Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Operations Research tools are not from any one discipline. Operations Research
takes tools from different discipline such as mathematics, statistics, economics,
psychology, engineering etc. and combines these tools to make a new set of
tool/technique for decision making. Today, O.R. has become a professional
discipline which deals with the application of scientific methods for making
decision, and especially to the allocation of scarce resources. The main purpose
of O.R. is to provide a rational basis for decisions making in the absence of
complete information, because the systems composed of human, machine, and
procedures may not have complete information. The emphasis on analysis of
operations as a whole distinguishes the O.R. from other research and
engineering. O.R. is an interdisciplinary discipline which provided solutions to
problems of military operations during World War II. Today business
applications are primarily concerned with O.R. analysis for evaluating various
possible alternative course of actions. The business and industry benefited from
O.R. in the areas of inventory, reorder policies, optimum location and size of
warehouses, advertising policies, etc.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 1
1.2 Evolution of OR
After the war, many of these team members continued their research in the field
of management science and their knowledge of OR was utilized by the industry
to solve complex managerial problems faced by the industrial managers.
•1950: O.R. began to develop in industrial field in US
•1953: Operations Research Society of America was formed
•1957: International Federation of Operational Research Society
Operational Research has been studied in Health Care settings since 1952.
Norman Bailey published “Operational Research in Medicine” in the June issue
in Operational Research Quarterly.
(Bailey June 1952).
Two developments that occurred during the post-World War II period led to the
growth and use of management science in nonmilitary applications. First,
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 2
continued research resulted in numerous methodological developments.
Probably the most significant development was the discovery by George
Dantzig, in 1947, of the simplex method for solving linear programming
problems. At the same time these methodological developments were taking
place, digital computers prompted a virtual explosion in computing power.
Computers enabled practitioners to use the methodological advances to solve a
large variety of problems. The computer technology explosion continues, and
personal computers can now be used to solve problems larger than those solved
on mainframe computers in the 1900s.
1.3 Defining OR
The definitions stressed by various experts and Societies on the subject together
enable us to know what O.R. is, and what it does. Some of the commonly found
definitions are as follows:
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 3
5. Miller and Starr state, “O.R. is applied decision theory, which uses any
scientific, mathematical or logical means to attempt to cope with the problems
that confront the executive, when he tries to achieve a thorough-going
rationality in dealing with his decision problem”.
Game Theory: This is used for making decisions under conflicting situations
where there are one or more players/opponents. In this the motive of the
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 4
players are dichotomized. The success of one player tends to be at the cost of
other players and hence they are in conflict.
Queuing Theory: This is used in situations where a queue is formed (for example
customers waiting for service, aircrafts waiting for landing, jobs waiting for
processing in the computer system, etc). The objective here is minimizing the
cost of waiting without increasing the cost of servicing.
Non-linear Programming: This is used when the objective function and the
constraints are not linear in nature. Linear relationships may be applied to
approximate non-linear constraints but limited to some range, because
approximation becomes poorer as the range is extended. Thus, the non-linear
programming is used to determine the approximation in which a solution lies
and then the solution is obtained using linear methods.
Markov Process:
Markov process permits to predict changes over time about the behavior of a
system. This is used in decision making in situations where the various states fo
a system can be defined. The probability of transition from one state to another
state is known and depends on the current state and is independent of how we
have arrived at a subsequent state.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 5
Network Scheduling: This technique is used extensively to plan, schedule, and
monitor large projects (for example computer system installation, R & D design,
construction, maintenance, etc.). The aim of this technique is minimize trouble
spots (such as delays, interruption, production bottlenecks, etc.) by identifying
the critical factors. The different activities and their relationships of the entire
project are represented diagrammatically with the help of networks and arrows,
which is used for identifying critical activities and path. There are two main types
of technique in network scheduling, they are: Program Evaluation and Review
Technique (PERT), used when activities time is not known accurately/ only
probabilistic estimate of time is available. Critical Path Method (CPM), used
when activities time is known accurately.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 6
Accounting: Assigning audit teams effectively, Credit policy analysis, Cash flow
planning, developing standard costs, establishing costs for by products, Planning
of delinquent account strategy.
