Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

DSME 6620

Decision Models and Applications


MGT578

Lecture Three

Linear Programming II
ecision Making

ptimization – Part I:
1
ion, and Math Software
DSME 6620
Decision Models and Applications
MGT578

Agenda

• Review of LP
• Special cases of LP solutions
– Infeasible solution
– Multiple optimal solutions
– Unbounded solution
• Blending problem
• Workforce scheduling problem
• Network models

2
DSME 6620
Decision Models and Applications
MGT578

Review of LP

• Decision variables: What must be


decided?
– Should describe the decisions to be made

In the microcomputer problem: production


quantities of Type I and Type II Microcomputer

• Objective function: What measure will the


decision maker want to optimize?
– The decision maker wants to maximize
(usually revenue or profit) or minimize
(usually cost) some function of decision
variables

In the microcomputer problem: maximize profit

• Constraints: What restrictions limit the


choices?
– Restrictions placed on a decision variable or
a group of decision variables

In the microcomputer problem: assembly time


constraint, inspection time constraint, storage
space constraint, and nonnegativity 3
DSME 6620
Decision Models and Applications
MGT578

Special Cases of LP Solutions

Consider the following linear program. We can show


there exists a unique optimal solution, which is (x1, x2)
= (240, 80) . Then, we will make some modifications on
the formulation and see special cases of optimality.

Max p = 7x1 + 6x2


subject to
x1 + 2x2 ≤ 400
x2 4x1 + 3x2 ≤ 1200
x2 ≤ 150
x1 ≥ 0, x2 ≥ 0
400

Line 2

200
B Line 3
A
C
Feasible
Region Line 1
x1
E D 400
DSME 6620
Decision Models and Applications
MGT578

Infeasible Solution

Assume that a new constraint, x1 ≥ 500, is added to


the formulation. Consider how the feasible region
will vary.

Max p = 7x1 + 6x2


subject to
x1 + 2x2 ≤ 400
4x1 + 3x2 ≤ 1200
x2 ≤ 150
x1 ≥ 500
x2 x1 ≥ 0, x2 ≥ 0

400

200 Line 2
B Line 3
A
C
Line 1
x1
E 5
D 400 500
DSME 6620
Decision Models and Applications
MGT578

Multiple Optimal Solutions

Assume the objective function is changed to 8x1 +


6x2. Consider what the slope of the objective
function line will be.

Max p = 8x1 + 6x2


subject to
x1 + 2x2 ≤ 400
4x1 + 3x2 ≤ 1200
x2 ≤ 150
x1 ≥ 0, x2 ≥ 0
x2

400
New objective function line

200 Line 2
B Line 3
A
C
Line 1
x1
E D 400
DSME 6620
Decision Models and Applications
MGT578

Unbounded Solution

Assume that the first two constraints are removed


from the formulation. Consider how the feasible
region will vary.

Max p = 7x1 + 6x2


subject to
x1 + 2x2 ≤ 400
4x1 + 3x2 ≤ 1200
x2 ≤ 150
x1 ≥ 0, x2 ≥ 0
x2

Objective function line

150

x1
0 7
DSME 6620
Decision Models and Applications
MGT578

Summary of Special Cases

Special Graphical Excel Managerial


cases method Solver Implication
Message
Infeasible Feasible region “Solver Too many
solution does not exist could not restrictions
find a
feasible
solution”
Multiple Slope of the Generate At least, we get an
optimal objective one optimal solution
solutions function is the optimal
same as that of solution
one constraint
line
Unbounde Optimal “The Set Problem may be
d solution objective Cell values improperly
function value is do not formulated:
infinite converge” - too few
constraints
- wrong objective
function (Max or
Min set correctly?)

8
DSME 6620
Decision Models and Applications
MGT578

Blending Problem

Chandler Oil has 5,000 barrels of cruid oil 1


and 10,000 barrels of cruid oil 2 available.
Chandler sells gasoline and heating oil. These
products are produced by blending together
the two cruid oils.

