mywbut.

com

Simplex Method

In the previous chapter, we discussed about the graphical method for solving linear programming
problems. Although the graphical method is an invaluable aid to understand the properties of
linear programming models, it provides very little help in handling practical problems. In this
chapter, we concentrate on the simplex method for solving linear programming problems with a
larger number of variables.

Many different methods have been proposed to solve linear programming problems, but simplex
method has proved to be the most effective. This technique will nurture your insight needed for a
sound understanding of several approaches to other programming models, which will be studied
in subsequent chapters. This method is applicable to any problem that can be formulated in terms
of linear objective function, subject to a set of linear constraints. Often, this method is termed
Dantzig's simplex method, in honour of the mathematician who devised the approach.

In the following section, we introduce you to the standard vocabulary of the simplex method.

Basic Terminology

Slack variable

It is a variable that is added to the left-hand side of a less than or equal to type constraint to
convert the constraint into an equality. In economic terms, slack variables represent left-over or
unused capacity.

Specifically:
ai1x1 + ai2x2 + ai3x3 + .........+ ainxn ≤ bi can be written as
ai1x1 + ai2x2 + ai3x3 + .........+ ainxn + si = bi

Where i = 1, 2, ..., m

Surplus variable

It is a variable subtracted from the left-hand side of a greater than or equal to type constraint to
convert the constraint into an equality. It is also known as negative slack variable. In economic
terms, surplus variables represent overfulfillment of the requirement.

Specifically:
ai1x1 + ai2x2 + ai3x3 + .........+ ainxn ≥ bi can be written as
ai1x1 + ai2x2 + ai3x3 + .........+ ainxn - si = bi

Where i = 1, 2, ..., m

1
mywbut.com

Artificial variable

It is a non negative variable introduced to facilitate the computation of an initial basic feasible
solution. In other words, a variable added to the left-hand side of a greater than or equal to type
constraint to convert the constraint into an equality is called an artificial variable.

Simplex Method - Maximization Case

Consider the general linear programming problem

Maximize z = c1x1 + c2x2 + c3x3 + .........+ cnxn

subject to

a11x1 + a12x2 + a13x3 + .........+ a1nxn ≤ b1

a21x1 + a22x2 + a23x3 + .........+ a2nxn ≤ b2
.........................................................................
am1x1 + am2x2 + am3x3 + .........+ amnxn ≤ bm
x1, x2,....., xn≥ 0

Where:
cj (j = 1, 2, ...., n) in the objective function are called the cost or profit coefficients.
bi (i = 1, 2, ...., m) are called resources.
aij (i = 1, 2, ...., m; j = 1, 2, ...., n) are called technological coefficients or input-output
coefficients.

Converting inequalities to equalities

Introducing slack variables to convert inequalities to equalities

a11x1 + a12x2 + a13x3 + .........+ a1nxn + s1 = b1

a21x1 + a22x2 + a23x3 + .........+ a2nxn + s2 = b2
..............................................................................
am1x1 + am2x2 + am3x3 + .........+ amnxn + sm = bm
x1, x2,....., xn ≥ 0
s1, s2,....., sm ≥ 0

An initial basic feasible solution is obtained by setting x1 = x2 =........ = xn = 0

s1 = b1
s2 = b2
..............
sm = bm

The initial simplex table is formed by writing out the coefficients and constraints of a LPP in a
systematic tabular form. The following table shows the structure of a simplex table.

2
mywbut.com

Structure of a simplex table

Where:
cj = coefficients of the variables (m + n) in the objective function.
cB = coefficients of the current basic variables in the objective function.
B = basic variables in the basis.
XB = solution values of the basic variables.
zj-cj = index row.

The Simplex Algorithm

Unquestionably, the simplex technique has proved to be the most effective in solving linear
programming problems. In the simplex method, we first find an initial basic solution (extreme
point). Then, we proceed to an adjacent extreme point. We continue this process until we reach
an optimal solution.

Don't be too perplexed, if you find certain details unclear in

this section - the vagueness will automatically disappear by
the time you have finished this chapter.

Steps (Maximization Case)

1. Formulate the Problem

• Formulate the mathematical model of the given linear programming problem.

• If the objective function is given in minimization form then convert it into maximization form in
the following way:
Min z = - Max (-z)

3
mywbut.com

Any minimization problem can be converted into an equivalent maximization

problem by multiplying the objective function with (-1)

• Convert every inequality constraint in the L.P. problem into an equality constraint by adding a
slack variable to each constraint.

2. Find out the Initial Solution

• Calculate the initial basic feasible solution by assigning zero value to the decision variables. This
solution is shown in the initial simplex table.

3. Test for Optimality

• Calculate the values of zj – cj

• If the values of zj – cj are positive, the current basic feasible solution is the optimal solution. If
there are one or more negative values, choose the variable corresponding to which the value of
zj – cj is least (most negative) as this is likely to increase the profit most.

4. Test for Feasibility

• Divide the values under XB column by the corresponding positive coefficient (aij) in the key
column, and compare the ratios. The row that indicates the minimum ratio is called the key row.
However, division by zero or negative coefficients in the key column is not allowed. In the case
of a tie, break the tie arbitrarily.

5. Identify the Pivot Element (Key Element)

• The number that lies at the intersection of the key column and key row of a given table is called
the key element. It is always a non-zero positive number.

