Linear Programming

Sensitivity Analysis
 Optimum solution summary

• Decision variable Optimum value Recommendation

• x1 3 Produce 3 tons of exterior paint

daily
• x2 3/2 Produce 1.5 tons of interior
paint daily
• z 21 Daily profit is $ 21000

• Optimum values of the variables and objective value are:

amount of exterior paint, x1 is 3 tons
amount of interior paint, x2 is 1.5 tons
the associated profit is $21000.
 Optimum solution summary

• It also provides information about the slack (-) and surplus (+)
variables associated with the constraint.

• Output shows that slacks of first two constraints are zero, which
means the raw materials M1 and M2 are consumed completely.

• Slack of the third constraint equals to 2.5 tons, which means that
the constraint is over satisfied (because interior production is less
than exterior production).

• Last slack associated with the fourth constraint shows that the
production of interior paint is half a ton lower than the maximum
limit specified by the maximum market demand.
• In LP problem

• Model parameters (objective coefficients and constraint

coefficients) are assume to be static values

• However, sensitivity analysis study the effect of making

changes in model parameters on a given optimum LP
solution

• Two types of sensitivity analysis we will discuss here

• 1. Changes in the objective coefficients

• 2. Changes in the right hand side of the constraints

• 1. Change in objective function coefficients

• Objective function in a two variables LP problem can be written

as Max or Min z = c1 x1 + c2 x2

• Changes in coefficient c1 and c2 will changes the slop of z that

leads to changing the optimum solution to a different corner
point of the solution space

• However there is a range of variation for both c1 and c2 that will

keep the current optimum unchanged. This information we find
from the sensitivity analysis.

• Determine the range of optimality for the ratio (c1/c2) or (c2/c1)

that will keep the optimum solution of a given model unchanged.
• In Reddy Mikks Model, optimum solution at corner point C provide
maximum value of z = 5x1 + 4x2

• If we change the coefficient of the objective function, then the

solution at C will remain optimum so long as the slope of z lies
between slope of the two line intersect at C

• 6x1 + 4x2 = 24 (raw material M1)

• 1x1 + 2x2 = 6 (raw material M2)

4 C2 2 1 C1 6
  , C1  0   , C2  0
6 C1 1 2 C2 4
• Here c1 ≠ 0 that is objective function line cannot be horizontal and
c2 ≠ 0 that is objective function line cannot be vertical.

• Therefore the optimality range in this model does not permit the
objective function to be horizontal and vertical.
• If objective function is coincide with x1 + 2x2 = 6 then alternative
optima can occurs any where on the line segment CD.

• Similarly if it coincides with 6x1 + 4x2 = 24, then all points on

the line segment BC provide alternative optima.

• Condition to determine, optimal range for one of the coefficients

assuming that the other coefficient remain same in z = 5x1 + 4x2.

• If c2 = 4, the associated optimal range for c1 is determined from

1 C1 6 1 C1 6 1 6
      4   C1  4   2  C1  6
2 C2 4 2 4 4 2 4

• If c1 = 5, the associated optimal range for c2 is determined from

4 C2 2 4 C2 2 4 2
      5   C 2  5   3.33  C2  10
6 C1 1 6 5 1 6 1
• 2. Unit worth of a resource

• In most LP models, the constraints are usually represent the uses of

limited resources, for such constraint the right hand side provides the
limit on the availability of the resource

• Sensitivity study the effect on optimal solution to changes in the

availability of the resources

• Unit worth of resources quantifies the rate of change in the optimum

value in the objective function as a result of making changes in the
availability of the resources.

• In Reddy Mikks Model first two constraints represent the limitations on

the usages of raw materials M1 and M2

• Let us determine the worth per unit for each resource

• Constraint for raw material M1

• Current optimum occurs at point C which is the intersection of the lines

associated with raw material M1 and M2.

• When availability of M1 changes (increases or decreases around the current

level 24 tons), the optimum solution point C will slide along the line segment
DG.

• Change in M1 outside the range of this segment will make the intersection
point C infeasible, so the end point D (2, 2) and G (6,0) delineated the
feasibility range for M1.

• Amount of M1 at D (2,2) is 6x1 + 4x2 = 6 x 2 + 4 x 2 = 20 tons.