Each barrel of cruid oil 1 has a “quality level”


of 10 and each barrel of cruid oil 2 has a
quality level of 5. Gasoline must have an
average quality level of at least 8, whereas
heating oil must have an average quality
level of 6. Gasoline sells for $25 per barrel,
and heating oil sells for $20 per barrel. We
assume that the demand for gasoline and
heating oil is unlimited, so that all of
Chandler’s production can be sold.

Chandler wants to maximize its revenue from


selling gasoline and heating oil.

9
DSME 6620
Decision Models and Applications
MGT578

Blending Problem

Analysis & Formulation

10
DSME 6620
Decision Models and Applications
MGT578

Blending Problem

-X1G gasoline from crude 1


-X2G gasoline from crude 2
-X1H heating oil from crude 1
-X2H heating oil from crude 2

OBJ: (X1G+X2G)*25+(X1H+X2H)*20

S.T.
10X1G+5X2G>=(X1G+X2G)*8
10X1H+5X2H>=(X1H+X2H)*6
These two constraints can be written as:
2X1G-3X2G>=0
4X1H-X2H>=0
And,
X1G+X1H<=5000
X2G+X2H<=10000

11
DSME 6620
Decision Models and Applications
MGT578

Workforce Scheduling Problem

A post office requires different numbers of full-time


employees on different days of the week. The
minimum number of full-time employees needed on
each day is:
Monday 17
Tuesday 13
Wednesday 15
Thursday 19
Friday 14
Saturday 16
Sunday 11

Union rules state that each full-time employee must


work five consecutive days and then receive two
days off. For example, an employee who works
Tuesday to Saturday must be off Sunday and
Monday. The post office wants to meet its daily
requirements using only full-time employees.
Formulate an LP that the post office can use to
minimize the number of full-time employees that must
be hired. 12
DSME 6620
Decision Models and Applications
MGT578

Workforce Scheduling Problem

Analysis & Formulation

13
DSME 6620
Decision Models and Applications
MGT578

Workforce Scheduling Problem

Analysis & Formulation

X1=Number of workers working from Sat to


Wed
X2=Number of workers working from Sun to
Thu
Similarly, X3, X4, X5, X6, X7
Obj: Min X1+X2+X3+…+X7
S.t.
X3+X6+X7+X1+X2>=17
X4+X7+X1+X2+X3>=13
X5+X1+X2+X3+X4>=15
X6+X2+X3+X4+X5>=19
X7+X3+X4+X5+X6>=14
X1+X4+X5+X6+X7>=16
X2+X5+X6+X7+X1>=11

14
DSME 6620
Decision Models and Applications
Examples
MGT578

Network Models Examples

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

• A network problem (in linear programming), is a


mathematical problem that can be described in a
diagram (also called graph) that includes arcs (or
arrows) and nodes (or circles).
Network Problems

• A network problem (in linear programming), is a mathematical


• problem
The decision
that canvariables are
be described in the levels(also
a diagram of directed
called graph) that
flows on
includes the
arcs (or arcs.
arrows) and nodes (or circles).
• The
h decision
d i i variables
i bl are the
h levels
l l off directed
di d flows
fl on the
h arcs.

– Solid lines = decision variables; Dashed lines = non decision (“external”)


OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

15
DSME 6620
Decision Models and Applications
MGT578

Network Problems (cont.)

• The objective function can be described


as a linear function of the decision
variables.

• The constraints in the problem are


described in terms of the arrows pointing in
and out of each node.
Network Problems (cont.)

