5.1 Lecture - 1f

Lecture 1f:
VARIANTS OF SIMPLEX METHOD

Jeff Chak-Fu WONG
Department of Mathematics
Chinese University of Hong Kong
jwongmath. uhk.edu.hk
MAT581SS
Mathematics for Logistics
Produced by Jeff Chak-Fu WONG
ABLE OF
TABLE OF C ONTENTS
1. GRAPHICAL SOLUTIONS TO A LINEAR PROGRAMMING PROBLEM
2. SIMPLEX METHOD
3. ARTIFICIAL VARIABLE TECHNIQUE
(a) The Big M method
(b) The Two-phase simplex method
A Worked Example
4. VARIANTS OF SIMPLEX METHOD
C ONTENTS
Example 1 Use two phase simplex method to

Miximise z = 2x1 x2
subject to
ABLE OF
x1 + x2 2
x1 + x2 4
x1 , x2 0
C ONTENTS
Solution:
Convert the LPP into maximisation by using
min z = max (z),
the LPP becomes
Maximise (z) = 2x1 + x2
subject to
ABLE OF
x1 + x2 2
x1 + x2 4
x1 , x2 0
C ONTENTS
By introducing slack variable s2 , surplus variable s1 , artificial variable

A1 , the standard form of LPP is
max (z) = 2x1 + x2
subject to
ABLE OF
x1
x2
s1
0s2
x1
x2
0s1
s2
A1
x1 , x2 , s1 , s2 , A1 0
C ONTENTS
The initial basis matrix is:

B=
sA1
s2
= I2 .
An obvious initial basis feasible solution is s2 = 4, A1 = 2, i.e.,
ABLE OF
XB = B 1 b = I2 b = [2 4]T
C ONTENTS
ABLE OF
y1 = B 1 a1 = I2 [1 1]T = [1 1]T
y2 = B 1 a2 = I2 [1 1]T = [1 1]T
y1 = B 1 s1 = I2 [1 0]T = [1 0]T
y2 = B 1 s2 = I2 [0 1]T = [0 1]T
yA1 = B 1 sA1 = I2 [1 0]T = [1 0]T
z1 c1 = CB x1 c1 = [1 0]1 1]T 0 = 1
z2 c2 = CB x2 c2 = [1 0][1 1]T 0 = 1
z3 c3 = CB s1 c3 = [1 0][1 0]T 0 = 1
z4 c4 = CB s2 c4 = [1 0][0 1]T 0 = 0
z5 c5 = CB yA1 c5 = [1 0][0 1]T (1) = 0
C ONTENTS
Phase I The objective function of the auxiliary LPP is

max (z ) = A1 .
Starting Tableau:
Tableau 6.1.1
cj :
CB
XB
(0
-1)
y1
y2
y3
y4
yA1
x1
x2
s1
s2
A1
Min. Ratio
-1
A1
-1
2/1=2
s2
4/1=4
-1
-1
zj cj
From Tableau 6.1.1 we observe that
ABLE OF
the non-basic variable x1 enters into the basis;

the basic variable A1 leaves the basis.
C ONTENTS
Starting Tableau:
Tableau 6.1.1
ABLE OF
cj :
CB
XB
(0
-1)
x1
x2
s1
s2
A1
Min. Ratio
-1
A1
-1
2/1=2
s2
4/1=4
R2 R1
zj cj
C ONTENTS
Starting Tableau:
Tableau 6.1.1
cj :
ABLE OF
CB
XB
(0
-1)
x1
x2
s1
s2
A1
x1
-1
s2
zj cj
C ONTENTS
10
The new basis matrix is:

h
B = sA1
s2
=
= I2 B
a1
s2
i.e., B
1 , we have
Find the inverse of B,
h
i h
i
1
I2 I2 B
.
B
That is,
ABLE OF
C ONTENTS
11
ABLE OF
1 a1 = [1 0]T
y1 = B
1 a2 = [1 0]T
y2 = B
1 s1 = [1 1]T
y3 = B
1 s2 = [0 1]T
y4 = B
z1 c1 = CB y1 c1 = [0 0][1 0]T 0 = 0
z2 c2 = CB y2 c2 = [0 0][1 0]T 0 = 0
z3 c3 = CB y3 c3 = [0 0][1 1]T 0 = 0
z4 c4 = CB y4 c4 = [0 0][0 1]T 0 = 0
C ONTENTS
12
First iteration:
Tableau 6.1.2
cj :
CB
XB
(0
0)
x1
x2
s1
s2
x1
-1
s2
zj cj
z (XB ) = 0
Since all zj cj 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.
ABLE OF
C ONTENTS
13
Phase II Assign the actual costs to the original variables. The new
objective function then becomes
Maximise (z) = 2x1 + x2 + 0s1 + 0s2
The initial basic feasible solution for this phase is the one obtained at
the end of Phase I.
ABLE OF
C ONTENTS
14
The initial basis matrix is:

=
B=B
a1
s2
An obvious initial basis feasible solution is x1 = 2, s2 = 2, i.e.,
ABLE OF
1 b = [2 2]T
XB = B
C ONTENTS
15
ABLE OF
y1 = [1 0]T
y2 = [1 0]T
y3 = [1 1]T
y4 = [0 1]T
z1 c1 = CB y1 c1 = [2 0][1 0]T (2) = 0

z2 c2 = CB y2 c2 = [2 0][1 0]T (1) = 1
z3 c3 = CB y3 c3 = [2 0][1 1]T 0 = 2
z4 c4 = CB y4 c4 = [2 0][0 1]T 0 = 0
C ONTENTS
16
The iterative simplex tableaus for this phase is:

Initial iteration:
Tableau 6.1.3
cj :
CB
XB
(2
0)
y1
y2
y3
y4
x1
x2
s1
s2
x1
-1
s2
-2
zj cj
z(XB ) = 4
Min. Ratio
2/1=2
From Tableau 6.1.3, we observe that s1 enters into the basis and s2
leaves the basis.
ABLE OF
C ONTENTS
17

Initial iteration:
Tableau 6.1.3
cj :
CB
XB
(2
0)
y1
y2
y3
y4
x1
x2
s1
s2
x1
-1
s2
-2
zj cj
z(XB ) = 4
Min. Ratio
2/1=2
leaves the basis.
ABLE OF
C ONTENTS
18

Initial iteration:
Tableau 6.1.3
cj :
CB
XB
(2
0)
y1
y2
y3
y4
x1
x2
s1
s2
x1
-1
s2
-2
zj cj
z(XB ) = 4
Min. Ratio
2/1=2
leaves the basis.
ABLE OF
C ONTENTS
19
Tableau 6.1.3
ABLE OF
cj :
CB
XB
(2
0)
x1
x2
s1
s2
x1
-1
s2
-2
zj cj
C ONTENTS
z(XB ) = 4
Min. Ratio
R1 + R2
2/1=2
R3 + 2 R2
20
Tableau 6.1.3
ABLE OF
cj :
CB
XB
(2
0)
x1
x2
s1
s2
x1
s1
zj cj
C ONTENTS
z(XB ) = 8
Min. Ratio
21
First iteration:
Tableau 6.1.4
cj :
CB
XB
(2
0)
x1
x2
s1
s2
x1
s1
zj cj
z(XB ) = 8
Since all zj cj 0, the current basic feasible solution is optimal.

The optimal solution is
x1 = 4, x2 = 0 with Maximum of (z) = 8
That is, x1 = 4, x2 = 0 with minimum of z = 8
ABLE OF
C ONTENTS
22
23
In this section we discuss some complications and variations which

are very often encountered during simplex procedure.
The following are some of them:
1. Degeneracy and cycling
2. Unbounded solution
3. Multiple solution or alternative optimum solution
4. No feasible solution
5. Unrestricted variables.
24
1. D EGENERACY (T IE FOR MINIMUM RATIO )
25
The concept of obtaining a degenerate basic feasible solution in a

LPP is known as degeneracy.
At the stage of improving the solution during simplex procedure,
minimum ratio

xBr
xBi
= min
yik > 0
i
yrk
yik
is determined in the last column of simplex table to find the key
row.
26
But sometimes
this ratio may not be unique
(i.e. more than one variable is eligible to leave the basis)
or
at the very first iteration, the value of one or more basic variables
in XB column becomes zero,
this causes the problem of degeneracy.
27
If the minimum ratio is zero, for two or more basic variables,

degeneracy may result and the simplex routine to cycle
indefinitely.
That is, the solution which we have obtained in one iteration
may repeat again after few iterations and therefore, no
optimum solution may be obtained.
This concept is known as cycling or circling.
28
M ETHOD TO RESOLVE DEGENERACY (T IE )
29
Step 1: First find out the rows for which the minimum non-negative
ratio is the same (tie), for example, suppose there is a tie between
first and second row.
Step 2: Now, rearrange the columns of the usual simplex table so
that the columns forming the original unit matrix come first.
Step 3: Find the minimum of the ratio

