Scribd
Upload
42 views

You Are Only Confined by The Walls You Built Yourself

The document discusses optimization methods in management science. It covers three key topics: 1) problems with no feasible solutions or unbounded variables, 2) post-optimality tests to analyze optimal solutions, and 3) sensitivity analysis to determine which parameters most impact the optimal solution. Special cases like negative variable bounds are addressed. Post-optimality tests like reoptimization and shadow prices provide insights. Sensitivity analysis identifies parameters that change the optimal solution if altered.

Uploaded by

Sohaib Arif
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
42 views

You Are Only Confined by The Walls You Built Yourself

The document discusses optimization methods in management science. It covers three key topics: 1) problems with no feasible solutions or unbounded variables, 2) post-optimality tests to analyze optimal solutions, and 3) sensitivity analysis to determine which parameters most impact the optimal solution. Special cases like negative variable bounds are addressed. Post-optimality tests like reoptimization and shadow prices provide insights. Sensitivity analysis identifies parameters that change the optimal solution if altered.

Uploaded by

Sohaib Arif
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 17

You are Only Confined by

Walls
You Built Yourself

Optimization Methods in Management Science


Muhammad Tayyab SDSB-LUMS
Key Learning Objective is to understand the:
1. theory of special cases with
1. no feasible solution to the problem
2. variables allowed to have negative bounds
3. variables which are not bounded
2. post-optimality tests for enhanced insights into the
model

Optimization Methods in Management Science


Muhammad Tayyab SDSB-LUMS
No Feasible Solutions
It may happen that you think you have reached an optimal solution
because “Optimality test” has passed
BUT
the solution is actually Not Feasible.
Optimality Test has
Passed

Solution is not feasible, as “artificial variable > 0”, i.e. penalty M is in


action
Muhammad Tayyab SDSB-LUMS
Variables Allowed to be Negative
In the Simplex Method, xj > 0 Ɐj is considered as a by default scenario.
What if any variable is allowed to have a negative value..???
In Wynder Glass Inc., let us consider that product 1 is already in production with a
rate of 10 and now the decision is to find an optimal product mix in which production
rate of product 1 is allowed to drop until 0 i.e. 0≤x1≤10.
Therefore, the constraint in the model will be x1 ≥ -10.
This constraint must be converted into the non-negativity constraint prior to the
application of the Simplex Method.

Muhammad Tayyab SDSB-LUMS


Variables Allowed to be Negative
In the Simplex Method, xj > 0 Ɐj is considered as a by default scenario.
What if any variable is allowed to have a negative value..???

In this case, x1 had an appropriate lower bound (even if it was a negative lower bound).

BUT, what if it had no lower bound, i.e. x1 is unbounded towards negative side…???
Muhammad Tayyab SDSB-LUMS
Variables Allowed to be Negative
What if any variable is allowed to have a negative value BUT with no
defined lower bound..???

If a variable xj has no lower bound constraint, then it can be considered


as a difference between two positive numbers xj+ and xj- as below

with the following properties.


+ive part of the decision variable xj

-ive part of the decision variable xj

Muhammad Tayyab SDSB-LUMS


Variables Allowed to be Negative
What if any variable is allowed to have a negative value BUT with no
defined lower bound..???

No constraint is restricting positive


or negative bound for x1

Muhammad Tayyab SDSB-LUMS


Post Optimality Analysis
Post optimality analysis is very important as it provides various insights into
the problem in hand

Theoretical testing

Muhammad Tayyab SDSB-LUMS


Post Optimality Analysis
1. Reoptimization

Many variants of a basic version of the problem can be studied through


Reoptimization.
Optimal solution for version 1 of the problem = BF solution for version 2
• e.g. Two-phase Method
• Final tableau of basic version is used as initial tableau of revised version after
modifications in objective function
• Advantage is that; the optimal solution to the revised version of the model can be
found in less iterations, as it lies near to the optimal solution of basic version rather
than initial basic solution.