6. Determine the New Solution

• The numbers in the replacing row may be obtained by dividing the key row elements by the
pivot element. The numbers in the remaining rows may be calculated by using the following
formula:

(corresponding no. of key row) X (corresponding no. of key column)

New number= old number- ------------------------------------------------------------------------------
pivot element

7. Revise the Solution

• Go to step 3 and repeat the procedure until all the values of zj – cj are either zero or positive.

4
mywbut.com

For the time being we assume that a feasible solution exists and the optimal value of the
objective function is finite. Later in the chapter, we will remove these restrictive
assumptions and consider several special cases like unbounded solution, multiple
optimum solution, no feasible solution, and degeneracy.

Example 1

Maximize z = 3x1 + 2x2

subject to

-x1 + 2x2 ≤ 4
3x1 + 2x2 ≤ 14
x1 – x2 ≤ 3

x1, x2 ≥ 0

Solution.

First, convert every inequality constraints in the LPP into an equality constraint, so that the
problem can be written in a standard from. This can be accomplished by adding a slack variable
to each constraint. Slack variables are always added to the less than type constraints.

Converting inequalities to equalities

-x1 + 2x2 + x3 = 4
3x1 + 2x2 + x4 = 14
x1 – x2 + x5 = 3
x1, x2, x3, x4, x5 ≥ 0

Where x3, x4 and x5 are slack variables.

Since slack variables represent unused resources, their contribution in the objective function is
zero. Including these slack variables in the objective function, we get

Maximize z = 3x1 + 2x2 + 0x3 + 0x4 + 0x5

5
mywbut.com

Initial basic feasible solution

Now we assume that nothing can be produced. Therefore, the values of the decision variables are
zero.
x1 = 0, x2 = 0, z = 0

When we are not producing anything, obviously we are left with unused capacity
x3 = 4, x4 = 14, x5 = 3

We note that the current solution has three variables (slack variables x3, x4 and x5) with non-zero
solution values and two variables (decision variables x1 and x2) with zero values. Variables with
non-zero values are called basic variables. Variables with zero values are called non-basic
variables.

Table 1

a11 = -1, a12 = 2, a13 = 1, a14 = 0, a15 = 0, b1 = 4

a21 = 3, a22 = 2, a23 = 0, a24 = 1, a25 = 0, b2 = 14
a31= 1, a32 = -1, a33 = 0, a34 = 0, a35 = 1, b3 = 3

Working

Calculating values for the index row (zj – cj)

z1 – c1 = (0 X (-1) + 0 X 3 + 0 X 1) - 3 = -3
z2 – c2 = (0 X 2 + 0 X 2 + 0 X (-1)) - 2 = -2
z3 – c3 = (0 X 1 + 0 X 0 + 0 X 0) - 0 = 0
z4 – c4 = (0 X 0 + 0 X 1 + 0 X 0) - 0 = 0
z5 – c5 = (0 X 0 + 0 X 0 + 0 X 1) – 0 = 0

Choose the smallest negative value from zj – cj (i.e., – 3). So column under x1 is the key
column.
Now find out the minimum positive value
Minimum (14/3, 3/1) = 3
So row x5 is the key row.

6
mywbut.com

Here, the pivot (key) element = 1 (the value at the point of intersection).
Therefore, x5 departs and x1 enters.

We obtain the elements of the next table using the following rules:

1. If the values of zj – cj are positive, the inclusion of any basic variable will not increase the value of
the objective function. Hence, the present solution maximizes the objective function. If there
are more than one negative values, we choose the variable as a basic variable corresponding to
which the value of zj – cj is least (most negative) as this will maximize the profit.
2. The numbers in the replacing row may be obtained by dividing the key row elements by the
pivot element and the numbers in the other two rows may be calculated by using the formula:

(corresponding no. of key row) x (corresponding no. of key column)

New number= old number- ------------------------------------------------------------------------------
pivot element

Calculating values for table 2

x3 row

a11 = -1 – 1 X ((-1)/1) = 0
a12 = 2 – (-1) X ((-1)/1) = 1
a13 = 1 – 0 X ((-1)/1) = 1
a14 = 0 – 0 X ((-1)/1) = 0
a15 = 0 – 1 X ((-1)/1) = 1
b1 = 4 – 3 X ((-1)/1) = 7

x4 row

a21 = 3 – 1 X (3/1) = 0
a22 = 2 – (-1) X (3/1) = 5
a23 = 0 – 0 X (3/1) = 0
a24 = 1 – 0 X (3/1) = 1
a25 = 0 – 1 X (3/1) = -3
b2 = 14 – 3 X (3/1) = 5

x1 row

a31 = 1/1 = 1
a32 = -1/1 = -1
a33 = 0/1 = 0
a34 = 0/1 = 0

7
mywbut.com

a35 = 1/1 = 1
b3 = 3/1 = 3

Table 2

Working

Calculating values for the index row (zj – cj)

z1 – c1 = (0 X 0 + 0 X 0 + 3 X 1) - 3 = 0
z2 – c2 = (0 X 1 + 0 X 5 + 3 X (-1)) – 2 = -5
z3 – c3 = (0 X 1 + 0 X 0 + 3 X 0) - 0 = 0
z4 – c4 = (0 X 0 + 0 X 1 + 3 X 0) - 0 = 0
z5 – c5 = (0 X 1 + 0 X (-3) + 3 X 1) – 0 = 3

Key column = x2 column

Minimum (7/1, 5/5) = 1
Key row = x4 row
Pivot element = 5
x4 departs and x2 enters.