elements of first column of unit matrix
corresponding elements of key column
Only for the rows for which the minimum ratio was not unique
(i.e. for tied rows, in our example for first and second rows).
(i) If this minimum is attained for second row (say) then this row will
determine the key element by intersecting with the key column.
(ii) If this minimum is also not unique, then go to the next step.
30
Step 4: Compute the minimum of the ratio

elements of second column of unit matrix
only for the rows for which minimum ratio is not unique in Step 3.
If the minimum is also not unique, then go to next step.
Step 5: Next compute the minimum of the ratio

elements of third column of unit matrix
If this minimum is still not unique, then go on repeating the
above procedure till the unique minimum ratio is obtained to
resolve the degeneracy.
After the resolution of this tie, simplex method is applied to
obtain the optimum solution.
31
Remark 1 If we have a tie for an artificial variable and some other

variable we can choose the artificial variable to leave the basis.
32
Example 2 Solve the LPP

Maximise z = 3x1 + 9x2
subject to
x1 + 4x2 8
x1 + 2x2 4
x1 , x2 0
33
Solution:
Introduce the slack variables s1 , s2 0, the LPP becomes
Maximise z = 3x1 + 9x2 + 0s1 + 0s2
subject to
x1
4x2
x1
2x2
s1
+
s2
x1 , x2 , s1 , s2 0
An obvious initial basic feasible solution is s1 = 8, s2 = 4.
34
Tableau 6.2.1
cj :
CB
XB
(3
0)
x1
x2
s1
s2
s1
s2
z(XB )
-3
-9
zj cj
Min. Ratio
From Tableau 6.2.1, we observe that the non-basic variable x2

enters into the basis.
Since the minimum ratio is 2 for both the slack variables s1 and
s2 , there is a tie for the variable to leave the basis.
This is an indication for the existence of degeneracy in the given
LPP.
Rearrange the columns of the simplex table so that the initial
identity matrix appears first.
35
Tableau 6.2.1
cj :
CB
XB
(3
0)
x1
x2
s1
s2
s1
s2
z(XB )
-3
-9
zj cj
Min. Ratio

LPP.
36
Tableau 6.2.1
cj :
CB
XB
(3
0)
x1
x2
s1
s2
Min. Ratio
s1
8/4=2
s2
4/2=2
z(XB )
-3
-9
zj cj

LPP.
37
Tableau 6.2.1
cj :
CB
XB
(3
0)
x1
x2
s1
s2
Min. Ratio
s1
8/4=2
s2
4/2=2
z(XB )
-3
-9
zj cj

LPP.
38
Tableau 6.2.1
cj :
CB
XB
(0
9)
y1
y2
y3
y4
s1
s2
x1
x2
Min. Ratio
s1
1/4
s2
0/2=0
-3
-9
zj cj
z(XB ) = 0
39
Now, using the Step 3 of the procedure for resolving degeneracy,

we find

elements of first column
min

1 0
,
= 0.
= min
4 2
which occurs for the second row. Hence, s2 must leave the basis
and the pivot element is 2.
Tableau 6.2.2
cj :
CB
XB
(0
9)
s1
s2
x1
x2
Min. Ratio
s1
1/4
s2
0/2=0
-3
-9
zj cj
z(XB ) = 0
40
Now, using the Step 3 of the procedure for resolving degeneracy,

we find

elements of first column
min

1 0
,
= 0.
= min
4 2
which occurs for the second row. Hence, s2 must leave the basis
and the pivot element is 2.
Tableau 6.2.2
cj :
CB
XB
(0
9)
s1
s2
x1
x2
Min. Ratio
s1
1/4
s2
0/2=0
-3
-9
zj cj
z(XB ) = 0
41
Tableau 6.2.2
cj :
CB
XB
(0
9)
s1
s2
x1
x2
Min. Ratio
s1
1/4
s2
0/2=0
-3
-9
zj cj
z(XB ) = 0
1
2
R2
Tableau 6.2.2
cj :
CB
XB
(0
9)
s1
s2
x1
x2
Min. Ratio
s1
-2
-1
1/4
s2
1/2
1/2
0/2=0
9/2
3/2
zj cj
z(XB ) = 18
R1 4 R2
R3 + 9 R2
42
First iteration:
Tableau 6.2.3
cj :
CB
XB
(0
9)
s1
s2
x1
x2
s1
-2
-1
x2
1/2
1/2
9/2
3/2
zj cj
z(XB ) = 18
Since all zj cj 0, an optimum solution has been reached.

