Decision 1:: Linear Programming

[email protected]

Last modified: 11th March 2021

Decision 1 Overview

1:: Algorithms 2:: Graphs and networks

Sorting and bin What is a graph and how they 3:: Algorithms on graphs
packing. represent things. What algorithms do I need to
be able to apply?
4:: Route inspection
5:: The Travelling Salesman
Find the shortest route 6:: Linear Programming
Find the shortest route
which travels along all which visits all places. How to find an optimal
roads solution graphically

7:: The simplex algorithm

8:: Critical path analysis
How to find an optimal
solution algebraically. How to plan a project.
Linear Programming
Linear Programming is an optimisation technique. It involves
making the most, or the least, of something when subjected to
some (linear) constraints.
In this chapter we will deal with
constrained problems in two
independent variables, but the
same process can be applied to
higher dimensional problems.

Rather than try and give you

examples of when linear
programming is useful, try to think
of any situation where you are
trying to make something and you
don’t have constraints?
https://www.geogebra.org/m/twb5mtec
Formulating the problem
Mr Blackett is thinking of taking a break from teaching Decision Maths and setting up a
business making and selling Mobius Strips and Klein Bottles hewn from solid gold. He
invests in a 3-D CNC milling machine and sets about defining the constraints as a linear
programming problem.

• Each Klein bottle takes 2 hours to mill and 3 hours to polish.

• Each Mobius strip takes 3 hours to mill and 1 hour to polish.

The milling machine can run for 80 hours per week. Mr Blackett can polish for 60 hours
a week.

He believes he can make £2,000 profit on every Klein bottle and £1,000 profit on every
Mobius Strip. He wishes to maximise his profit!
First define the decision variables. Let be the number of Klein bottles and be the number
of Mobius strips produced in a week.
Next state the objective. Maximise

Then identify the constraints. Milling,

Polishing,
Non Negativity,
Formulating the problem – trickier example
A manufacturer supplies specialist stainless steel to the constructions industry.
The manufacturer has supplies of two types of stainless steel in store.

A) 82% Iron, 17% Chromium, 1% Carbon - Costs £1.20/kg, 15,000kg in stock

B) 88% Iron, 11.5% Chromium, 0.5% Carbon - Costs £1.00/kg, 10,000kg in stock

A customer wants 20,000kg of stainless steel which is

• No more than 85% Iron
• More than 15% Chromium
• No more than 0.9% Carbon

The manufacture will alloy type A with type B to produce the steel for the customer.

The manufacturer wishes to minimise costs.

Cost/
Steel Iron Chrome Carbon Maximum kg
A 82% 17% 1% 15,000 £1.20
B 88% 11.5% 0.5% 10,000 £1
Required?
Note the for “more than”, for “less than” for “no more than” and for “no less than”
Formulating the problem – trickier example ctd.

Steel Iron Chrome Carbon Maximum Cost/kg A customer wants 20,000kg

of stainless steel which is
• No more than 85% Iron
A () 82% 17% 1% 15,000 £1.20 • More than 15%
Chromium
B () 88% 11.5% 0.5% 10,000 £1 • No more than 0.9%
Required? Carbon

Define the decision variables State (and simplify) the constraints (ctd.)

Let be the number of kg of type A used.

Let be the number of kg of type B used.

State the objective

Minimise

Note we define this constraint as

State (and simplify) the constraints even though it is obvious that the
optimal solution will be , this lets
us define a valid solution region
which we may use for later
adjustments to the problem.
Test your understanding
Exercise 6A
Pearson Decision 1, Page 143
Graphical Methods
The region of a graph that The convention for linear programming
problems is to leave the feasible region
satisfies all the constraints of (which satisfies all inequalities) unshaded.
a linear programming Graphing software (such as Desmos),
problem is called the always shade the region which satisfies
the inequality, so be careful if using
feasible region. graphing software to check your answers.

It is a well known rule of

flower arranging that the
overall height of the

Height of arrangement (cm)

arrangement should be
between 1.5 and twice the
height of the vase. We can
visualise this on the plot on
the right, with the white
region representing the
feasible region for the
vase/arrangement
combinations.
Height of vase (cm)
Graphing the problem
Recall that Mr Blackett is thinking of branching out, earlier we formulated his problem
as the following inequalities.
Clicky graph

Let be the number of Klein bottles

Let be the number of Mobius strips
produced in a week.

Maximise

Milling,
Polishing,
Non Negativity,

First let
First
Put alet
solid line on
Then let axis and
the and
Then let
shade the negative
Plot with a solid line as it
regions.
isPlot with a solid line as it is
Shade the unwanted region
Test your understanding
Again recall that we had a customer wanting to buy 20,000kg of stainless steel. We
formulated the problem with the following constraints.
Let be the number of kg of type A Clicky graph

used.
Let be the number of kg of type B
used.

