Operations research

Linear Programming
Prof: Francisco Fraile
email: [email protected]
Contents
Definition
Characteristics
Problem Definition
Example 1
Assumptions
Example 2
General Problem
Solution: graphic method (2 variables)
 Definition
“Mathematical process to 𝑍 = 500𝑋1 + 350𝑋2
determine the optimal
Robot 1
allocation of scarce resources”
+
Linear Optimisation: involves
maximising or minimising a linear
objective function, subject to a set of
linear inequality or equality constraints.

Programming: Planning activities to

organise limited resources, not computer Robot 2
programming
 Origins
Leonard Kantorovich
Nobel prize in Economics 1978
optimal allocation of resources using
linear programming

Great economic impact suitable to

model classic economic problems

Large number of applications in

many fields, impossible to list them all
Problem types
that can be solved with linear programming include:

Resourcing Problems (Production Mix) Optimal allocation of

resources to maximise profits

Blending Problems Minimum mix of resources to provide a required

level of product or service
Problem Characteristics

Maximize or Minimize an objective function

Subject to a set of constraints
Linear relationships (objective function and constraints)
Continuous variables (positive real numbers)
Problem Formulation
Terminology

Decision Variables (x) Factors to be selected by the decision maker

Technological coefficients (c) Environment conditions, limits
imposed by the environment to the objective function and out of the
control of the decision maker
Objective function (Z) represents the relation between decision
variables and result variables. Objective could be the maximisation
or minimization of the objective
Constraints (Ax ≤ b) limitations imposed to the decision system,
independent of the objective function, owing to the relationships to
the environment
Problem Formulation
A linear programming model is defined by a set of linear
equations (linear algebra):

Objective function (Z) linear equation 𝒄𝑻 𝒙 = 𝒄𝟏 𝒙𝟏 + 𝒄𝟐 𝒙𝟐 + ⋯ + 𝒄𝟐 𝒙𝟐

that represents the (linear) relationship between the decision and
the uncontrollable variables and that we want to minimise or
maximise

Constraints Set of (linear) equations that represent the limitations

imposed by the environment in the management
Problem Formulation Decisions
Decision variables
𝒙𝟏
𝒙𝟐
Vector 𝒙 = …
𝒙𝟐
Environment
Coefficients
𝒄𝟏
Effect on 𝒄𝟐
decisions Vector 𝒄 = …
𝒄𝟐

Objective
𝒄𝑻 𝒙 =
Environment Maximise or 𝒄𝟏 𝒙 𝟏 + 𝒄𝟐 𝒙 𝟐
minimise + ⋯ + 𝒄𝟐 𝒙 𝟐
Constraints
Limitations on
𝒂𝟏𝟏 𝒙𝟏 + 𝒂𝟏𝟐 𝒙𝟐 ≤ 𝒃𝟏
decisions
𝒂𝟐𝟏 𝒙𝟏 ≤ 𝒃𝟐
Problem Formulation Guidelines
Identify objective (max/min objective function and objective
variable)
Identify the decision variables
Express the objective function in terms of the decision variables
Identify the constraints and the type of constraint:
Limited resources (<=): Less or equal
Requirements (=>): Greater or equal
Requirements (=): Equal

Express the constraints in terms of the decision variables

Check for consistency of units
 Production Mix Problems
Create products to satisfy demand using resources
Each product will have a different selling price or profit per unit
Each resource will have a limited availability

The objective is to maximize profits

Simplest form of optimization problem in OR.

Normally, it is a component of a largest problem (seldom
encountered in practical problems)
Example 1: Production Mix
A company produces two kinds of products, P1 and P2.

The selling price of P1 is 300

The selling price of P2 is 250
P1 requires 2 operator man hours and 1 hour of machine operating
time
P2 requires 1 operator man hours and 3 hours of machine time
Operator man hours cannot be higher than 40 hours per day
Machine operating time cannot be higher than 45 hours per day
Marketing states that P1 units per day cannot be higher than 12

How many units of P1 and P2 should the company produce every

day to maximise revenue?
Example 1: Production Mix
Objective?

Decision variables?

Objective function?

Constraints?
 Production Mix Problems
Easy examples:
Production Mix
Ski Patrol
The good carpenter
Blending Problems
Blending materials or resources to create products that satisfy a
given demand.
Each resource or material will have a different cost and/or
availability
Each type of product will have a different selling price and demand
There might by additional requirements (e.g. quality
requirements)
The objective is to find the optimal blending that minimizes cost or
maximizes profits while satisfying the constraints (availability,
demand, requirements)
Blending problems arise in Process Manufacturing: Metal
processing, food processing, chemical industry, pharmaceuticals,
textiles.
Example 2: Blending
The production of a paint requires two materials, M1 and M2

The cost per gram of M1 is 45 cents

The cost per gram M2 is 12 cents

The final product has two properties that depend on M1 and M2, the brightness
and the hue.
To meet the quality requirements, the final product must have a brightness of at
least 300 and a hue of at most 250
The brightness is directly proportional to the amount of M1 and M2 in grams used
in the mix
The hue depends only on the amount of M1. The hue/gram of M1 is 3

How many kgs of M1 and M2 should be used in every paint pot?

Example 2: Blending
Decision variables?