• Similarly amount of M1 at G (6, 0) is 6 x 6 + 4 x 0 = 36 tons.

• Thus the range of feasibility for M1 is 20 ≤M1≤36 and it can increase by 12

units or decrease by 4 units
• As the amount of M1 changes within the range 20 ≤ 24 ≤ 36,
the value of z associated with the optimum point C changes at
a constant rate as the DG line is a linear segment

• z = 5x1 + 4x2 at D = (2, 2) is 5 x 2 + 4 x 2=18 ($1000) and at G

= (6, 0) is 5 x 6 + 4 x 0 =30 ($1000)
change in z from D to G
• Let y1 
change in M1 from D to G
30  18 3
• It follows that y1  36  20  4  0(1000$
.75 per ton of M1)

• i.e. 1 tons change in M1 in the range 20≤M1≤36 will change the

optimum value of z by $750.
• Constraint for raw material M2

• Range of feasibility for M2 is delineated by the end points B (4, 0) and H (8/3, 2)
and the point H is defined by intersection of lines ED and BC

• Amount of M2 at B (4, 0) is x1 + 2x2 = 4 + 2 x 0 = 4 tons.

• Similarly amount of M2 at H (8/3, 2) is 8/3 + 2 x 2 = 6.6 tons.
• The range of feasibility for M2 is 4 ≤M2≤ 6.6 and it can increase by 0.6 units or
decrease by 2 units

• z = 5x1 + 4x2 at B = (4, 0) is 5 x 4 + 4 x 0 = 20 ($1000) and at H = (8/3, 2) is 5 x

8/3 + 4 x 2 = 21.33 ($1000)

• i.e. ($1000 per ton of M1)

• So 1 tons change in M2 in the range 4≤M2≤6.66 will change the optimum value
of z by
21.33  20 1.33
y $500.
2    0.5
6.66  4 2.66
• Sensitivity Analysis of Computer Solution

• Changes in the coefficients of the objective function within the

specific limits will not change the value of x1 and x2 but will affect
the value of the objective function z.

• However, changes in the right-hand side within the specified limits

will alter the values of both the variables and the objective function.

• For example, the current optimum solution will remain unchanged

so long as the profit per ton, of exterior paint lies between $2000
and$6000.

• Feasibility of current solution is not affected so long as daily

availability of raw material M1 stays between 20 and 36 tons.
• Reduced cost of a variable

• An LP variable is regarded as an economic activity that consumes

(input) resources for the purpose of producing (output) profit.

• Reduced cost per unit of activity = (Cost of consumed resources per

unit of activity j) - (Profit per unit of activity j)

• If activity’s reduced cost per unit is positive, means the value of its
associated variable in the optimum solution should be zero, then the
cost of its consumed resources per unit is higher than its profit per unit
and the activity should not be undertaken.

• Alternatively, an activity that is economically attractive will have a

zero reduced cost in the optimum solution, signifying an equilibrium
point has been reached at which the output (unit profit) equals the
input (unit cost of the resources).
• Dual price of a constraint

• Dual price represents the unit worth of a resource, it gives the

contribution to the objective function resulting from a unit increase or
decrease in the availability of a resource.

• Dual prices for raw materials, M1 and M2 are 0.75 and 0.5, or $750 and
$500 per ton, respectively. These are valid for the respective ranges 20
≤ M1 ≤ 36 and 4 ≤ M2 ≤ 6.6 only.

• If availability of M1 increase from 24 to 28 tons, then optimum value

of objective function will increase by 750x(28 -24) = $3000

• Dual prices for 3rd and 4th constraints of Reddy Mikks model are zero
for the ranges -1.5 to ∞ and 1.5 to ∞, indicate that increases in
resources representing the market limits for production of interior and
exterior paints have no effect on optimum solution.
Simultaneous changes in all coefficients

• Sensitivity analysis information is based on the assumption

that only one coefficient changes; it is assumed that all other
coefficients will remain as stated in the original problem.

• Analysis of simultaneous changes is possible with the help of

the 100 percent rule.

• To observe the changes in the profit contributions of the

exterior and interior paint to $4 and $6, we see that the lower
limit for the objective function coefficient of exterior paint is
2; thus the allowable decrease is 3 = 5 - 2.
 Simultaneous changes in all objective coefficients
• In term of percentage changes the decrease of $1 in the objective
function coefficient (from 5 to 4) for the exterior paint is (1/3) (100) =
33.33% of the allowable decrease.