1 Operations Research in Health Care: Perspectives from an engineer, with examples from emergency medicine and cancer therapy Timothy Chan Steven
Brooks University of Toronto St. Michael’s Hospital, Clinical and Population Research Rounds St. Michael’s Hospital, October 20, 2011
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 7
• Adjusting nurse schedules in light of daily fluctuations in demand
• Operating theatres allocation, fitness plans optimisation
Macro “For the provider” - Resource allocation, Utilization, throughput
• Create clinic schedule that balances shared resource utilization like the
Upstream blood lab, Downstream chemo day care, Nursing, Rooms
• BASED ON STRUCTURE
o (1) Physical (Iconic or Analogue) (2) Symbolic(LP) (Verbal or
Mathematical)
• BASED ON PURPOSE AND NATURE
o (1)Descriptive (2)Predictive (3) Normative or Optimisation Models(LP)
• BASED ON CERTAINITY
o (1)Deterministic(LP) (2) Probabilistic or Stochastic Models
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 8
• BASED ON TIME REFERENCE
o Static(LP) (2) Dynamic Model
• BASED ON METHOD OF SOLUTION
o Analytical(LP) (2) Iterative or Heuristic (3) Simulation Models
Operations Research has certain limitations which are mostly related to the
model building, money and time factors problems involved in its application.
Some of them are as given below:
ii) Magnitude of Calculations (time and money): The aim of the O.R. is to find out
optimal solution taking into consideration all the factors. In this modern world
these factors are enormous and expressing them in quantitative model and
establishing relationships among these require voluminous calculations, which
can be handled only by machines(computers).
iii) Dynamic nature of a business environment: The basic data are subjected to
frequent changes, incorporating these changes into the operations research
models is very expensive. However, a fairly good solution at present may be
more desirable than a perfect operations research solution available in future or
after sometime.
iv) Non-quantifiable Factors: When all the factors related to a problem can be
quantifiable only then operations research provides solution otherwise not. The
non-quantifiable factors are not incorporated in O.R. models. Importantly O.R.
models do not take into account emotional factors or qualitative factors.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 9
1.9 Linear Programming
• The problem variables X and Y are called decision variables and they
represent the solution or the output decision.
• The profit/cost function that the manufacture wishes to
increase/decrease, represents the objective of making the decisions and
is called the objective function.
• The conditions matching the resource availability and resources
requirement are called constraints. These usually limit (or restrict) the
values the decision variable can take.
• Explicitly stated all decision variable should take non negative values. This
is true for all Linear Programming problems and is called non negativity
restriction.
• The problem that has been written down in algebraic form represents the
mathematical model of the given system and is called the Problem
Formulation.
• This abstract representation of reality is called a mathematical model.
• Optimal Solution A solution to the model that optimizes (maximizes or
minimizes) some measure of merit over all feasible solutions.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 11
3. An LP model usually includes a set of constraints known as non-negativity
constraints. These constraints ensure that the variables in the model can take
only nonnegative values (i.e., > 0). This is logical because negative values of
physical quantities are impossible; one simply cannot produce a negative
number of chairs or computers.
6. You will see the term inequality quite often when we discuss LP problems.
By inequality we mean that not all LP constraints need be of the form A + B = C.
This particular relationship, called an equation, implies that the sum of the term
A and term B exactly equals term C. In most LP problems, we see inequalities of
the form A + B < C or A + B > C. This implies that A plus B is less than or equal to
C. This concept provides a lot of flexibility in defining problem limitations.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 12
a particular resource, then making 10 units of that product uses 30 hours of the
resource.
3. Third assumption deals with additivity, meaning that the total of all
activities equals the sum of the individual activities. For example, if an objective
is to maximize profit which is equal to $8 per unit of the first product made plus
$3 per unit of the second product made, and if 1 unit of each product is actually
produced, the profit contributions of $8 and $3 must add up to produce a profit
sum of $11.
Maximize/Minimize Z = ∑ cjxj
Subject to
∑ aijxj i where i = 1,2,!!m
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 13
1.14 Guidelines to developing a correct LP model
1. Formulation
Translating a problem scenario from words to a
mathematical model
2. Solution
Solving the model to obtain the optimal solution
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 14
1.16 Method of solving LP problem, graphical method
There are many ways of solving a LP problem. They are:
Graphically methods
Algebraic method
Simplex method
Using Software:
Excel Solver
CoinMP (open source)
LINDO, LINGO
lpsolve, AMPL, MPL, CPLEX
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 15
Iso-profit line approach.
