College of Management, NCTU Operation Research I Fall, 2008

Chap6 Duality Theory and Sensitivity Analysis

‰ The rationale of duality theory

Max 4x1 + x2 + 5x3 + 3x4
S.T. x1 – x2 – x3 + 3x4 ≤ 1
5x1 + x2 + 3x3 + 8x4 ≤ 55
–x1 + 2x2 + 3x3 – 5x4 ≤ 3
x1~x4 ≥ 0
9 If we multiply a certain number (y ≥ 0) on both sides of any constraint, the
result won’t be affected.
9 The arbitrary y is called the multiplier. y1, y2, and y3 are the multipliers of
constraints 1, 2, and 3, respectively.
9 How can we tell the range of the optimal objective value? (Let Z* be the
optimal value)
Consider 4x1 + x2 + 5x3 + 3x4 ≤ 5/3 (5x1 + x2 + 3x3 + 8x4)

Î Z* ≤ 5/3(55) = 275/3
Similarly
4x1 + x2 + 5x3 + 3x4 ≤ constraint 2 + constraint 3
= (5x1+ x2 + 3x3 + 8x4) + (–x1 + 2x2 + 3x3 – 5x4)

Î Z* < 55 + 3 = 58
9 That is
(1) Z* ≤ 0．b1 + 5/3．b2 + 0．b3
(2) Z* ≤ 0．b1 + 1． b2 + 1．b3
……………

9 In general, we want to determine the values of yi such that the following

statement holds.
Z* ≤ 4x1 + x2 + 5x3 + 3x4 ≤ y1(x1 – x2 – x3 + 3x4) + y2(5x1 + x2 + 3x3 + 8x4) +
y3(–x1 + 2x2 + 3x3 – 5x4)
≤ y1 + 55y2 + 3y3

9 Issue: which is the best estimate of the optimal value?

Jin Y. Wang 6-1

College of Management, NCTU Operation Research I Fall, 2008

9 So, we can conclude that

(1) y1, y2, y3 ≥ 0
(2) 4 ≤ y1 + 5y2 – y3
1 ≤ –y1 + y2 + 2y3
5 ≤ –y1 + 3y2 + 3y3
3 ≤ 3y1 + 8y2 – 5y3
(3) Since we want to estimate Z* as best as possible, i.e. we want to have
y1 + 55y2 + 3y3 as less as possible. Therefore, we have
Min y1 + 55y2 + 3y3

9 The above formulation is called the Dual problem.

9 The original formulation is called the Primal problem.
9 Note: (1) Primal: coefficients of the objective functionÅÆ Dual: RHS
(2) Primal: RHS ÅÆ Dual: coefficients of the objective function
(3) Primal: coefficients of a variable in the functional constraint ÅÆ
Dual: coefficients in a functional constraint
(4) Max ÅÆ Min
(5) ≤ ÅÆ ≥
‰ Example of finding the dual problem
Max 5x1 + 4x2 + 3x3
S.T. 2x1 + 3x2 + x3 ≤ 5
4x1 + x2 + 2x3 ≤ 11
3x1 + 4x2 + 2x3 ≤ 8
x1, x2, x3 ≥ 0

‰ Matrix representation of the primal and dual problems

Jin Y. Wang 6-2

College of Management, NCTU Operation Research I Fall, 2008

9 Primal Problem (P)

Max Z = ∑ j =1 c j x j
n
Max Z = cx

∑ j =1 aij x j ≤ bi , for i = 1, 2, …, m
n
S.T. S.T. Ax ≤ b
xj ≥ 0, for j = 1, 2, …, n x ≥ 0

9 Dual Problem (D)

Min W = ∑ j =1bi yi
m
Min W = yb

∑i =1 aij yi ≥ c j , for j = 1, 2, …, n
m
S.T. S.T. yA ≥ c
yi ≥ 0, for i = 1, 2, …, m y ≥ 0
‰ Weak duality property
If x is a feasible primal solution, and y is a feasible dual solution, then cx ≤ yb
<Proof>

9 Corollary: If the primal is feasible and unbounded, the dual is infeasible and
vice versa.

‰ Strong duality property

If x* is an optimal solution for the primal problem and y* is an optimal solution
for the dual problem, then cx* = y*b.
‰ A corollary for weak and strong duality property
9 These two properties implies that cx < yb for feasible solutions if one or both
of them are not optimal for their respective problems, whereas equality holds
when both are optimal.

