MATH302: OPERATIONS RESEARCH

INTRODUCTION & LP MODELS

MR. MAXWELL SUOBOGBIREE

0205022309
0242564707
0273265888
ACCRA INSTITUTE OF TECHNOLOGY
(AIT)
 Introduction
2

Definition of Operations Research (OR)

Operations Research is a scientific approach to
decision-making, which seeks to determine how best
to design & operate a system, under conditions
requiring the allocation of scarce resources
 Algorithms of OR
3

The algorithms of OR act as tools for effective systematic

problem solving & decision making

How can managers decide well over time & in the

presence of uncertainties, what to produce & charge?

How can they assign & schedule tasks & manage

resources?
How can banks reduce queues? Such decisions are

made every day

 History of OR
4

 Initiated during the World War II (1940s) by British (& US) Military

when scientists were asked to analyze military problems.

 Applying mathematical methods to military was Operational

Research. Today, It's at times called Management Science or

Decision Science
 After the war, extensive applications in: Engineering,
Manufacturing, Transport, Health, Business, Marketing, Finance,
Telecom, Serv. etc
 Operations Research Technique-Brief Overview
5

LP maximizes or minimizes a linear function

subject to certain constraints that limit the
degree to which the goal can be pursued.

 Waiting-line or queuing models helps managers

understand & make better decisions concerning
the operation of systems involving waiting lines.

 DEA is an LP approach for assessing relative

efficiency & productivity of homogenous firms
using multiple inputs to produce multiple outputs.
 Award-winning Applications (1)
6

OR group at American Airlines use forecasting &

optimization models to offer discount & full fares in

revenue mgt. Make $1b incremental revenue/annum.

 China optimally selected & scheduled massive

projects for meeting the country's future energy

needs (1995). Saved $425 million
 Award-winning Applications (2)
7

South African defense force optimally redesigned the size

& shape of the defense force & its weapons systems

(1997). Saved $1.1 billion

A bank used DEA to improve branch productivity and

profits. It identified over $6million of annual expense

savings not identifiable with traditional financial &
operating ratio analysis in its 33-branch system
 Impact of Operations Research on everyday
living
8

 If you schedule a holiday in London & use skyscanner to

book your fights, an algorithm developed by operations

researchers will search among millions of options to find
the cheapest fare. Another algorithm will schedule the fight
crews & aircraft used by the airline.
 If you rent a car in London, the price you pay for the car is

determined by a mathematical model that seeks to

maximize revenue for the car rental firm.
 Lecture 2: Linear Programming

LP is a widely used mathematical modeling technique

designed to help managers to plan & decide given

limited resources.

LP solves a variety of problems related with

management, scheduling, media selection, financial

planning, capital budgeting, transportation etc
 Brief History of Linear Programming
10

LP is a model made up of linear relationships representing

a firm's decisions, given an objective & resource

constraint(s).
LP was first formalized & applied to problems in economics

in 1939 by L. Kantorovich. Hidden until first conceived by

George B Dantzig around 1947. Koopman coined “Linear
Programming” in 19483, but Dantzig published the Simplex
Method 4 in 1949.
 Linear Programming Formulation
11

The components of an LP are:

decision variables (the problem solution)

 parameters (cost, profit, price per unit, etc.)

objective (maximise or minimise)

constraints (restrictions: time, labour, material,

money) We shall explain LP using a simple example.
First, let's list the properties of LPPs.
 Basic components of the model
12

Decision Variables

It is the unknown to determined from the

solution of a model (what does the model
seek to determine). It is one of the specific
decisions made by a decision maker (DM).
 Basic components of the model
13

Objective Function
It is the end result (goal) desired to be
achieved by the system. A common objective is
to maximize profit or minimize cost. It is
expressed as a mathematical function of the
system decision variables
 Basic components of the model
14

Constraints
These are the limitations imposed on the variables
to satisfy the restriction of the modeled system. They
must be expressed as mathematical functions of the
system decision variables (D.V).
 Basic components of the model
15

Parameters
Numerical coefficients and constants used in the
objective function and constraints.
 Properties of Linear Programs
16

 One objective function:- Only one objective function

 One or more constraints :- Constraints can be one or more

 Alternative courses of action :- A decision is required

 Proportionality:-means the slope of a constraint or objective function line is

constant
 Additivity:-the objective function terms and the constraint terms be additive

 Divisibility:-the solution values (of the decision variables) cannot be restricted

to integer values; the decision variables can take on any fractional value

 Certainty:-All model parameters are assumed to be known with certainty.

