br />

1 1<br />

1.93<br />

g<br />

4<br />

= −1<br />

≤<br />

3 4<br />

mzl d<br />

2<br />

2<br />

0,<br />

0,<br />

0,<br />

1<br />

2<br />

1<br />

1<br />

2<br />

2<br />

28<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


( 745l<br />

/ mz)<br />

2<br />

6<br />

1<br />

+ 1<strong>6.</strong>9 × 10<br />

g<br />

5<br />

=<br />

−1<br />

≤ 0 ,<br />

3<br />

110d<br />

( 745l<br />

/ mz)<br />

2<br />

1<br />

6<br />

1<br />

+ 157.5×<br />

10<br />

g<br />

6<br />

=<br />

−1<br />

≤ 0,<br />

3<br />

85d<br />

g<br />

7<br />

= mz / 40 −1<br />

≤ 0,<br />

g = 5m<br />

/ b −1<br />

0, ,<br />

8<br />

≤<br />

g<br />

9<br />

= b /12m −1<br />

≤ 0,<br />

g<br />

10<br />

= 2.6 / b −1<br />

≤ 0,<br />

g<br />

11<br />

= b / 3.6 −1<br />

≤ 0,<br />

g<br />

12<br />

= 0.7 / m −1<br />

≤ 0,<br />

g<br />

13<br />

= m / 0.8 −1<br />

≤ 0,<br />

g = 17 / z −1<br />

0 ,<br />

14<br />

≤<br />

g = z / 28 −1<br />

0,<br />

15<br />

≤<br />

g = .3/ l −1<br />

0,<br />

16<br />

7<br />

1<br />

≤<br />

g = l / 8.3 −1<br />

0,<br />

17 1<br />

≤<br />

g = .3/ l −1<br />

0,<br />

18<br />

7<br />

2<br />

≤<br />

g = l / 8.3 −1<br />

0,<br />

19 2<br />

≤<br />

g = .9 / d −1<br />

0,<br />

20<br />

2<br />

1<br />

≤<br />

g = d / 3.9 −1<br />

0,<br />

21 1<br />

≤<br />

g = .0 / d −1<br />

0 ,<br />

22<br />

5<br />

2<br />

≤<br />

2<br />

<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

29<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

g = d / 5.5 −1<br />

0,<br />

23 2<br />

≤<br />

1.5d1<br />

+ 1.9<br />

g<br />

24<br />

= −1<br />

≤ 0 ,<br />

l<br />

1<br />

1.1d<br />

2<br />

+ 1.9<br />

g<br />

25<br />

= −1<br />

≤ 0 . (<strong>6.</strong>29)<br />

l<br />

2<br />

There are 7 <strong>design</strong> variables and 25 constraints. The <strong>optimization</strong> model required by<br />

