61

Step 2 : Construct the auxiliary LPP in which the new objective function Z * is to be
maximized subject to the given set of constraints.
Step 3 : Solve the auxiliary LPP by simplex method until either of the following three
possibilities arise.
(i) Max Z * < 0 and atleast one artificial variable appears in the optimum basis at a
non-zero level. In this case the given LPP does not possess any feasible solution,
stop the procedure.
(ii) Max Z * = 0 and atleast one artificial variable appears in the optimum basis at zero
level. In this case proceed to phase-II.
(iii) Max Z * = 0 and no artificial variable appears in the optimum basis. In this case
proceed to phase-II.
Phase-II
Use the optimum basic feasible solution to phase-I as a starting solution for the original
LPP. Assign the actual costs to the variables in the objective function and a 0 cost to every
artificial variable that appears in the basis at the zero level. Use simplex method to the
modified simplex table obtained at the end of phase-I, till an optimum basic feasible solution (if
any) is obtained.
Note 1 : In phase -I , the iterations are stopped as soon as the value of the new objective
function becomes zero because this is its maximum value. There is no need to continue till the
optimality is reached if this value becomes zero earlier than that.
Note 2 : The new objective function is always of maximization type regardless of whether
the original problem is of maximization or minimization type.
Note 3 : Before starting phase —II, remove all artificial variables from the table which
were non-basic at the end of phase-I.
* Example 17 : Use Two-phase simplex method to solve
Maximize Z = 5 x 1 + 8 x 2
subject to the constraints
3 x1 + 2 x 2 ³ 3
x1 + 4 x 2 ³ 4
x1 + x 2 £ 5
x 1 , x 2 ³ 0.
Solution : By introducing the non-negative slack, surplus and artificial variables,the
standard from of the LPP becoems
Max Z = 5x 1 + 8x 2 + 0 s 1 + 0 s 2 + 0 s 3
subject to
3x 1 + 2x 2 - s 1 + 0 s 2 + 0 s 3 + R1 = 3
x 1 + 4 x 2 + 0 s 1 - s 2 + 0 s 3 + R2 = 4
x 1 + x 2 + 0 s1 + 0 s 2 + s 3 = 5
and x 1 , x 2 , s 1 , s 2 , s 3 , R1 , R2 ³ 0 .
(Here : s 1 , s 2 — surplus, s 3 —slack, R1 , R2 — artificials )
 62

The initial basic feasible solution is given by

R1 = 3, R2 = 4, s 3 = 5 (basic)
( x 1 = x 2 = s 1 = s 2 = 0, non-basic)
Phase-I
Assigning a cost —1 to the artificial variables and costs 0 to all other variables, the
objective function of the auxiliary LPP becomes
Max Z * = - R1 - R2
subject to the given constraints.
The iterative simplex tables for the auxiliary LPP are :
Initial iteration :
Cj (0 0 0 0 0 -1 -1)
CB YB XB x1 x2 s1 s2 s3 R1 R2 q
3
-1 R1 3 3 2 -1 0 0 1 0
2
4
-1 R2 4 1 (4) 0 -1 0 0 1
4
5
0 s3 5 1 1 0 0 1 0 0
1
(Z *j - C j ) -7 -4 -6 1 1 0 0 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 2 and drop R2 .

Cj (0 0 0 0 0 -1 -1)

CB YB XB x1 x2 s1 s2 s3 R1 R2 q

æ 5ö 1 -1 2
-1 R1 1 ç ÷ 0 -1 0 1
è 2ø 2 2 5
1 -1 1
0 x2 1 1 0 0 0 4
4 4 4
3 1 -1 16
0 s3 4 0 0 1 0
4 4 4 3
-5 -1 3
(Z *j - C j ) -1 0 1 0 0
2 2 2

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
 63

Second iteration : Introduce x 1 and drop R1 .

Cj (0 0 0 0 0 -1 -1)

CB YB XB x1 x2 s1 s2 s3 R1 R2

2 -2 1 2 -1
0 x1 1 0 0
5 5 5 5 5
9 1 -3 -1 3
0 x2 0 1 0
10 10 10 10 10
37 3 1 -3 -1
0 s3 0 0 1
10 10 10 10 10

(Z *j - C j ) 0 0 0 0 0 0 1 1

Since all ( Z *j - C j ) ³ 0, the current basic feasible solution is optimum. Furthermore, no

artificial variable appears in the optimum basis so we proceed to phase- II.
Phase-II :
Here, we consider the actual costs associated with the original variables. The new
objective function then becomes
Max. Z = 5x 1 + 8x 2 + 0 s 1 + 0 s 2 + 0 s 3
The initial basic feasible solution for this phase is the one obtained at the end of phase-I.
The iterative simplex tables for this phase are :
Initial iteration :
Cj (5 8 0 0 0)

CB YB XB x1 x2 s1 s2 s3 q

2 -2 æ1ö
5 x1 1 0 ç ÷ 0 2
5 5 è 5ø
9 1 -3
8 x2 0 1 0 -
10 10 10
37 3 1
0 s3 0 0 1 37
10 10 10
46 -6 -7
(Z j - C j ) 0 0 0
5 5 5

Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.
 64

First iteration : Introduce s 2 and drop x 1

Cj (5 8 0 0 0)
CB YB XB x1 x2 s1 s2 s3 q

0 s2 2 5 0 -2 1 0 —
3 3 -1
8 x2 1 0 0 —
2 2 2
7 -1 æ1ö
0 s3 0 ç ÷ 0 1 7
2 2 è 2ø
(Z j - C j ) 12 7 0 -4 0 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
Second iteration : Introduce s 1 and drop s 3

Cj (5 8 0 0 0)
CB YB XB x1 x2 s1 s2 s3

0 s2 16 3 0 0 1 4
8 x2 5 1 1 0 0 1
0 s1 7 -1 0 1 0 2
(Z *j - C j ) 40 3 0 0 0 8

Since all ( Z J - C j ) ³ 0, the current basic feasible solution is optimal.