1. Graph the
constraints.
2. Identify the
feasible solution space.
3. Set the objective
function equal to some
amount that is divisible by
each of the objective
function coefficients and
one that falls in the
feasible region.
4. After identifying the optimal point, determine which two
constraints intersect there.
5. Substitute the values obtained in the previous step into the
objective function to determine the value of the objective function at the
optimum.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 16
1.18 Algebraic Method of solving LP problem
Certain terminologies:
Slack and surplus
The term slack is used for the amount of resource that is not used.
For less-than-or-equal to constraint.
Slack = (Amount of resource available) - (Amount of resource used)
The term surplus is used with greater-than-equal-to constraints to indicate the
amount by which the right-hand side of a constraint is exceeded.
For a greater-than-or-equal-to constraint.
Surplus = (Actual amount) – (Minimum amount)
This system has two equations, we can select any two of the four variables as
basic variables. The remaining two variables are then non-basic variables.
A solution found by setting the two non-basic variables equal to 0 and solving
for the two basic variables is a basic solution.
If a basic solution has no negative values, it is a basic feasible solution.
3T + 4C + S1 = 2400
2T + 1C + S2 = 1000
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 17
Four variable & 2 equation.
But with two equation we can solve only for two variables.
So we fix two of the variables with some arbitrary value and solve for
the remaining two.
Two variable that we want to fix out of 4 variables can be done in 4C2
was.
In general if there are n decision variables & m equations (constraints),
fixing can be done in nCm (n!/m!(n-m)!) ways. (In this case we can fix in 6
different ways)
Since we can fix the two variables with any arbitrary value we can in
fact have infinite solutions to start with.
Fixing the variable to 0, is a good option to start with (But why), we will
see that.
Variable that are fix to 0 are called non-basic variables, variable that we
solve for are called basic variables.
Solution that are obtained by fixing non-basic variable to 0, are called
basic solution
We have, in this case six solutions
Of these some are feasible & some are not. This, can be ascertained by
non-negativity constraint.
In this specific case we have 4 Basic feasible and two basic infeasible
solution (infeasible)
The solution which return maximum value for Z in the optimum
solution.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 18
1.19 Simplex Method of solving LP problem
The graphical method is useful only for problems involving two decision variables
and relatively few problem constraints. The algebraic method is useful only for
problems involving fewer decision variables and relatively few problem
constraints as it involves testing all the case for arriving at an optimal solution.
The number of cases to be tested being given by nCm (n!/m!(n-m)!) combinations
where n represents decision variables & m represents equations (constraints).
Hence it can be seen that as the variable and constraints increases, number of
cases to be tested increases many folds making it very difficult to solve the
problem using the algebraic method. Accordingly, for such cases we use a
method called the simplex method, which was developed by George B. DANTZIG
(1914-2005) in 1947 while on assignment with the U.S. Department of the air
force.
The first step of the simplex method requires that each inequality be converted
into an equation. ”Less than or equal to” inequalities are converted to equations
by introducing slack variables.
Suppose S1 carpentry hours and S2 painting hours are the slack variable
introduced which primarily represents unused hours in a week. The constraints
become;
4T + 3C + S1 = 2400
2T + 1C + S2 = 1000
As unused hours result in no profit, the slack variables can thus be included in
the objective function with zero coefficients:
Z = 7T + 5C + 0s1 + 0s2
Z - 7T – 5C – 0s1 – 0s2 = 0
The problem can now be considered as solving a system of 3 linear equations
involving the 5 variables T,C,S1,S2,Z in such a way that Z has the maximum value;
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 19
Now, the system of linear equations can be written in matrix form or as a 3x6
augmented matrix.
Note that T should enter into the solution mix because each unit of T (a table)
contributes a profit of Rs 7 compared with only Rs 5 for each unit of C (a chair)
Select the pivot row (determine which variable to replace in the solution mix).
Divide the last element in each row by the corresponding element in the pivot
column. The pivot row is the row with the smallest non-negative result. Which
implies that S2 Should be replaced by T in the solution mix.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 20
Now calculate the new values for the pivot row. Divide every number in the
row by the pivot number to obtain the new tableau, i.e divide the pivot row
with 2 to obtain the new tableau
Use row operations to make all numbers in the pivot column equal to 0 except
for the pivot number which remains as 1.