‰ Complementary solutions property

Jin Y. Wang 6-3

College of Management, NCTU Operation Research I Fall, 2008

9 At each iteration, the simplex method simultaneously identifies a CPF solution

x for the primal problem and a complementary solution y for the dual
problem (in tabular form, the coefficients of the slack variables in row 0),
where cx = yb.
9 If x is not optimal for the primal problem, then y is not feasible for the dual
problem.
P: Max Z = 5x1 + 4x2 + 3x3
S.T. 2x1 + 3x2 + x3 ≤ 5
4x1 + x2 + 2x3 ≤ 11
3x1 + 4x2 + 2x3 ≤ 8
x1, x2 x3 ≥ 0
D: Min W = 5y1 +11y2 + 8y3
S.T. 2y1 + 4y2 + 3y3 ≥ 5
3y1 + y2 + 4y3 ≥ 4
y1 + 2y2 + 2y3 ≥ 3
y1, y2, y3 ≥ 0

Z - 5x1 - 4x2 - 3x3 =0

2x1 + 3x2 + x3 + x4 =5
4x1 + x2 + 2x3 + x5 = 11
3x1 – 4x2 + 2x3 + x6 = 8
Æ y1 = , y2 = , y3 =

Z - 1/2 x1 + 2x2 + 3/2x6 = 12

1/2 x1 + x2 + x4 + 1/2x6 = 1
x1 – 3x2 + x5 + x6 = 3
3/2x1 + 2x2 + x3 + 1/2x6 = 4
Æ y1 = , y2 = , y3 =

Z + 3x2 + x4 + x6 = 13
x1 + 2x2 + 2x4 – x6 =2
–5x2 – 2x4+ x5 =1
–x2 + x3 – 3x4 + 2x6 =1
Æ y1 = , y2 = , y3 =

9 Recall that

Jin Y. Wang 6-4

College of Management, NCTU Operation Research I Fall, 2008

Coefficient of:
Iteration Basic Variable Right Side
Z Original Variable Slack Variables
Z 1 cBB-1A – c cBB-1 cBB-1b
Any
xB 0 B-1A B-1 B-1b
9 Let y (dual variables) = cBB-1 (the shadow price). Obviously, Z (=cx) = yb.
9 The coefficients of original and slack variables in the objective row are yA–c
and y, respectively.
9 Note that yA–c represents the value of surplus variables in the dual.

9 If the optimality hasn’t yet reached, at least one coefficient in the objective
row is < 0. Thus, at lease one constraint of the dual is violated.

9 This also explains the strong duality property in some sense.

‰ Complementary optimal solutions property

9 At the final iteration, the simplex method simultaneously identifies an optimal
solution x* for the primal problem and a complementary optimal solution y*
for the dual problem (the coefficients of the slack variables in row 0), where
cx* = y*b
9 In fact, y* = CBB-1 are the shadow prices.
‰ Symmetry property
9 For any primal problem and its dual problem, all relationships between them
must be symmetric because the dual of this dual problem is this primal
problem.
‰ Duality theorem – All possible relationships between the primal and dual
problems
9 If one problem has feasible solutions and a bounded objective function (and so
has an optimal solution), then so does the other problem.
9 If one problem has feasible solutions and an unbounded objective function
(and so no optimal solution), then the other problem has no feasible solutions.