• Given the upper limit of 10 for interior paint, the allowable increase for
interior paint is 6 = 10 - 4.

• In terms of percentage change, the increase of $4 in the objective

function coefficient (from 4 to 8) for interior paint is (4/6) (100) =
66.66% of the allowable increase.

• Sum of percentage changes of the allowable increase (33.33%) and the

percentage change of the allowable decrease (66.66%) is 99.99% which
is less than 100% implies no change in optimal solution, only change in
profit contribution.
Simultaneous changes in all constraints coefficients

• To observe the changes by taking 3 tons of raw material M1 and

0.25 tons of raw material M2, we see that the allowable increase
for M1 is 36 – 24 = 12 and the allowable increase for M2 is 6.6 – 6
= 0.6.

• The percentage changes in 3 tons of raw material M1 are (3/12)

(100) = 25% of the allowable increase in the constraint’s right
hand side. Similarly, 0.25 tons of raw material M2 are (0.25/0.6)
(100) = 41.67% of allowable increase in constraint’s right hand
side.

• Sum of percentage changes does not exceed 100% therefore we

can conclude that the dual prices are applicable and that objective
function will bb (4) (3) + (8) (1.5) = 24
• Simultaneous changes in all objective function coefficients

• Changes in objective coefficients in the objective function of Reddy Mikks

model is given as z = (5 + d1) x1 + (4 + d2) x2

• Resulting conditions (associated with slacks s1and s2) given by

• 0.75 + 0.25 d1 – 0.125 d2 ≥ 0
• 0.50 - 0.50 d1 + 0.750 d2 ≥ 0
• are actually the same as the condition 1/2 ≤ c1/c2 ≤ 6/4

• If we set c1 = 5 + dl and c2 = 4 + d2, then only the optimum value of the

objective function will change, but the values of the variables will remain
unchanged

• If changes are such that the given conditions are not satisfied, a new
optimum must be recomputed using the simplex procedure.
• Simultaneous changes in right hand side of all constraints.

• Right-hand sides of constraints are assumed to be 24 + D1, 6+ D2, 1

+ D3 and 2 + D4.

• These Dl, D2, D3 and D4 provide the new solution values, which
must be nonnegative to satisfy feasibility condition.

• For example

• Suppose, D1 = 5, D2 = -1, D3 = 1, and D4 =2, then

• x1 = 3 + 0.25 x 5 - 0.5 x -1 = 4.75

• x2 = 1.5 - 0.125 x 5 + 0.75 x -1 = 0.125
• s1 = 2.5 + 0.375 x 5 -1.25 x -1 + 1 x 1 = 6.625
• s2 = 0.5 + 0.125 x 5- 0.75 x -1 + 1 x 2 = 3.875
• All values are nonnegative; hence the changes will yield new
feasible solution x1 = 4.75 tons, x2 = 0.125 ton, and z = 5 x
4.75 + 4 x 0.125 = 24.25 or $24,250

• Resulting value of z exceeds current value by ∆z = $24,250 -

$21000 = $3,250.

• Amount can be computed as ∆z = D1y1 + D2y2 + D3y3 + D4y4,

where y1, y2, y3 and y4 are dual prices of respective resources.

• Thus, ∆z = 5 x $750 + ( -1) x $500 + 1 x 0 + 2 x 0 = $3,250

• This type of computation is valid if only the changes Di, i = 1,

2, 3, 4 leads to a feasible solution.
• If changes result in a negative value of any of the variables,
then a new solution must be recomputed.

• Both graphical and computer solution gives the ranges (20,

36) and (4, 6.67) for resources 1and 2, respectively (these
ranges will be derived algebraically also).

• Thus, raw material, M1, can be increased from its present

level of 24 tons to a maximum of 36 tons with a
corresponding increase in profit of 12 X $750 = $9000.
• Similarly, the limit on raw material, M2, can be increased
from 6 tons to a maximum of 6.67 tons with a corresponding
increase in profit of .67 X $500 = $335.

• Similar interpretations can be given if the raw material levels

are decreased below the current levels within the indicated
ranges of applicability.

