ENGINEERING OPTIMIZATION

ME F320/MF F320

BITS Pilani
Pilani Campus
 LP Model in
Equation Form

2
From Graphical to
Algebraic Solution

3
From Graphical to Algebraic Solution

4
From Graphical to Algebraic Solution

5
From Graphical to Algebraic Solution
• Algebraic solution space
– m equations and n variables
– n>m
– basic solutions  set n−m variables equal to zero
– Maximum number of basic solutions: ?

6
From Graphical to Algebraic Solution
• Algebraic solution space
– m equations and n variables
– n>m
– basic solutions  set n−m variables equal to zero
n n!
– Maximum number of basic solutions: Cm = m! n−m !

– basic solutions correspond to corner points in graphical

7
From Graphical to Algebraic Solution
• Consider the problem:

Maximize:
z = 2x1 + 3x2
subject to
2x1 + x2 ≤ 4
x1 + 2x2 ≤ 5
x1 , x2 ≥ 0

8
From Graphical to Algebraic Solution
Maximize:
z = 2x1 + 3x2
subject to
2x1 + x2 + s1 = 4
x1 + 2x2 + s2 = 5
x1 , x2 , s1 , s2 ≥ 0

Clearly n = 4 and m = 2
⇒ Maximum basic solutions = 6 (note corner points are 4)

9
10
From Graphical to Algebraic Solution
• Which n−m variables should be set equal to zero to
target a specific corner point?
n
– target all basic solutions: Cm points

– problem size increases  enumerating all corner points

12
Simplex Method
• Investigate only a “select few” basic solutions.
• Iterations are required…
• Example:
Maximize:
z = 2x1 + 3x2
subject to
2x1 + x2 ≤ 4
x1 + 2x2 ≤ 5
x1 , x2 ≥ 0

13
Simplex Method
Step 1: Convert to standard form
Maximize:
z = 2x1 + 3x2
subject to
2x1 + x2 + s1 = 4
x1 + 2x2 + s2 = 5
x1 , x2 , s1 , s2 ≥ 0

Clearly n = 4 and m = 2

14
15
Simplex Method
• Iteration #1
– Start at the origin: all decision variables are zero.
– x1 = 0, x2 = 0 ⇒ s1 = 4, s2 = 5
– z=0
– check whether an increase in the values of non-basic
(variables that are zero) x1 and x2 above their current values
can improve (increase) the value of z  investigate the
objective function (z)  increase in x1 or x2 (or both) will
increase z.
– prepare for next iteration

16
Simplex Method
• Simplex method doesn’t allow simultaneous increases in
variables  targets one variable at a time.
• The variable to increase: one with largest rate of
improvement in z.

z = 2x1 + 3x2

• Elect to increase x2

17
Simplex Method
• x2 must be increased until corner point B is reached

18
 Simplex Method
• x2 must be increased until corner point B is reached

𝟏𝟓 𝟏 𝟑
𝐁: 𝐳 𝐱𝟏 , 𝐬𝟐 = + 𝐱 𝟏 − 𝐬𝟐
𝟐 𝟐 𝟐

𝐀: 𝐳(𝐱𝟏 , 𝐱𝟐 ) = 𝟐𝐱𝟏 + 𝟑𝐱𝟐

19
Simplex Method
• Iteration #2
– Reached point B.
– x1 = 0, x2 = 2.5 ⇒ s1 =? , s2 = 0
15
– z= 2
– x2 is now basic variable (entered the solution), while s2 is non-
basic (exited the solution).
– prepare for next iteration.

20
 Simplex Method
• x1 must be increased until corner point C is reached

𝟏𝟓 𝟏 𝟑
𝐁: 𝐳 𝐱𝟏 , 𝐬𝟐 = + 𝐱 𝟏 − 𝐬𝟐 𝟏 𝟒
𝟐 𝟐 𝟐 𝐂: 𝐳 𝐬𝟏 , 𝐬𝟐 = 𝟖 − 𝐬 𝟏 − 𝐬𝟐
𝟑 𝟑

𝐀: 𝐳(𝐱𝟏 , 𝐱𝟐 ) = 𝟐𝐱𝟏 + 𝟑𝐱𝟐

21
Simplex Method
• Iteration #3
– Reached point C.
– x1 = 1, x2 = 2 ⇒ s1 =? , s2 =?
– z=8
– x1 is now basic variable (entered the solution), while s1 is non-
basic (exited the solution).
– Iterations complete.