Jin Y. Wang 6-5

College of Management, NCTU Operation Research I Fall, 2008

9 If one problem has no feasible solutions, then the other problem has either no
feasible solutions or an unbounded objective function.
Dual
Optimal Infeasible Unbounded
Optimal
Primal Infeasible
Unbounded
Example for primal infeasibleÆdual infeasible
Max 2x1 – x2 Min y1 – 2y2
S.T. x1 – x2 ≤ 1 S.T. y1 – y2 ≥ 2
–x1 + x2 ≤ –2 –y1 + y2 ≥ –1
x1, x2 ≥ 0 y1, y2 ≥ 0
‰ Applications of dual problem
9 Dual problem can be solved directly by the simplex method.
¾ The number of functional constraints affects the computational effort of
the simplex method far more than the number of variables does.

9 It is useful for evaluating a proposed solution for the primal problem.

‰ Economic interpretation of duality

9 Primal Problem (P)
Max Z = ∑ j =1 c j x j
n

∑ j =1 aij x j ≤ bi , for i = 1, 2, …, m
n
S.T.
xj ≥ 0, for j = 1, 2, …, n

9 Dual Problem (D)

Min W = ∑ j =1bi yi
m

∑ i =1 aij yi ≥ c j , for j = 1, 2, …, n
m
S.T.
yi ≥ 0, for i = 1, 2, …, m

Jin Y. Wang 6-6

College of Management, NCTU Operation Research I Fall, 2008

9 If resource i is not used for production, we rent it out with unit rental of yi.
9 ∑
m
i =1
aij y i means the rental income if we reduce the production amount of
product j by one.

∑i =1 aij yi ≥ c j means the rental income be greater than the sales income.
m
9
9 The objective of dual is to minimize the rental income in order to obtain the
reasonable rental price.

‰ Primal-Dual relationships
9 Dual is a LP problem and also has corner-point solutions.
9 By using the augmented form, we can express these corner-point solutions as
basic solutions.
9 Because the functional constraints have the ≥ form, this augmented form is
obtained by subtracting the surplus from the left-hand side of each constraint.
‰ Complementary basic solutions property
9 Each basic solution in the primal problem has a complementary basic solution
in the dual problem, where their respective objective function values are equal.
9 Recall again:
Coefficient of:
Iteration Basic Variable Right Side
Z Original Variable Slack Variables
Any Z 1 cBB-1A – c cBB-1 cBB-1b

9 Given row 0 (objective row) of the simplex tableau for the primal basic
solution, the complementary dual basic solution can be easily found.
Primal Variable Associated Dual Variables
(Original variable) xj (Surplus Variable) ym+j (j = 1, 2, …, n)
(Slack Variable) xn+i (Original Variable) yi (i = 1, 2, …., m)

Jin Y. Wang 6-7

College of Management, NCTU Operation Research I Fall, 2008

9 The m basic variables for the primal problem are required to have a coefficient
of zero in the objective row, which thereby requires the m associated dual
variables to be zero, i.e., nonbasic variables for the dual problem.

9 The values of the remaining n (basic) variables will be the simultaneous

solution to the system of equations.
9 The dual solution read from row 0 must also be a basic solution.
‰ Complementary slackness property
Primal Variable Associated Dual Variable
Basic Nonbasic (m variables)
Nonbasic Basic (n variables)
9 If one of (primal) variables is a basic variable (> 0), then the corresponding
one (dual) must be a nonbasic variable (= 0).
9 Wyndor Example:
Max Z = 3x1+ 5x2 Min W = 4y1 + 12y2 + 18y3
S.T. x1 ≤ 4 S.T. y1 + 3y3 ≥ 3
2x2 ≤ 12 2y2 + 2y3 ≥ 5
3x1 + 2x2 ≤ 18
x1, x2 ≥ 0 y1, y2, y3 ≥ 0

Max Z = 3x1+ 5x2 Min W = 4y1 + 12y2 + 18y3

S.T. x1 + x3 = 4 S.T. y1 + 3y3 – y4 =3
2x2 + x4 = 12 2y2 + 2y3 – y5 = 5
3x1 + 2x2 + x5 = 18

¾ Given x = (4, 6, 0, 0, -6), we know that x1, x2, and x5 are basic variables.
The coefficients of these variables in row 0 are zero. Thus, y4, y5, and y3
are nonbasic variables (=0).
y1 =3
2y2 =5

Jin Y. Wang 6-8

College of Management, NCTU Operation Research I Fall, 2008

¾ Given x = (0, 6, 4, 0, 6), we have

9 Let x1*~xn* be primal feasible solution, y1*~ym* be dual feasible solution.

Then, they are both optimal if and only if
m
(1) xj* > 0 implies ∑a
i =1
ij yi* = c j .