the fuzzy PD controller <strong>optimization</strong> engine is:<br />

min.<br />

F<br />

+ + + + + + +<br />

( b , m , z , l , l , b b )<br />

− − −<br />

( b , m z )<br />

s.t. g<br />

1<br />

, ,<br />

− − −<br />

( b , m z )<br />

g ,<br />

2<br />

,<br />

− − + −<br />

( m , z , l b )<br />

3 1<br />

,<br />

1 2 1<br />

,<br />

g ,<br />

− − + −<br />

( m , z , l b )<br />

g ,<br />

4 2<br />

,<br />

− − + −<br />

( m , z , l b )<br />

g ,<br />

5 1<br />

,<br />

− − + −<br />

( m , z , l b )<br />

g ,<br />

6 2<br />

,<br />

+ +<br />

( m z )<br />

g ,<br />

7<br />

,<br />

( b<br />

− m + )<br />

g ,<br />

8<br />

,<br />

+ −<br />

( b m )<br />

g ,<br />

9<br />

,<br />

−<br />

( b )<br />

g 10<br />

,<br />

+<br />

( b )<br />

g 11<br />

,<br />

−<br />

( m )<br />

g 12<br />

,<br />

+<br />

( m )<br />

g 13<br />

,<br />

1<br />

1<br />

2<br />

2<br />

2<br />

30<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

−<br />

( z )<br />

g 14<br />

,<br />

+<br />

( z )<br />

g 15<br />

,<br />

−<br />

( )<br />

g ,<br />

16 l 1<br />

+<br />

( )<br />

g ,<br />

17 l 1<br />

−<br />

( )<br />

g ,<br />

18 l 2<br />

+<br />

( )<br />

g ,<br />

19<br />

l 2<br />

−<br />

( )<br />

g ,<br />

20 b 1<br />

+<br />

( )<br />

g ,<br />

21 b 1<br />

−<br />

( )<br />

g ,<br />

22<br />

b 2<br />

+<br />

( )<br />

g ,<br />

23 b 2<br />

+<br />

( l )<br />

−<br />

b<br />

g ,<br />

24 1 −<br />

,<br />

− +<br />

( l )<br />

g . (<strong>6.</strong>30)<br />

25 2<br />

, b2<br />

<strong>6.</strong>5.2 Monotonicity Analysis<br />

Table <strong>6.</strong>10 is the corresponding monotonicity table for Equation (<strong>6.</strong>30). Figure <strong>6.</strong>14<br />

shows the complete output from MONO. Note that there are 7 conditionally critical sets in<br />

this problem.<br />

31<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>10 Monotonicity table of the speed reducer <strong>design</strong> problem<br />

b m z l 1 l 2 b 1 b 2<br />

f + + + + + + +<br />

g 1 - - - . . . .<br />

g 2 - - - . . . .<br />

g 3 . - - + . - .<br />

g 4 . - - . + . -<br />

g 5 . - - + . - .<br />

g 6 . - - . + . -<br />

g 7 . + + . . . .<br />

g 8 - + . . . . .<br />

g 9 + - . . . . .<br />

g 10 - . . . . . .<br />

g 11 + . . . . . .<br />

g 12 . - . . . . .<br />

g 13 . + . . . . .<br />

g 14 . . - . . . .<br />

g 15 . . + . . . .<br />

g 16 . . . - . . .<br />

g 17 . . . + . . .<br />

g 18 . . . . - . .<br />

g 19 . . . . + . .<br />

g 20 . . . . . - .<br />

g 21 . . . . . + .<br />

g 22 . . . . . . -<br />

g 23 . . . . . . +<br />

g 24 . . . - . + .<br />

g 25 . . . . - . +<br />

32<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

----------------- PC MONO V3.1 : speed03.out ------------------<br />

TABLE 0 .<br />

b m z l1 l2 b1 b2<br />

F + + + + + + +<br />

G1 - - - . . . .<br />

G2 - - - . . . .<br />

G3 . - - + . - .<br />

G4 . - - . + . -<br />

G5 . - - + . - .<br />

G6 . - - . + . -<br />

G7 . + + . . . .<br />

G8 - + . . . . .<br />

G9 + - . . . . -<br />

G10 - . . . . . .<br />

G11 + . . . . . .<br />

G12 . - . . . . .<br />

G13 . + . . . . .<br />

G14 . . - . . . .<br />

G15 . . + . . . .<br />

G16 . . . - . . .<br />

G17 . . . + . . .<br />

33<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

G18 . . . . - . .<br />

G19 . . . . + . .<br />

G20 . . . . . - .<br />

G21 . . . . . + .<br />

G22 . . . . . . -<br />

G23 . . . . . . +<br />

G24 . . . - . + .<br />

G25 . . . . - . +<br />

******************************<br />

* GLOBAL FACTS *<br />

******************************<br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

None<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

34<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

* Conditionally Critical Sets :<br />

One of ( G1 G2 G8 G10 ) must be critical for b.<br />

One of ( G1 G2 G3 G4 G5 G6 G9 G12 ) must be critical for m.<br />

One of ( G1 G2 G3 G4 G5 G6 G14 ) must be critical for z.<br />

One of ( G16 G24 ) must be critical for l1.<br />

One of ( G18 G25 ) must be critical for l2.<br />

One of ( G3 G5 G20 ) must be critical for b1.<br />

One of ( G4 G6 G9 G22 ) must be critical for b2.<br />

Figure <strong>6.</strong>14 The output of MONO program of the speed reducer example<br />