Making 500 tables results in the profit being increased by ₹3500; the value
changes from₹0 to ₹3500. Also T=500, C=0, S1=900 and S2=0. Now repeat the
steps until there are no negative numbers in the last row. Select the new pivot
column. C should enter into the solution mix. Select the new pivot row. S1 should
be replaced by C in the solution mix.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 21
Now calculate new values for the pivot row. Divide every number in the row by
the pivot number 5/2 (in this case).
Next use row operations to make all numbers in the pivot column equal to 0
except for the pivot number.
As the last row contains no negative numbers, this solution gives the maximum
value of Z.
This simplex tableau represents the optimal solution to the LP problem. The
optimal solution (maximum profit to be made) is to make 320 tables and 360
chairs and profit of Z= $4040. Note that the solution matches with other two
processes.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 22
1.20 Special cases of LP
Four special cases and difficulties arise at times when using the graphical
approach to solving LP problems:
Infeasibility
Unboundedness
Redundancy and
Alternate optimal solutions.
No feasible Solution: - lack of a feasible
solution region can occur if constraints
conflict with one another.
As a further graphic illustration of this, let
us consider the following three constraints:
X1 + 2X2 < 6
2X1 + X2 < 8
X1 > 7
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 23
Alternate Optimal Solutions:- Multiple
optimal solutions are possible in LP problems.
Max Profit = $3X1 + $2X2
Subject to
6X1 + 4X2 < 24
X1 < 3
X1,X2 > 0
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 24
Per Tie
(Yards)
All Silk $6.70 6,000 7,000 0.125 100% Silk
All $3.55 10,000 14,000 0.08 100%
Polyester Polyester
Poly- $4.31 13,000 16,000 0.10 50%
cotton Polyester/50%
Blend 1 Cotton
Poly- $4.81 6,000 8,500 0.10 30%
Cotton Polyester/70%
Blend 2 Cotton
(b) Example 6: Win Big Gambling Club promotes gambling junkets from a
large mid western city to Casinos in The Bahamas. The club has budgeted
up to $8,000 per week for local advertising. The money is to be allocated
among four promotional media: TV spots, newspaper ads, and two types
of radio advertisements. Win Big’s goal is to reach the largest possible
high potential audience through the various media. Table 3 presents the
number of potential gamblers reached by making use of an advertisement
in each of the four media. It also provides the cost per advertisement
placed and the maximum number of ads that can be purchased per week.
Win Big’s contractual arrangements require that at least five radio spots
be placed each week. To ensure a broad-scoped promotional campaign,
management also insists that no more than $1,800 be spent on radio
advertising every week.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 25
(c) Example 7: Hong Kong Bank of Commerce and Industry is a busy bank that
has requirements for between 10 and 18 tellers, depending on the time
of day. The afternoon time, from noon to 3 P.M. is usually heaviest. Table
4 indicates the workers needed at various hours that the bank is open.
The bank now employs 12 full-time tellers but also has several people
available on its roster of part-time employees. A part-time employee must
put in exactly 4 hours per day but can start anytime between 9 A.M. and
1 P.M. Part-timers are a fairly inexpensive labor pool because no
retirement or lunch benefits are provided for them. Full-timers, on the
other hand, work from 9 A.M. to 5 P.M. but are allowed 1 hour for lunch.
(Half of the full-timers eat at 11 A.M. and the other half at noon.) Each
full-timer thus provides 35 hours per week of productive labor time.
The bank’s corporate policy limits part-time hours to a maximum of 50%
of the day’s total requirement. Part-timers earn $7 per hour (or $28 per
day) on average, and full timers earn $90 per day in salary and benefits,
on average. The bank would like to set a schedule that would minimize its
total personnel costs. It is willing to release one or more of its full-time
tellers if it is cost-effective to do so.
(d) Example 8: The Whole Food Nutrition Center uses three different types
of bulk grains to blend a natural breakfast cereal that it sells by the
pound. The store advertises that each 2-ounce serving of the cereal,
when taken with ½ cup of whole milk, meets an average adult’s
minimum daily requirement for protein, riboflavin, phosphorus, and
magnesium. The cost of each bulk grain and the protein, riboflavin,
phosphorus, and magnesium units per pound of each are shown in Table
5.