∑a
*
(2) ij x *j < bi implies yi = 0.
j =1

9 Ex: Max 5x1 + 6x2 + 9x3 + 8x4 Min 5y1 + 3y2

S.T. x1 + 2x2 + 3x3 + x4 ≤ 5 S.T. y1 + y2 ≥ 5
x1 + x2 + 2x3 + 3x4 ≤ 3 2y1 + y2 ≥ 6
x1,…, x4 ≥ 0 3y1 + 2y2 ≥ 9
y1 + 3y2 ≥ 8
y1, y2 ≥ 0

9 Ex: Max 8x1 – 9x2 + 12x3 + 4x4 + 11x5

Jin Y. Wang 6-9

College of Management, NCTU Operation Research I Fall, 2008

S.T. 2x1 – 3x2 + 4x3 + x4 + 3x5 ≤ 1

x1 + 7x2 + 3x3 – 2x4 + x5 ≤ 1
5x1 + 4x2 – 6x3 + 2x4 + 3x5 ≤ 22
x1,.., x5 ≥ 0

‰ Complementary optimal basic solutions property

9 Each optimal basic solution in the primal problem has a complementary
optimal basic solution in the dual problem, where their respective objective
functions are equal.
¾ The coefficients in the primal row 0 are all ≥ 0 Æ All dual variables
(including decision and surplus variables) are all ≥ 0 .
¾ Nonnegative surplus variables imply that functional constraints are
satisfied.
¾ Nonnegativity constraints are satisfied.
‰ How to write the dual if the primal is not in the standard form
Max 3x1 + x2 – x3
S.T. x1 + x2 + 4x3 ≤ 5
x1 – x2 – 7x3 = 7
x1, x2 ≥ 0, x3 is free

Jin Y. Wang 6-10

College of Management, NCTU Operation Research I Fall, 2008

‰ Summary
Primal (Dual) Dual (Primal)
Max Min
th
i constraint is “ ≤ ” (sensible) ith variable ≥ 0 (sensible)
ith variable ≥ 0 (sensible) ith constraint is “ ≥ ” (sensible)
ith constraint is “=” (odd) ith variable is unsigned (odd)
ith variable is unsigned (odd) ith constraint is “=” (odd)
ith constraint is “ ≥ ” (bizarre) ith variable ≤ 0 (bizarre)
ith variable ≤ 0 (bizarre) ith constraint is “ ≤ ” (bizarre)
9 For “ ≥ ” constraint in primal, we can convert “ ≥ ” to “ ≤ ” by multiplying both
sides by -1.
9 For negative variable x (< 0) in primal, we can use another variable, whose
value is (-1)*x, to replace x.
9 Ex: Max 5x1 + 4x2
S.T. x1 – x2 ≥ 5
3x1 + x2 ≥ 2
–x1 + 2x2 = 9
x1 ≥ 0, x2 ≤ 0

9 The radiation therapy example

Min Z = 0.4x1 + 0.5x2

Jin Y. Wang 6-11

College of Management, NCTU Operation Research I Fall, 2008

S.T. 0.3x1 + 0.1x2 ≤ 2.7

0.5x1 + 0.5x2 = 6
0.6x1 + 0.4x2 ≥ 6
x1, x2 ≥ 0

‰ Dual simplex method (section 7.1)