Hence, the optimum basic feasible solution is
x1 = 0, x2 = 2 with maximum of z = 18
43
2. Unbounded solutions:
The case of unbounded solutions occur when the feasible
region is unbounded such that the value of the objective
function can be increased indefinitely.
It is not necessary, however, that an unbounded feasible region
will give rise to an unbounded value of the objective function.
The following example will illustrate these points.
2. Unbounded solutions:
44
U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

Example 3 Solve the following LPP
Maximise z = 3x1 x2
subject to
x1 x2
10
x1
20
x1 , x2 0
45
Solution:
By introducing the slack variables s1 , s2 0, the LPP becomes
Maximise z = 3x1 x2 + 0s1 + 0s2
subject to
x1
x1
x2
s1
+
s2
10
20
x1 , x2 , s1 , s2 0
46
First Iteration
Tableau 6.3.1
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
s1
10
-1
s2
20
-3
zj c j
z(XB ) = 0
Min. Ratio
Tableau 6.3.2
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
s1
10
-1
10/1 = 10
s2
20
20/1 = 20
z(XB ) = 0
-3
zj c j
R2 R1
R3 + 3 R1
47
Tableau 6.3.3
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
x1
10
-1
s2
10
-1
-2
zj c j
z(XB ) = 30
Min. Ratio
Second Iteration
Tableau 6.3.4
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
x1
10
-1
s2
10
-1
10/1 = 10
z(XB ) = 30
-2
zj c j
48
Tableau 6.3.5
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
x1
10
-1
s2
10
-1
10/1 = 10
z(XB ) = 30
-2
zj c j
R1 + R2
R3 + 2 R2
Tableau 6.3.6
cj :
CB
XB
(3
-1
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
x1
20
-1
x2
10
-1
zj c j
z(XB ) = 50
Min. Ratio
The optimum solution is:

x1 = 20, x2 = 10 with maximum of z = 50
49
The graphical solution of the LPP is given in Figure 1.

x2
x2
x2
x 1 = 20
x 1 - x 2= 10
15
10
B(20,10)
x 1 = 20
x 1 - x 2= 10
15
10
B(20,10)
10
15
(a)
x 1 - x 2= 10
15
10
A(10,0)
20
x1
B(20,10)
A(10,0)
x 1 = 20
A(10,0)
10
15
20
x1
(b)
10
15
20
x1
(c)
Figure 1:
We observe that the solution space is unbounded. But the optimal
solution occurs at the vertex (20, 10).
50
NBOUNDED
S OLUTION S PACE AND U NBOUNDED O PTIMAL S OLUTION

Maximise z = 2x1 + x2
subject to
x1 x2
10
2x1 x2
40
x1 0, x2 0
U NBOUNDED S OLUTION S PACE AND U NBOUNDED O PTIMAL S OLUTION
51
Solution:
Introducing the slack variables s1 , s2 0 the LPP in standard form is
Maximise z = 2x1 + x2 + 0s1 + 0s2
subject to
x1
x2
2x1
x2
s1
+
s2
10
40
x1 , x2 , s1 , s2 0
An obvious IBFS is s1 = 10, s2 = 40.
52
Starting Table:
Tableau 6.4.1
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
s1
10
-1
s2
40
-1
-2
-1
zj c j
z(XB ) = 0
Min. Ratio
Tableau 6.4.2
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
s1
10
-1
s2
40
-1
z(XB ) = 0
-2
-1
zj c j
Min. Ratio
10
1
40
2
= 10
= 20
53
Tableau 6.4.3
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
s1
10
-1
s2
40
-1
R2 2 R1
z(XB ) = 0
-2
-1
R3 + 2 R1
zj c j
Tableau 6.4.4
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
x1
10
-1
s2
20
-2
-3
zj c j
z(XB ) = 20
Min. Ratio
54
First iteration:
Tableau 6.4.5
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
x1
10
-1
s2
20
-2
-3
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
zj c j
z(XB ) = 20
Tableau 6.4.6
cj :
CB
XB
x1
10
-1
s2
20
-2
20/1 = 20
z(XB ) = 20
-3
zj c j
55
Tableau 6.4.7
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
x1
10
-1
s2
20
-2
z(XB ) = 20
-3
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
zj c j
R1 + R2
R3 + 3 R2
Tableau 6.4.8
cj :
CB
XB
x1
30
-1
x2
20
-2
-4
zj c j
z(XB ) = 80
Min. Ratio
56
Second iteration:
Table 6.4.8
cj :
CB
XB
(2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
x1
30
-1
x2
20
-2
-4
zj c j
z(XB ) = 80
Min. Ratio
From Table 6.4.8 we see that s1 column is the pivotal column,