The minimum adult daily requirement (called the U.S. Recommended
Daily Allowance [USRDA]) for protein is 3 units, for riboflavin is 2 units,
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 26
for phosphorus is 1 unit, and for magnesium is 0.425 units. Whole Food
wants to select the blend of grains that will meet the USRDA at a
minimum cost.
(f) Example 10: The international City Trust (ICT) invests in short-term trade
credits, corporate bonds, gold stocks, and construction loans. To
encourage a diversified portfolio, the board of directors has placed limits
on the amount that can be committed to any one type of investment. ICT
has $5 million available for immediate investment and wishes to do two
things: (1) maximize the return on the investments made over the next six
months and (2) satisfy the diversification requirements as set by the board
of directors.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 27
The specific of investment possibilities are as follows:
INVESTMENT INTEREST RETURN MAXIMUM INVESTMENT
($ MILLIONS)
Trade credit 7% 1.0
Corporate bonds 11% 2.5
Gold stocks 19% 1.5
Construction loans 15% 1.8
In addition, the board specifies that at least 55% of the funds invested
must be in gold stocks and construction loans, and that no less than 15%
be invested on trade credit.
1.22 References
1.(B) Balakrishnan, N., Render, B. and Stair Jr., R.M.: Managerial decision
modelling with spreadsheets, Pearson education.
2.Anderson, Sweeney, Williams and Martin: An Introduction to Management
Science – Quantitative Approach to Decision Making, 13th Edition.
3.Barry R., Stair R.M. Jr., Hanna M.E., Hale T.S., Badri T.N., Quantitative Analysis
for Management, 12th Edition.
4. Youtube video on LP by Prof G Srinivasan, Dept of management studies, IIT
Madras (open courseware)- https://nptel.ac.in/courses/112106134/1
5. Case study available online - Hillier-Lieberman Textbook, Introduction to Operations
Research, INFORMS, Interfaces.
6. Class notes, assignments, homework and handouts. (Note PPT would not be
shared)
7. http://www.orsi.in/index.php - Operations research society of India
http://ifors.org/ - International Federation of Operational Research Societies
http://www.apors.asia/ - Asia Pacific Operational Research Societies
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 28
1.23 Journals
Management Science International Journal of Production Research
Operations Research Omega
European Journal of Operational Research Journal of the American Statistical Association
Interfaces International Journal of Production Economics
Naval Research Logistics INFORMS Journal on Computing
Manufacturing and Service Operations SIAM Review
Management (INFORMS journal)
Mathematics of Operations Research Computers and Operations Research
Decision Sciences Decision Support Systems
Operations Research Letters Networks
Production and Operations Management Computers and Industrial Engineering
IIE Transactions Mathematical and Computer Modelling
Journal of Operations Management Journal of the Operational Research Society
Annals of Operations Research Mathematical Programming
Journal of the Operational Research Society Transportation Science
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 29
1.24 Sensitivity analysis
Objective function coefficient (OFC). The OFCs are the coefficients of the
decision variables in the objective function (such as the $7 and $5 for T and C,
respectively, in Flair’s model as shown above). In many business-oriented LP
models, OFCs typically represent unit profits or costs, and they are measured in
monetary units such as dollars, Euros, and rupees. Are OFCs likely to have any
uncertainty in their values? Clearly the answer is yes, because in many real-
world situations, selling and cost prices are seldom likely to be static or fixed.
For this reason, we will study how the optimal solution may be affected by
changes in OFC values.
Constraint coefficient: The constraint coefficients are the coefficients for the
decision variables in a model’s constraints (such as the 3 and 4 in the carpentry
constraint in Flair’s model). In many problems, these represent design
parameter with regard to the decision variable. For example, needing 3 hrs of
carpentry per table is a product design issue that has probably been specified by
design engineers. Although we could think of specific situation where these
types of input parameters could also be subject to uncertainty in their values,
such changes are less likely as compared to OFC and RHS values. For this reason,
we do not usually study the impact of changes in constraint coefficient values
on the optimal solution and we will study only the impact of changes in an
objective function coefficient and that due to changes in a constraint’s Right-
hand-side value.