22
 Simplex Method
• How point F can be discarded without looking at the
graph?
𝐅: 𝐳 𝐱𝟏 , 𝐬𝟏 = 𝟏𝟐 − 𝟒𝐱𝟏 − 𝟑𝐬𝟏

𝟏𝟓 𝟏 𝟑
𝐁: 𝐳 𝐱𝟏 , 𝐬𝟐 = + 𝐱 𝟏 − 𝐬𝟐
𝟐 𝟐 𝟐 𝟏 𝟒
𝐂: 𝐳 𝐬𝟏 , 𝐬𝟐 = 𝟖 − 𝐬𝟏 − 𝐬𝟐
𝟑 𝟑

𝐀: 𝐳(𝐱𝟏 , 𝐱𝟐 ) = 𝟐𝐱𝟏 + 𝟑𝐱𝟐

23
Simplex Method
• Always connects corner points.
• Path to optimum: ABC.
• Each corner point along path is associated with an
iteration.
• Method always moves alongside edges of solution
space  method doesn’t cut across solution space.
• Cannot go from A to C directly.

24
10th ed. 3-18, 9th ed. set 3.3A (Q. 3)
Consider the three-
dimensional LP solution
space, whose feasible
extreme points are A, B, . . .
, and J.
(a) Which of the following
pairs of corner points
cannot represent successive
simplex iterations: (A, B),
(H, I ), (E, H ), and (A, I )?
Explain why.

25
10th ed. 3-18, 9th ed. set 3.3A (Q. 3)
(b) Suppose that simplex
iterations start at A and that
optimum occurs at H.
Indicate whether any of the
following paths are not
legitimate for the algorithm,
and state the reason.
(i) AB GH
(ii) ACIH
(iii)ADFCABG
H
26
 10th ed. 3-20, 9th ed. set 3.3A (Q. 5)
For each of the given objective
functions and the solution space,
select the non-basic variable that
leads to the next simplex corner
point, and determine the associated
improvement in z.
(a) Maximize 𝐳 = 𝐱 𝟏 − 𝟐𝐱 𝟐 + 𝟑𝐱 𝟑
(b) Maximize 𝐳 = 𝟓𝐱 𝟏 + 𝟐𝐱 𝟐 + 𝟒𝐱 𝟑
(c) Maximize 𝐳 = −𝟐𝐱 𝟏 + 𝟕𝐱 𝟐 + 𝟐𝐱 𝟑
(d) Maximize 𝐳 = 𝐱 𝟏 + 𝐱 𝟐 + 𝐱 𝟑

27
 Simplex
Computations

28
Simplex Computations
Reddy Mikks model
Maximize:
z = 5x1 + 4x2 + 0s1 + 0s2 + 0s3 + 0s4
subject to
6x1 + 4x2 + s1 + 0s2 + 0s3 + 0s4 = 24
x1 + 2x2 + 0s1 + s2 + 0s3 + 0s4 = 6
−x1 + x2 + 0s1 + 0s2 + s3 + 0s4 = 1
0x1 + x2 + 0s1 + 0s2 + 0s3 + s4 = 2
x1 , x2 , s1 , s2 , s3 , s4 ≥ 0

Clearly n = 6 and m = 4
29
Simplex Tableau
• Starting Simplex Tableau (starts at origin)
z − 5x1 − 4x2 = 0
Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 −5 −4 0 0 0 0 0
s1 row 0 6 4 1 0 0 0 24
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2

• Do we have a maxima?
30
Entering Variable
• Increase x1 (most positive objective coefficient)
or
• Simplex tableau: select non-basic variable with most
negative coefficient in objective equation (simplex
optimality condition).

