⑤ Special Cases in the Simplex Method

While
applying
the
simplex algorithm to solve linear
programming problem,
one can

notice the
following special cases:

1.
Degeneracy.
2. Alternative Optima.
3. Nonexisting/Infeasible solutions.

1.
Degeneracy:
-

During the application ofsimplex algorithm, it is

possible to have tie a in minimum

ratio.
Though the tie be broken values to leave from the basis
can
by choosing the
lead to the situation where take
arbitrarily, it can least
at one basic variable will

obtained solution
on value zero in next iteration. The new is said to be
degenerate.
-
The drawback of
degeneracy:The algorithm cannot be terminated if it leads to

indefinite cycling during iterations.

The one redundant
ofatleast constraint be revealed.
possibility can
-
Example:Maximize z 3x,+972
=

subjectto x, 4x2
+
8
=

x
+2x2 14
x1,Rz 0

Sh:Standard form:
Minimize *-3x,-ax
to
subject
x+4x2 8

E
x3
also canonical form.
+ =

This is in
x+ 2R2 x4 4
+
=

Re,cr, y, 12 0
 Min ratio
814 2 ratio
3 the
=

min
Tie in

4/2 2
for leaving variable.
=

choose
We arbitrarily

as to leave the basis.

<1,0 =

-
x=
0

As noticed in the table above, there is a tie in the minimum ratios. So, we

choose us to from the basis and solution.

arbitrarily leave is enters the

It is due
observed that to a tie in minimum ratio in initial solution, there is

a
degeneracy in first iteration since the basic variable my assumes a value of
Zero.

Optimal soln is Iman=18 at (0,2).

the value ofI due
Note that remained same as in iteration 1 to
degeneraly
 redundant
I constraint

Fig Degeneracy
1: in Ex. 1.
2. Alternative Optima:It is
possible an IP
that problem can have
infinitely many
solutions (alternative optimal. This
happens when the objective function parallels
the nonredundant and
binding constraint.

Example 2:Maximize 2 2x, +422

=

subject
to
x, 2xz 5
+
=

x, x=
+
4
=

x,x2 ? 0

S:Canonical form ofLP model:

Minimize I* =-2x, -4K2
subject to x, + 2x2 tes 5
=

x, +
R2
+
x4 4
=

2,R2, 13, 14 0.

Initial solution is obtained

by setting xx0
=
= n 5,4x
=>
=
4
=

and Min 20 =
 the first alumn

3
Since in

corresponds
variable
the nonbasic
x, e, can now

be made basic without

altering the value of

2.
treated
So x, is as
entering
variable and 24 as the

leaving variable.

The obtained solution is x, 3,

=

K2=1, and Max 2 10.

=

Infactthe solutions (, x2) (0,52) = and (nx, x2) (3,1)

= are the corner

optima only.
point
The rest of
the solutions can be
mathematically expressed as;

x 0.0 + (1-0). 3 3-30

=

1+
(1-0).1= 20, where 2011.
E 0.
=

+ 0
 ↑

↑
Infinitely ⑳

solutions
many
line
along this

= 12

Fig Alternative
2: optima in Ex2.
 3. Infeasible solution:

Linear
Programming models with constraints result
inconsistant in no feasible solution.
This situation occurs when the constraints are of
E) type equations, and
or

the artificial variables are used to the basic initial feasible solution. If at
give
least one artificial variables has a positive solution value with the
optimality
condition satisfied, then the
given problem has no feasible solution.

Examples: Consider the following LP:

Maximize 2 3x,+2R2
=

subject
to
=
2x, + x2 2

3x, +442112

x, N2 0

fificial variables
S:Canonical form: Minimize 2*=-3x,-2xm by introducing
as

and
25 2 x5
=

subjectto 2
2x, +xz + xy
=

E
12
3x, +4K x5
=

-
x4 +

e
(*) 2
3x, 4x2 o +

x4
- -
+ =

x , ,
. .
.,
5 =
0
 x5 4
=>
=

and is in basis.

be
Note that
we have reached a
pointwhere no further pivoting can

done on I but the value ofus is

positive.
This shows that the problem has no
feasible solution.