To summarize, only two things can occur due to a change in an OFC: (1) if the
current optimal corner point continues to remain optimal, the decision variable
values do not change, even though the objective function value may change;
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 31
and (2) if the current corner point is no
longer optimal, the values of the
decision variables change, as does the
objective function value.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 32
larger, as shown by the region marked R1 in figure. As a consequence of this
increase in the size of the feasible region, the locations of corner points 4 and 5
has shifted to new locations 4A and 5A respectively. However, the level profit
line approach indicates that the intersection of the carpentry and painting
constraints is still the optimal solution. That is, the “same” corner point (in the
sense that the same two constraints intersect at this point) is still optimal. But it
now has a new location and hence, there are new values for T, C and profit. The
values at corner point 4A can be computed to be T = 560 and C = 180, for a profit
of $4,820. This implies that if Flair is able to obtain an additional 300 hours of
painting time, it can increase profit by $780 for 300 additional hours of painting
time translates to a profit increase of $2.60 per additional hour of painting time.
Observe that the profit increases by $2.60 per each additional hour of painting
time gained and it decreases by the same $2.60 per each hour of painting time
lost from the current level. This value, known as the shadow price, is an
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 33
important concept in LP models. The shadow price of a constraint can be
defined as the change in the optimal objective function value for a one unit
increase in the RHS value of that constraint. In the case of painting time, the
shadow price is $2.60; this implies that each hour of painting time (with respect
to the current availability) affects Flair’s profit by $2.60. Because painting time
is a binding < constraint, each additional hour obtained increases profit by $2.60,
while each hour lost decreases profit by $2.60.
Is this shadow price of $2.60 valid for any level of change in the painting time
availability? That is, for example, can Flair keep obtaining additional painting
time and expect its profits to keep increasing endlessly by $2.60 for each hour
obtained? Clearly, this cannot be true and we illustrate the reason for this in the
following section.
Consider, for
example, what
happens if Flair can
increase the painting
time availability even
further, to 1,700
hours. Under this
scenario as shown in
figure, the feasible
region increases by
the region marked
R3. However, due to
the presence of the
non-negativity constraint > 0, the corner point defined by the intersection of the
carpentry and painting constraints is no longer feasible. In fact, the painting
constraint has now become a redundant constraint. Obviously, in such a case,
the optimal solution has shifted to a new corner point. The level profit lines
approach indicates that the optimal solution is now at corner point 5C (T=800,
C=0, profit=$5,600). Note that this translates to a profit increase of $1,560
(=$5,600-$4,040) for 700 additional hours, or $2.23 per hour, which is different
from the shadow price of $2.60. That is, the shadow price of $2.60 is not valid
for an increase of 700 hours in the painting time availability.
What happens if the painting time availability is decreased all the way down to
700 hours? Here again, as shown in figure, the intersection point of the
carpentry and painting constraint is no longer even feasible. The carpentry
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 34
constraint is now
redundant, and the
optimal solution has
switched to a new
corner point given by
corner point 3A
(T=125, C=450,
profit=$3,125). This
translates to a profit
decrease of $915
(=$4,040 - $3,125) for
a decrease of 300
hours or $3.05 per hour, which is again different from the shadow price of $2.60.
That is, the shadow price of $2.60 is not valid for a decrease of 300 hours in the
painting time availability.
The preceding discussion based on figures shows that a certain range of change
in the RHS value of a binding constraint, the “same” corner point will continue
to remain optimal.
That is, the constraints that are currently binding at the optimal solution will
continue to remain the binding constraints. The location of this optimal corner
point will, however, change, depending on the change in the RHS value. In fact,
it turns out that as long as this corner point exists in the feasible region, it will
continue to remain optimal. In Flair’s case, this means that the corner point
where the carpentry and painting constraints intersect will remain the optimal
solution as long as it exists in the feasible region. Also, the shadow price of $2.60
measures the impact on profit for a unit change in painting time availability as
long as this corner points continues to exist in the feasible region. Once this RHS
value changes to such an extent that the current binding constraints no longer
intersect in the feasible region, the shadow price of $2.60 is no longer valid and
changes to a different value. It is algebraically possible to use the graphical
solution to determine the RHS range within which the current optimal corner
point continues to exist, albeit at a new location. We will, however, use the
information provided in the solver sensitivity report to further discuss this issue
in a subsequent section.
A similar analysis can be conducted with the RHS value for the other binding
constraint in Flair’s example – the carpentry constraint.