but there is no positive element in that column.
Hence, there exists an unbounded solution to the given LPP.
If we solve the LPP by graphical method we can see that the
feasible region is unbounded. See Figure 2.
57
x 1 - x 2= 10
20
10
5 10
20
x1
30
10
20
P(30,20)
30
2 x 1 - x 2 = 40
40
Figure 2:
58
M ULTIPLE S OLUTIONS OR A LTERNATE O PTIMAL S OLUTIONS
59
3. M ULTIPLE S OLUTIONS OR A LTERNATE O PTIMAL S OLUTIONS

While solving the LPP by simplex method, in the optimum simplex
table, if the net evaluation zj cj 6= 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 zj cj = 0 for at least
one non-basic variable, then the problem is said to have an
alternative or infinite number of solutions.
60
In a graphical method if the optimal solution occurs at a vertex

of the solution space, then the problem said to have
a unique optimal solution.
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.
61

Maximise z = x1 +
1
x2
2
subject to
2x1
x2
x1
2x2
x1 , x2 0
62
Solution:
By introducing the slack variables s1 , s2 0, the LPP becomes
Maximise z = x1 +
1
x2 + 0s1 + 0s2
2
subject to
2x1
x2
x1
2x2
s1
+
s2
x1 , x2 , s1 , s2 0
63
First Iteration
Tableau 6.5.1
cj :
CB
XB
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
s1
s2
-1
12
zj c j
z(XB ) = 0
Min. Ratio
Tableau 6.5.2
cj :
CB
XB
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
s1
s2
z(XB ) = 0
-1
12
zj c j
Min. Ratio
4
2
3
1
=2
=4
1
2
R1
64
Tableau 6.5.3
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
cj :
CB
XB
s1
1/2
1/2
s2
z(XB ) = 0
-1
12
zj c j
Min. Ratio
4
2
3
1
=2
=4
R2 R1
R3 + R1
Tableau 6.5.4
cj :
CB
XB
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
x1
1/2
1/2
s2
3/2
-1/2
1/2
zj c j
z(XB ) = 2
Min. Ratio
65
Tableau 6.5.5
cj :
CB
XB
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
2
=4
1/2
1
3/2 = 2/3
x1
1/2
1/2
s2
3/2
-1/2
z(XB ) = 2
1/2
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
Min. Ratio
2
1/2 = 4
1
3/2 = 2/3
zj c j
2
3
R2
Tableau 6.5.6
cj :
CB
XB
x1
1/2
1/2
s2
2
3
-1/3
2/3
1/2
zj c j
z(XB ) = 2
R1
1
2
R2
66
Tableau 6.5.7
cj :
CB
XB
x1
1
2
x2
zj c j
5
3
2
3
z(XB ) = 2
(1
1
2
0)
y1
y2
ys1
ys2
x1
x2
s1
s2
2/3
-1/3
-1/3
2/3
1/2
Min. Ratio
67
N O FEASIBLE SOLUTION OR N ON - EXISTING FEASIBLE SOLUTION
68
In an LPP, where there is no point in the solution space satisfying

all the constraints, then the problem is said to have no feasible
solution.
In simplex method, if there exists at least one artificial variable in
the basis at positive level, and even though optimality
conditions are satisfied which is the indication of non-feasible
solution.
69
C.f. Example 2 in Lecture Note 1d.

subject to
2x1 + x2 2
3x1 + 4x2 12
x1 , x2 0.
70
First iteration:
Tableau 5.4.2
cj :
CB
XB
(3
M )
x1
x2
s1
s2
A1
x2
A1
-5
-4
-1
5M + 1
4M + 2
zj cj
z(XB ) = 4M + 4
Here, the coefficient of M in each zj cj 0, and an artificial variable