Calculating values for table 3

x3 row

a11 = 0 – 0 X (1/5) = 0
a12 = 1 – 5 X (1/5) = 0
a13 = 1 – 0 X (1/5) = 1
a14 = 0 – 1 X (1/5) = -1/5
a15 = 1 – (-3) X (1/5) = 8/5
b1 = 7 – 5 X (1/5) = 6

8
mywbut.com

x2 row

a21 = 0/5 = 0
a22 = 5/5 = 1
a23 = 0/5 = 0
a24 = 1/5
a25 = -3/5
b2 = 5/5 = 1

x1 row

a31 = 1 – 0 X (-1/5) = 1
a32 = -1 – 5 X (-1/5) = 0
a33 = 0 – 0 X (-1/5) = 0
a34 = 0 – 1 X (-1/5) = 1/5
a35 = 1 – (-3) X (-1/5) = 2/5
b3 = 3 – 5 X (-1/5) = 4

Don't convert the fractions into decimals, because

many fractions cancel out during the process while
the conversion into decimals will cause unnecessary
complications.

Final Table

Since all the values of zj – cj are positive, this is the optimal solution.
x1 = 4, x2 = 1
z = 3 X 4 + 2 X 1 = 14.

The largest profit of Rs.14 is obtained, when 1 unit of x2 and 4 units of x1 are produced. The
above solution also indicates that 6 units are still unutilized, as shown by the slack variable x3 in
the XB column.

Real life complex applications usually involve hundreds of constraints and

thousands of variables. So virtually these problems can not be solved
manually. For solving such problems, you will have to rely on employing an

9
mywbut.com

electronic computer

Example 2

Luminous Lamps produces three types of lamps - A, B, and C. These lamps are processed on
three machines - X, Y, and Z. The full technology and input restrictions are given in the
following table.

Machine
Product Profit per unit
X Y Z

A 10 7 2 12

B 2 3 4 3

C 1 2 1 1

Available Time 100 77 80

Find out a suitable product mix so as to maximize the profit.

Solution.

The decision problem can be formulated as

Maximize z = 12x1 + 3x2 + x3

subject to

10x1 + 2x2 + x3 ≤ 100

7x1 + 3x2 + 2x3 ≤ 77
2x1+ 4x2 + x3 ≤ 80

x1, x2, x3 ≥ 0

Converting inequalities to equalities

10x1 + 2x2 + x3 + x4 = 100

7x1 + 3x2 + 2x3 + x5 = 77

10
mywbut.com

2x1+ 4x2 + x3 + x6 = 80
x1, x2, x3, x4, x5, x6 ≥ 0

Where x4, x5 and x6 are slack variables.

Including these slack variables in the objective function, we get

Maximize z = 12x1 + 3x2 + x3 + 0x4 + 0x5 + 0x6

Initial basic feasible solution

x1 = 0, x2 = 0, x3 = 0, z = 0
x4 = 100, x5 = 77, x6 = 80

Table 1

cj 12 3 1 0 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5 x6
B b (=XB)

0 x4 10 2 1 1 0 0 100

0 x5 7 3 2 0 1 0 77

0 x6 2 4 1 0 0 1 80

zj-cj -12 -3 -1 0 0 0

Key column = x1 column.

Minimum (100/10, 77/7, 80/2) = 10
Key row = x4 row
Pivot element = 10
x4 departs and x1 enters

Now we are assuming that you can easily calculate the values yourself.

11
mywbut.com

Table 2
cj 12 3 1 0 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5 x6
B b (= XB)

12 x1 1 1/5 1/10 1/10 0 0 10

0 x5 0 8/5 13/10 -7/10 1 0 7

0 x6 0 18/5 4/5 -1/5 0 1 60

zj-cj 0 -3/5 1/5 6/5 0 0

Final Table
cj 12 3 1 0 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5 x6
B b (= XB)

12 x1 1 0 -1/16 3/16 -1/8 0 73/8

3 x2 0 1 13/16 -7/16 5/8 0 35/8

0 x6 0 0 -17/8 11/8 -9/4 1 177/4

zj-cj 0 0 11/16 15/16 3/8 0

An optimal policy is x1 =73/8, x2 = 35/8, x3 = 0.

The associated optimal value of the objective function is z = 12 X (73/8) + 3 X (35/8) + 1 X 0 =
981/8.

Minimization Case

In the previous section, the simplex method was applied to linear programming problems where
the objective was to maximize the profit with less than or equal to type constraints. In many
cases, however, constraints may of type ≥ or = and the objective may be minimization (e.g., cost,
time, etc.). Thus, in such cases, simplex method must be modified to obtain an optimal policy.

Consider the general linear programming problem

Minimize z = c1x1 + c2x2 + c3x3 + .........+ cnxn

12
mywbut.com

subject to

a11x1 + a12x2 + a13x3 + .........+ a1nxn ≥ b1

a21x1 + a22x2 + a23x3 + .........+ a2nxn ≥ b2
.........................................................................
am1x1 + am2x2 + am3x3 + .........+ amnxn ≥ bm
x1, x2,....., xn≥ 0

Changing the sense of the optimization

Any linear minimization problem can be viewed as an equivalent linear maximization problem,
and vice versa.