Now consider the case where the marketing department wants to make this
production limit even more restrictive (i.e. the 450 limit is decreased). As long
as we are permitted to make at least 360 chairs, figure indicates that a corner
point 4 is feasible and still optimal. That is, as long as the change in the RHS value
for the chairs constraint is within the slack of 90 units, the current optimal corner
point continues to exist and remains optimal. However, if the chairs production
limit is reduced below 360, corner point 4 is no longer feasible, and a new corner
point becomes optimal. A similar analysis can be conducted with the other
nonbinding constraint in the model (i.e. T > 100).
The preceding discussion illustrates that for nonbinding constraints, the
allowable change limit on one side is infinity. On the opposite side the allowable
change limit equals the slack (or surplus).
The Reduced Cost values (in an Excel solver sensitivity analysis) show the
difference between the marginal contribution of a decision variable to the
objective function value (profit, in Flair’s example) and the marginal worth of
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 36
the resources it would consume if produced. A property of LP models is that if a
variable has a nonzero value at optimality, its marginal contribution to the
objective function value will equal the marginal worth of the resources it
consumes. For instance, in the Flair furniture problem each table produced uses
3 hours of carpentry time, uses 2 hours of painting time. Counts 1 unit towards
the 100-unit minimum tables’ requirement, and counts 0 units toward the 450-
maximum chairs limit.
Based on the preceding discussion of the shadow price, the marginal worth of
these resources can be calculated as
= 3 X shadow price of carpentry constraint +
2 X shadow price of painting constraint +
1 X shadow price of minimum tables required constraint +
0 X shadow price of maximum chairs allowed constraint
= 3 X $0.6 + 2 X $2.6 + 1 X $0 + 0 X $0
= $7
Note that this is equal to the profit contribution per table. The same
calculation will hold for chairs also. The profit contribution per chair is $5, and
the marginal worth of the resources it consumes is calculated as
= 4 X shadow price of carpentry constraint +
1 X shadow price of painting constraint +
0 X shadow price of minimum tables required constraint +
1 X shadow price of maximum chairs allowed constraint
= 4 X $0.6 + 1 X $2.6 + 0 X $0 + 1 X $0
= $5
Therefore the reduced cost of table and chair are zero.
Large number of
software are
available to carry
out sensitivity
analysis. For
discussion, we
will use the excel
solver, which is
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 37
widely available. Let us take the example of Flair furniture. The screen shot show
how the problem is set up in a excel sheet.
Thereafter, the problem is solved and the sensitivity report is obtained as
indicated in the snapshot below.
The sensitivity report snap shot of the Flair furniture problem is shown below.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 38
in the Allowable Increase and Allowable Decrease columns of the Excel
Sensitivity Report. In the Flair furniture problem, for tables, their profit
contribution per table can range anywhere from a low of $3.75 (=$7 - $3.25) to
a high of $10 (= $7 + $3), and the current production plan (T = 320, C = 360) will
continue to remain optimal. The total profit will, of course, change, depending
on the actual profit contribution per table. For example, if the profit contribution
is $6 per table, the total profit is $3,720 (= $6 X 320 + $5 X 360). However, for
any profit contribution below $3.75 or over $10 per table, a different corner
point solution will become optimal.
For chairs, the profit contribution per chair can range anywhere from a low of
$3.50 (=$5 - $1.50) to a high of $9.33 (=$5 + $4.33), and the current production
plan will continue to remain optimal. Here again, the total profit contribution is
$8 per chair, the total profit is $5,120(=$7 X 320 + $8 X 360). Any profit
contribution below $3.50 or over $9.33 per chair will result in a different corner
point solution being optimal.
The answer report, snapshot shown below, provides details about the method
used for the solution, no of iteration, time taken and optimal value of objective
function, resources used, available slack(surplus) and the maximized
profit/minimize cost
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 39
Problem on sensitivity analysis
Burn-off, a manufacture of diet drinks, is planning to introduce a miracle drink
that will magically burn away fat. The drink is a bit expensive, but Burn-Off
guarantees that a person using this diet plan will lose up to 50 pounds in just
three weeks. The drink is made up of four “mystery” ingredients (which we will
call ingredients A, B, C and D). The plan calls for a person to consume at least
three 12-ounce doses per day (i.e. at least 36 ounces per day) but no more than
40 ounces per day. Each of the four ingredients contains different levels of three
chemical compounds (which we will call chemicals X, Y, and Z). Health
regulations mandate that the dosage consumed per day should contain
minimum prescribed levels of chemical X and Y and should not exceed maximum
prescribed levels for the third chemical, Z. The composition of the four
ingredients in terms of the chemical compounds (units per ounce) is shown in
table along with the unit cost prices of the ingredients. Burn-Off wants to find
the optimal way to mix the ingredients to create the drink, at minimum cost per
daily dose.