A1 appears in the basis at non-zero level.
Thus, the given LPP does not possess any feasible solution.
We can say the LPP possess a pseudo-optimal solution.
71
U NRESTRICTED VARIABLE
72
In an LPP, if any variable is unrestricted (it can have positive

value or negative value or zero value) it can be expressed as
the difference between two non-negative variables.
The problem can be converted into an equivalent one involving
only non-negative variables.
73
Example 6 Solve the LPP

subject to
x1 + 2x2
x1 + x2
x1 + 3x2
and x1 , x2 are unrestricted.
74
Solution:
Since x1 , x2 are unrestricted, we put x1 = x1 x1 and x2 = x2 x2 so
that x1 , x1 , x2 , x2 0.
The given LPP becomes
Maximise z = 2x1 2x1 + 3x2 3x2
subject to
x1
x1
2x2
2x2
x1
x1
x2
x2
x1
x1
3x2
3x2
x1 , x1 x2 , x2 0
75
The given LPP becomes

Maximise z = 2x1 2x1 + 3x2 3x2 + 0s1 + 0s2 + 0s3
subject to
x1
x1
2x2
2x2
x1
x1
x2
x2
x1
x1
3x2
3x2
s1
+
s2
+
s3
x1 , x1 , x2 , x2 , s1 , s2 , s3 0
76
Tableau 6.6.1
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
s1
-1
-2
s2
-1
-1
s3
-1
-3
-2
-3
zj c j
z(XB ) = 0
Min Ratio
Tableau 6.6.2
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
Min Ratio
s1
-1
-2
4/2 = 2
s2
-1
-1
6/1 = 6
s3
-1
-3
9/3 = 3
z(XB ) = 0
-2
-3
zj c j
77
Tableau 6.6.3
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
Min Ratio
s1
-1/2
1/2
-1
1/2
4/2 = 2
s2
-1
-1
6/1 = 6
s3
-1
-3
9/3 = 3
z(XB ) = 0
-2
-3
zj c j
1
2
R1
Tableau 6.6.4
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
Min Ratio
4/2 = 2
x2
-1/2
1/2
-1
1/2
s2
3/2
-3/2
-1/2
R2 R1
s3
5/2
-5/2
-3/2
R3 3 R1
-7/2
7/2
3/2
R4 + 3 R1
zj c j
z(XB ) = 6
78
Tableau 6.6.5
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
x2
-1/2
1/2
-1
1/2
s2
3/2
-3/2
-1/2
s3
5/2
-5/2
-3/2
-7/2
7/2
3/2
zj c j
z(XB ) = 6
Min Ratio
Tableau 6.6.6
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
Min Ratio
x2
-1/2
1/2
-1
1/2
s2
3/2
-3/2
-1/2
4/(3/2) = 2.6667
s3
5/2
-5/2
-3/2
3/(5/2) = 1.2
z(XB ) = 6
-7/2
7/2
3/2
zj c j
79
2
5
R3
Tableau 6.6.7
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
M. R.
R3
R2
1
2
3
2
R4 +
7
2
R3
x2
-1/2
1/2
-1
1/2
R1 +
s2
3/2
-3/2
-1/2
s3
6/5
-1
-3/5
2/5
-7/2
7/2
3/2
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
zj c j
z(XB ) = 6
Tableau 6.6.8
cj :
CB
XB
x2
13/5
-1
1/5
1/5
s2
11/5
2/5
-3/5
x1
6/5
-1
-3/5
2/5
-3/5
7/5
zj c j
z(XB ) = 51/5
M. R.
80
R3
Tableau 6.6.9
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
M. R.
x2
13/5
-1
1/5
1/5
s2
11/5
2/5
-3/5
x1
6/5
-1
-3/5
2/5
-3/5
7/5
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
M. R.
zj c j
z(XB ) = 51/5
Tableau 6.6.10
cj :
CB
XB
x2
13/5
-1
1/5
1/5
(13/5)/(1/5) = 13
s2
11/5
2/5
-3/5
(11/5)/(2/5) = 5.5
x1
6/5
-1
-3/5
2/5
-3/5
7/5
zj c j
z(XB ) = 51/5
81
5
2
R2
Tableau 6.6.11
cj :
CB
XB
(2
-2
-3
0)
y1
y2
y3
y4
ys1
ys2
ys3
x1
x
1
x2
x
2
s1
s2
s3
R1
1
5
R2
2/5
R3 +
R2
7/5
R4 +
3
5
3
5
0)
y4
ys1
ys2
ys3
x2
x
2
s1
s2
s3
x2
13/5
-1
1/5
1/5
s2
11/2
5/2
-3/2
x1
6/5
-1
-3/5
-3/5
(2
-2
-3
y1
y2
y3
x1
x
1
zj c j
z(XB ) = 51/5
M. R.
Tableau 6.6.12
cj :
CB
XB
x2
3/2
-1
-1/2
1/2
s1
11/2
5/2
-3/2
x1
9/2
-1
3/2
-1/2
3/2
1/2
zj c j
z(XB ) = 27/2
M. R.
82
R2
If we solve the LPP by ordinary simplex method, the optimum

