g

e e rin
g in
En
Mathematical Modeling
t ems ón
S y s L e
Dr. Vincent A. L. Boyer in/ Dr. M. AngélicaueSalazar vo A.
ra m e N
r o gProgram a d (PISIS)
P
Graduate
m
in Systems Engineering

u a te Facultad u t ó o de Nuevo
de Ingenierı́anMecánica y Eléctrica

ra d Universidad Autónoma
dA
León

G
rs ida
ive
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling
 Content

g
e e rin
n
E ngi
1 Introduction to Linear Programming
t ems ón
Sy s L e
in o
2 Examples of Optimization Problems
m N uev
r o gra a de
P m
óno
3 Modeling
u a te ut
d A
Gra s i d ad
iv er
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling
 Introduction to Linear Programming

Agenda

g
e e rin
n
E ngi
1 Introduction to Linear Programming
t ems ón
Sy s L e
in o
2 Examples of Optimization Problems
m N uev
r o gra a de
P m
óno
3 Modeling
u a te ut
d A
Gra s i d ad
iv er
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling
 Introduction to Linear Programming

Main components of an optimization model

g
e e rin
gin
n decisions to be
Ethe
Decision variables: these are used to described
s
made tem ón
ys Le
in S of the decision
Objective function: it is a function
u e o
vvariables, it

o g ram maker wantsdetoNminimize (cost) or

establishes what the decision
r
maximize (profits)
P a m
Constraints:tethese representuthe o
ónlimitations, conditions or
d u a t
ra
requirements of the problemA (example, vehicle capacity, demand to be
G i d ad
rs
satisfied, manpower, etc.)
e
iv
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 1 / 19
 Introduction to Linear Programming

Definitions

g
e e rin
in
A function f (x1 , x2 , ..., xn ) of x1 , x2 , ..., xn is a E ng function if and
linear
s
only if for some set of constants c1 , c2 , ...,
emcn ,
f (x1 , x2 , ..., xn ) = c1 x1 + c2 x2 + ...y+sctn xn .
n
S Leó
in
For example, f (x1 , x2 ) = 2x1 + x2 is a linear function
u evoof x1 and x2 , but
o g ram
f (x1 , x2 ) = x 2 1 x2 is not a linear function of xN
1 and x2 .
de
e Pr nom
a
at function f (x1 ,uxt2ó, ..., xn ) and any number b, the
For anyulinear
d
r a A
inequalities
G i d adxn ) ≤ b and f (x1 , x2 , ..., xn ) ≥ b are linear
f (x1 , x2 , ...,
inequalities. ers
iv
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 2 / 19
 Introduction to Linear Programming

Definitions

e r ing
A linear programming problem (LP) is an optimization i e for which
nproblem
g
we do the following: s En
1We attempt to maximize (or minimize)
y s teamlinear functionLeofónthe
in Sthat is to be maximized
decision variables. This function
u e vo or minimized
o g ram
is called objective function
d eN
The values of ther decision variables must
ma satisfy a set of constraints.
2

e Pmust be a linearnequation
o
d u at
Each constraint
u tó or linear inequality
3
G d a d A with each decision variable. For any
a restriction is associated
A rsign
i
s restriction specifies either that xi must be
variable xi , the sign
er
ivi ≥ 0) or that xi may be unrestricted in sign.
nonnegativen(x
U

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 3 / 19
 Introduction to Linear Programming

Definitions
The fact that the objective function for an LP must be a linear function has
two implications:
e r ing
1 Proportionality assumption: The contribution to the
g i neobjective
En to the value of the
function for each decision variable is proportional
s
decision variable
y s tem L e ón
S
2 Additivity assumption: The contribution to the objective
o function
in of the values of uthe
for any variable in independent evother decision
variables. o g ram d eN
Analogously, theefactPrtha each LP constraint
n o ma must be a linear inequality or
a d uathas two implications:
linear equation A utó
r d
1 G
Proportionality
r ida
assumption:
s
The contribution of each variable to
the left-handiv e of each constraint is proportional to the value of
side
Un
the variable
2 Additivity assumption: The contribution of a variable to the

left-hand side of each constraint is independent of the values of the

other variables.
Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)
Mathematical Modeling 4 / 19
 Introduction to Linear Programming

Definitions

g
e e rin
n
E ngi
1
e
The feasible region for an LP is the tset mif sall points satisfying
n all the
ys restrictions. o Leó
LP’s constraints and all the LP’sSsign
inan optimal solution uevto an LP is a point
2 For a maximization problem,m
a the largest objective N
o
in the feasible region
r grwith ad
e function value.
P m
no an optimal solution is a point in
u a tea minimizationuproblem,
Similarly, for
t ó
d region withdtheA smallest objective function value.
the feasible
Gra sid
a
iv er
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 5 / 19
 Introduction to Linear Programming