Objective function?

Constraints?
General Problem (Canonical form)
𝐦𝐚𝐱𝐢𝐦𝐢𝐬𝐞 𝒄𝑻 ∙ 𝒙
𝐬𝐮𝐛𝐣𝐞𝐜𝐭 𝐭𝐨 𝑨𝒙 ≤ 𝒃 𝒂𝒏𝒅 ∀𝒊, 𝒙𝒊 > 𝟎

Decision variables
𝒙 = (𝒙𝟏 , … , 𝒙𝒏 )

Coefficients
𝒄 = (𝒄𝟏 , … , 𝒄𝒏 )

Constraints
𝑨 𝒑𝒙𝒏 𝒎𝒂𝒕𝒓𝒊𝒙 𝒘𝒉𝒆𝒓𝒆 𝑨𝒊 𝒂𝒓𝒆 𝒍𝒊𝒏𝒆𝒂𝒓𝒍𝒚 𝒊𝒏𝒅𝒆𝒑𝒆𝒏𝒅𝒆𝒏𝒕
𝒃 = (𝒃𝟏 , … , 𝒃𝒑 )
 Graphical representation
Represent decision variables in a coordinate system (2 decision variables = 2 dimensions)

Mathematical representation Graphical representation

Decision variables Let us start with the first constraint
𝒙𝟏 Units of P1 per day
𝒙𝟐 Units of P2 per day

Objective function
Maximise 𝒁 = 𝟑𝟎𝟎𝒙𝟏 + 𝟐𝟎𝟎𝒙𝟐

Constraints
𝟐𝒙𝟏 + 𝒙𝟐 ≤ 𝟒𝟎
𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟒𝟓 Feasibility region (gray): area containining all
feasible solutions
𝒙𝟏 ≤ 𝟏𝟐
𝒙𝟏 , 𝒙𝟐 >0 Feasible solution: Solution that satisfies the
constraint
 Graphical representation
Represent decision variables in a coordinate system (2 decision variables = 2 dimensions)

Mathematical representation Graphical representation

Decision variables When we enter a new constraint, the feasibility
𝒙𝟏 Units of P1 per day region changes (acts as a cutting line)
𝒙𝟐 Units of P2 per day

Objective function
Maximise 𝒁 = 𝟑𝟎𝟎𝒙𝟏 + 𝟐𝟎𝟎𝒙𝟐

Constraints
𝟐𝒙𝟏 + 𝒙𝟐 ≤ 𝟒𝟎
𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟒𝟓
𝒙𝟏 ≤ 𝟏𝟐
𝒙𝟏 , 𝒙𝟐 >0
 Graphical representation
Represent decision variables in a coordinate system (2 decision variables = 2 dimensions)

Mathematical representation Graphical representation

Decision variables When we enter a new constraint, the feasibility
𝒙𝟏 Units of P1 per day region changes (acts as a cutting line)
𝒙𝟐 Units of P2 per day

Objective function
Maximise 𝒁 = 𝟑𝟎𝟎𝒙𝟏 + 𝟐𝟎𝟎𝒙𝟐

Constraints
𝟐𝒙𝟏 + 𝒙𝟐 ≤ 𝟒𝟎 (irrelevant)
𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟒𝟓
𝒙𝟏 ≤ 𝟏𝟐
𝒙𝟏 , 𝒙𝟐 >0
 Graphical representation
Represent decision variables in a coordinate system (2 decision variables = 2 dimensions)

Mathematical representation Graphical representation

Decision variables For any value of z, the objective function is also a
𝒙𝟏 Units of P1 per day line. Intersections with the feasible region
represent feasible solutions that provide the same
𝒙𝟐 Units of P2 per day value of the objective function

Objective function
Maximise 𝒁 = 𝟑𝟎𝟎𝒙𝟏 + 𝟐𝟎𝟎𝒙𝟐

Constraints
𝟐𝒙𝟏 + 𝒙𝟐 ≤ 𝟒𝟎
𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟒𝟓
𝒙𝟏 ≤ 𝟏𝟐
𝒙𝟏 , 𝒙𝟐 >0
 Graphical representation
Represent decision variables in a coordinate system (2 decision variables = 2 dimensions)

Mathematical representation Graphical representation

Decision variables The optimal value is always at a vertex of the
𝒙𝟏 Units of P1 per day feasibility region
𝒙𝟐 Units of P2 per day

Objective function
Maximise 𝒁 = 𝟑𝟎𝟎𝒙𝟏 + 𝟐𝟎𝟎𝒙𝟐

Constraints
𝟐𝒙𝟏 + 𝒙𝟐 ≤ 𝟒𝟎
𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟒𝟓
𝒙𝟏 ≤ 𝟏𝟐
𝒙𝟏 , 𝒙𝟐 >0
Assumptions (wrap-up)
1. Deterministic problems are considered. In other words, all the
data are known with certainty
2. The objective function is linear
3. Constraints are also linear
4. Decision variables cannot take negative values
5. Additivity of resources: the total use of each resource, is
obtained by summing partial usages of this resource
6. Divisibility of decision variables: these variables can take
fractional values
7. Decision and uncontrollable variables are independent
8. Both the quantity of the resource employed and the objective
function value are proportional to the values of the decision
variables