<strong>6.</strong>5.3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the constraint functions (g 1 , g 2 , g 3 , g 4 , g 5 , g 6 , g 8 , g 9 , g 10 , g 12 ,<br />

g 14 , g 16 , g 18 , g 20 , g 22 , g 24 , and g 25 ). The quantitative definitions for the error change inputs<br />

(Δe) are the change of the values of the constraint functions (Δg 1 , Δg 2 , Δg 3 , Δg 4 , Δg 5 , Δg 6 ,<br />

Δg 8 , Δg 9 , Δg 10 , Δg 12 , Δg 14 , Δg 16 , Δg 18 , Δg 20 , Δg 22 , Δg 24 , and Δg 25 ).<br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

b = 3.6, m = 0.8, z = 28, l 1 = 8.3,<br />

l 2 = 8.3, b 1 = 3.9, b 2 , = 5.5. (<strong>6.</strong>31)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

initial<br />

g 1<br />

initial<br />

g 5<br />

initial<br />

g 10<br />

g 2<br />

initial<br />

initial<br />

initial<br />

initial<br />

= -0.58, = -0.78, = -0.79, = -0.95,<br />

g 6<br />

initial<br />

initial<br />

= -0.37, = -0.11, = -0.44, = -0.11,<br />

g 12<br />

g 3<br />

g 8<br />

initial<br />

g 14<br />

g 4<br />

initial<br />

g 9<br />

initial<br />

= -0.28, = 0.00, = -0.40, = -0.12,<br />

g 16<br />

35<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

initial<br />

g 18<br />

initial<br />

g 25<br />

initial<br />

initial<br />

initial<br />

= -0.12, = -0.26, = -0.09, = -0.07,<br />

g 20<br />

g 22<br />

= -0.04. (<strong>6.</strong>32)<br />

g 24<br />

In this example, the values of the move limits are:<br />

(Δb) max = 0.1, (Δm) max = 0.01, (Δz) max = 1.0, (Δl 1 ) max = 0.1,<br />

(Δl 2 ) max = 0.01, (Δb 1 ) max = 0.01, (Δb 2 ) max = 0.01. (<strong>6.</strong>33)<br />

Therefore, the values of the quantization level of error change inputs are evaluated by<br />

Equation (5.10) as follows:<br />

Δg 1 (x 0 ) max = 0.04, Δg 2 (x 0 ) max = 0.03, Δg 3 (x 0 ) max = 0.02,<br />

Δg 4 (x 0 ) max = 5.2×10 -3 , Δg 5 (x 0 ) max = 5.1×10 -3 , Δg 6 (x 0 ) max = 4.9×10 -3 ,<br />

Δg 8 (x 0 ) max = 4.6×10 -2 , Δg 9 (x 0 ) max = 1.5×10 -2 , Δg 10 (x 0 ) max = 2.1×10 -2 ,<br />

Δg 12 (x 0 ) max = 1.1×10 -2 , Δg 14 (x 0 ) max = 2.3×10 -2 , Δg 16 (x 0 ) max = 1.1×10 -2 ,<br />

Δg 18 (x 0 ) max = 1.1×10 -2 , Δg 20 (x 0 ) max = 1.9×10 -3 , Δg 22 (x 0 ) max = 1.7×10 -3 ,<br />

Δg 24 (x 0 ) max = 1.3×10 -2 , Δg 25 (x 0 ) max =1.3×10 -2 . (<strong>6.</strong>34)<br />

<strong>6.</strong>5.4 The <strong>optimization</strong> results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 302 iterations, when the change in<br />

objective function value in consecutive iterations is less than 0.01%, and all constraints are<br />

less than the initial constraints with a tolerance of 10 -4 . Table <strong>6.</strong>11 shows the comparison of<br />

the results for the speed reducer <strong>design</strong> <strong>optimization</strong> problem. Figure <strong>6.</strong>15 shows the<br />

iteration history of the objective function.<br />

36<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>11 Comparison of the results for the speed reducer example<br />