Max –8x1 – 8x2 – 9x3 Min y1 – 8y2 – 2y3
S.T. x1 + x2 + x3 ≤ 1 S.T. y1 – 2y2 + y3 ≥ –8
–2x1 – 4x2 – x3 ≤ –8 y1 – 4y2 – y3 ≥ –8
x1 – x2 – x3 ≤ –2 y1 – y2 – y3 ≥ –9
x1 ~ x3 ≥ 0 y1 ~ y3 ≥ 0
9 Note that the current primal does not have a feasible solution. However, the
dual does have one.
9 Working on the dual is one way of tackling this situation.
9 Another way is working the complementary counterpart on the primal.
9 Convert the dual to standard form and add slack variables (currently, the basic
solution is feasible).
W + y1 – 8y2 – 2y3 =0
–y1 + 2y2 – y3 + y4 =8
–y1 + 4y2 + y3 + y5 =8
–y1 + y2 + y3 + y6 = 9

9 Add slack variables to the primal (currently, the basic solution is infeasible).
Z + 8x1 + 8x2 + 9x3 =0
x1 + x2 + x3 + x4 =1
–2x1 – 4x2 – x3 + x5 = –8
x1 – x2 – x3 + x6 = –2
9 In the dual problem, pick y3 to enter and pick y5 to leave.

Jin Y. Wang 6-12

College of Management, NCTU Operation Research I Fall, 2008

9 The coefficients of the dual objective function correspond to the current value
of primal’s basic variable. The dual current values of basic variables
correspond to the coefficients of the primal objective function.
9 The dual entering column corresponds to the primal leaving row.
9 In dual: Find a nonbasic variable with a negative coefficient in row 0 (positive
coefficient in algebraic form) to enter.
ÍÎ
In Primal: Find a basic variable with a negative current value to leave.
9 In dual: Find the variable corresponding to the entering variable with a positive
coefficient and smallest ratio to leave.
ÍÎ
In Primal: Find the variable with a negative coefficient in the leaving row and
smallest ratio to enter.
9 For dual, pick y3 to enter and pick y5 to leave.
W – y1 – 0y2 + 2y5 = 16
–2y1 + 6y2 + y4 + y5 = 16
–y1 + 4y2 + y3 + y5 =8
0y1 – 3y2 – y5 + y6 = 1
9 For primal, pick x6 to leave, x2 to enter.
Z + 16x1 + x3 + 8x6 = –16
2x1 + 0x3 + x4 + x6 = –1
–6x1 + 3x3 + x5 – 4x6 = 0
–x1 + x2 + x3 – x6 =2

‰ Conditions for adopting the dual simplex method

Jin Y. Wang 6-13

College of Management, NCTU Operation Research I Fall, 2008

9 The prerequisite for using the dual simplex is the dual has initial feasible
solution. Thus, the primal coefficient of row 0 must be nonnegative.
‰ What if both primal and dual do not have a feasible initial solution?
9 Big M or two-phase method.
‰ Another example for dual simplex
Z + 4x2 + x4 + x5 = 12
x1 – 3x2 + 11x4 – x5 = –4
x2 + x3 – 3x4 + 2x5 = –8

Jin Y. Wang 6-14

College of Management, NCTU Operation Research I Fall, 2008

‰ Sensitivity Analysis
Max Z = cx
Ax = b
x ≥ 0
Recalled that we can rewrite it as the following
Max Z = cBxB + cNxN
S.T. BxB + NxN = b
xB, xN ≥ 0
Î Max Z + (cBB-1N - cN)xN = cBB-1b
xB = B-1b – B-1NxN

What happens to the optimal solution if

(1) b is changed
(2) c is changed
(3) A is changed
(4) add a new variable
(5) add a new constraint
‧
‰ b is changed to b
Max cx
‧
S.T. Ax = b
x ≥ 0
9 In this case, the original optimal basic solution is no more feasible, since b is
‧
changed to b.
‧
9 In order to make sure that the basic won’t change, we need to check B-1 b.

‧
If B-1 b ≥ 0 Î The basic won’t change, we are at optimality since
(cBB-1N - cN ≥ 0)

‧
Otherwise (B-1 b < 0) Î The new basic solution is infeasible. That is, the basic
change. Need to solve the problem by using dual simplex.