Min. z = c1x1 + c2x2 + c3x3 + .........+ cnxn

It can be written as
Max. z = -(c1x1 + c2x2 + c3x3 + .........+ cnxn)

Converting inequalities to equalities

Introducing surplus variables (negative slack variables) to convert inequalities to equalities

a11x1 + a12x2 + a13x3 + .........+ a1nxn - s1 = b1

a21x1 + a22x2 + a23x3 + .........+ a2nxn - s2 = b2
..............................................................................
am1x1 + am2x2 + am3x3 + .........+ amnxn - sm = bm
x1, x2,....., xn≥ 0
s1, s2,....., sm≥ 0

An initial basic feasible solution is obtained by setting x1 = x2 =........ = xn = 0

-s1 = b1 or s1 = -b1
-s2 = b2 or s2 = -b2
..............................
-sm = bm or sm = -bm

which is not feasible because it violates the non-negativity stipulation, (i.e., s1 ≥ 0). Therefore,
we need artificial variables.

After introducing artificial variables, the set of constraints can be written as

a11x1 + a12x2 + a13x3 + .........+ a1nxn - s1 + A1 = b1

a21x1 + a22x2 + a23x3 + .........+ a2nxn - s2 + A2 = b2
..............................................................................
am1x1 + am2x2 + am3x3 + .........+ amnxn - sm + Am = bm

13
mywbut.com

x1, x2,....., xn≥ 0

s1, s2,....., sm≥ 0
A1, A2,....., Am≥ 0

Now, an initial basic feasible solution can be obtained by setting all the decision and surplus
variables to zero. Thus, an initial basic feasible solution to LPP is
A1 = b1 , A2 = b2, ....., Am = bm

Now to obtain an optimal solution, we must drive out the artificial variables. Following are the
two methods to solve linear programming problems in such cases.

• Two Phase method

• M method

Don't worry if you are not able to understand the above

mathematical formulation. In the following sections, we
illustrate the concept with the help of some "well-behaved"
examples.

Two Phase Method

In this method, the whole procedure of solving a linear programming problem involving artificial
variables is divided into two phases. In phase I, we form a new objective function by assigning
zero to every original variable (including slack and surplus variables) and -1 to each of the
artificial variables. Then we try to eliminate the artificial varibles from the basis. The solution at
the end of phase I serves as a basic feasible solution for phase II. In phase II, the original
objective function is introduced and the usual simplex algorithm is used to find an optimal
solution.

Example 1

Minimize z = -3x1 + x2 - 2x3

subject to

14
mywbut.com

x1 + 3x2 + x3 ≤ 5
2x1 – x2 + x3 ≥ 2
4x1 + 3x2 - 2x3 = 5

x1, x2, x3 ≥ 0

Solution.

If the objective function is in minimization

form, then convert it into maximization form.

Changing the sense of the optimization

Any linear minimization problem can be viewed as an equivalent linear maximization problem,
and vice versa. Specifically:

Minimize cjxj = Maximize (- cj)xj

If z is the optimal value of the left-hand expression, then -z is the optimal value of the right-hand
expression.

Maximize z = 3x1 – x2 + 2x3

subject to

x1 + 3x2 + x3 ≤ 5
2x1 – x2 + x3 ≥ 2
4x1 + 3x2 - 2x3 = 5

x1, x2, x3 ≥ 0

Converting inequalities to equalities

x1 + 3x2 + x3 + x4 = 5
2x1 – x2 + x3 – x5 = 2
4x1 + 3x2 - 2x3 = 5

x1, x2, x3, x4, x5 ≥ 0

Where:
x4 is a slack variable
x5 is a surplus variable

15
mywbut.com

The surplus variable x5 represents the extra

units.

Now, if we let x1, x2 and x3 equal to zero in the initial solution, we will have x4 = 5 and x5 = -2,
which is not possible because a surplus variable cannot be negative. Therefore, we need
artificial variables.

x1 + 3x2 + x3 + x4 = 5
2x1 – x2 + x3 – x5 + A1 = 2
4x1 + 3x2 - 2x3 + A2 = 5

x1, x2, x3, x4, x5, A1, A2 ≥ 0

Where A1 and A2 are artificial variables.

Phase 1

In this phase, we remove the artificial variables and find an initial feasible solution of the
original problem. Now the objective function can be expressed as

Maximize 0x1 + 0x2 + 0x3 + 0x4 + 0x5 + (–A1) + (–A2)

subject to

x1 + 3x2 + x3 + x4 = 5
2x1 – x2 + x3 – x5 + A1 = 2
4x1 + 3x2 - 2x3 + A2 = 5

x1, x2, x3, x4, x5, A1, A2 ≥ 0

Initial basic feasible solution

The intial basic feasible solution is obtained by setting

x1 = x2 = x3 = x5 = 0

Then we shall have A1 = 2 , A2 = 5, x4 = 5

Table 1
cj 0 0 0 0 0 -1 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 A1 A2
B b (= XB)

0 x4 1 3 1 1 0 0 0 5

16
mywbut.com

-1 A1 2 -1 1 0 -1 1 0 2

-1 A2 4 3 -2 0 0 0 1 5

zj–cj -6 -2 1 0 1 0 0

Key column = x1 column

Minimum (5/1, 2/2, 5/4) = 1
Key row = A1 row
Pivot element = 2
A1 departs and x1 enters.