Fuzzy PD Golinski [1973]<br />

Objective function value 3007.8 2985.2<br />

Iteration 302 -<br />

x 1 3.5197 3.5<br />

x 2 0.7039 0.7<br />

x 3 17.3831 17.0<br />

x 4 7.3000 7.3<br />

x 5 7.7152 7.3<br />

x 6 3.3498 3.35<br />

x 7 5.2866 5.29<br />

g 1 -0.1095 -0.0739<br />

g 2 -0.2458 -0.1980<br />

g 3 -0.5127 -0.4990<br />

g 4 -0.9073 -0.9194<br />

g 5 0.0000 0.0001<br />

g 6 0.0000 -0.0020<br />

g 7 -0.6941 -0.7025<br />

g 8 0.0000 0.0000<br />

g 9 0.5833 -0.5833<br />

g 10 -0.2613 -0.2571<br />

g 11 -0.0223 -0.0278<br />

g 12 -0.0056 0.0000<br />

g 13 -0.1201 -0.1250<br />

g 14 -0.0220 0.0000<br />

g 15 -0.3792 -0.3929<br />

g 16 0.0000 0.0000<br />

g 17 -0.1205 -0.1205<br />

g 18 -0.0538 0.0000<br />

g 19 -0.0705 -0.1205<br />

g 20 -0.1343 -0.1343<br />

g 21 -0.1411 -0.1410<br />

g 22 -0.0542 -0.0548<br />

g 23 -0.0388 -0.0382<br />

g 24 -0.0514 -0.0514<br />

g 25 0.0000 0.0574<br />

37<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Figure <strong>6.</strong>15 The iteration history of objective function of the speed reducer example<br />

<strong>6.</strong>6 Heat exchange <strong>design</strong> <strong>optimization</strong> example<br />

<strong>6.</strong><strong>6.</strong>1 Formulating the <strong>optimization</strong> model<br />

A fluid having a given flow rate W and specific heat C p is heated from temperature T 0<br />

(100 o F) to T 3 (500 o F) by passing three heat exchangers in series and WC p = 10 5<br />

(B.t.u/hr- o F), as shown in Figure <strong>6.</strong>16 [Avriel and Williams, 1971]. In each heat exchanger<br />

(stage) the cold stream is heated by a hot fluid having the same flow rate W and specific<br />

heat Cp as the cold stream. The temperatures of the hot fluid entering the heat exchangers,<br />

t 11 (300 o F), t 21 (400 o F) and t 31 (600 o F) the overall heat transfer coefficients U 1 (120<br />

B.t.u/hr-ft 2 - o F), U 2 (80 B.t.u/hr-ft 2 - o F), U 3 (40 B.t.u/hr-ft 2 - o F) of the exchangers are known<br />

constants. Optimal <strong>design</strong> involves minimizing the sum of the heat transfer areas of the<br />

three exchangers (A 1 +A 2 +A 3 ). The <strong>design</strong> <strong>optimization</strong> model can be summarized in<br />