Data for Burn-Off Diet Drink
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 40
Based on these reports, answer the questions given below.
Q1 What is the impact on cost if Burn-off insists on using 1 ounce of ingredient
B to make the drinks?
A1. The reduced cost indicates that each ounce of ingredient B used to make the
drink will cause the total cost per daily dosage to increase by $0.069 (-$0.07).
The new cost will be $13.06 + $0.07 = $13.13.
Q2 There is some uncertainty in the cost of ingredient C. How sensitive is the
current optional solution to this cost?
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 41
A2. The current cost of ingredient C is $0.60 per ounce. The range for the cost
coefficient of this ingredient shows an allowable increase of $1.50 and an
allowable decrease of $0.073 in order for the current corner point solution to
remain optimal. The cost per ounce of ingredient C could therefore fluctuate
between $0.527 (=$0.60 - $0.073) and $2.10 (=$0.60 + $1.50) without affecting
the current optimal mix.
The total cost will, however, change depending on the actual unit cost of
ingredient C. For example, if the cost of ingredient C increases to $1.00 per
ounce, the new total cost will be
= $13.06 + ($0.40 extra per ounce X 4.125 ounces of C)
= $13.06 + $1.65
= $14.71
Q3. What do the shadow prices for chemical X and chemical Z imply in this
problem?
A3. The shadow price for chemical X is $0.088. Because the constraint for
chemical X is a > constraint, an increase by 1 unit in the RHS (from 280 to 281)
makes the problem solution even more restrictive. That is, the feasible region
becomes smaller. The optimal objective function value could, therefore, worsen.
The shadow price indicates that for each additional unit of chemical X required
to be present in the drink, the overall cost will increase by $0.088. This value is
valid for an increase for an increase of up to 41 units and a decrease by 11 units
in the requirement for chemical X.
In contrast, the constraint for chemical Z is a < constraint. An increase in the RHS
of the constraint (from 1,050 to 1,051) will cause the feasible region to become
bigger. Hence, the optimal objective function value could possibly improve. The
negative value of the shadow price for this constraint indicates that each unit
increase in the maximum limit allowed for chemical Z will cause the total cost to
decrease by $0.024. This value is valid for an increase of up to 47.143 units.
Likewise, the total cost will increase by $0.024 for each unit decrease within the
maximum limit allowed for chemical Z. This is valid for a decrease of up to 346
units in the maximum limit for chemical Z.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 42
Q4. Burn-off can decrease the minimum requirements for chemical X by 5 units
(from 280 to 275), provided that the maximum limit allowed for chemical Z is
reduced to 1,000 units (i.e, reduced by 50 units), Is this trade-off cost-effective
for Burn-off to implement?
A4. Because we are dealing with simultaneous changes in RHS values, we first
verify whether the 100% rule is satisfied. To do so, we take the ratio of each
proposed change to its maximum allowable change. The calculation is
Sum of ratios = (5/11) + (50/346) = 0.599 < 1
Because the sum does not exceed 1, we can use the shadow price information
in the sensitive report above. The reduction of 5 units in the requirement for
chemical X will cause the feasible region to increase in size. The total cost will
therefore improve (i.e. go down) by $0.44 (= 5 units, at a shadow price of $0.088
per unit).
In contrast, the reduction of 50 units in the maximum allowable limit for
chemical Z makes the feasible region shrink in size. The total cost will therefore
be adversely affected (i.e. go up) by $1.20 (= 50 units at a shadow price of $0.024
per unit).
The net impact of this trade-off is therefore an increase in total cost $0.76 (=
$1.20 - $0.44). The new cost will be $13.82. Clearly, this trade-off not cost-
effective from Burn-offs perspective and should be rejected.
Class notes, Session 8, 9, 10 and 11 – QMM, Only for internal class room purpose during the Jul-Nov 2018 Session Page 43