Table 2
cj 0 0 0 0 0 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 A2
B b (= XB)

0 x4 0 7/2 1/2 1 1/2 0 4

0 x1 1 -1/2 1/2 0 -1/2 0 1

-1 A2 0 5 -4 0 2 1 1

zj-cj 0 -5 4 0 -2 0

A2 departs and x2 enters.

Here, Phase 1 terminates because both the artificial variables have been removed from the
basis.

Phase 2

The basic feasible solution at the end of Phase 1 computation is used as the initial basic feasible
solution of the problem. The original objective function is introduced in Phase 2 computation
and the usual simplex procedure is used to solve the problem.

Table 3
cj 3 -1 2 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

0 x4 0 0 33/10 1 -9/10 33/10

17
mywbut.com

3 x1 1 0 1/10 0 -3/10 11/10

-1 x2 0 1 -4/5 0 2/5 1/5

zj-cj 0 0 -9/10 0 -13/10

Table 4
cj 3 -1 2 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

0 x4 0 9/4 3/2 1 0 15/4

3 x1 1 3/4 -1/2 0 0 5/4

0 x5 0 5/2 -2 0 1 1/2

zj-cj 0 13/4 -7/2 0 0

Don't be impatient. The next table is the last table.

Table 5
cj 3 -1 2 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

2 x3 0 3/2 1 2/3 0 5/2

3 x1 1 3/2 0 1/3 0 5/2

0 x5 0 11/2 0 4/3 1 11/2

zj-cj 0 17/2 0 7/3 0

18
mywbut.com

An optimal policy is x1 = 5/2, x2 = 0, x3 = 5/2. The associated optimal value of the objective
function is z = 3 X (5/2) – 0 + 2 X (5/2) = 25/2.

Two Phase Method

Example 2

Maximize z = 12x1 + 15x2 + 9x3

subject to

8x1 + 16x2 + 12x3 ≤ 250

4x1 + 8x2 + 10x3 ≥ 80
7x1 + 9x2 + 8x3 = 105

x1, x2, x3 ≥ 0

Solution.

Introducing slack, surplus & artificial variables

8x1 + 16x2 + 12x3 + x4 = 250

4x1 + 8x2 + 10x3 – x5 + A1 = 80
7x1 + 9x2 + 8x3 + A2 = 105

Where:
x4 is a slack variable.
x5 is a surplus variable.
A1& A2 are artificial variables.

Phase 1

Maximize 0x1 + 0x2 + 0x3 + 0x4 + 0x5 + (–A1) + (–A2)

subject to

8x1 + 16x2 + 12x3 + x4 = 250

4x1 + 8x2 + 10x3 – x5 + A1 = 80
7x1 + 9x2 + 8x3 + A2 = 105

x1, x2, x3, x4, x5, A1, A2 ≥ 0

Equating x1, x2, x3, x5 to zero.

19
mywbut.com

Initial basic feasible solution

x4 = 250, A1= 80 , A2 = 105

Table 1
cj 0 0 0 0 0 -1 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 A1 A2
B b (= XB)

0 x4 8 16 12 1 0 0 0 250

-1 A1 4 8 10 0 -1 1 0 80

-1 A2 7 9 8 0 0 0 1 105

zj–cj -11 -17 -18 0 1 0 0

Table 2
cj 0 0 0 0 0 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 A2
B b (= XB)

0 x4 16/5 32/5 0 1 6/5 0 154

0 x3 2/5 4/5 1 0 -1/10 0 8

-1 A2 19/5 13/5 0 0 4/5 1 41

zj-cj -19/5 -13/5 0 0 -4/5 0

Here, Phase 1 terminates because both the artificial variables have been removed from the
basis.

Phase 2

Table 3
cj 12 15 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

20
mywbut.com

0 x4 0 80/19 0 1 10/19 2270/19

9 x3 0 10/19 1 0 -7/38 70/19

12 x1 1 13/19 0 0 4/19 205/19

zj-cj 0 -39/19 0 0 33/38

Table 4
cj 12 15 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

0 x4 0 0 -8 1 2 90

15 x2 0 1 19/10 0 -7/20 7

12 x1 1 0 -13/10 0 9/20 6

zj-cj 0 0 39/10 0 3/20

The optimal solution is:

x1 = 6, x2 = 7, x3 = 0
z = 12 X 6 + 15 X 7 + 9 X 0 = 177.

Well, after going through this section you definitely deserve

some food. Before moving to the next section, you must take a break because you really
deserve it.

21
mywbut.com

The Big M Method

It is a modified version of the simplex method, in which we assign a very large value (M) to each
of the artificial variables. We illustrate this method with the help of following examples.

Example 1

Maximize z = x1 + 5x2

subject to

3x1 + 4x2 ≤ 6
x1 + 3x2 ≥ 2

x1, x2 ≥ 0

Solution.