Types of solutions

g
e e rin
n
E ngi
t emsfor an LP. ón
Sythes feasible solutions
Feasible solution: it satisfies all constraints
Le
1

2
i
Optimal solution: it is the bestn of e v o for an LP,
m N u
r o gra function. a de
according to the objective

e P it does nottósatisfy
Infeasible solution:
nom
at least one constraint of the
3

LP. uat
r a d d Au
G a
rsid
n ive
U

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 6 / 19
 Introduction to Linear Programming

Types of LPs

g
e e rin
n
E ngi
1 Some LPs have a unique optimal solution
t ems ón
Some LPs have an infinite number y s
S of optimal solutions L e
(alternative or
evo
2

multiple optimal solutions) in u

Some LPs have no o g ram solutions (infeasible
feasible d e N LPs)
3
P r a
mpoints
e n o
4 Some LPs
d u at are unbounded: there
utó are in the feasible region with
A
Gra
arbitrarily large (in a maximization problem) objective function values.
i d ad
s er
iv
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 7 / 19
 Examples of Optimization Problems

Agenda

g
e e rin
n
E ngi
1 Introduction to Linear Programming
t ems ón
Sy s L e
in o
2 Examples of Optimization Problems
m N uev
r o gra a de
P m
óno
3 Modeling
u a te ut
d A
Gra s i d ad
iv er
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling
 Examples of Optimization Problems

Facility Location Problem

e r i ng
g i ne besites
The facility location problem consists of a set of potential facility where
a facility can be opened, and a set of customers that n must serviced.
The goal is to pick a subset of facilities tomopen, s E in order to minimize
s
the operative cost (transportation cost,ycost te of opening/building
L e óna facility,
S o
distance to customers, ...). in uev
m N
Decision: Facility o
r gra
Localization
ad
e
P m
u a e
Objective: tMinimize
u t ónocost
the operative
d
Constraints:
dA
GraAll customers smust
ida be served
ve
Locationiconstraint
r
n
U of the facility
Capacity

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 8 / 19
 Examples of Optimization Problems

Fleet Sizing Problem

e r ing
e of customers
The fleet sizing problem consists of a set of vehicles and anset
i
g
that must be serviced. The goal is to pick a subset
s En of the vehicles (fleet
size), in order to minimize the operative costm
s t e (transportationecost,
ó n leasing
cost, customer satisfaction, ...).
n Sy v oL
i e
ra m
Decision: Number of vehicles
e Nu
Objective: Minimize
P rogthe operative cost
m ad
o
d u ate must be served
Constraints:
u tón
A
Gra
All customers
ad id routes
The set of possible
e rsthe
n iv
Shift length of Driver
U
Vehicle capacity

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 9 / 19
 Examples of Optimization Problems

Shift Scheduling Problem

in a set of g
The shift scheduling problem consists of a set of employeeserand
e
n gin
activities to be fulfilled during a planning horizon (days/week/month). The
E
goal is to assign the activities to the employees,s in order to minimize the
y s tem over or under
operative cost (cost of contracting an employee,
L e n
ócovering an
activity, inactive employees, ...). n S vo
i ue
m N
gra of activities) of ean employee
Decision: Shift (Sequence
o d
Pr the operativenocost
Objective: Minimize
e ma
t utó
dua
Constraints:
A
GraAll the activities
sid
ad be covered
should
i v er have some breaks
Employees should
ShiftU n
length
Employee’s skill

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 10 / 19
 Modeling

Agenda

g
e e rin
n
E ngi
1 Introduction to Linear Programming
t ems ón
Sy s L e
in o
2 Examples of Optimization Problems
m N uev
r o gra a de
P m
óno
3 Modeling
u a te ut
d A
Gra s i d ad
iv er
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling
 Modeling

General Linear Programming Problem

g
Objective
e e rin
n
ngi
Min c1 x1 + c2 x2 + ... + cn xn
E
Constraints t ems ón
Sy s L e
s.t. a11 x1 + a12 x2 + ... in+ a1n xn (≤, =, e
u ≥)vo b1
a21 x1 + a22 x2 m
r+a ... + a2n xn (≤, N
e =, ≥) b2
.. .. rog .
.. m ad ..
. .P o .
t e ó n
1 a+ am2 x2 + ...ut+ amn xn
am1 xu (≤, =, ≥) bm
G rad d ad
A
rs i
ive
Nature of the variables
n
xi ≥ 0, U
∀i ∈ {1, ..., n}

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 11 / 19
 Modeling

The carpenter’s problem

ing
Suppose that a carpenter makes tables and chairs and sells all the tables
and chairs he makes in a market. e r
He does not have a steady income and he wishes to optimize g i ne this situation.
So, the carpenter needs to determine how many stables En and chairs he should
m knows that theónincome he
teHe
make in order to maximize his net income. s
Schair sold is $300.vo Le
y
receives per table sold is $500 andnper
The carpenter works 8 hoursaam
i
day from Monday toN ue and takes 2 hours
Friday
gr to make a chair.
to make a table and 1rohour
e
a dAlso, each week he receives
e P
50 units of rawtmaterial, of which ó o m
henrequires 1 unit for each table and 2
u a
d chair he makes. A
units foraeach u t
Gr sid
ad
i ver
How canUnwe fomulate this problem?