9 Ex: Max Z = 3x1 + 5x2

Jin Y. Wang 6-15

College of Management, NCTU Operation Research I Fall, 2008

S.T. x1 + x3 = 4
2x2 +x4 = 12
3x1 + 2x2 +x5 = 18

⎡ 2⎤ ⎡ x3 ⎤
x = ⎢⎢6⎥⎥ , xB =
* ⎢ x ⎥ , x = ⎡ x4 ⎤
⎢ 2⎥ N ⎢x ⎥
⎢⎣2⎥⎦ ⎢⎣ x1 ⎥⎦ ⎣ 5⎦

⎡4⎤ ⎡4⎤
b changes from ⎢12⎥ to ⎢⎢15⎥⎥
⎢ ⎥
⎢⎣18⎥⎦ ⎢⎣18⎥⎦

⎡4⎤ ⎡4⎤
b changes from ⎢12⎥ to ⎢⎢24⎥⎥
⎢ ⎥
⎢⎣18⎥⎦ ⎢⎣18 ⎥⎦

9 The final tableau can be obtained based on the given information.

Based on Z + (cBB-1N - cN)xN = cBB-1b and xB = B-1b – B-1NxN or
recall the fundamental insight table.

Jin Y. Wang 6-16

College of Management, NCTU Operation Research I Fall, 2008

Coefficient of:
Iteration Basic Variable Right Side
Z Original Variable Slack Variables
Z 1 cBB-1A – c cBB-1 cBB-1b
Any
xB 0 B-1A B-1 B-1b

Final tableau after RHS changing

Basic Coefficient of: Right Side
Variable x1 x2 x3 x4 x5
Z
x3
x2
x1
Apply dual simplex. We pick x1 to leave and x4 to enter.
Basic Coefficient of:
Variable x1 x2 x3 x4 x5 Right Side
Z
x3
x2
x4

‰ Allowable range for RHS given that the basic won’t change
Max Z = 3x1 + 5x2
S.T. x1 +x3 = 4
2x2 +x4 = 12
3x1 + 2x2 +x5 = 18
⎡4⎤ ⎡ 4 ⎤
b changes from ⎢12⎥ to ⎢⎢12 + Δ ⎥⎥ .
⎢ ⎥ We need to guarantee B-1b ≥ 0
⎢⎣18⎥⎦ ⎢⎣ 18 ⎥⎦

‰ Simultaneous changes in Right-Hand Sides

Jin Y. Wang 6-17

College of Management, NCTU Operation Research I Fall, 2008

9 Adopt the 100 percent rule.

9 Calculate for each change percentage of the allowable change (increase or
decrease) for that right-hand side to remain within its allowable range.
9 If the sum of the percentage changes does not exceed 100 percent, the shadow
prices definitely will still be valid (basic remains the same).
9 If the sum does exceed 100 percent, then we cannot be sure.
9 Example (Wyndor)

Constraint Shadow Price Current RHS Allowable Allowable

Increase Decrease
Plant 1 0 4 Infinite 2
Plant 2 1.5 12 6 6
Plant 3 1 18 6 6

⎡4⎤ ⎡4⎤
Change b from ⎢12⎥ to ⎢⎢15⎥⎥ .
⎢ ⎥
⎢⎣18⎥⎦ ⎢⎣15⎥⎦

b2: percentage of allowable increase =

b3: percentage of allowable decrease =

Conclusion:

‧
‰ c is changed to c
‧
Supposed c is changed to c
‧
Î Max c x
Ax = b
x ≥ 0
9 Notice that the optimal solution computed before is still .
9 In order to check the optimality, look at the objective row and be sure
.
9 That is, if the above inequality holds, we are still at optimality. Otherwise,
use (revised) simplex method starting with x* as initial feasible solution.
9 Ex: Max Z = 3x1 + 5x2

Jin Y. Wang 6-18

College of Management, NCTU Operation Research I Fall, 2008

S.T. x1 + x3 = 4
2x2 + x4 = 12
3x1 + 2x2 + x5 = 18
At the final iteration, we have

If c changes from [3 5] to [4 5] .