solution is
x1 = x1 x1 = 9/2 0 = 9/2
x2 = x2 x2 = 3/2 0 = 3/2
with maximum of z = 27/2
83

5.1 Lecture - 1f

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

5.1 Lecture - 1f

Transféré par

Droits d'auteur :

Formats disponibles

Lecture 1f:

VARIANTS OF SIMPLEX METHOD

Example 1 Use two phase simplex method to

By introducing slack variable s2 , surplus variable s1 , artificial variable

The initial basis matrix is:

An obvious initial basis feasible solution is s2 = 4, A1 = 2, i.e.,

Phase I The objective function of the auxiliary LPP is

From Tableau 6.1.1 we observe that

the non-basic variable x1 enters into the basis;

The new basis matrix is:

Since all zj cj 0 and no artificial variable appears in the

The initial basis matrix is:

An obvious initial basis feasible solution is x1 = 2, s2 = 2, i.e.,

z1 c1 = CB y1 c1 = [2 0][1 0]T (2) = 0

The iterative simplex tableaus for this phase is:

The iterative simplex tableaus for this phase is:

The iterative simplex tableaus for this phase is:

Since all zj cj 0, the current basic feasible solution is optimal.

VARIANTS OF SIMPLEX METHOD

VARIANTS OF SIMPLEX METHOD

In this section we discuss some complications and variations which

VARIANTS OF SIMPLEX METHOD

1. D EGENERACY (T IE FOR MINIMUM RATIO )

1. D EGENERACY (T IE FOR MINIMUM RATIO )

The concept of obtaining a degenerate basic feasible solution in a

1. D EGENERACY (T IE FOR MINIMUM RATIO )

1. D EGENERACY (T IE FOR MINIMUM RATIO )

If the minimum ratio is zero, for two or more basic variables,

1. D EGENERACY (T IE FOR MINIMUM RATIO )

M ETHOD TO RESOLVE DEGENERACY (T IE )

M ETHOD TO RESOLVE DEGENERACY (T IE )

M ETHOD TO RESOLVE DEGENERACY (T IE )

Step 4: Compute the minimum of the ratio

Step 5: Next compute the minimum of the ratio

M ETHOD TO RESOLVE DEGENERACY (T IE )

Remark 1 If we have a tie for an artificial variable and some other

M ETHOD TO RESOLVE DEGENERACY (T IE )

Example 2 Solve the LPP

M ETHOD TO RESOLVE DEGENERACY (T IE )

M ETHOD TO RESOLVE DEGENERACY (T IE )

From Tableau 6.2.1, we observe that the non-basic variable x2

M ETHOD TO RESOLVE DEGENERACY (T IE )

From Tableau 6.2.1, we observe that the non-basic variable x2

M ETHOD TO RESOLVE DEGENERACY (T IE )

From Tableau 6.2.1, we observe that the non-basic variable x2

M ETHOD TO RESOLVE DEGENERACY (T IE )

From Tableau 6.2.1, we observe that the non-basic variable x2

M ETHOD TO RESOLVE DEGENERACY (T IE )

M ETHOD TO RESOLVE DEGENERACY (T IE )

Now, using the Step 3 of the procedure for resolving degeneracy,

M ETHOD TO RESOLVE DEGENERACY (T IE )

Now, using the Step 3 of the procedure for resolving degeneracy,

M ETHOD TO RESOLVE DEGENERACY (T IE )

M ETHOD TO RESOLVE DEGENERACY (T IE )

Since all zj cj 0, an optimum solution has been reached.

M ETHOD TO RESOLVE DEGENERACY (T IE )

U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

U NBOUNDED SOLUTION SPACE BUT BOUNDED OPTIMAL SOLUTION

The optimum solution is: