Special Cases in Simplex Method

Maximize z  x1  5 x2  3x3

Subject to x1  2 x2  x3  3
2 x1  x2 4
x1 , x2 , x3  0

One artificial variable is sufficient here !!!!

Maximize z  2 x1  4 x2  4 x3  3 x4

Subject to
x1  x2  x3 4
x1  4 x2  x4  8
x1 , x2 , x3 , x4  0

Artificial variable not required !!!!

 Some exceptional cases that may occur while
solving an LPP

 Degeneracy

 Alternative optimal Solutions

 Unbounded Solutions

 Infeasible Solutions
DEGENERACY
Degeneracy
During the application of Simplex method, a tie for the
minimum ratio may be broken arbitrarily.

When a tie happens at least one basic variable will be

zero in the next iteration and the new solution is said to
be degenerate.

One of the reasons for degeneracy may be due to the

presence of some redundant constraint.

Implication: Cycling or circling, the simplex table starts

repeating itself after some iterations, that is to say that
the value of the objective function do not improve after
a few iterations.
 Solving the problem graphically

x2
Example 1 Constraint x1 + 4x2  8 is
redundant as it is not essential
Max z = 3x1 + 9x2 for finding the optimal solution

Subject to
x1 + 4x2  8
x1 + 2x2  4 x1 + 4x2  8
x1, x2≥ 0
x1
x1 + 2x2  4

Optimal solution occurs at (0,2) with z = 18.

 Solving by simplex method
Basic x1 x2 s1 s2 Sol.
Min. ratio
z -3 -9 0 0 0
8/4 = 2 Tie: Make
s1 1 4 1 0 8 arbitrary
s2 1 2 0 1 4 4/2 = 2 selection

Basic x1 x2 s1 s2 Sol.
z -3/4  0 9/4 0 18
x2 1/4 1 1/4 0 2 8

s2 1/2 0 -1/4 1 0 0

Basic x1 x2 s1 s2 Sol.
z 0 0 3/2 3/2 18
Optimality conditions are
x2 0 1 ½ -½ 2 satisfied
x1 1 0 -1/2 2 0
 How to remove degeneracy

 In two variable problems the redundant constraint can be

identified easily by graphical method.

 There are no reliable techniques for identifying redundant

constraints directly from the simplex table.
ALTERNATIVE OPTIMUM SOLUTION
ALTERNATIVE OPTIMUM SOLUTION

When the objective function is parallel to the binding

constraint (that is a constraint that is satisfied as an
equation at the optimal solution) then the objective
function will assume the same optimal value at more
then one solution point.
 Optimal occurs at (3,1) and (0,2.5)
Example with optimum value as 10

Max z = 2x1 + 4x2

Subject to
x1 + 2x2  5 (0,2.5)
x1 + x2  4 (3,1)

x1, x2≥ 0

(4,0)

Graphical Solution
 Solving by simplex method

Basic x1 x2 s1 s2 Sol. Min. ratio

z -2 -4 0 0 0
s1 1 2 1 0 5 5/2

s2 1 1 0 1 4 4/1
Basic x1 x2 s1 s2 Sol. Optimality conditions are
satisfied, however it may be
z 0 0 2 0 10
noted that the z-row
x2 ½ 1 ½ 0 5/2 coefficient corresponding to
s2 ½ 0 -1/2 1 3/2 the non basic variable is zero.

Basic x1 x2 s1 s2 Sol. This indicates that x1 can

z 0 0 2 0 10 enter the basic solution
x2 0 1 1 -1 1 without changing the value
of z, but causing a change
x1 1 0 -1 2 3 in the values of variables
UNBOUNDED SOLUTION
Unbounded solution

 In some linear programming problems, the value of some of

the variables may be increased indefinitely without violating the
constraints.

 This means that the solution space is unbounded in at least

one direction.

 As a result the objective function value may increase or

decrease indefinitely

 Unbounded solutions sometimes occur due to the poorly

constructed model. Some irregularities that may occur are
ignoring the non redundant constraints or incorrect estimation
of parameters
 The solution space is unbounded
in the direction of x2 and the
solution is also unbounded.
x2
Example 1

Max z = 2x1 + x2
Subject to Unbounded
x1 - x2  10 optimal
space
2x1  40
x1, x2≥ 0
x1
Graphical Solution
 Solving by simplex method

Basic x1 x2 s1 s2 Sol.
z -2 -1 0 0 0
s1 1 -1 1 0 10
s2 2 0 0 1 40

Since x1 has the most negative coefficient, it is selected as

the entering variable.

But all the constraints under x2 are either negative or zero.

This means that x2 can be increased indefinitely without
violating any of the constraints. The feasible region is
unbounded in x2 direction.
 Basic x1 x2 s1 s2 Sol.
z 0 -3 2 0 20
x1 1 -1 1 0 10
s2 0 2 -2 1 20

Basic x1 x2 s1 s2 Sol.
z 0 0 -1 3/2 50
x1 1 0 0 1/2 20
x2 0 1 -1 1/2 10

We have an entering variable but there is no leaving variable ???