Converting inequalities to equalities

By introducing surplus variables, slack variables and artificial variables, the standard form of
LPP becomes

Maximize x1 + 5x2 + 0x3 + 0x4 – MA1

subject to

3x1 + 4x2 + x3 = 6
x1 + 3x2 – x4 + A1 = 2

x1 ≥ 0, x2 ≥ 0, x3 ≥ 0, x4 ≥ 0, A1 ≥ 0

Where:
x3 is a slack variable
x4 is a surplus variable.
A1 is an artificial variable.

22
mywbut.com

Initial basic feasible solution

x1 = x2 = x4 = 0
A1 = 2, x3 = 6

Table 1
cj 1 5 0 0 –M

Basic variables Solution values

cB x1 x2 x3 x4 A1
B b (= XB)

0 x3 3 4 1 0 0 6

–M A1 1 3 0 –1 1 2

zj–cj –M–1 –3M–5 0 M 0

Working

Calculating values for index row (zj – cj)

z1 – c1 = 0 X 3 + (–M) X 1 – 1 = –M – 1
z2 – c2 = 0 X 4 + (–M) X 3 – 5 = –3M – 5
z3 – c3 = 0 X 1 + (–M) X 0 – 0 = 0
z4 – c4 = 0 X 0 + (–M) X (–1) – 0 = M
z5 – c5 = 0 X 0 + (–M) X 1 – (–M) = 0

As M is a large positive number, the coefficient of M in the zj–cj row would decide the
incoming basic variable.
As -3M < -M, x2 becomes a basic variable in the next iteration.
Key column = x2 column.
Minimum (6/4, 2/3) = 2/3
Key row = A1 row
Pivot element = 3.
A1 departs and x2 enters.

Note that in the iteration just completed, artificial variable A1 was eliminated from the
basis. The new solution is shown in the following table.

23
mywbut.com

Table 2
cj 1 5 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (= XB)

0 x3 5/3 0 1 4/3 10/3

5 x2 1/3 1 0 –1/3 2/3

zj–cj 2/3 0 0 –5/3

Table 3
cj 1 5 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (= XB)

0 x4 5/4 0 3/4 1 5/2

5 x2 3/4 1 1/4 0 3/2

zj–cj 11/4 0 5/4 0

The optimal solution is:

x1 = 0, x2 = 3/2
z = 0 + 5 X 3/2 =15/2

Example 2

Maximize z = -4x1 - 2x2

subject to

3x1 + x2 ≥ 27
x1 + x2 ≥ 21
x1 + 2x2 ≥ 30

x1, x2 ≥ 0

24
mywbut.com

Solution.

By introducing surplus and artificial variables, the standard form of LPP becomes

Maximize z = -4x1 - 2x2 + 0x3 + 0x4 + 0x5 – MA1 – MA2 – MA3

3x1 + x2 - x3 + A1 = 27
x1 + x2 - x4 + A2 = 21
x1 + 2x2 - x5 + A3 = 30

x1 ,x2, x3, x4, x5, A1, A2 ≥ 0

Where:
x4 and x5 are surplus variables
A1 and A2 are artificial variables.

Table 1
cj -4 -2 0 0 0 –M –M –M

Basic variables Solution values

cB x1 x2 x3 x4 x5 A1 A2 A3
B b (= XB)

–M A1 3 1 -1 0 0 1 0 0 27

-M A2 1 1 0 -1 0 0 1 0 21

–M A3 1 2 0 0 -1 0 0 1 30

zj–cj -5M + 4 -4M + 2 M M M 0 0 0

As -5M < -4M, x1 becomes a basic variable in the next iteration

Key column = x1 column.
Minimum (27/3, 21/1, 30/1) = 9
Key row = A1 row
Pivot element = 3.
A1 departs and x1 enters.

Table 2
cj -4 -2 0 0 0 –M –M

Basic variables Solution values

cB x1 x2 x3 x4 x5 A 2 A 3
B b (= XB)

–4 x1 1 1/3 -1/3 0 0 0 0 9

25
mywbut.com

-M A2 0 2/3 1/3 -1 0 1 0 12

–M A3 0 5/3 1/3 0 -1 0 1 21

zj–cj 0 -7M/3 + 2/3 -2M/3 - 4/3 M M 0 0

Table 3
cj -4 -2 0 0 0 –M

Basic variables Solution values

cB x1 x2 x3 x4 x5 A2
B b (= XB)

–4 x1 1 0 -2/5 0 1/5 0 24/5

–M A2 0 0 1/5 -1 2/5 1 18/5

-2 x2 0 1 1/5 0 -3/5 0 63/5

zj–cj 0 0 -M/5 + 6/5 M -2M/5 + 2/5 0

Table 4
cj -4 -2 0 0 0

Basic variables Solution values

cB x1 x2 x3 x4 x5
B b (= XB)

–4 x1 1 0 -1/2 1/2 0 3

0 x5 0 0 1/2 -5/2 1 9

-2 x2 0 1 1/2 -3/2 0 18

zj–cj 0 0 1 1 0

The optimal solution is

x1 = 3, x2 = 18
z = -4 X 3 - 2 X 18 = -48

26
mywbut.com

Some Special Cases

• Unrestricted Variables
• Unbounded Solution
• No Feasible Solution
• Multiple Optimum Solutions
• Degeneracy

1. Unrestricted (unconstrained) Variables

Sometimes decision variables are unrestricted in sign (positive, negative or zero). In all such
cases, the decision variables can be expressed as the difference between two non-negative
variables. For example, if x1 is unrestricted in sign, then

Put x1 = x1' - x1''

Example

Maximize z = 2x1 + 3x2

subject to

-x1 + 2x2 ≤ 4
x1 + x2 ≤ 6
x1 + 3x2 ≤ 9

x1, x2 are unrestricted in sign

Solution.