 Example 1
17

 Fan Milk Ghana Limited produces FanYogo & FanChoco.

Some Ghanaians love these. I don't know about you. The 2
key resources used by the company are milk & skilled
labour. Given the limited resources, the company desires to
know how many sachets of FanYogo & FanChoco to
produce each day in order to maximize profit. There are 40
hours of labour & 120 gallons of milk available each day for
production
 FanYogo requires 1 hr of labour time & FanChoco requires
2 hrs. Also, FanYogo requires 4 gallons of milk whilst Fan
Choco requires 3 gallons. The profit from a sachet of
FanYogo is GH¢40 per unit, & from FanChoco is GH¢50 per
unit. Formulate a linear programming model for this
product mix problem & solve using graphical analysis
 Example 1 Solution table
18

(hrs) (gl)
Product Labour Milk Profit(GHC/unit)
FanYogo 1 4 40
FanChoco 2 3 50
Resource 40 120
available
 Eg 1 solution: Formulation steps
19

Define the decision variables - what to

produce?

 Construct the objective function - maximise

profit or maximise revenue or minimise cost?

Maximize profit Z Formulate the constraints -

available resources?

Milk and labour

 Eg1 solution: DV & Z
20

What are the decision variables?

x1 = number of FanYogo sachet to
produce
x2 = number of FanChoco sachet to
produce .
What is the objective function?
Z , i.e. Total Revenue from each x1 & x2
price of 1 unit of FanYogo, x1, is GH¢40
Max Z = 40x1 + 50x2
The company should find x1 & x2 so Zmax
 Eg 1 solution: Constraints
21

Resources constraints? See table below

How many resources? Are they limited?
How many hours of labour & how many
gallons of milk do you need to produce
FanYogo?
What about FanChoco? But what is the upper
limit of resources available?
What is the total labour used by FMGL
 Recall Example 1: Solution table
22

(hrs) (gl)
Product Labour Milk Profit(GH¢/unit)
FanYogo 1 4 40
FanChoco 2 3 50
Resources 40 120
Example 1: Solution constraints Mathematically
23

Labour constraint:
1x1 + 2x2 ≤ 40hr.

Milk constraint:
4x1 + 3x2 ≤ 120gl.
 Example 1: Solution contraints ≤ vs =

24

Why ≤ sign and not =?

Because the 40 hours of labour & the 120 gl
of milk are the maximum limitations that can
be used, not the amount that must be used.
Can the company use less? more?
 Example 1 solution
25

Final restriction: nonnegativity constraints x1

≥ 0 , x2 ≥ 0 .
meaning?

number of FanYogo and FanChoco produced

must be 0 or positive.
 Example 1 Complete Model
26

Max : Z = GH¢40x1 + GH¢50x2......(1)

s.t.
1x1 + 2x2 ≤ 40..................(2)
4x1 + 3x2 ≤ 120................(3)
 x1,x2 ≥ 0..........................(4)
Next, is to solve for the x's using (a) graphs
(b) simplex method (math/algebra) (c) PC
(based on SM)
 Feasible solution
27

A feasible solution (convex set) does not violate any

of the above constraints: for e.g. If x1 = 5 & x2 =

10 then Z=?
 Z∗ = 40x1 + 50x2 = 40(5) + 50(10) = GH¢700

Labor constraint check: 1(5) + 2(10) = 25 < 40

hours (binding constraint, LHS=RHS)