\The optimal solution is Max z = 40, x 1 = 0, x 2 = 5.
* Example 18 : Solve by two phase simplex method
Maximize X 0 = - 4 x 1 - 3 x 2 - 9 x 3
subject to
2 x 1 + 4 x 2 + 6 x 3 - s1 + R1 = 15
6 x 1 + x 2 + 6 x 3 - s 2 + R2 = 12
x 1, x 2 , x 3 , s1 , s 2 , R1 , R2 ³ 0 .
Solution : The initial basic feasible solution is given by
R1 = 15, R2 = 12 (basic) (x 1 = x 2 = x 3 = s 1 = s 2 = 0 , non-basic)
Phase I :
Assigning a cost —1 to the artificial variables and costs 0 to all other variables, the
objective function of the auxiliary LPP becomes
Max Z * = - R1 - R2
The iterative simplex tables for the auxiliary LPP are :
 65

Initial iteration :
Cj (0 0 0 0 0 -1 -1)
CB YB XB x1 x2 s1 s2 s3 R1 R2 q
15
-1 R1 15 2 4 6 -1 0 1 0
6
12
-1 R2 12 6 1 (6) 0 -1 0 1
6
(Z *j - C j ) -27 -8 -5 -12 1 1 0 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 3 and drop R2 .

Cj (0 0 0 0 0 -1 -1)
CB YB XB x1 x2 x3 s1 s2 R1 R2 q
3
-1 R1 3 -4 (3) 0 -1 1 1 -1
3
1 -1 1
0 x3 2 1 1 0 0 12
6 6 6
(Z *j - C j ) -3 4 -3 0 1 -1 0 2

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
Second iteration : Introduce x 2 and drop R1 .
Cj (0 0 0 0 0 -1 -1 )
CB YB XB x1 x2 x3 s1 s2 R1 R2
-4 -1 1 1 -1
0 x2 1 1 0
3 3 3 3 3
11 22 1 -4 -1 4
0 x3 0 1
6 8 18 18 18 18
(Z *j - C j ) 0 0 0 0 0 0 1 1

Since all ( Z *j - C j ) ³ 0, the current basic feasible solution is optimal. Further, no

artificial variable appears in the basis, so we proceed to phase-II.
Phase-II
Here, we consider the actual costs associated with the original variables. The new
objective function then becomes
Max X 0 = - 4 x 1 - 3x 2 - 9 x 3 + 0 s 1 + 0 s 2
The initial basic feasible solution for this phase is the one obtained at the end of phase-I.
 66

The iterative simplex tables for this phase are :

Initial iteration :
Cj (-4 -3 -9 0 0)

CB YB XB x1 x2 x3 s1 s2 q
-4 -1 1
-3 x2 1 1 0 -
3 3 3
11 æ 22 ö 1 -4 3
-9 x3 ç ÷ 0 1
6 è 18 ø 18 18 2
-39 1
(X 0 - C j ) -3 0 0 1
2 2

Since there are some ( X 0 - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 1 and drop x 3

Cj (-4 -3 -9 0 0)
CB YB XB x1 x2 x3 s1 s2
12 -3 1
-3 x2 3 0 1
11 11 11
3 18 1 -4
-4 x1 1 0
2 11 22 22
27 7 5
(X 0 - C j ) - 15 0 0
11 11 11

Since all ( X 0 - C j ) ³ 0, the current basic feasible solution is optimal.

3
\ The optimal solution is Max X 0 = - 15, x 1 = , x 2 = 3, x 3 = 0
2
* Example 19 : Use two-phase method to
1
Maximize Z = 2 x 1 + x 2 + x 3
4
subject to constraints
4 x1 + 6 x 2 + 3 x 3 £ 8
3 x1 - 6 x 2 - 4 x 3 £ 1
2 x1 + 3 x 2 - 5 x 3 ³ 4
x 1 , x 2 , x 3 ³ 0.
Solution : By introducing slack, surplus and artificial variables, the standard from of
the LPP becomes
1
Max Z= 2x 1 + x 2 + x 3
4
 67

subject to 4 x 1 + 6 x 2 + 3x 3 + s 1 + 0 s 2 + 0 s 3 = 8
3x 1 - 6 x 2 - 4 x 3 + 0 s 1 + s 2 + 0 s 3 = 1
2x 1 + 3x 2 - 5x 3 + 0 s 1 + 0 s 2 - s 3 + R1 = 4
and x 1 , x 2 , x 3 , s 1 , s 2 , s 3 , R1 ³ 0
(Here s 1 , s 2 -slack, s 3 —surplus, R1 —artificial)
The initial basic feasible solution is given by
s 1 = 8, s 2 = 1, R1 = 4 (basic) ( x 1 = x 2 = x 3 = s 3 = 0, non-basic)
Phase-I
The objective function of the auxiliary LPP is
Max Z * = - R1
The iterative simplex tables for the auxiliary LPP are :
Initial iteration :
Cj (0 0 0 0 0 0 -1)
CB YB XB x1 x2 x3 s1 s2 s3 R1 q
8
0 s1 8 4 (6) 3 1 0 0 0
6
0 s2 1 3 -6 -4 0 1 0 0 —
4
-1 R1 4 2 3 -5 0 0 -1 1
3
(Z *j - C j ) -4 -2 -3 5 0 0 1 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 2 and drop s 1 .
Cj (0 0 0 0 0 0 -1)
CB YB XB x1 x2 x3 s1 s2 s3 R1
4 2 1 1
0 x1 1 0 0 0
3 3 2 6
0 s2 9 7 0 -1 1 1 0 0
-13 -1
-1 R1 0 0 0 0 -1 1
2 2
13 1
(Z *j - C j ) 0 0 0 0 1 0
2 2