Unbounded Solution.
INFEASIBLE SOLUTION
 Infeasible solution

LP models with inconsistent constraints have no feasible

solution.

This situation can never occur when all the constraints are
of the type ().

If one of the artificial variable remains positive in the

optimum iteration, then the solution space will be infeasible
 No feasible solution space
Example
Max z = 3x1 + 2x2
Subject to
3x1 + 4x2 ≥ 12
2x1 + x2  2
3x1 + 4x2 ≥ 12
x1, x2 ≥0

2x1 + x2 2
 Solving by simplex method

B.V x1 x2 x4 x3 R Sol. Min. ratio

z -3 - 3M -2-4M  M 0 0 -12M
x3 2 1 0 1 0 2 2/1

R 3 4 -1 0 1 12 12/4

B.V x1 x2 x4 x3 R Sol.n
z 1+5M 0 M 2+4M 0 4 – 4M
x2 2 1 0 1 0 2
R -5 0 -1 -4 1 4

The conditions for optimality are satisfied, however the artificial

variable R is positive in the final iteration. This indicates that the
problem is infeasible.
Points to Remember
• During the application of simplex method, a tie for a leaving variable
will give rise to degeneracy (break the tie arbitrarily).

• If in the optimal simplex table a non basic variable is at zero level in

the z-row (objective function-row) then the problem has an alternate
optimal solution.

• In a simplex table if we have an entering variable but there is no

leaving variable then the problem has an unbounded solution.

• If in an optimal simplex table an artificial variable remains at non

zero level then the problem has no feasible solution.
Consider the LPP
Max z  2 x1  x2

Subject to
x1  x2  10
2 x1  x2  40
x1 , x2  0
Basic z x1 x2 s1 s2 Sol. Ratio
z 1 -2 1 0 0 0
s1 0 1 -1 1 0 10 10
s2 0 2 -1 0 1 40 20
z 1 0 -1 2 0 20
x1 0 1 -1 1 0 10 -
s2 0 0 1 -2 1 20 20
z 1 0 0 0 1 40
x1 0 1 0 -1 1 30
x2 0 0 1 -2 1 20
Note that we have got the optimal tableau.
(30,20)

All points on this line from

(30,20) onwards are
Optimal nonbasic solutions.
Consider the LPP

Minimize x + 1/3 y

Subject to: x + y  20
-2 x + 5 y  150
x5
x 0 y 0

Dr. Manoj Kumar Pandey

 MATRIX FORMULATION
of
LINEAR PROGRAMMING PROBLEM
Consider the following LPP (standard form)

Max (or Min) z  c1 x1  c2 x2  ...  cn xn

Subject to
a11 x1  a12 x2  ...  a1n xn  b1
a21 x1  a22 x2  ...  a2 n xn  b2
.
.
am1 x1  am 2 x2  ...  amn xn  bm
x1 , x2 ,..., xn  0, b1 , b2 ,..., bm  0
Using the notations
 x1   b1 
x  b 
 2  2
C  c1 c2 . . cn  X  .  , b . 
,    
.  . 
 xn  bm 

 a11 a12 . . a1n 

a a22 . . a2 n 
 21
A . 
 
 . 
 am1 am 2 . . amn 
To formalize the definition of a basic solution, the system AX =b
can be expressed in vector form as follows

n
 Aj x j  b
j 1
Where Aj is the jth column of A. A subset of m vectors is
said to form a basis, B, if and only if, the selected m vectors
are linearly independent. In this case, the matrix B is nonsingular.
If XB is the set of m variables associated with the vectors of
nonsingular B, then XB must be a basic solution. In this case,
we have
1
BX B  b  XB  B b
The problem can be written as

 1 c   z   0 
 
   
 0 A X
    b
Let B is a m x m non-singular sub-matrix of the
coefficient matrix A.
Let XB be the corresponding set of basic variables
with cB as its associated objective vector. The
corresponding solution (and the objective function
value ) may be computed as follows:
1 -cB   z  0
  X    
0 B   B b 
1
 z  1  CB  0 
 X   0 B   b 
 B   
  z  1 CB B  0
1

X   1   
 B  0 B  b 

 z  C B B b  C B X B 
1

X    1    B 1 b 
 B  B b   
The general simplex tableau:

 1 cB B   1 c   z   1 cB B   0 
1 1

        1   
 0 B   0 A   X   0 B  b 
1

Multiplying out the matrices, we get

 1 cB B A - C 
1
 z   cB B b 1

    
X  B b 
1 1
0 B A 
Thus in the new simplex tableau the z column
will be 1 
0
 
 . 
 
 . 