Equation (<strong>6.</strong>35).<br />

38<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

T 0 T 1 T 2 T 3<br />

Stage 1<br />

Stage 2<br />

Stage 3<br />

A 1 A 2 A 3<br />

t 12<br />

t 11<br />

t 22<br />

t 21<br />

t 32<br />

t 31<br />

Figure <strong>6.</strong>16 Three stage heat exchanger system<br />

min.<br />

( A1<br />

, A2<br />

, A3<br />

) = A1<br />

+ A2<br />

A3<br />

f +<br />

( T + ) −1<br />

0<br />

s.t. g = 0.0025<br />

1<br />

t12<br />

1<br />

≤<br />

( − T + T + ) −1<br />

0<br />

g<br />

2<br />

= 0.0025<br />

1 2<br />

t22<br />

≤ ,<br />

( − T + ) −1<br />

0<br />

g<br />

3<br />

= 0.01<br />

2<br />

t32<br />

≤ ,<br />

,<br />

g<br />

4<br />

= 100A1<br />

83333.333 − A1t<br />

12<br />

83333.33 + 0.00999T1<br />

−1<br />

≤ 0 ,<br />

g<br />

5<br />

= A2<br />

1250 − A2t<br />

22<br />

1250T1<br />

+ T2<br />

T1<br />

−1<br />

≤ 0 ,<br />

g<br />

6<br />

= A3T2<br />

1250000 − A3t<br />

32<br />

1250000 − T2<br />

500 −1<br />

≤ 0 ,<br />

g<br />

7<br />

= 100 / A1<br />

−1<br />

≤ 0,<br />

g = A /10000 −1<br />

0,<br />

8 1<br />

≤<br />

g<br />

9<br />

= 1000 / A2<br />

−1<br />

≤ 0,<br />

g = A /10000 −1<br />

0,<br />

10 2<br />

≤<br />

g<br />

11<br />

= 1000 / A3<br />

−1<br />

≤ 0,<br />

g = A /10000 −1<br />

0,<br />

12 3<br />

≤<br />

g<br />

13<br />

= 10 / T1<br />

−1<br />

≤ 0,<br />

g = T /1000 −1<br />

0 ,<br />

14 1<br />

≤<br />

g<br />

15<br />

= 10 / T2<br />

−1<br />

≤ 0,<br />

39<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

g = T /1000 −1<br />

0,<br />

16 2<br />

≤<br />

g = / t −1<br />

0,<br />

17<br />

10 12<br />

≤<br />

g = t /1000 −1<br />

0,<br />

18 12<br />

≤<br />

g = / t −1<br />

0,<br />

19<br />

10 22<br />

≤<br />

g = t /1000 −1<br />

0,<br />

20 22<br />

≤<br />

g = / t −1<br />

0,<br />

21<br />

10 32<br />

≤<br />

g = t /1000 −1<br />

0<br />

(<strong>6.</strong>35)<br />

22 32<br />

≤<br />

There are 8 <strong>design</strong> variables, 5 of them do not appear in the objective function. There<br />

are 22 constraints, mostly side constraints. The <strong>optimization</strong> model required by the fuzzy<br />

PD controller <strong>optimization</strong> engine is:<br />

min.<br />

s.t.<br />

f<br />

+ + +<br />

( A , A A )<br />

1 2<br />

,<br />

+ +<br />

( T t )<br />

g ,<br />

1 1<br />

,<br />

12<br />

− + +<br />

( T , T t )<br />

g ,<br />

2 1 2<br />

,<br />

− +<br />

( T )<br />

3 2<br />

, t32<br />

3<br />

22<br />

g ,<br />

− + +<br />

( A , T t )<br />

g ,<br />

4 1 1<br />

,<br />

12<br />

− + + −<br />

( A , T , T t )<br />

g ,<br />

5 2 1 2<br />

,<br />

− + −<br />

( A , T t )<br />

g ,<br />

6 3 2<br />

,<br />

−<br />

( )<br />

g ,<br />

7 A 1<br />

g<br />

+<br />

8<br />

( A 1<br />

)<br />

−<br />

( )<br />

9 A 2<br />

32<br />

22<br />

g http://mech.<strong>design</strong>er.yzu.edu.tw/<br />

40


g<br />

+<br />

10<br />

( A 2<br />

)<br />

g<br />

−<br />

11( A 3<br />

)<br />

g<br />

+<br />

12<br />

( A 3<br />

)<br />

−<br />

g<br />

13<br />

( T 1<br />

)<br />

+<br />

g<br />

14<br />

( T 1<br />

)<br />

−<br />

g<br />

15<br />

( T 2<br />

)<br />

+<br />

g<br />

16<br />

( T 2<br />

)<br />

g<br />

−<br />

17<br />

( t 12<br />

)<br />

g<br />

+<br />

18<br />

( t 12<br />

)<br />

g<br />

−<br />

19<br />

( t 22<br />

)<br />

g<br />

+<br />

20<br />

( t 22<br />

)<br />

g<br />

−<br />

21( t 32<br />

)<br />

+<br />

( )<br />

22 t 32<br />

<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

g (<strong>6.</strong>36)<br />

<strong>6.</strong><strong>6.</strong>2 Monotonicity Analysis<br />

Table <strong>6.</strong>12 is the corresponding monotonicity table for Equation (<strong>6.</strong>36). Figure <strong>6.</strong>17<br />