Since x1 and x2 are unrestricted in sign, we can replace them by non-negative variables x1' , x1'',
x2' , x2'' .

Put x1 = x1' - x1''

x2 = x2' - x2''

The given problem can be written as

Max. z = 2(x1' - x1'') + 3(x2' - x2'')

27
mywbut.com

subject to

-(x1' - x1'') + 2(x2' - x2'') ≤ 4

(x1' - x1'') + (x2' - x2'') ≤ 6
(x1' - x1'') + 3(x2' - x2'') ≤ 9

Introducing slack variables

Max. z = 2x1' - 2x1'' + 3x2' - 3x2''

subject to

-x1' + x1'' + 2x2' - 2x2'' + x3 = 4

x1' - x1'' + x2' - x2'' + x4 = 6
x1' - x1'' + 3x2' - 3x2'' + x5 = 9

Where x3, x4 and x5 are slack variables

Table 1
cj 2 -2 3 -3 0 0 0

Basic variables Solution values

cB x1' x1'' x2' x2'' x3 x4 x5
B b (=XB)

0 x3 -1 1 2 -2 1 0 0 4

0 x4 1 -1 1 -1 0 1 0 6

0 x5 1 -1 3 -3 0 0 1 9

zj-cj -2 2 -3 3 0 0 0

Key column = x2' column.

Minimum (4/2, 6/1, 9/3) = 2
Key row = x3 row.
Pivot element = 2
x3 departs and x2' enters.

Table 2
cj 2 -2 3 -3 0 0 0

Basic variables Solution values

cB x1' x1'' x2' x2'' x3 x4 x5
B b (=XB)

28
mywbut.com

3 x2' -1/2 1/2 1 -1 1/2 0 0 2

0 x4 3/2 -3/2 0 0 -1/2 1 0 4

0 x5 5/2 -5/2 0 0 -3/2 0 1 3

zj-cj -7/2 7/2 0 0 3/2 0 0

Table 3
cj 2 -2 3 -3 0 0 0

Basic variables Solution values

cB x1' x1'' x2' x2'' x3 x4 x5
B b (=XB)

3 x2' 0 0 1 -1 1/5 0 1/5 13/5

0 x4 0 0 0 0 2/5 1 -3/5 11/5

2 x1' 1 -1 0 0 -3/5 0 2/5 6/5

zj-cj 0 0 0 0 -3/5 0 7/5

Table 4
cj 2 -2 3 -3 0 0 0

Basic variables Solution values

cB x1' x1'' x2' x2'' x3 x4 x5
B b (=XB)

3 x2' 0 0 1 -1 0 -1/2 1/2 3/2

0 x3 0 0 0 0 1 5/2 -3/2 11/2

2 x1' 1 -1 0 0 0 3/2 -1/2 9/2

zj-cj 0 0 0 0 0 3/2 1/2

The optimal solution is:

x1' = 9/2, x1'' = 0, x2' = 3/2, x2'' = 0.

Solution of the original problem is:

x1 = x1' - x1'' = 9/2 - 0 = 9/2

29
mywbut.com

x2 = x2' - x2'' = 3/2 - 0 = 3/2

z = 2 X 9/2 + 3 X 3/2 = 27/2.

2. Unbounded Solution

If in course of simplex computation zj - cj < 0, but minimum positive value is ≤ 0 then the
problem has an unbounded solution.

Example

Maximize 5x1 + 4x2

subject to

x1 ≤ 7
x1 - x2 ≤ 8

x1, x2 ≥ 0.

Solution.

Converting inequalities to equalities

x1 + x3 = 7
x1 - x2 + x4 = 8
x1, x2, x3, x4 ≥ 0.

Where x3 and x4 are slack variables.

Table 1
cj 5 4 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

0 x3 1 0 1 0 7

0 x4 1 -1 0 1 8

zj-cj -5 -4 0 0

30
mywbut.com

Table 2
cj 5 4 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

5 x1 1 0 1 0 7

0 x4 0 -1 -1 1 1

zj-cj 0 -4 5 0

Since minimum positive value is infinity, it is not possible to proceed with the simplex
computation any further. This is the criterion for unbounded solution.

3. No Feasible Solution
If in course of simplex computation, one or more artificial variables remain in the basis at positive level
at the end of phase 1 computation, the problem has no feasible solution. For example, let us consider
the following problem.

Example

Maximise -200x1 - 300x2

subject to

2x1 + 3x2 ≥ 1200

x1 + x2 ≤ 400
2x1 + 3/2x2 ≥ 900

x1, x2 ≥ 0

Solution.

After introducing slack, surplus and artificial variables the problem can be presented as

Maximise -200x1 - 300x2

subject to

2x1 + 3x2 – x3 + A1 = 1200

x1 + x2 + x4 = 400
2x1 + 3/2x2 – x5 + A2 = 900

31
mywbut.com

Where:
A1 and A2 are artificial variables.
x4 is a slack variable.
x3 and x5 are surplus variables.