Since all ( Z *j - C j ) ³ 0, the current basic feasible solution is optimal for the auxiliary
LPP.
But at the same time the artificial variable R1 appears in the optimum basis at the zero
level. This optimal solution may or may not be optimal to the given (original) LPP. So we
proceed to phase-II.
 68

Phase-II
Here, we consider the actual costs associated with the original variables and assign a cost
0 to the artificial variable R1 , which appeared at zero level in phase—1, in the objective
function. The new objective function then becomes.
1
Max Z = 2x 1 + x 2 + x 3 + 0 s 1 + 0 s 2 + 0 s 3
4
The initial basic feasible solution for this phase is the one obtained at the end of
phase—I.
The iterative simples tables for this phase are :
Initial iteration :
1
Cj (2 1 0 0 0 0)
4
CB YB XB x1 x2 x3 s1 s2 s3 R1 q
4 2 1 1 4
1 x2 1 0 0 0
3 3 2 6 2
9
0 s2 9 (7) 0 -1 1 1 0 0
7
13 -1
0 R1 0 0 0 - 0 -1 1 —
2 2
4 -4 1 1
(Z j - C j ) 0 0 0 0
3 3 4 6

Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.

First iteration : Introduce x 1 and drop s 2

1
Cj (2 1 0 0 0 0)
4
CB YB XB x1 x2 x3 s1 s2 s3 R1
10 25 1 -2
1 x2 0 1 0 0
21 42 14 21
9 -1 1 1
2 x1 1 0 0 0
7 7 7 7
-13 -1
0 R1 0 0 0 0 -1 1
2 2
64 5 5 4
(Z j - C j ) 0 0 0 0
21 84 14 21
Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.
64 9 10
\ The optimal solution is Max Z = , x1 = , x 2 = , x 3 = 0.
21 7 21
 69

* Example 20 : Use two phase simplex method to

Maximize Z = 5 x 1 + 3 x 2
subject to 2 x1 + x 2 £ 1
x1 + 4 x 2 £ 6
x 1 , x 2 ³ 0.
Solution : By introducing slack, surplus and artificial variables, the standard from of
the LPP becomes
Max Z = 5x 1 + 3x 2
subject to 2x 1 + x 2 + s 1 + 0 s 2 = 1
x 1 + 4 x 2 + 0 s 1 - s 2 + R1 = 6
and x 1 , x 2 , s 1 , s 2 , R1 ³ 0.
(Here : s 1 — slack, s 2 —surplus, R1 —artificial )
The initial basic feasible solution is given by s 1 = 1, R1 = 6 (basic) (x 1 = x 2 = s 2 = 0,
non-basic)
Phase-I
The objective function of the auxiliary LPP is
Max Z * = - R1
The iterative simplex tables for the auxiliary LPP are :
Initial iteration :
Cj (0 0 0 0 -1)
CB YB XB x1 x2 s1 s2 R1 q
0 s1 1 2 (1) 1 0 0 1
6
-1 R1 6 1 4 0 -1 1
4
(Z *j - C j ) -6 -1 -4 0 1 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 2 and drop s 1 .
Cj (0 0 0 0 -1)
CB YB XB x1 x2 s1 s2 R1
0 x2 1 2 1 1 0 0
-1 R1 2 -7 0 -4 -1 1
(Z *j - C j) -2 7 0 4 1 0

Since all ( Z *j - C j ) ³ 0, the current basic feasible solution is optimal to the auxiliary LPP
But since, Max Z * < 0 and one artificial variable R1 appears in the optimum basis at
non-zero level, the given (original) LPP has no feasible solution.
 70

* Example 21 : Using simplex algorithm.

Minimize 2 x 1 - x 2 subject to the constraints
x1 + x 2 ³ 2
x1 + x 2 £ 4
x 1 , x 2 ³ 0.
Solution : Let Z = - 2x 1 - x 2
\ Min Z = - 2x 1 - x 2
subject to x1 + x 2 ³ 2
x1 + x 2 £ 4
x1 , x 2 ³ 0.
That is Max Z * = 2x 1 + x 2
subject to x1 + x 2 ³ 2
x1 + x 2 £ 4
x 1 , x 2 ³ 0.
By introducing slack, surplus and artificial variables, the standard from of the LPP
becomes
Max Z * = 2x 1 + x 2
subject to x 1 + x 2 - s 1 + 0 s 2 + R1 = 2
x 1 + x 2 + 0 s1 + s 2 = 4
x 1 , x 2 , s 1 , s 2 , R1 ³ 0.
(Here : s 1 - surplus, s 2 —slack, R1 —artificial)
The initial basic feasible solution is given by R1 = 2, s 2 = 4 (basic) (x 1 = x 2 = s 1 = 0,
non-basic)
Phase-I
The objective function of the auxiliary LPP is
Max Z * = - R1
The iterative simplex tables for the auxiliary LPP are :
Initial iteration :
Cj (0 0 0 0 -1)
CB YB XB x1 x2 s1 s2 R1 q
2
-1 R1 2 (1) 1 -1 0 1
1
4
0 s2 4 1 1 0 1 0
1
(Z *j - C j ) -2 -1 -1 1 0 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
 71

First iteration : introduce x 1 and drop R1 .

Cj (0 0 0 0 -1)
CB YB XB x1 x2 s1 s2 R1
0 x1 2 1 1 -1 0 1
0 s2 2 0 0 1 1 -1
(Z *j - C j) 0 0 0 0 0 1

Since all ( Z *j - C j ) ³ 0, and no artificial variable appears in the optimum basis, the
current basic feasible solution is optimal to the auxiliary LPP and we proceed to Phase—II.
Phase-II
Here, we consider the actual costs associated with the original variable. The new
objective function then becomes
Max Z * = 2x 1 + x 2 + 0 s 1 + 0 s 2
The initial basic feasible solution for this phase is the one obtained at the end of phase-I.
The iterative simplex tables for this phase are :
Initial iteration :
Cj (2 1 0 0)
CB YB XB x1 x2 s1 s2 q
2 x1 2 1 1 -1 0 —
2
0 s2 2 0 0 (1) 1
1
(Z *j - C j ) 4 0 1 -2 0

Since there are some ( Z *j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce s 1 and drop s 2 .

Cj (2 1 0 0)
CB YB XB x1 x2 s1 s2
2 x1 4 1 1 0 1
0 s1 2 0 0 1 1
(Z *j - C j) 8 0 1 0 2

Since all ( Z *j - C j) ³ 0, the current basic feasible solution is optimal.

\ The optimal solution is max Z * = 8, x 1 = 4, x 2 = 0
But Min Z = - Max (-Z) = - Max Z * = -8
\ Min Z = - 8, x 1 = 4, x 2 = 0
 72

Disadvantage of Big—M Method over Two—phase method :

Even though Big—M method can always be used to check the existence of a feasible
solution, it may be computationally inconvenient especially when a digital computer is used
because of manipulation of the constant M. On the other hand,two-phase method eliminates
the constant M from calculations.
Variants of the Simplex Method
Here we present certain complications and variations encountered in the application of
the simplex method and how they are resolved. These are called the variants of the simplex
method. The following variants are being considered.
1. Degeneracy and Cycling
2. Unbounded Solution
3. Multiple Solutions
4. Non-existing Feasible Solution
5. Unrestricted Variables.
1. Degeneracy and Cycling : The concept of obtaining a degenerate basic feasible
solution in a LPP is known as Degeneracy. Degeneracy in a LPP may arise
(i) at the initial stage when atleast one basic variable is zero in the initial basic
feasible solution.
(ii) at any subsequent iteration when more than one basic variable is eligible to leave
the basis and hence one or more variables becoming zero in the next iteration and
the problem is said to degenerate. There is no assurance that the value of the
objective function will improve, since the new solutions may remain degenerate. As
a result, it is possible to repeat the same sequence of simplex iterations endlessly
without improving the solutions. This concept is known as cycling or circling.
Perturbation rule to avoid cycling :
(a) Divide each element in the tied rows by the positive co-efficients of the key column
in that row.
(b) Compare the resulting ratios, column by column, first in the unit matrix and then
in the body matrix from left to right.
(c) The row which first contains the smallest algebraic ratio contains the leaving
variable.
* Example 22 : Solve the following LPP by simplex method :
Maximize F = x 1 + 2 x 2 + x 3
subject to 2 x 1 + x 2 - x 3 £ 2
-2 x 1 + x 2 - 5 x 3 ³ - 6
4 x1 + x 2 + x 3 £ 6
x 1 , x 2 , x 3 ³ 0.
Solution : Given Maximize F = x 1 + 2x 2 + x 3
subject to 2x 1 + x 2 - x 3 £ 2
 73

2x 1 - x 2 + 5x 3 £ 6
4x1 + x 2 + x 3 £ 6
x 1 , x 2 , x 3 ³ 0.
By introducing the non-negative slack variables s 1 , s 2 , s 3 the standard from of LPP
becomes
Maximize F = x 1 + 2x 2 + x 3
subject to 2x 1 + x 2 + x 3 + s 1 + 0 s 2 + 0 s 3 = 2
2x 1 - x 2 + 5x 3 + 0 s 1 + s 2 + 0 s 3 = 6
4 x 1 + x 2 + x 3 + 0 s1 + 0 s 2 + s 3 = 6
and x 1 , x 2 , x 3 , s 1 , s 2 , s 3 ³ 0.
The initial basic feasible solution is given by
s 1 = 2, s 2 = 6, s 3 = 6 (basic) (x 1 = x 2 = x 3 = 0, non-basic)
Initial iteration :
Cj (1 2 1 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3 q
2
0 s1 2 2 (1) -1 1 0 0 = 1*
1
0 s2 6 2 -1 5 0 1 0 -
6
0 s3 6 4 1 1 0 0 1 =6
1
(F j - C j ) 0 -1 -2 -1 0 0 0

Since there are some ( F j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 2 and drop s 1 .

Cj (1 2 1 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3 q

2 x2 2 2 1 -1 1 0 0 —
8
0 s2 8 4 0 4 1 1 0 =2
4
4
0 s3 4 2 0 (2) -1 0 1 = 2*
2
(F j - C j ) 4 3 0 -3 2 0 0

Since there are some ( F j - C j ) < 0, the current basic feasible solution is not optimal.
The non-basic variable x 3 enters in to the basis. Since both the basic variable s 2 and
s 3 having the same minimum ratio 2, there is a tie in selecting the leaving variable. To resolve
 74

this degeneracy, we divide each entry corresponding to basic variables s 2 and s 3 and then
corresponding to non-basic variables x 1, x 2 , x 3 and s 1 . We get the following quotients.
s2 s3 x1 x2 x3 s1
1 0 4 0 4 1
Row 2 :
4 4 4 4 4 4
0 1 2 0 2 1
Row 3 : -
2 2 2 2 2 2

The columnwise comparison of quotients starting with basic variables s 2 and s 3 , we find
that column s 2 gives algebraically smaller ratio : 0, for Row 3 and as such Row 3 is selected as
key row. So the basic variable s 3 leaves the basis.
Second iteration : Introduce x 3 and drop s 3 .