0 

The column of xj (in the constraint equations)

1
will be B A j ,
where A j is the jth column of A.
i.e., the column corresponding to xj.
1
The solution column will be B b.
And in the z-row, the coefficient of xj will be
1
cBB Aj  c j  z j  c j
Finally the solution entry in the z-row will
be c B b1
B

In matrix form the new tableau will look like

Basic z X Solution
1
z 1
1
cB B A  C cB B b
1
XB 0 1
B A B b
In terms of the components xj , the new
tableau will look like

Basic z xj Solution
1
z 1 cB B1 Aj  c j cB B b
1
XB 0 1
B Aj B b
Consider the following LPP
Maximize z  x1  x2  2 x3
subject to 2 x1  2 x2  3x3  5
x1  x2  x3  3
x1  x2  x3  2
x1 , x2 , x3  0
 After introducing the slack variables s1 , s2 , s3 to the
respective constraints, the problem becomes

Maximize z  x1  x2  2 x3
Subject to 2 x1  2 x2  3x3  s1 5
x1  x2  x3  s2 3
x1  x2  x3  s3  2
x1 , x2 , x3 , s1 , s2 , s3  0
So the starting simplex tableau is

Basic z x1 x2 x3 S1 S2 S3 Sol

z 1 -1 1 -2 0 0 0 0
S1 0 2 -2 3 1 0 0 5
S2 0 1 1 -1 0 1 0 3

S3 0 1 -1 1 0 0 1 2
 After we apply the Simplex method a portion of
the tableau is as follows:

Basic z x1 x2 x3 s1 s2 s3 Sol

z 1
1 1 0
x2 3
0 1 0
s3 0 0 1 1
x3 0 1 2 0

Find the missing numbers.

Simplex Table

Basic z xj Solution
1
z 1 cB B1 Aj  c j cB B b
1
XB 0 1
B Aj B b
Since x2 , s3 , x3 are the basic variables in the
final tableau in that order, the basic matrix B is
formed by the columns corresponding to these
variables in the starting tableau in that order.

 2 0 3 

B   1 0  1 
  1 1 1 
 1 3 0
1 
B  0 1 1 
1 2 0 

The new column corresponding to x1 in the

final tableau is
1 3 0   2   5 
1     
B A1  0 1 1  1    2  
1 2 0   1   4 
 The new column corresponding to x2 in the final
tableau is
1 
0
 
0 
as it is the first basic variable in the final tableau .
The new column corresponding to x3 in the final
tableau is 0
0
 
1 
as it is the third basic variable in the final tableau.
The solution column (in the constraint equations) in
the final tableau is
1 3 0   5  14
1     
B b  0 1 1  3   5  Now C B   1 0 2 
1 2 0   2  11
Thus entry below x1 in z-row is CB B1 A1  c1 = 3-1=2

Entries below x2, x3 in z-row are zero as they are basic

variables.
14 
1
solution z = CB B b    1 0 2   5  =8
 
11
 All these are shown below.
Answer:

Basic z x1 x2 x3 s1 s2 s3 Sol

z 1 2 0 0 1 1 0 8
x2 0 5 1 0 1 3 0 14
s3 0 2 0 0 0 1 1 5

x3 0 4 0 1 1 2 0 11
Consider the following LPP
Maximize z  4 x1  3x2  x3  2 x4
subject to

4 x1  2 x2  x3  x4  5
3x1  x2  2 x3  x4  4
x1 , x2 , x3 , x4  0
After introducing the slack variables s1 , s2
to the respective constraints, the problem becomes

Maximize z  4 x1  3 x2  x3  2 x4
subject to
4 x1  2 x2  x3  x4  s1 5
3 x1  x2  2 x3  x4  s2  4
x1 , x2 , x3 , x4 , s1 , s2  0
 After we apply the Simplex method a portion of
the final tableau is as follows:

Basic z x1 x2 x3 x4 s1 s2 Sol

z
x2 1 -1
x4 -1 2

Find the missing entries.

Answer:

Basic z x1 x2 x3 x4 s1 s2 Sol

z 1 3 0 2 0 1 1 9
x2 0 1 1 -1 0 1 -1 1
x4 0 2 0 3 1 -1 2 3
Tutorial-5
(1) The optimal table of the following LPP is given below,
find the missing entries

Min z = 4x1 + 6x2 + 5x3

Subject to 2x1 + 4x2 + 3x3 ≥ 32

x1 + 2x2 + 4x3 ≥ 28
x1, x2, x3 ≥ 0
Basic x1 x2 x3 s1 s2 solution

Z 0 0

X2 1 0 -2/5 3/10

X3 0 0 1 1/5 -2/5

66
(2) Consider the LPP

Min z = 4x1 + 3x2

Subject to
2x1 + x2 ≥ 10
x 1 + x2 ≥6
-3x1 + 2x2 6
x1≥0, x2≥0
 x1 
X  x 
Show that B  2  , is an optimal basis for the above LPP.
 s3 

67
 (3) Following is an optimal table of a LPP with all ≤ type
constraints and non-negative right-hand side vector, find the
optimal basis and the construct the original LPP.

BV x1 x2 x3 x4 s1 s2 s3 Sol

z 0 4 0 1 4 2 0 40
0 1/2 1 0 1/2 0 0 5/2
1 -1/2 0 1/3 -1/6 1/3 0 5/2
0 0 0 2/3 -1/3 -1/3 1 3