‰ How much can c change and still have the same optimal solution?
9 The coefficients of nonbasic variables are changed Æ Let cBB-1N - cN remains
≥ 0

Example 1: xB = [ x1, x2, x3]T, xN = [x4, x5, x6]T, cB = [1 2 1 ], cN = [2 –3 –3]

⎡1 3 − 1⎤ ⎡ 2 0 0⎤
B = ⎢⎢0 1 4 ⎥⎥ , N = ⎢− 1 0 1⎥
⎢ ⎥
⎢⎣0 2 2 ⎥⎦ ⎢⎣ 2 1 0⎥⎦

Jin Y. Wang 6-19

College of Management, NCTU Operation Research I Fall, 2008

9 The coefficients of basic variables are changed Æ same reasoning.

Example 2 (Wyndor): At the final iteration, xB = [x3, x2, x1]T, xN = [x4, x5]T,
⎡1 0 1 ⎤ ⎡0 0 ⎤
cB = [0 5 3], cN = [0 0], B = ⎢⎢0 2 0⎥⎥ , N = ⎢1 0 ⎥
⎢ ⎥
⎢⎣0 2 3⎥⎦ ⎢⎣0 1⎥⎦
How much can we change c2 given the original optimal is still optimal?

‰ Simultaneous changes on objective function coefficients

Variable Value Reduced cost Current coefficient Allowable Allowable
increase decrease
x1 0 4.5 3 4.5 Infinite
x2 9 0 5 Infinite 3
9 Apply the 100 percent rule.
9 In the Wyndor example, if c1 is increased by 1.5, then c2 can be decreased by
as much as .

9 If c1 is increased by 3, then c2 can only be decreased by as much as .

‰ Reduce cost
9 For any nonbasic decision variable xj, the associated coefficient in row 0 is
referred to as the reduced cost for xj.
9 It is the minimum amount by which the unit cost of activity j would have to be
reduced to make it worthwhile to undertake activity j (increase xj from zero).

Jin Y. Wang 6-20

College of Management, NCTU Operation Research I Fall, 2008

9 Interpreting cj as the unit profit of activity j, the reduce cost for xj is the
maximum allowable increase in cj to keep the current BF solution optimal.

‧
‰ A is changed to A
Max cx
‧
S.T. A x = b
x ≥ 0
9 Similar to the procedures when b is changed; but a bit more complex.
9 Change the coefficients of nonbasic variables.
¾ Current optimal basic solution is still feasible.
¾ Only need to check the optimality.
¾ An easy way: If the complementary dual basic solution y* still satisfies
the single dual constraint that has changed, then the original optimal
solution remains optimal.
¾ Or use our usual way of checking whether cBB-1N - cN ≥ 0.
9 Change the coefficients of basic variables
¾ Check the feasibility first.

¾ If feasible, perform the optimality check.

¾ If infeasible, compute the coefficients of row 0 and reoptimize with the

dual simplex or two phase/big M methods.

‰ Sometimes, coefficients of c and A may be changed simultaneously.

9 The solutions techniques are similar as discussed before. Let’s go to the
examples directly.

Jin Y. Wang 6-21

College of Management, NCTU Operation Research I Fall, 2008

9 Example 1: Changes in the coefficients of a nonbasic variable

¾ Given the following LP problem and the final tableau.
Max Z = 3x1 + 5x2
S.T. x1 + x3 = 4
2x2 + x4 = 24
3x1 + 2x2 + x5 = 18
Basic Coefficient of: Right Side
Variable x1 x2 x3 x4 x5
Z 9/2 0 0 0 5/2 45
x3 1 0 1 0 0 4
x2 3/2 1 0 0 1/2 9
x4 -3 0 0 1 -1 6
¾ Changes are made to the coefficients of x1. Let c1 = 4 and a31 = 2.
¾ Applying the dual theory.

¾ Or compute the value of cBB-1N - cN.

9 Example 2: Changes in the coefficients of a basic variable

¾ Use the same problem as in Example 1.
¾ Changes are made to the coefficients of x2. Let c2 = 3, a22 = 3, and a32=
4.
¾ Compute the new current basic solution.

¾ Optimality check.

Jin Y. Wang 6-22

College of Management, NCTU Operation Research I Fall, 2008

¾ Reoptimize using revised simplex method, if needed.