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 12 / 19
 Modeling

The carpenter’s problem

g
e e rin
Decision n
Number of tables and chairs to be produced Engi
t ems ón
Objective Sy s L e
in o
uev
ram
Maximize the net income
o g d eN
Constraints e Pr n o ma
t
uaworking utónot exceed the available hours for
r a d
The total time should
d A
Gworking
rs ida
ive of raw material must be used to produce the tables
At most 50nunits
U
and chairs.

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 13 / 19
 Modeling

The carpenter’s problem

Decision variables
g
Let t be the number of tables to be produced.
e e rin
n
Let c be the number of chairs to be produced.
Engi
t ems ón
Objective
Sy s L e
in o
Max z = 500t + 300c
uev
ram eN
Constraints
P rog m ad
te
The totalaworking time shouldtó
not o
n exceed the available hours for
d
working:
a u A u
Gr 2t + c ≤ 40 sid
ad
ver
i of raw material must be used to produce the tables
At most 50nunits
U
and chairs:
t + 2c ≤ 50
Nature of the variables: t ≥ 0, c ≥ 0
Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)
Mathematical Modeling 14 / 19
 Modeling

The carpenter’s problem

g
e e rin
n
E ngi
The optimization model t ems ón
Sy s L e
Max 500tin + 300c = z evo



s.t.ram2t + c
u
≤ N40
e


Prog t + m
d
2c a ≤ 50
o
ate

tócn≥ 0

t ≥ 0,

d u A u
Gra dad rsi
n ive
U

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 15 / 19
 Modeling

Dorian’s problem

Dorian Auto manufactures luxury cars and trucks. The companygbelieves

that its most likely customers are high-income women and high-income
e e rin men.
In order to reach these groups, Doran Auto has embarked n ginon an ambitious
TV advertising campaign and has decided to purchase
m s E 1-minute commercials
spots on two types of program: comedysshows
y te and footballLegames.ón Each
comedy commercial is seen by 7 million
in S women andu2evmillion o men. Each
football commercial is seenaby m 2 million women
r costs N
and 12 million men. A
e1-minute
1-minute comedy show r o g
ad $50000, a d
and football ad costs
P o m
$100000. Dorian
d u ate would u
n
like the commercials
tóleast
to be seen by at least 28
million high-income
ra women
d A
and at 24 million high-income men. Use
a
linearGprogramming to determine
e rsid how Dorian Auto can meet its advertising
iv
requirements at minimum cost.
Un

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 16 / 19
 Modeling

Dorian’s problem

g
e e rin
n
Decision
E ngi
msfootball game to purchase.
Number of 1-minute ads on comedy shows and
e n
S
oyst
L eó
Objective in
uev
m N
r o gra of dollars).a de
Minimize the cost (in thousands

a t eP t ó n om
Constraints u
r a d d Au
G a
At least 28 million ihigh-income women must see the ads.
e r sd
iv high-income men must see the ads.
At least 24 nmillion
U

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 17 / 19
 Modeling

Dorian’s problem

g
Decision variables
e e rin
x1 : number of 1-minute comedy ads to purchase.
gin
n
sE
x2 : number of 1-minute football ads to purchase.
em n
Objective S
oyst
L eó
in
uev
Min z = 50x1 + 100x2 a
r m
(in N
thousands of dollars)
e
P rog m ad
Constraints ate
u u t óno
d 28 million high-income
A women must see the ads:
Gra7x1 + 2x2 ≥ 28sidad
At least
r
At least 24 n ive high-income men must see the ads:
million
U
2x1 + 12x2 ≥ 24
Nature of the variables: x1 ≥ 0, x2 ≥ 0

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 18 / 19
 Modeling

The Dorian’s optimization problem

g
e e rin
n
E ngi
The optimization model t ems ón
S y s L e
Min 50x1in+ 100x2 evo



m
s.t. ra7x1 + 2x2 N u
g 2x + 12xa de≥ 28


Pr o 1 m2 ≥ 24
e o

n
uat

utó 2
x1 ≥ 0, x ≥0

a d A
Gr ad
rsid
n ive
U

Dr. Vincent A. L. Boyer / Dr. M. Angélica Salazar A. (PISIS-FIME)

Mathematical Modeling 19 / 19