shows the complete output from MONO. Note that there are 3 conditionally critical sets.<br />

The monotonicity sign of <strong>design</strong> variables T 1 T 2 , t 12 , t 22 and t 32 do not appear in the<br />

objective function.<br />

41<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Table <strong>6.</strong>12 Monotonicity table of the heat exchange problem<br />

A 1 A 2 A 3 T 1 T 2 t 12 t 22 t 32<br />

F + + + . . . . .<br />

g 1 . . . + . + . .<br />

g 2 . . . - + . + .<br />

g 3 . . . . - . . +<br />

g 4 - . . + . - . .<br />

g 5 . - . + + . - .<br />

g 6 . . - . + . . -<br />

g 7 - . . . . . . .<br />

g 8 + . . . . . . .<br />

g 9 . - . . . . . .<br />

g 10 . + . . . . . .<br />

g 11 . . - . . . . .<br />

g 12 . . + . . . . .<br />

g 13 . . . - . . . .<br />

g 14 . . . + . . . .<br />

g 15 . . . . - . . .<br />

g 16 . . . . + . . .<br />

g 17 . . . . . - . .<br />

g 18 . . . . . + . .<br />

g 19 . . . . . . - .<br />

g 20 . . . . . . + .<br />

g 21 . . . . . . . -<br />

g 22 . . . . . . . +<br />

----------------- PC MONO V3.1 : Heat02.out ------------------<br />

TABLE 0 .<br />

1 2 3 4 5 6 7 8<br />

F + + + . . . . .<br />

G1 . . . + . + . .<br />

G2 . . . - + . + .<br />

42<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

G3 . . . . - . . +<br />

G4 - . . + . - . .<br />

G5 . - . + + . - .<br />

G6 . . - . + . . -<br />

G7 - . . . . . . .<br />

G8 + . . . . . . .<br />

G9 . - . . . . . .<br />

G10 . + . . . . . .<br />

G11 . . - . . . . .<br />

G12 . . + . . . . .<br />

G13 . . . - . . . .<br />

G14 . . . + . . . .<br />

G15 . . . . - . . .<br />

G16 . . . . + . . .<br />

G17 . . . . . - . .<br />

G18 . . . . . + . .<br />

G19 . . . . . . - .<br />

G20 . . . . . . + .<br />

G21 . . . . . . . -<br />

G22 . . . . . . . +<br />

******************************<br />

43<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

* GLOBAL FACTS *<br />

******************************<br />

* Irrelevant Variables :<br />

None<br />

* Unbounded Variables :<br />

None<br />

* Directed Equality Constraints :<br />

None<br />

* Critical Constraints :<br />

None<br />

* Uncritical (Relaxed) Constraints :<br />

None<br />

* Conditionally Critical Sets :<br />

One of ( G4 G7 ) must be critical for 1.<br />

One of ( G5 G9 ) must be critical for 2.<br />

One of ( G6 G11 ) must be critical for 3.<br />

Figure <strong>6.</strong>17 The output of MONO program of the heat exchange example<br />

<strong>6.</strong><strong>6.</strong>3 Defining the initial values and move limits of the <strong>design</strong> variables<br />

The inputs of the fuzzy PD controller <strong>optimization</strong> engine are the constraint function<br />

values and the change of the constraint function values. The quantitative definitions for the<br />

error inputs (e) are the values of the all constraint functions. The quantitative definitions<br />

for the error change inputs (Δe) are the change of the values of the all constraint functions.<br />