Cj (1 2 1 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3
1 1
2 x2 4 3 1 0 0
2 2
0 s2 0 0 0 0 3 1 -2
-1 1
1 x3 2 1 0 1 0
2 2
1 3
(F j - C j ) 10 6 0 0 0
2 2

Since all ( F j - C j ) ³ 0, the current basic feasible solution is optimal.

\ The optimal solution is Max F = 10, x 1 = 0, x 2 = 4, x 3 = 2.
* Example 23 : Solve the LPP Maximize Z = 5 x 1 - 2 x 2 + 3 x 3
subject to 2 x 1 + 2 x 2 - x 3 ³ 2
3 x1 - 4 x 2 £ 3
x2 + 3x3 £ 5
x1 , x 2 , x 3 ³ 0
Solution : By introducing the non-negative surplus variable s 1 , slack variables
s 2 , s 3 and an artificial variable R1 , the standard form of the LPP becomes
Maximize Z = 5x 1 - 2x 2 + 3x 3 + 0 s 1 + 0 s 2 + 0 s 3 - MR1
Subject to 2x 1 + 2x 2 - x 3 - s 1 + 0 s 2 + 0 s 3 + R1 = 2
3x 1 - 4 x 2 - 0 x 3 + 0 s 1 + s 2 + 0 s 3 = 3
0 x 1 + x 2 + 3x 3 + 0 s 1 + 0 s 2 + s 3 = 5
and x 1 , x 2 , x 3 , s 1 , s 2 , s 3 , R1 ³ 0
The initial basic feasible solution is given by
R1 = 2, s 2 = 3, s 3 = 5 (basic) ( x 1 = x 2 = x 3 = s 1 = 0, non-basic)
 75

Initial iteration :
Cj (5 -2 3 0 -M 0 0)
CB YB XB x1 x2 x3 s1 R1 s2 s3 q
2
-M R1 2 (2) 2 -1 -1 1 0 0 *
2
3
0 s2 3 3 -4 0 0 0 1 0
3
0 s3 5 0 1 3 0 0 0 1 —
(Z j - C j ) -2 M -2M - 5 -2M + 2 M -3 M 0 0 0
Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal. The
non-basic variable x 1 enters into the basis.
Since there is a tie in selecting the leaving variable among R1 and s 2 , it is an indication
of the existence of degeneracy. But since R1 is an artificial basic variable, we select R1 as the
leaving variable.
First iteration : Introduce x 1 and drop R1 .
Cj (5 -2 3 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3 q
-1 -1
5 x1 1 1 1 0 0 —
2 2
3
æ ö 3
0 s2 0 0 -7 ç ÷ 1 0 0*
è 2ø 2
5
0 s3 5 0 1 3 0 0 1
3
-11 -5
(Z j - C j ) 5 0 7 0 0
2 2
Since there are some ( Z j - C j ) £ 0, the current basic feasible solution is not optimal.
Second iteration : Introduce x 3 and drop s 2 .
Cj (5 -2 3 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3 q
-4 1
5 x1 1 1 0 0 0 -
3 3
-14 2
3 x3 0 0 1 1 0 —
3 3
5
0 s3 5 0 (15) 0 -3 -2 1
15
-56 11
(Z j - C j ) 5 0 0 3 0
3 3
 76

Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.
Third iteration : Introduce x 2 and drop s 3 .

Cj (5 -2 3 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3 q
13 -4 7 4
5 x1 1 0 0 -
9 15 45 45
14 æ 1 ö 2 14 14 15
3 x3 0 0 1 ç ÷ ´ *
9 è 15 ø 45 45 9 1
1 -1 -2 1
-2 x2 0 1 0 —
3 5 15 15
101 -11 53 56
(Z j - C j ) 0 0 0
9 15 45 45

Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.
Fourth iteration : Introduce s 1 and drop x 3 .
Cj (5 -2 3 0 0 0)
CB YB XB x1 x2 x3 s1 s2 s3
23 1 4
5 x1 1 0 4 0
3 3 3
70 2 14
0 s1 0 0 15 1
3 3 3
-2 x2 5 0 1 3 0 0 1
85 5 14
(Z j - C j ) 0 0 11 0
3 3 3
Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.
85 23
\ The optimal solution is Max Z = , x1 = , x 2 = 5, x 3 = 0.
3 3
2. Unbounded Solution : In some linear programming problems, the solution space
becomes unbounded so that the value of the objective function also can be increased indefinitely
without a limit. But it is wrong to conclude that just because the solution space is unbounded
the solution also is unbounded. The solution space may be unbounded but the solution may be
finite.
* Example 1 : (Unbounded solution space but bounded optimal solution)
Max Z = 3 x 1 - x 2
subject to x 1 - x 2 £ 10
x 1 £ 20
and x 1 , x 2 ³ 0,
 77

Solution : By introducing the slack variables s 1 , s 2 , the standard form of the LPP
becomes
Max Z = 3x 1 - x 2 + 0 s 1 + 0 s 2
subject to x 1 - x 2 + s 1 + 0 s 2 = 10
x 1 + 0 x 2 + 0 s 1 + s 2 = 20
x 1 , x 2 , s 1 , s 2 ³ 0.
The initial basic feasible solution is given by
s 1 = 10, s 2 = 20 (basic) ( x 1 = x 2 = 0, non-basic)
Initial iteration :
Cj (3 -1 0 0)
CB YB XB x1 x2 s1 s2 q
10
0 s1 10 (1) -1 1 0 *
1
20
0 s2 20 1 0 0 1
1
(Z j - C j ) 0 -3 1 0 0
Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.
First iteration : Introduce x 1 and drop s 1 .
Cj (3 -1 0 0)
CB YB XB x1 x2 s1 s2 q
3 x1 10 1 -1 1 0 -
10
0 s2 10 0 (1) -1 1 *
1
(Z j - C j ) 30 0 -2 3 0
Since ( Z j - C j ) = - 2 < 0, the current basic feasible solution is not optimal.
Second iteration : Introduce x 2 and drop s 2 .

Cj (3 -1 0 0)
CB YB XB x1 x2 s1 s2
3 x1 20 1 0 0 1
-1 x2 10 0 1 -1 1
(Z j - C j ) 50 0 0 1 2

Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.

\ The optimal solution is Max Z = 50, x 1 = 20, x 2 = 10.
The above problem is represented graphically in the following figure to indicate that
there is a bounded optimal solution even through the solution space is unbounded.
 78
x2

20

Solution
Space x1 + x2 = 10
10
B
x1 = 0 (20, 10)
x1 = 10

-10 O 20 x1
10
x2 = 0

-10
Fig. 1
Here the solution space is unbounded above, but the optimal solution occurs at the vertex
(20, 10).
* Example 25 : Solve the following LPP by simplex method.
Max Z = 2 x 1 + x 2
subject to the constraints x 1 - x 2 £ 10
2 x 1 - x 2 £ 40
and x1 , x 2 ³ 0
Solution : By introducing the non-negative slack variables s 1 and s 2 , the standard from
of the LPP becomes
Max Z = 2x 1 + x 2 + 0 s 1 + 0 s 2
Subject to x 1 - x 2 + s 1 + 0 s 2 = 10
2x 1 - x 2 + 0 s 1 + s 2 = 40
and x 1 , x 2 , s1 , s 2 ³ 0
The initial basic feasible solution is given by
s 1 = 10, s 2 = 40 (basic) (x 1 = x 2 = 0, non-basic)
Initial iteration :
Cj (2 1 0 0)
CB YB XB x1 x2 s1 s2 q
10
0 s1 10 (1) -1 1 0 *
1
40
0 s2 40 2 -1 0 1 *
2
(Z j - C j ) 0 -2 -1 0 0

Since there are some ( Z j - C j ) < 0, the current basic feasible solution is not optimal.
The non-basic variable x 1 enters in to the basis and the basic variable s 1 leaves the basis.
 79

First iteration :
Cj (2 1 0 0)
CB YB XB x1 x2 s1 s2 q
2 x1 10 1 -1 1 0 —
20
0 s2 20 0 (1) -2 1 *
1
(Z j - C j ) 20 0 -3 2 0
Since ( Z j - C j ) = - 3 < 0, the current basic feasible solution is not optimal.
The non-basic variable x 2 enters in to the basis and the basic variable s 2 leaves the basis.
Second iteration :
Cj (2 1 0 0)
CB YB XB x1 x2 s1 s2
2 x1 30 1 0 -1 1
1 x2 20 0 1 -2 1
(Z j - C j ) 80 0 0 -4 3
Since ( Z 3 - C 3 ) = - 4 < 0,. the current basic feasible solution is not optimal.
æX ö
Also, since all a ir < 0, it is not possible to find the positive ratio q = min çç Bi , a ir > 0 ÷÷ i. e.
è a ir ø
it is not possible to find the leaving variable. The solution of this problem is unbounded.
The above problem is represented graphically in the following figure to indicate that
there is a unbounded optimal solution with unbounded solution space.
x2

30
x1 - x2 = 10
20 Solution
Space
x1 = 0
10 2x1 - x2 = 40

x1
0
-20 -10 10 20 30 40
-10 x2 = 0

-20

-30

-40

Fig. 2
Here the solution space is unbounded and the optimal solution is also unbounded.
 80

3. Multiple solutions (or) Alternate optimal solutions :

In some linear programming problems the optimal solution need not be unique. There
may be alternative or infinite number of solutions, i. e., with different product mixes, we have
the same value of the objective function.
While dealing with the graphical method, if the optimal solution occurs at a vertex of the
solution space, then the problem is said to have a unique optimal solution. On the other hand, if
the optimum solution occurs on an edge of the solution space, then the problem is said to have
an alternative or infinite number of solutions.
While dealing with the simplex method, in the optimum simplex table, if the net
evaluation (Z j - C j ) ¹ 0 for all non-basic variables, then the problem is said to have a
unique optimal solution. On the other hand, if the net evaluation (Z j - C j ) = 0 for
atleast one non-basic variable, then the problem is said to have an alternative or
infinite number of solutions.
* Example 26 : Solve the LPP
Maximize Z = 6 x 1 + 4 x
subject to the constraints
2 x1 + 3 x 2 £ 30
3 x1 + 2 x 2 £ 24
x1 + x 2 ³ 3
and x 1, x 2 ³ 0.
Solution : By introducing the non-negative slack variables s 1 , s 2 , surplus variable s 3
and an artificial variable R1 , the standard from of the LPP becomes
Max Z = 6 x 1 + 4 x + 0 s 1 + 0 s 2 - MR1
Subject to 2x 1 + 3x 2 + s 1 + 0 s 2 + 0 s 3 = 30
3x 1 + 2x 2 + 0 s 1 + s 2 + 0 s 3 = 24
x 1 + x 2 + 0 s 1 + 0 s 2 - s 3 + R1 = 3
and x 1, x 2 , s 1 , s 2 , s 3 , R1 ³ 0
The initial basic feasible solution is given by
s 1 = 30, s 2 = 24, R1 = 3 (basic) (x 1 = x 2 = s 3 = 0, non-basic)
Initial Iteration :
Cj (6 4 0 0 0 - M)
CB YB XB x1 x2 s1 s2 s3 R1 q
30
0 s1 30 2 3 1 0 0 0
2
24
0 s2 24 3 2 0 1 0 0
3
3
-M R1 3 (1) 1 0 0 -1 1 *
1
(Z j - C j ) -3 M -M - 6 -M - 4 0 0 M 0
 81

First iteration : Introduce x 1 and drop R1 .

Cj (6 4 0 0 0)
CB YB XB x1 x2 s1 s2 s3 q
24
0 s1 24 0 1 1 0 2
2
15
0 s2 15 0 -1 0 1 (3) *
3
6 x1 3 1 1 0 0 -1 -
(Z j - C j ) 18 0 2 0 0 -6

Second iteration : Introduce s 3 and drop s 2 .

Cj (6 4 0 0 0)
CB YB XB x1 x2 s1 s2 s3
5 -2
0 s1 14 0 1 0
3 3
-1 1
0 s3 5 0 0 1
3 3
2 1
6 x1 8 1 0 0
3 3
(Z j - C j ) 48 0 0 0 2 0

Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.

The optimal solution is Max z = 48, x 1 = 8, x 2 = 0.
Alternative Solutions :
Since ( z2 - c 2 ) = 0, corresponding to the non-basic variable x 2 , the alternative solutions
also exist. Therefore the solution as obtained above is not unique.
Thus we can bring x 2 into the basis in place of s 1 or s 3 . So introducing x 2 into the basis in
place of s 1 , the new optimum simplex table is obtained as follows :
Cj (6 4 0 0 0)
CB YB XB x1 x2 s1 s2 s3
42 3 2
4 x2 0 1 0
3 5 5
39 1 1
0 s3 0 0 1
5 5 5
12 -2 3
6 x1 1 0 0
5 5 5
(Z j - C j ) 48 0 0 0 2 0
 82

Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.

12 42
\ Another optimal solution is Max Z = 48, x 1 = , x2 = .
5 5
Thus, if two alternative optimal solutions can be obtained, then infinite number of
optimum solutions can be obtained .
\ The given LPP possess an infinite number of optimum solutions or multiple
optimal solutions.
4. Non-existing feasible solution :
In an LPP, when there is no point belonging to the solution space satisfying all the
constraints, then the problem is said to have no feasible solution. In other words, in the
optimum simplex table, if atleast one artificial variable appears in the basis at non-zero level
(with positive value in X B column) and the optimality condition is satisfied, then the problem is
said to have no feasible solution.
Example (1) : See the example (1) under the Big—M method.
5. Unrestricted or unconstrained variables :
In an LPP, if any variable is unconstrained (without specifying its sign) it can be
expressed as the difference between two non-negative variables. The problem can be converted
in to an equivalent one involving only non-negative variables.
* Example 27 : Solve the LPP :
Maximize Z = 2 x 1 + 3 x 1
subject to -x1 + 2 x 2 £ 4
x1 + x 2 £ 6
x1 + 3 x 2 £ 9
and x 1 , x 2 are unrestricted.
Solution : Since x 1 , x 2 are unrestricted, we put x 1 = x 1 ¢ - x 1 ¢ ¢ and x 2 = x 2 ¢ - x 2 ¢ ¢ so
that x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ ³ 0.
Then given LPP becomes
Maximize Z = 2x 1 ¢ - 2x 1 ¢ ¢ + 3x 2 ¢ - 3x 2 ¢ ¢
Subject to -x 1 ¢ + x 1 ¢ ¢ + 2x 2 ¢ - 2x 2 ¢ ¢ £ 4
x1 ¢ - x1 ¢ ¢ + x 2 ¢ - x 2 ¢ ¢ £ 6
x 1 ¢ - x 1 ¢ ¢ + 3x 2 ¢ - 3x 2 ¢ ¢ £ 9
x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ ³ 0.
By introducing the non-negative slack vaiables s 1 , s 2 , s 3 , the standard from of the LPP
becomes
Maximize Z = 2x 1 ¢ - 2x 1 ¢ ¢ + 3x 2 ' - 3x 2 ' ' + 0 s 1 + 0 s 2 + 0 s 3
subject to - x 1 ¢ + x 1 ¢ ¢ + 2x 2 ¢ + 2x 2 ¢ ¢ + s 1 + 0 s 2 + 0 s 3 = 4
x 1 ¢ - x 1 ¢ ¢ + x 2 ¢ - x 2 ¢ ¢ + 0 s1 + s 2 + 0 s 3 = 6
x 1 ¢ - x 1 ¢ ¢ + 3x 2 ¢ - 3x 2 ¢ ¢ + 0 s 1 + 0 s 2 + s 3 = 9
x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ , s 1, s 2 , s 3 ³ 0
 83

The initial basic feasible solution is given by

s 1 = 4, s 2 = 6, s 3 = 9 (basic) (x 1 ¢ = x 1 ¢ ¢ = x 2 ' = x 2 ¢ ¢ = 0 non-basic)
Initial iteration :
Cj (2 -2 3 -3 0 0 0)
x1 ¢ ¢ x2¢¢
CB YB XB x1 ¢ x2¢ s1 s2 s3 q

4
0 s1 4 -1 1 (2) -2 1 0 0 = 2*
2
6
0 s2 6 1 -1 1 -1 0 1 0 =6
1
9
6 s3 9 1 -1 3 -3 0 0 1 =3
3
(Z j - C j ) 0 -2 2 -3 3 0 0 0

First iteration : Introduce x 2 ¢ and drop s 1 .

Cj (2 -2 3 -3 0 0 0)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2 s3 q
-1 1 1
3 x 2¢ 2 1 -1 0 0 —
2 2 2
3 -3 -1 8
0 s2 4 0 0 1 0
2 2 2 3
5 -5 -3 6
6 s3 3 ( ) 0 0 0 1 *
2 2 2 5
-7 7 3
(Z j - C j ) 6 0 0 0 0
2 2 2

Second iteration : Introduce x 1 ¢ and drop s 3 .

Cj (2 -2 3 -3 0 0 0)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2 s3 q
13 1 1
3 x 2¢ 0 0 1 -1 0 13
5 5 5
11 æ 2ö -3 11
0 s2 0 0 0 0 ç ÷ 1 *
5 è 5ø 5 2
6 -3 2
2 x1 ¢ 1 -1 0 0 0 —
5 5 5
51 -3 7
(Z j - C j ) 0 0 0 0 0
5 5 5
 84

Third iteration : Introduce s 1 and drop s 2 .

Cj (2 -2 3 -3 0 0 0)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2 s3
3 -1 1
3 x 2¢ 0 0 1 -1 0
2 2 2
11 5 -3
0 s1 0 0 0 0 1
2 2 2
9 3 -1
2 x1 ¢ 1 -1 0 0 0
2 2 2
27 3 1
(Z j - C j ) 0 0 0 0 0
2 2 2

Since all ( Z j - C j ) ³ 0, the current basic feasible solution is optimal.

27 9 3
\ The optimal solution is Max Z = , x 1 ¢ = , x 2 ¢ = , x 1 ¢ ¢ = 0, x 2 ¢ ¢ = 0
2 2 2
9 9 3 3
But x 1 = x 1 ¢ - x 1 ¢ ¢ = - 0 = ; x 2 = x 2 ¢ - x 2 ¢ ¢ = - 0 =
2 2 2 2
27 9 3
The optimal solution is Max Z = , x1 = , x 2 = .
2 2 2
* Example 28 : Solve the LPP :
Maximize Z = 8 x 2
subject to x1 - x 2 ³ 0
2 x1 + 3 x 2 £ - 6
and x 1, x 2 are unrestricted.
Solution : Since x 1 , x 2 are unrestricted, we put x 1 = x 1 ¢ - x 1 ¢ ¢ and x 2 = x 2 ¢ - x 2 ¢ ¢ so
that x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ ³ 0.
\The given LPP becomes
Maximize Z = 8x 2 ¢ - 8x 2 ¢ ¢
subject to x1 ¢ - x1 ¢ ¢ - x 2 ¢ + x 2 ¢ ¢ ³ 0
-2x 1 ¢ + 2x 1 ¢ ¢ - 3x 2 ¢ + 3x 2 ¢ ¢ ³ 6
x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ ³ 0.
By introducing the surplus variables s 1 , s 2 and artificial variables R1 and R2 the
standard from of LPP becomes
Maximize Z = 0 x 1 ¢ + 0 x 1 ¢ ¢ + 8x 2 ¢ - 8x 2 ¢ ¢ + 0 s 1 + 0 s 2 - MR1 - MR2
subject to x 1 ¢ - x 1 ¢ ¢ + x 2 ¢ + x 2 ¢ ¢ - s 1 + 0 s 2 + R1 = 0
- 2x 1 ¢ + 2x 1 ¢ ¢ - 3 x 2 ¢ + 3x 2 ¢ ¢ + 0 s 1 - s 2 + R2 = 6
and x 1 ¢ , x 1 ¢ ¢ , x 2 ¢ , x 2 ¢ ¢ , s 1 , s 2 , R1 , R2 ³ 0.
The initial basic feasible solution is given by R1 = 0, R2 = 6. (basic) (x 1 ¢ = x 1 ¢ ¢ = x 2 ¢
= x 2 ¢ ¢ = s 1 = s 2 = 0, non-basic)
 85

Initial iteration :
Cj (0 0 8 -8 0 0 -M -M)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2 R1 R2 q

-M R1 0 1 -1 -1 (1) -1 0 1 0 0*
6
-M R2 6 -2 2 -3 3 0 -1 0 1 =2
3
(Z j - C j ) -6 M M -M 4M - 8 -M + 8 M M 0 0

First iteration : Introduce x 2 ¢ ¢ and drop R1 ,

Cj (0 0 8 -8 0 0 -M)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2 R2 q

-8 x 2¢ ¢ 0 1 -1 -1 1 -1 0 0 —
6
-M R2 6 -5 (5) 0 0 3 -1 1 *
5
(Z j - C j ) -6M 5M - 8 -5M + 8 0 0 -3M + 8 M 0

Second iteration : Introduce x 1 ¢ ¢ and drop R2 .

Cj (0 0 8 -8 0 0)
CB YB XB x1 ¢ x1 ¢ ¢ x2¢ x2¢¢ s1 s2
6 -2 -1
-8 x 2¢ ¢ 0 0 -1 1
5 5 5
6 3 1
0 x1 ¢ ¢ -1 1 0 0 -
5 5 5
-48 16 8
(Z j - C j ) 0 0 0 0
5 5 5

Since all ( Z j - C j ) ³ 0, and no artificial variable appears in the basis, the current basic
feasible solution is optimal.
-48 6 6
\The optimal solution Max Z = , x 1 ¢ = 0, x 2 ¢ = 0, x 1 ¢ ¢ = , x 2 ¢ ¢ =
5 5 5
6 -6
But x 1 = x 1 ¢ = 0 - =
5 5
6 -6
and x 2 = x 2 ¢ - x 2 ¢ ¢ = 0 - =
5 5
-48 -6 -6
The optimal solution is Max Z = , x1 = , x2 = .
5 5 5