• The objective function can be described as a linear function of the


decision variables.
• The constraints in the problem are described in terms of the arrows
pointing
i i in i andd out off each
h node.
d

Ext. Input (E.Inp)

Y1
X1

Ni Y2
X2

Y3 Ext. Output (E.Out)

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Network Problems
16
Types of Flow Constraints
• Two types of flow constraints are common:
DSME 6620
Decision Models and Applications
MGT578

Network Problems
Types of Flow Constraints

• Two types of flow constraints are common:


– Flow in = Flow out (referred to as the
no‐“inventory” case):
X1 + X2 + E.Inp – Y1 – Y2 – Y3 – E.Out = 0

– Flow in >= Flow out (referred to as the


“possible inventory” case):
X1 + X2 + E.Inp – Y1 – Y2 – Y3 – E.Out >= 0

• In general, we may have a linear


combination of inputs/outputs:
• a1*X1 + a2*X2 + E.Inp – b1*Y1 – b2*Y2 – b3*Y3 – E.Out >= 0

• Of course, you can augment a network


problem with other constraints (e.g.,
maximum or minimum on the flows).

17
DSME 6620
Decision Models and Applications
MGT578

Example 1:
A Transportation Problem

• A manufacturer has three warehouses (1, 2, 3),


and three retail stores (4, 5, 6). The following
network diagram indicates the shipping routes
and shipping costs. The values on the arcs
indicate the per-unit shipping costs between
the various
Example 1: cities. The supply of material at
each warehouseProblem
A Transportation is indicated by the number
next to nodes 1, 2, and 3. The demand for
A manufacturer has three warehouses (1, 2, 3), and three retail stores (4, 5, 6). The follow
furniture
etwork diagram is the
indicates indicated byand
shipping routes the number
shipping next
costs. The toon the arcs
values
ndicate thenodes 4, 5, 6.costs between the various cities. The supply of material at
per-unit shipping
• Objective: Formulate an LP to determine the
warehouse is indicated by the number next to nodes 1, 2, and 3. The demand for furniture
ndicated by the number next to nodes 4, 4 5,
5 6.
6
least costly shipping plan for Furniture Max.
Objective: Formulate an LP to determine the least costly shipping plan for Furniture M

18
DSME 6620
Decision Models and Applications
M5705 Quantitative Decision Making Prof. Aviv MGT578
Olin Business School, Washington U

Problem Formulation

Problem Formulation

X13 X36
X34

X14
X46
X24
X54
X56
X25

M5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington U

19
DSME 6620
Decision Models and Applications
MGT578

Problem Formulation

Problem Formulation

• The objective function is stated as follows:


Min $20*X13 + $30*X14 + ...
• The constraints are stated node‐by‐node.
node by node For example,
example

(N1) 30 ‐ X13 + X14

‐ inequality is to reflect a situation in which:


• The “‐”
– Some inventory might be left at node 1, or
– Not all of the supply is used (think of the supply as manufacturing capacity).
– If all of the 30 units need to be delivered, then we need to set:
(N1) 30 = X13 + X14
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Problem Formulation

• Can you write the constraint for node #4? 20


– Some inventory might be left at node 1, or
DSME 6620
– Models
Decision Not all and
of the supply is used (think of the supply as manufacturing capacity).
Applications
MGT578
– If all of the 30 units need to be delivered, then we need to set:
(N1) 30 = X13 + X14
Problem Formulation
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Problem Formulation

• Can you write the constraint for node #4?

(N4) X14 + X24 + X34 + X54 ‐ 10 + X46

• To make the constraint “ready for Excel,” we want to move all


decision variables to one side of the equation. For example:
(N4) X14 + X24 + X34 + X54 – X46 – 10 ‐ 0
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

21
DSME 6620
Decision Models and Applications
MGT578

Problem Formulation: Modeling in


Excel

Problem Formulation: Modeling in Excel

• How do you set the decision variables in a spreadsheet?