44<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

The values of the error inputs at different quantized levels are defined by the initial<br />

values of the constraint functions. In this example, the initial values of the <strong>design</strong> variables<br />

are:<br />

A 1 = 1000, A 2 = 2500, A 3 = 5000, T 1 = 100,<br />

T 2 = 400, t 12 = 200, t 22 = 300 t 32 = 500. (<strong>6.</strong>37)<br />

The constraint function values can be evaluated at the initial <strong>design</strong> point, which give:<br />

initial<br />

g 1<br />

initial<br />

g 5<br />

initial<br />

g 9<br />

initial<br />

g 13<br />

initial<br />

g 17<br />

initial<br />

g 21<br />

g 2<br />

initial<br />

initial<br />

initial<br />

initial<br />

= -0.25, = 0.50, = 0.00, = -1.20,<br />

initial<br />

= -1.00, = -0.20, = -0.90, g = -0.90,<br />

g 6<br />

initial<br />

initial<br />

= -0.60, = -0.75, = -0.80, g = -0.50,<br />

g 10<br />

initial<br />

initial<br />

initial<br />

= -0.90, = -0.90, = -0.98, g = -0.60,<br />

g 14<br />

initial<br />

initial<br />

initial<br />

= -0.95, = -0.80, = -0.97, g = -0.70,<br />

g 18<br />

g 22<br />

g 3<br />

initial<br />

g 7<br />

initial<br />

g 11<br />

g 15<br />

g 19<br />

initial<br />

= -0.98, = -0.50. (<strong>6.</strong>38)<br />

g 4<br />

8<br />

12<br />

16<br />

20<br />

In this example, the values of the move limits are:<br />

(ΔA 1 ) max = 1, (ΔA 2 ) max = 1, (ΔA 3 ) max = 1, (ΔT 1 ) max = 1,<br />

(ΔT 2 ) max = 1, (Δt 12 ) max = 1, (Δt 22 ) max = 1, (Δt 32 ) max = 1.<br />

(<strong>6.</strong>39)<br />

Therefore, the values of the quantization level of error change inputs are evaluated by<br />

Equation (5.10) as follows:<br />

Δg 1 (x 0 ) max = 0.005, Δg 2 (x 0 ) max = 0.008, Δg 3 (x 0 ) max = 0.020,<br />

Δg 4 (x 0 ) max = 0.023, Δg 5 (x 0 ) max = 0.051, Δg 6 (x 0 ) max = 0.006,<br />

Δg 7 (x 0 ) max = 1.0×10 -4 , Δg 8 (x 0 ) max = 1.0×10 -4 , Δg 9 (x 0 ) max = 1.6×10 -4 ,<br />

Δg 10 (x 0 ) max = 1.0×10 -4 , Δg 11 (x 0 ) max = 4.0×10 -5 , Δg 12 (x 0 ) max = 1.0×10 -4 ,<br />

Δg 13 (x 0 ) max = 1.0×10 -3 , Δg 14 (x 0 ) max = 1.0×10 -3 , Δg 15 (x 0 ) max = <strong>6.</strong>3×10 -5 ,<br />

Δg 16 (x 0 ) max = 1.0×10 -3 , Δg 17 (x 0 ) max = 2.5×10 -4 , Δg 18 (x 0 ) max = 1.0×10 -3 ,<br />

Δg 19 (x 0 ) max = 1.1×10 -4 , Δg 20 (x 0 ) max = 1.0×10 -3 , Δg 21 (x 0 ) max = 4.0×10 -5 ,<br />

Δg 22 (x 0 ) max = 1.0×10 -3 . (<strong>6.</strong>40)<br />

<strong>6.</strong><strong>6.</strong>4 The <strong>optimization</strong> results<br />

Using the initial <strong>design</strong> point and move limits described in the previous section, the<br />

fuzzy PD controller <strong>optimization</strong> engine terminates after 1963 iterations, when the change<br />

in objective function value in consecutive iterations is less than 0.01%, and all constraints<br />