• The discussion does not imply that the given resources cannot
be changed to levels outside the given ranges. It simply says
that the worth per unit for each of resources 1 and 2 are
guaranteed only within the specified ranges.
 Post optimal or sensitivity analysis
• Sensitivity analysis is carried out after the (current) optimum
solution of an LP model is obtained.

• Objective is to determine whether changes in the model's coefficients

will leave the current solution unchanged and if not, how a new
optimum (assuming one exists) can be obtained efficiently.

• The changes in the model can result in one of four cases.

• Current (basic) solution remains unchanged.

• Current solution becomes infeasible.
• Current solution becomes non-optimal.
• Current solution becomes both non-optimal and infeasible.
• Changes Affecting Feasibility:

• Feasibility of the current optimum solution may be affected

only if the right hand side of the constraints is changed or a
new constraint is added to the model.

• In both cases, infeasibility occurs when one or more of the

current basic variables become negative.

• Another way of looking at the effect of changing the

availability of the resources (right-hand-side) is to determine
the range for which the current solution remains feasible.
• Addition of new constraints: It leads to one of two cases.

• New constraint is redundant, meaning that it is satisfied by the

current optimum solution.

• New constraint is violated, in which case the dual simplex

method must be used to (attempt to) recover feasibility.

• Notice that the addition of a non redundant constraint can only

worsen the current optimum value of the objective function.

• Changes Affecting Optimality: Current solution will cease to

be optimal only when we change the objective coefficients or
the unit resource usage.
• Changes in the objective coefficients cj

• Effect of making changes in cj on optimality entails re-computing zj - cj for

the non-basic variables only, but not the basic variables as they will always
equal zero regardless of changes made in cj.

• Computational procedure is summarized as:

– Compute the dual prices vector using new vector CB if it has been changed.
– Compute zj – cj = YPj – cj for all current Y  C B B1xj, two cases will result.
non-basic

• If optimality condition is satisfied, current solution will remain same but at

a new optimum value of the objective function.
– (If CB is unchanged, the optimum objective value will remain the same).

• If the optimality condition is not satisfied, we apply the (primal) simplex

method to recover optimality.
• Addition of a new activity.

• Addition of a new activity in an LP model is equivalent to

adding a new variable and it is desirable only if it is profitable
that is, if it improves the optimal value of the objective function.

• This condition can be checked by computing z j - cj = YPj – cj for

the new activity, where Y are the current optimal dual values
and Pj and cj represent the resource usages and the profit per unit
of the new activity.

• If computed zj - cj not satisfies the optimality condition, then the

new activity is not desirable. Otherwise, the new activity is
profitable and must be brought into the basic solution.
 Popeye Canning

• Popeye Canning is contracted to receive 60000 lb of ripe

tomatoes at 7 per cent per pound from which it produces both
canned tomato juice and tomato paste. The canned products
are packaged in 24 can cases. A can of juice requires 1 lb of
fresh tomatoes and a can of paste requires 1/3 lb only. The
company’s share of the market is limited to 2000 cases of
juice and 6000 cases of paste. The wholesale prices per cases
of juice and paste are $18 and $9 respectively. Develop an
optimum production program for Popeye. Determine the ratio
of the price per case of juice to the price per case of paste that
will allow Popeye to produce more cases of juice than of
paste.
 Popeye Canning

• Objective Function

• Maximize Z = 18 x1 + 9 x2
• Subject to
• 24 x1 + 8 x2 ≤ 60000
• x1 ≤ 2000
• x2 ≤ 6000
• x1, x2 ≥ 0
• x1 and x2 are the number of cases of juice and paste of the
company should produce.
 Popeye Canning
• (a) Determine the worth per unit of additional pound of tomato.
• (b) Is it worthwhile for the company to raise its market share for juice?
For paste?
• (c) If the company reduces the contracted amount of tomato to 50000 lb,
determine the new optimal solution.
• (d) If the profit per case of juice is reduced to $15 and the profit per case
of paste is increased to $10, will the current optimum solution change?
• (e) If the amount of purchased tomato is increased to 80000 lb and the
market share of paste is increased to 7000, determine the new optimum
solution.
• (f) If the amount of purchased tomato is decreased to 30000 lb and the
market share of paste is increased to 8000 cases would it be possible to
determine the new solution.