Milk constraint check: 4(5) + 3(10) = 70 < 120

gallons (binding constraint, LHS=RHS)

 Infeasible solution
28

An infeasible solution violates at least one of

the constraints
If x1 = 10 FanYogo & x2 = 20 FanChoco then

Z=?
Z∗ = 40x1 + 50x2 = GH¢1400

Labor constraint check: 1(10) + 2(20) = 50 >

40 hours (not binding, LHS ≠RHS)

 Graphical Solution of LP Model
29

Graphical solution is limited to LP models

containing only 2 decision variables (can be used

with 3 variables but with great difficulty).
Graphical solution provides visualization of how a

solution for LPP is obtained.

Treat each constraint as an equation & plot it

(remember we need only 2 points to draw a

straight line);
Identifying Optimal Solution: Step (1)
30

Identify each constraint area. Find area

common to all the constraints (feasible

solution area, FSA). Add the objective
function (OF) line for an arbitrarily selected
level of profit to the graph;
For maximisation problems, we need the

point where a level curve or OF line is

farthest from the origin & still touches the
Identifying Optimal Solution: Step (2)
31

For minimisation, we need the point where a

level curve is closest to the origin & still

touches the FSA.
This is an intersection point of any 2

constraints. Understand the problem

thoroughly.
 Solve these constraints simultaneously to

obtain the optimal solution.

Graph of non-negativity constraint
32
 Graphing
33

To plot the constraints, treat each as

equations (=) .
1st=labour constraint : 1x1 + 2x2 = 40

when x1 = 0, x2 =?? = 20, (0,20)

when x2 = 0, x1 =?? = 40 (0,40) .

Do the same for the 2nd = milk constraint

Graph of Labour Constraint
34
Labour constraint area
35
Graph of Milk Constraint
36
Both Constraints, feasible solution
area
37
 Feasible Solution Area or feasible
region
38

 FSA is an area on the graph bounded by the constraint

equations. Any point on this polygon or in its interior is

in the feasible region. Some points in FSA lead to
FMGL's maximum profit.
 Now we must find those points. First, we include the

profit line. But since we don't know actual profit value

we assume arbitrary number, say, Z = 800, put that into
Z = 40x1 + 50x2 = GH¢800
 Then, find x1 =? if x2 = 0 & x2 =? if x1 = 0
Searching for optimal solution
39
OB, isoprofit line =GHc800=Z
40
Alternative Objective Function Solution Lines
41
 Feasible Solution Area
42

What we are doing is shifting the isoprofit

line (in max, or isocost line in a min) parrallel
outwards from the origin as far as possible
without losing the FSA.
Maximum profit is reached at a point where
the Z line is farthest from the origin & still
tangent/touching to FSA.
This point is B called optimal (best) solution
Identification of Optimal Solution Point
43
Optimal Solution Coordinates
44
 Optimal Solution Coordinates -
features
45

 Now we must pinpoint the values at the vertices of this FSA.

 B(x1,x2) above is indeed the optimal solution. The solution

point should be on the boundary of FSA & at one of the

corners of the boundary where the 2 constraint lines intersect
(extreme points)
 But your graph must be perfect to get the optimal values of x1

= 24 & x2 = 8. So why ain't A & C optimal solution points?

Extreme (Corner) Point Solutions

46
 Simultaneous Equation solutions
47

Solve the 2 constraints simultaneously &

optimal x1 & x2:

1x1 + 2x2 = 40..............(1)

4x1 + 3x2 = 120................(2)

Using substitution method,

 x1 = 40−2x2..............(3)

 (3) into (2)

 Simultaneous Equation solutions..
48

4(40−2x2) + 3x2 = 120

160−8x2 + 3x2 = 120

x2 = 8

x2 = 8 into (3) x1 = 40−2(8) = 24

 Simultaneous Equation: Finding Z
49

Solving the for profit yields:

 Z = GH¢40x1 + GH¢50x2

Z = GH¢40(24) + GH¢50(8)

Z = GH¢1,360

If FMGL produces 24 FanYogo and 8

FanChoco, it gets GH¢1,360

 Characteristics of the Solution
50

 Altering the objective function coefficients results in a new

solution.