‰ Add xn+1ÎAdd a new variable (xn+1) to the objective function and each
constraint
9 Assume the value of this new variable be zero. That is, let the new solution
⎡ x *1 ⎤
⎢ ⎥
.
be ⎢ * ⎥ .
⎢x n ⎥
⎢ ⎥
⎢⎣ 0 ⎥⎦
9 This new solution is still feasible.
9 Check optimality
-1
¾ If cBB N - cN ≥ 0 Î done, this new solution is the optimal solution.

¾ If cBB-1N - cN < 0 Î basic changed Î adopt (revised) simplex starting

from the above solution.
9 Example:
Max x1 + x2
S.T. 2x1 + x2 + x3 =4
x1 + 2x2 + x4 = 4
x1 ~ x4 ≥ 0
⎡4 / 3⎤
⎢4 / 3⎥
x* = ⎢ ⎥
⎢ 0 ⎥
⎢ ⎥
⎣ 0 ⎦
add x5

Jin Y. Wang 6-23

College of Management, NCTU Operation Research I Fall, 2008

Max x1 + x2 + 2x5
S.T. 2x1 + x2 + x3 + 3x5 = 4
x1 + 2x2 + x4 – 3x5 = 4
x1 ~x5 ≥ 0
Keep x1, x2 as basic variables, x3, x4, and x5 as nonbasic variables.
⎡4 / 3⎤
⎢4 / 3⎥
⎢ ⎥ -1
x = ⎢ 0 ⎥ , check cBB N - cN
⎢ ⎥
⎢ 0 ⎥
⎢⎣ 0 ⎥⎦

Jin Y. Wang 6-24

College of Management, NCTU Operation Research I Fall, 2008

‰ Add a new constraint

9 If new slack ≥ 0 (current constraint satisfies the new constraint)
Î done, at optimality
¾ If the current optimal solution satisfies the new constraint, then it would
still be the optimal solution.
¾ The reason is that a new constraint can only eliminate some previously
feasible solutions without adding any new ones.
9 If new slack ≤ 0 (current constraint does not satisfy the new constraint)
Î basis changed
¾ Introduce this new constraint into the final simplex tableau, where the
usual additional variable (slack) is designated to be the new basic
variable for this new row.
¾ Apply elementary operations to convert to proper formÆ resolve the
problem by dual simplex.
9 Example:
Max Z = 3x1+ 5x2
S.T. x1 ≤ 4
2x2 ≤ 24
3x1 + 2x2 ≤ 18
x1, x2 ≥ 0

¾ Add a new constraint 2x1 + 3x2 ≤ 24.

¾ The previous optimal solution (0, 9) violates the new constraint. So the
previous optimal solution is no longer feasible.
¾ Add the new constraint to the current final simplex tableau.
Revised final tableau
Basic Coefficient of: RHS
Variable x1 x2 x3 x4 x5 x6
Z 9/2 0 0 0 5/2 0 45
x3 1 0 1 0 0 0 4
x2 3/2 1 0 0 1/2 0 9
x4 -3 0 0 1 -1 0 6
x6 2 3 0 0 0 1 24

¾ Convert to proper form

Jin Y. Wang 6-25

College of Management, NCTU Operation Research I Fall, 2008

Converted to proper form

Basic Coefficient of: RHS
Variable x1 x2 x3 x4 x5 x6
Z 9/2 0 0 0 5/2 0 45
x3 1 0 1 0 0 0 4
x2 3/2 1 0 0 1/2 0 9
x4 -3 0 0 1 -1 0 6
x6 -5/2 0 0 0 -3/2 1 -3
¾ Apply the dual simplex

New final tableau

Basic Coefficient of: RHS
Variable x1 x2 x3 x4 x5 x6
Z 1/3 0 0 0 0 5/3 40
x3 1 0 1 0 0 0 4
x2 3/2 1 0 0 0 1/3 8
x4 -4/3 0 0 1 0 -2/3 8
x6 5/3 0 0 0 1 -2/3 2

Jin Y. Wang 6-26