Muhammad Tayyab SDSB-LUMS


Unlearn first, to learn fur

Optimization Methods in Management Science


Muhammad Tayyab SDSB-LUMS
Post Optimality Analysis
2. Shadow prices

• Managers decide resources (bi) for each activity by analyzing various factors
• Managers want to utilize the allocated resources at their maximum
• Still some resources remain underutilized (expressed by non-binding constraints)
• the basic variable associated to that resource constraint remains > 0 at optimal solution
• Whereas, some resources are fully utilized (expressed by the binding constraints)
• the basic variable associated to this resource constraint = 0 at optimal solution
• Shadow price tries to look beyond the walls of these fully utilized resources to answer
the question that what will happen if the upper limit of availability of this resource is
increased slightly?
Muhammad Tayyab SDSB-LUMS
Post Optimality Analysis
2. Shadow prices

By definition:
The shadow price (yi*) for resource (i) is the marginal benefit of this resource: i.e. the
rate of improvement in Z by slightly increasing the availability (bi) of this resource.
As a metaphor; re-consider the initial Wynder Glass Co. problem
what will happen if we
increase any of the 3
= b1, (availability of resource 1) resources in this
= b2, (availability of resource 2) problem? Will the
profit (Z) improve?
= b3, (availability of resource 3)

Muhammad Tayyab SDSB-LUMS


Post Optimality Analysis
Slack variables
Maximum profit = Z = $36,000

Non-Binding constraint

n
Binding

io
ut
constraints
l
so
al
im
pt
O

Non-Binding
constraint

Binding constraints

Muhammad Tayyab SDSB-LUMS


Post Optimality Analysis
2. Shadow prices

where b1 = 4 (availability of resource 1)

where b2 = 12 (availability of resource 2)

where b3 = 18 (availability of resource 3)


For instance, when b2 = 13, then the maximum profit will be;
Z = 36 + y2* i.e., Z = 36 +3/2, i.e., Z = 37.5 i.e., $37,500
Therefore, The increase in Z is $1500 per unit increase in resource 2.
This fact will remain true until we reach the boundary of constraint 3, after which, any
increase in resource 2 will be useless as it will not improve Z, unless resource 3 is
increased.
Muhammad Tayyab SDSB-LUMS
Post Optimality Analysis
2. Shadow prices

Information about “shadow prices” is when b2 = 13


valuable for managers;
• to allocate resources to different
activities within the organization
when b2 = 12
• to acquire extra resources from
outside the organization at regular Resources with +ive shadow prices
cost for any activity are called “scarce goods”
• to decide maximum premium to be and
surplus resources (shadow price = 0)
paid to acquire any extra resource
are called “free goods”
from outside

Muhammad Tayyab SDSB-LUMS


Post Optimality Analysis
Sensitive parameters are those “that cannot be changed
3. Sensitivity Analysis
without changing the optimal solution”
• if yi* > 0, then the optimal solution changes as bi changes, therefore, bi is a sensitive
Sensitivity
parameter (it has a shadow price) of
• if yi* = 0, then the optimal solution does not change as bi changes, therefore, bi is not available
resources
a sensitive parameter (it does not have a shadow price)
• constraint 1 is not binding, so a small change in coefficient (also called Sensitivity
technological coefficient) of variable in this constraint will not affect the optimal of
resource
solution and thus this coefficient is not sensitive, whereas constraint 2 and 3 are utilization
other way round
• the parameters in objective function (coefficients of decision variables) can be Sensitivity
of
changed up to the extent where the slope of objective function stays between the objective
slopes of binding constraints without affecting the optimal solution [as (2,6) in this parameters
case], hence these parameters are sensitive
Muhammad Tayyab SDSB-LUMS
Post Optimality Analysis
It studies “how the optimal solution changes, when many variables are
4. Parametric Linear varied simultaneously over some range”,
Programming (it can be thought of a two-way sensitivity analysis)

• Investigation of the trade-off among parametric values


• Trade-off among costs and benefits
• For instance, should we cut down resources from activity 2 and allocate those resources to
activity 1? (In constraints)
• if yes, then how much? Eg. Manpower, electricity power
• Should we bear more tax in order to operate a new electricity generator which will cost us
less than the rental power? (In objective function)
• One of the factors is incorporated in the objective function, and the second one is put in the
constraint, thus, the factor in objective function attains some value at the expense of the
second one which is in the constraint

Muhammad Tayyab SDSB-LUMS

You might also like