 Hence the solution is sensitive to the values of the coefficients in

the OF

 Altering the constraint coefficients, the solution space and solution

points may change also.

 Evaluating the effects of changes in model parameters is called

sensitivity analysis
 Slack & Surplus Variables
51

Slack = unused resource = leftover = resource

available - resource used

A slack variable contributes nothing to the

objective function value, so yield no profit. So its

given 0 value in the OF
Changing the constraints from ≤ to = requires the

addition of a variable (slack) leading to the

standard form LPP.
LPP: Standard Form
52
 Slack & Surplus Variables
53

s1 = amount of unused labour

s2 = amount of unused milk
What do you think are the s1 and s2 at the origin?
What do the slack variables, s1 & s2 contribute to in
the
Z = GH¢40x1 + GH¢50x2 fx? Why?
Nothing because they are unused resources
 FMGL - Standard form model
54

Max : Z = 40x1 + 50x2 + 0s1 + 0s2

s.t.
1x1 + 2x2 + s1 = 40
4x1 + 3x2 + s2 = 120
x1,x2,s1,s2 ≥ 0
s1 & s2 are added to the constraints (in max) &
but each is multiplied by 0, then added to the
OF.
 Example 2: Minimization
55

 Ogboo wants to plant maize this month and needs to fertlize the field. He can fertilize with NPK

brand or Urea. Each brand yields specific amounts of nitrogen and phosphate per bag as:

 Ogboo's field requires at least 16 pounds of nitrogen and 24 pounds of phosphate. NPK costs GH

¢6 per bag. Urea costs GH¢3 per bag. The farmer wants to know how many bags of each brand

to purchase to minimize total cost of fertilizer given the data above. Identify the DVs &

formulate an LPP. Find the optimal solution grapphically

 Example 2: Soln (1)
56

Decision Variables?:
x1 = bags of NPK
x2 = bags of Urea .
Objective Function?:
Minimize Z = GH¢6x1 + 3x2
where:
GH¢6x1= total cost of bags of NPK
GH¢3x2= total cost of bags of Urea
 Example 2: Soln (2)
57

Model Constraints?: .

 2x1 + 4x2 ≥ 16lb (nitrogen constraint) .

4x1 + 3x2 ≥ 24lb (phosphate constraint) .

x1,x2 ≥ 0 (non-negativity constraint)

Constraint Graph - Minimization
58
Convex Feasible Region Minimization
59
 Example 2: soln (2)
60

Draw an isocost line that intersects the FSA.

On that line, all points have the same z-value
(or same cost). 2x1 + 4x2 −s1 = 16lb
4x1 + 3x2 −s2 = 24lb
Z = 6x1 + 3x2 + 0s1 + 0s2
 Example 2: Soln (2), Lowest cost
61

 Point C, gives x1=8, x2=0, Z=48, s1=2(8)+4(0)-

16=0, s2=4(8)+3(0)-24=8
 Point B, gives x1=4.8, x2=1.6, Z=33.6,

s1=2(4.8)+4(1.6)-16=0, s2=4(4.8)+3(1.6)-24=0
 Point A, gives x1=0, x2=8, Z=24, s1=2(0)+4(8)-

16=16, s2=4(0)+3(8)-24=0
Graphical Solutions Min: Fertilizer EG
62
Point in FSA with smallest z-value
63
 Surplus Variables -Minimization (1)
64

A surplus variable is subtracted from a ≥

constraint to convert it to =, standard form
LPP.
 A surplus is an excess above a constraint
requirement.
Surpluses contribute nothing to the OF;
hence they are assigned 0 coefficients in the
OF. In our eg,
 2x1 + 4x2 - s1 = 16 (nitrogen)
4x1 + 3x2 - s2 = 24 (phosphate)
Graphical Solutions - Min: Fertilizer
EG
65
 Irregular Types of LPPs
66