Plot the line

Put
Pick
Plot
Plot
Plot
Whenaathese
the
the solid
point, line
say
linetwo
line with ona solid
lines, andline.
the
to help
Check
shade
When and
aabove axisPlot
plot.
point andwith
either
them. sidedotted
of it
toshade
line,
Check
Plot asa the
itpoint
determine
this is negative
with which
a solidside
either lineto
ofand
it
toregions.
Check
shade.a point either side of it to determine
shade
determine
below
which side to shade.which
it. side to
shade.
Test your understanding
Clicky graph
Exercise 6B
Pearson Decision 1, Page 147
Finding a minimum point – sliding ruler method
Recall that we had a customer wanting to buy 20,000kg of stainless steel, earlier we
formulated his problem as inequalities, and graphed them as shown below.

This graph shows

us the feasible
region – what
values of type A
and type B that the
customer would
accept in the mix.

It doesn’t tell us
which of these
points is optimal,
that is, which is
going to minimise
cost.

https://www.desmos.com/calculator/vnfvgt3lcr
Finding a minimum point – sliding ruler method
We earlier, stated our objective function to be Minimise
The objective function is
not one line, but an infinite
family of parallel lines.

We can plot the objective

function if we assume a
value of C.

Lets look at a few examples

When
When
Finding a minimum point – sliding ruler method
We earlier, stated our objective function to be Minimise

Notice that all the cost lines are

𝑥+
𝑦=
parallel, and as the cost increases,
20
the line moves ‘up’. Put a ruler on 00
your cost lines, slide it ‘up’ and you 0
will see that the first time it
intersects the feasible region is at the
intersection of 𝑦
5
1.7
𝑥=
and

Solving simultaneously, gives

Substituting into to calculate cost

Gives
 Finding a maximum point - sliding ruler method
For a maximum point, look for the last point covered by an objective line as it
leaves the feasible region.
For a minimum point, look for the first point covered by an objective line as it
enters the feasible region.
A linear programming problem is formulated
such that, Clicky graph

Maximum Profit
Show the feasible region on the axis opposite.

The objective function is to maximise

By plotting a profit line for a value of P, then

sliding the ruler, find the coordinates of the
point of maximum profit and calculate at this
point.

𝑃=
𝑃 = 36
𝑃 = 30
https://www.desmos.com/calculator/qbkbkcphh6 20
Test your understanding
Clicky graph
 Vertex testing method
To find an optimal point using the vertex testing method: https://www.
desmos.com/
1) Find the coordinates of each vertex of the feasible region. calculator/o7
hauownx9

2) Evaluate the objective function at each of these points.

3) Select the vertex that gives the optimal value of the objective function..
A feasible region is
defined by the following
Vertex
Vertex Value of
inequalities A 0 25 4250
B 6 7 2030
The objective function is C 11 2 1880
to minimise D 13 1 1990

Hence the
vertex which
Sketch the graph to identify minimises the
the feasible region objective
then solve the related function is C,
simultaneous equations to with a cost of
find the coordinates of 1880.
each vertex.
Test your understanding
The feasible region for a linear
programming problem is defined by the A B 𝑦 =10
following inequalities.

𝑦 =− 5
The objective function is to minimise

By the vertex testing method determine

𝑥 + 13

𝑥=4
the minimum feasible value of C.

𝑦=
−𝑥
D

+4
Vertex
Vertex C 𝑦 =1
AA 0.6 10 147.8
BB 3 10 179
CC 3 1 53
Therefore the min. value of
DD 2.25 1.75 53.75 at
Exercise 6C
Pearson Decision 1, Page 158
 Solutions with integer values
Recall that Mr Blackett is thinking of branching out into golden Klein bottles and Mobius strips ,
earlier we formulated his problem as the following inequalities and graphed it as shown, with
the objective function in orange.
Maximise

Milling,
Polishing,
Non Negativity,

Mr Blackett decides that he should make an integer amount of

each product each week.
1) Find the coordinates of the optimal point.

2) Imagine an integer ‘box’ around the optimal ( 1 4 ,1 8 ) ( 15 ,1 8 )

point, determine which of the corners are in the
feasible region, and which of those optimises
the objective function. ( 14 , 17 ) ( 15 ,17 )
Solutions with integer values – ctd.

In
In R?
R?

Yes £45,000
- No
- No
- No

If one constraint failed, you don’t

need to check the other.
Exercise 6D
Pearson Decision 1, Page 165