• x1: entering variable because it enters basic solution.

31
Leaving Variable
• Compute ratios of the right-hand side of equations
(Solution column) to corresponding (strictly) positive
constraint coefficients under the entering variable (x1).
Basic Entering x1 Solution Ratio (intercept)
24
s1 row 6 24 x1 = = 4 → minimum
6
6
s2 row 1 6 x1 = = 6
1
1
s3 row −1 1 x1 = = −1 (ignore)
−1
2
s4 row 0 2 x1 = = ∞ (ignore)
0
32
Leaving Variable
Basic Entering x1 Solution Ratio (intercept)
24
s1 row 6 24 x1 = = 4 → minimum
6
6
s2 row 1 6 x1 = = 6
1
1
s3 row −1 1 x1 = = −1 (ignore)
−1
2
s4 row 0 2 x1 = = ∞ (ignore)
0
Conclusion: x1 enters (x1 = 4) and s1 leaves (s1 = 0)

Minimum Ratio Test (MRT)

34
35
Leaving Variable
How do the computed ratios determine the leaving
variable and the value of the entering variable?
• Computed ratios are actually the intercepts of the
constraint lines with the (entering variable) x1-axis.
• x1 must be increased to smallest nonnegative intercept
with the x1-axis (= 4) to reach corner point B.
• Point B: s1 associated with constraint equals zero 
leaving variable (simplex feasibility condition) 
guarantees feasibility of new solution.

36
 New Feasible Basic Solution
• Solution point B determined by “swapping” entering
variable x1 and leaving variable s1 in simplex tableau.
• Swapping using Gauss-Jordan row operations
Pivot element
Enter
Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 −5 −4 0 0 0 0 0
Leave s row 0 6 4 1 0 0 0 24
1
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2

37
Gauss-Jordan row operations
1. Pivot row
– Basic column: leaving variable replaced with entering variable.
Current pivot row
– New pivot row =
Pivot element

Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 −5 −4 0 0 0 0 0
x1 row 0 6/6 4/6 1/6 0 0 0 24/6
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2
38
Gauss-Jordan row operations
1. Pivot row
– Basic column: leaving variable replaced with entering variable.
Current pivot row
– New pivot row =
Pivot element

Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 −5 −4 0 0 0 0 0
x1 row 0 1 2/3 1/6 0 0 0 4
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2
39
Gauss-Jordan row operations
2. Other rows (including z)
– New row = Current row − (Pivot column coefficient) ×
(New pivot row)

Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 −5 −4 0 0 0 0 0
x1 row 0 1 2/3 1/6 0 0 0 4
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2
40
Gauss-Jordan row operations
2. Other rows (including z)
– New z-row = Current z-row −(−5)×(New x1-row)
= (1, −5, −4, 0, 0, 0, 0, 0, 0) −(−5)×(0, 1, 2/3, 1/6, 0, 0, 0, 4)
= (1, 0, −2/3, 5/6, 0, 0, 0, 20)

Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 0 −2/3 5/6 0 0 0 20
x1 row 0 1 2/3 1/6 0 0 0 4
s2 row 0 1 2 0 1 0 0 6
s3 row 0 −1 1 0 0 1 0 1
s4 row 0 0 1 0 0 0 1 2
41
New Simplex Tableau
Basic z x1 x2 s1 s2 s3 s4 Solution
z row 1 0 −2/3 5/6 0 0 0 20
x1 row 0 1 2/3 1/6 0 0 0 4
s2 row 0 0 4/3 −1/6 1 0 0 2
s3 row 0 0 5/3 1/6 0 1 0 5
s4 row 0 0 1 0 0 0 1 2
• New tableau structure similar to that of starting tableau
– constraint coefficients of basic variable form an identity
matrix.
– non-basic variables (x2, s1) are zero  solution-column yields
new basic solution (x1 = 4, s2 = 2, s3 = 5, s4 = 2)
– This “conditioning” is due to Gauss-Jordan row operations.
42
Thank you!

43