For some LPPs, the general rules do not

apply.
Special types of problems include those with:
Multiple/alternate optimal solutions
Infeasible LPs
Unbounded LPs
Redundant solution
 Multiple/Alternate Optimal Solution
67

Graphically, this is when the objective function's

isoprofit or isocost line is parallel to one of the
constraints - i.e., they have the same slope.
For e.g, Management saw more than 1 optimal
solution when they formulated this simple LPP:
 Min. Z = GH¢3x1 + GH¢2x2
subject to:
6x1 + 4x2 ≤ 24
x1 ≤ 3
x1,x2 ≥ 0
Infinite Optimal Solution
68
 Multiple Optimal Solution (AOS).
69

 From graph, the 1st isoprofit of GH8 runs parallel

to the constraint equation.

 If π = GH12, isoprofit line rest directly on

constraint 1.
 Any point along |AB| gives an optimal x1 & x2

combination. NB: AOS allows managers greater

flexibility to decide which combination to select.
 Multiple Optimal Solution (AOS)...
70

Z = 3(0) + 2(6) = 12 &

 Z = 3(4) + 6(0) = 12
The profit remains the same at each AOS.
Goal programming is often used to choose
among alternative optimal solutions.
Infeasibility (1), conflicting constraints
71

It's when there is No solution to an LPP that

satisfies all of the constraints.
That is, there is no intersection of all the
constraints. At least 1 constraint is violated.
For e.g. if a marketing manager provides a
constraint different from what a production
manager provides, an infeasibility can occur
(leading to no solution to the problem) unless
one revises his restriction.
Infeasibility (2), empty feasible region
72
 Infeasibility (3)
73

 Infeasibility is a frequent occurrence in real-life, large-scale LPPs

that involve 100s of constraints
 Unboundedness (1)
74

Sometimes an LP won't have a finite solution.

The FSA is not closed (is open), has no point
Since unlimited profits or costs are not
possible in the real world, this may be due to
an error in formulating the LP.
As below, since this is a max problem &
feasible region extends infinitely to the right,
there is an unbounded solution
Unboundedness (2)
75
 Unboundedness (3)
76

For a minimization problem, an LP is

unbounded if there are points in the feasible
region with arbitrarily small z-values.
 A minimization problem is unbounded if we
never leave the feasible region when moving
in the direction of decreasing z.
In short, for unboundedness, there are points
in the feasible region with arbitrarily large z-
values (max problem) or arbitrarily small z-
values (min problem).
 Redundancy
77

Redundant constraints occur in large LP

formulations.
It causes no major difficulties in solving LPP
graphically, but you should identify its
occurrence.
It's simply 1 that does not affect the FSA.
That is, one constraint may be more binding
or restrictive than another & thereby negate
its need to be considered.
Consider the LPP below with the 3
constraints
Redundancy..
78
 Redundancy...
79

Max. Z = GH¢1x1 + GH¢2x2

 subj to:
 x1 + x2 ≤ 20
2x1 + x2 ≤ 30
x1 ≤ 25
x1,x2 ≥ 0
Constraint 3, x1 ≤ 25, is redundant &
unnecessary as it has no effect on the FSA
from the first two more restrictive/binding
constraints.
 Assignment 1
80

A company produces two products, A and B,

which have profits of $9 and $7, respectively.
Each unit of product must be processed on
two assembly lines, where the required
production times are as follows:
 Assignment 1 Con’t
81

(a) Formulate a linear programming model to

determine the optimal product mix that will maximize
profit.
(b) Solve Problem graphically.
(c) Transform this model into standard form.
(d) Identify the amount of unused resources (i.e., slack)
at each of the graphical extreme points.
(e) What would be the effect on the optimal solution if
the production time on line 1 was reduced to 40 hours?
(f) What would be the effect on the optimal solution if
the profit for product B was increased from $7 to $15?
to $20?