Engineering Optimization: Theory and Practice, Fourth Edition

(Martin Jones) #1
6.7 Simplex Method 333

SOLUTION


Iteration 1


Step 1: The function value at each of the vertices of the current simplex is given by


f 1 =f(X 1 )= 4. 0 − 4. 0 + 2 ( 16. 0 )+ 2 ( 16. 0 )+ 16. 0 = 80. 0
f 2 =f(X 2 )= 5. 0 − 4. 0 + 2 ( 25. 0 )+ 2 ( 20. 0 )+ 16. 0 = 107. 0
f 3 =f(X 3 )= 4. 0 − 5. 0 + 2 ( 16. 0 )+ 2 ( 20. 0 )+ 25. 0 = 96. 0

Therefore,

Xh=X 2 =

{

5. 0

4. 0

}

, f (Xh) = 107. 0 ,

Xl=X 1 =

{

4. 0

4. 0

}

, and f (Xl) = 80. 0

Step 2: The centroidX 0 is obtained as


X 0 =

1

2

(X 1 +X 3 )=

1

2

{

4. 0 + 4. 0

4. 0 + 5. 0

}

=

{

4. 0

4. 5

}

with f (X 0 ) = 87. 75

Step 3: The reflection point is found as


Xr= 2 X 0 −Xh=

{

8. 0

9. 0

}


{

5. 0

4. 0

}

=

{

3. 0

5. 0

}

Then

f (Xr)= 3. 0 − 5. 0 + 2 ( 9. 0 )+ 2 ( 15. 0 )+ 25. 0 = 71. 0

Step 4: Asf (Xr) < f (Xl) we find, Xeby expansion as


Xe= 2 Xr−X 0 =

{

6. 0

10. 0

}


{

4. 0

4. 5

}

=

{

2. 0

5. 5

}

Then
f (Xe)= 2. 0 − 5. 5 + 2 ( 4. 0 )+ 2 ( 11. 0 )+ 30. 25 = 56. 75

Step 5: Sincef (Xe) < f (Xl) we replace, XhbyXeand obtain the vertices of the new
simplex as


X 1 =

{

4. 0

4. 0

}

, X 2 =

{

2. 0

5. 5

}

, and X 3 =

{

4. 0

5. 0

}

Step 6: To test for convergence, we compute


Q=

[

( 80. 0 − 87. 75 )^2 + 6 ( 5. 75 − 87. 75 )^2 + 6 ( 9. 0 − 87. 75 )^2

3

] 1 / 2

= 91. 06

As this quantity is not smaller thanε, we go to the next iteration.
Free download pdf