The problem is solved by two phase method.

Phase 1

Maximise – A1 – A2

subject to

2x1 + 3x2 – x3 + A1 = 1200

x1 + x2 + x4 = 400
2x1 + 3/2x2 – x5 + A2 = 900

x1, x2, x3, x4, x5, A1, A2 ≥ 0

Table 1
cj 0 0 0 0 0 -1 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 A1 A2
B b (=XB)

-1 A1 2 3 -1 0 0 1 0 1200

0 x4 1 1 0 1 0 0 0 400

-1 A2 2 3/2 0 0 -1 0 1 900

zj-cj -4 -9/2 1 0 1 0 0

Table 2
cj 0 0 0 0 0 -1

cB Basic variables x1 x2 x3 x4 x5 A2 Solution values

B b (=XB)

0 x2 2/3 1 -1/3 0 0 0 400

0 x4 1/3 0 1/3 1 0 0 0

-1 A2 1 0 1/2 0 -1 1 300

32
mywbut.com

zj-cj -1 0 -1/2 0 1 0

Table 3
cj 0 0 0 0 0 -1

Basic variables Solution values

cB x1 x2 x3 x4 x5 x6
B b (=XB)

0 x2 0 1 -1 -2 0 0 400

0 x1 1 0 1 3 0 0 0

-1 A2 0 0 -1/2 -3 -1 1 300

zj-cj 0 0 1/2 3 1 0

In the above table, the optimum solution still contains the artificial variable A2 in the basis at
positive level, this indicates that the problem has no feasible solution.

4. Multiple Optimal Solutions

The optimal solution may not be unique, if the non basic variables have a zero coefficient in the
index row (zj-cj). This implies that bringing the non basic variable into the basis will neither
increase nor decrease the value of the objective function. Thus, the problem has multiple
optimal solutions (Alternative optimal solutions).

Example

Maximize 2000x1 + 3000x2

subject to

6x1 + 9x2 ≤ 100

2x1 + x2 ≤ 20

x1, x2 ≥ 0

Solution.

After introducing slack variables, the corresponding equations are

33
mywbut.com

6x1 + 9x2 + x3 = 100

2x1 + x2 + x4 = 20
x1, x2, x3, x4 ≥ 0

Where x3 and x4 are slack variables.

Table 1
cj 2000 3000 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

0 x3 6 9 1 0 100

0 x4 2 1 0 1 20

zj-cj -2000 -3000 0 0

Table 2
cj 2000 3000 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

3000 x2 2/3 1 1/9 0 100/9

0 x4 4/3 0 -1/9 1 80/9

zj-cj 0 0 3000/9 0

Since zj-cj ≥ 0 for all variables, x1 = 0, x2 = 100/9 is an optimum solution of the problem. The
maximum value of the objective function is 100000/3. However, the zj-cj value corresponding to
the non basic variable x1 is zero. This indicates that there is more than one optimal solution of the
problem. Thus, by entering x1 into the basis we may obtain another alternative optimal solution.

cj 2000 3000 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

3000 x2 0 1 1/6 -1/2 20/3

2000 x1 1 0 -1/12 3/4 20/3

34
mywbut.com

zj-cj 0 0 1000/3 0

However, the value of the objective function will not be improved, although the present solution
x1 = 0, x2 = 100/9 is shifted to x1 = 20/3 and x2 = 20/3.

5. Degeneracy

In some cases, there may be ambiguity in selecting the variable that should be introduced into the
basis, i.e., there is a tie between the replacement ratio of two variables. To resolve degeneracy,
we select one of them arbitrarily.

Consider the following example.

Example

Maximize 3x1 + 9x2

subject to

x1 + 4x2 ≤ 8
x1 + 2x2 ≤ 4

x1, x2 ≥ 0

Solution.

After introducing slack variables, the corresponding equations are:

x1 + 4x2 + x3 = 8
x1 + 2x2 + x4 = 4
x1, x2, x3, x4 ≥ 0

Where x3 and x4 are slack variables.

Table 1
cj 3 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

0 x3 1 4 1 0 8

35
mywbut.com

0 x4 1 2 0 1 4

zj-cj -3 -9 0 0

Minimum positive value (8/4, 4/2) = (2, 2)

There is a tie between the two values. So you are at liberty to break the tie arbitrarily.

In the following material, we will consider both the cases.

Case I

x4 departs & x2 enters.

Table 2
cj 3 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

0 x3 -1 0 1 -2 0

9 x2 1/2 1 0 1/2 2

zj-cj 3/2 0 0 9/2

x1 = 0, x2 = 2, z = 18

Case II

x3 departs & x2 enters.

Table 2
cj 3 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

9 x2 1/4 1 1/4 0 2

0 x4 1/2 0 -1/2 1 0

zj-cj -3/4 0 9/4 0

36
mywbut.com

x4 departs & x1 enters.

Table 3
cj 3 9 0 0

Basic variables Solution values

cB x1 x2 x3 x4
B b (=XB)

9 x2 0 1 1/2 -1/2 2

3 x1 1 0 -1 2 0

zj-cj 0 0 3/2 3/2

x1 = 0, x2 = 2, z = 18

37