45<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

are satisfied with a tolerance of 0.01% of the initial values of the constraints.. Table <strong>6.</strong>13<br />

shows the comparison of the results for the speed reducer <strong>design</strong> <strong>optimization</strong> problem.<br />

Figure <strong>6.</strong>18 shows the iteration history of the objective function.<br />

Table <strong>6.</strong>13 Comparison of the results for the heat exchange example<br />

Fuzzy PD Avriel and Williams [1971]<br />

Objective function value 7288.8 7049<br />

Iteration 1963<br />

A 1 951.8 567<br />

A 2 1529.5 1357<br />

A 3 4807.3 5125<br />

T 1 20<strong>6.</strong>6 181<br />

T 2 307.9 295<br />

t 12 193.4 219<br />

t 22 298.7 286<br />

t 32 407.8 395<br />

g 1 2.4×10 -5 0.0000<br />

g 2 1.4×10 -7 0.0000<br />

g 3 1.1×10 -3 0.0000<br />

g 4 -1.2×10 -4 3.2×10 -4<br />

g 5 -5.5×10 -2 <strong>6.</strong>6×10 -5<br />

g 6 0.0000 0.0000<br />

g 7 -0.8949 -0.8236<br />

g 8 -0.9048 -0.9433<br />

g 9 -0.3462 0.2631<br />

g 10 -0.8470 -0.8643<br />

g 11 -0.7920 -0.8049<br />

g 12 -0.5193 -0.4875<br />

g 13 -0.9516 -0.9448<br />

g 14 -0.7934 -0.8190<br />

g 15 -0.9675 -0.9661<br />

g 16 -0.6921 -0.7050<br />

g 17 -0.9483 -0.9543<br />

g 18 -0.8066 -0.7810<br />

g 19 -0.9665 -0.9650<br />

g 20 -0.7013 -0.7140<br />

g 21 -0.9755 -0.9747<br />

g 22 -0.5922 -0.6050<br />

46<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Figure <strong>6.</strong>18 The iteration history of objective function of the heat exchange example<br />

Reference<br />

Arora, J. S., 1989, “Introduction to optimum <strong>design</strong>,” McGraw-Hill, New York.<br />

Avriel, M., and Williams, A. C., 1971, “An extension of geometric programming with<br />

application in engineering <strong>optimization</strong>,” Journal of <strong>Engineering</strong> Mathematics, Vol. 5, pp.<br />

187-194.<br />

Belegundu, A. D., 1982, “A study of mathematical programming methods for<br />

structural <strong>optimization</strong>,” Department of Civil and Environmental <strong>Engineering</strong>, University<br />

of Iowa City, Iowa.<br />

Coello, C. A. C., and Montes, E. M., 2002, “Constraint-handling in genetic algorithm<br />

through the use of dominance-based tournament selection,” Advanced <strong>Engineering</strong><br />

Informatics, Vol. 16, pp. 193-203.<br />

Golinski, J., 1973, “An adaptive <strong>optimization</strong> system applied to machine synthesis,”<br />

Mechanism and Machine Synthesis, Vol. 8, pp. 419-43<strong>6.</strong><br />

47<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/


<strong>Chapter</strong> <strong>6.</strong> <strong>Engineering</strong> <strong>design</strong> <strong>optimization</strong> <strong>examples</strong><br />

Ragsdell, K. M. and Phillips, D. T., 1976, “Optimal <strong>design</strong> of a class of welded<br />

structures using geometric programming,” ASME Journal of <strong>Engineering</strong> for Industry, Vol.<br />

98, pp. 1021-1025.<br />

Rao, S. S., 1996, “<strong>Engineering</strong> Optimization: Theory and Practice, 3rd Ed.”, John<br />

Wiley & Sons, Inc.<br />

Rao, S. S., 1987, “Multiobjective <strong>optimization</strong> of fuzzy structure system,”<br />

International Journal for Numerical Methods in <strong>Engineering</strong>, Vol. 24, pp. 1157-1171.<br />

48<br />

http://mech.<strong>design</strong>er.yzu.edu.tw/

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!