• There are three main approaches:
– Alternative 1: construct a row that includes all variables:

– Alternative 2: a source‐destination matrix with only feasible flows allowed:

– Alternative 3: a source‐destination matrix:

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington Univers

22
Alternative #1
DSME 6620
– Alternative 3: a source‐destination matrix:
Decision Models and Applications
MGT578

Alternative #1
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Alternative #1

Interpretation:
p

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

23
DSME 6620
Decision Models and Applications
MGT578

Alternative #2

Alternative #2

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Alternative #3 24
DSME 6620
Decision Models and Applications
MGT578

Alternative #3
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

Alternative #3

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

25
DSME 6620
Decision Models and Applications
MGT578

Example 2:
A Multi‐period Investment Problem

• Parents Patti and Russ want to provide for their


daughter’s college education with some of the
$100,000 they have recently inherited. They hope to
set aside part of the money (at the beginning of year
Example 2:
1) and establish an account that would cover the
needs A Multi‐period Investment college
of their daughter’s Problem education, which
begins
Parents fourwant
Patti and Russ years from
to provide now
for their (i.e., college
daughter’s at the beginning
education with some ofofthe
year
$100,000 5). recently
they have Theirinherited.
estimateThey hopeisto set aside part
that of the money
first-year college
(at the
beginning of year 1) and establish an account that would cover the needs of their daughter’s
expenses
college education, whichwillbegins
come four to $24,000
years from now and(i.e., at will increase
the beginning $2000
of year 5). Their
per
estimate year
is that during
first-year collegeeach
expensesof willthe
comeremaining
to $24,000
$24 000 and willthree years
increase $2000 ofper
college.
year during The
each of the following
remaining three yearsinvestments are available
of college. The following investments are to
available to them.
them.

Objective: Faced with this prospect, Patti and Russ wish to set aside the minimum amount of
• Objective: Faced with this prospect, Patti and Russ
initial investment that will “guarantee” that the college expenses will be met.
wish to set aside the minimum amount of initial
Assumeinvestment
no uncertainty that willwith
in dealing “guarantee”
this problem. that the college
expenses will be met.
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University
• Assume no uncertainty in dealing with this problem.
26
6620investment that will “guarantee” that the college expenses will be met.
initial
DSME
Decision Models and Applications
MGT578
Assume no uncertainty in dealing with this problem.

Problem Formulation
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington

Problem Formulation

now
Planning horizon:
years of study

time
1 2 3 4 5 6 7 8

Decision variables:

V = initial investment

A-type investment s: A1,A2,...,A7

B type investment s: B1,B


B-type B3,B
B5

C-type investment s: C1,C4

D-type investment s: D1

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington

27
DSME 6620
Decision Models and Applications
MGT578

Problem Formulation
Problem Formulation

Describe the problem as a network flow model:

x = the beginning of year x. 24 26 28 30

D1

B1 B3 B5

1 2 3 4 5 6 7 8
A1 A2 A3 A4 A5 A6 A7

C1 C4

initial investment (V)

A Key observation: The investment arrows carry yield (return on investment). For example,
B3 leaves node #3, but gets into node #5 as 1.11*B3.

•OMM5705
A Key observation: The investment
Quantitative Decision Making Prof. Aviv
arrows carry
Olin Business School, Washington Univers

yield (return on investment). For example, B3


leaves node #3, but gets into node #5 as
1.11*B3.
Problem Formulation

Constraints: A couple of examples


28
DSMEA6620
Key observation: The investment arrows carry yield (return on investment). For exam
Decision Modelsnode
B3 leaves and #3,
Applications
but gets into node #5 as 1.11*B3. MGT578
OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washing

Problem Formulation

Problem Formulation

Constraints: A couple of examples

(Y1) V = A1 + B1 + C1 + D1

(Y7) 1.05*A6 + 1.11*B5 + 1.16*C4 = A7 + 28

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washing

29
DSME 6620
Decision Models and Applications
MGT578

Problem Formulation

Problem Formulation

OMM5705 Quantitative Decision Making Prof. Aviv Olin Business School, Washington University

30

You might also like