OPERATIONS RESEARCH

PART 5: INTEGER LINEAR PROGRAMMING

TRUNG-HIEP BUI
scv.udn.vn/buitrunghiep | [email protected] | 0935-743-555
 DISTRIBUTION AND NETWORK MODELS

CONTENT

 Types of Integer Linear Programming Models

 Graphical and Computer Solutions for an All-integer LP
 Application involving 0-1 Variables
 Modeling Flexibility Provided by 0-1 Integer Variables

LEARNING OUTCOMES
 DISTRIBUTION AND NETWORK MODELS

TYPES OF INTEGER LINEAR PROGRAMMING MODELS

All-integer linear program LP RELAXATION Mixed-integer linear program

0-1 linear integer program

 DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

BT has $2 million to purchase rental property (townhouse and apartment buildings).

- Each townhouse can be purchased for $282,000 and 05 townhouses are available.
- Each apartment building can be purchased for $400,000 and the is no limit quantity of
apartments.
BT can devote up to 140 hours per month to managing these new properties.
- Each townhouse requires 4 hours per month.
- Each apartment building requires 40 hours per month.
The annual cash flow is estimated at $10,000 per townhouse and $15,000 per apartment.
BT would like to determine the number of townhouses (T) and the number of apartment buildings
(A) to purchase to maximise annual cash flow.
DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

T, A 0 and integer
 DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

T, A 0 and integer

LP Relaxation
 DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

T, A 0 and integer

T = 2.479 ; A = 3.252
10(2.479) + 15(3.252) = 73.574 (k$)

Round the solution:

T = 2.479 ≈ 2; A = 3.252 ≈ 3
10(2) + 15(3) = 65 (k$)

(T = 3; A = 3): Infeasible solution

282(3) + 400(3) = 2046 (k$) > 2.000 (k$)
 DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

T, A 0 and integer

T = 2.479 ≈ 2; A = 3.252 ≈3
10(2.479) + 15(3.252) = 73.574 (k$)

Round the solution:

T = 2.479 ≈ 2; A = 3.252 ≈3
10(2) + 15(3) = 65 (k$)

Optimal Integer solution:

(T = 4; A = 2):
10(4) + 15(2) = 70 (k$)
 DISTRIBUTION AND NETWORK MODELS

GRAPHICAL AND COMPUTER SOLUTIONS FOR AN ALL-INTEGER LP

T, A 0 and integer

T = 2.479 ≈ 2; A = 3.252 ≈3
10(2.479) + 15(3.252) = 73.574 (k$)

Round the solution:

T = 2.479 ≈ 2; A = 3.252 ≈3
10(2) + 15(3) = 65 (k$)

Optimal Integer solution:

(T = 4; A = 2):
10(4) + 15(2) = 70 (k$)
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

03 materials are used to produce 03 products: Fuel additive, Solvent base, and Carpet cleaning fluid.
The following decision variables are used:
F = tons of fuel additive produced
S = tons of solvent base produced
C = tons of carpet cleaning fluid produced

Quantity of material to produce a ton of product

Profit
PRODUCT
per ton of product Material 1 Material 2 Material 3

Fuel additive 40 0.4 0.6

Solvent base 30 0.5 0.2 0.3
Carpet cleaning fluid 50 0.6 0.1 0.3
Maximum available material 20 5 21
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

03 materials are used to produce 03 products: Fuel additive, Solvent base, and Carpet cleaning fluid.
Quantity of material to produce a ton of product
Profit
PRODUCT
per ton of product Material 1 Material 2 Material 3

Fuel additive 40 0.4 0.6

Solvent base 30 0.5 0.2 0.3
Carpet cleaning fluid 50 0.6 0.1 0.3
Maximum available material 20 5 21

This LP formulation does not include a fixed cost for production setup of the products.
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

03 materials are used to produce 03 products: Fuel additive, Solvent base, and Carpet cleaning fluid.

Quantity of material
Profit to produce a ton of product Maximum
Setup
PRODUCT per ton of production
cost
product Material 1 Material 2 Material 3 (tons)

Fuel additive 40 0.4 0.6 200 50

Solvent base 30 0.5 0.2 0.3 50 25
Carpet cleaning fluid 50 0.6 0.1 0.3 400 40
Maximum available
20 5 21
material
The 0-1 variables can be used to incorporate the fixed setup costs into the production model.
SF = 1 if the fuel additive is produced; 0 if not
SS = 1 if the solvent base is produced; 0 if not
SC = 1 if the carpet cleaning fluid is produced; 0 if not

Using these setup variables, the total setup cost is: 200.SF + 50.SS + 400.SC
The objective function to include the setup cost: Max (40.F + 30.S + 50.C – 200.SF - 50.SS – 400.SC)
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

03 materials are used to produce 03 products: Fuel additive, Solvent base, and Carpet cleaning fluid.

Quantity of material
Profit to produce a ton of product Maximum
Setup
PRODUCT per ton of production
cost
product Material 1 Material 2 Material 3 (tons)

Fuel additive 40 0.4 0.6 200 50

Solvent base 30 0.5 0.2 0.3 50 25
Carpet cleaning fluid 50 0.6 0.1 0.3 400 40
Maximum available material 20 5 21
Using these setup variables, the total setup cost is: 200.SF + 50.SS + 400.SC

The objective function to include the setup cost: Max (40.F + 30.S + 50.C – 200.SF - 50.SS – 400.SC)
The constraints:

F ≤ 50.SF S ≤ 25.SS C ≤ 40.SC

SF, SS, SC = 0 or 1
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

Quantity of material
Profit to produce a ton of product Maximum
Setup
PRODUCT per ton of production
cost
product Material 1 Material 2 Material 3 (tons)

Fuel additive 40 0.4 0.6 200 50

Solvent base 30 0.5 0.2 0.3 50 25
Carpet cleaning fluid 50 0.6 0.1 0.3 400 40
Maximum available material 20 5 21
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – FIXED COST PROBLEM

 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – DISTRIBUTION SYSTEM DESIGN

Shipping cost per unit ($)

Annual fixed cost Annual capacity from Plant to Distribution center
Proposed plant
($) (unit)
Boston Atlanta Houston

Detroit 175.000 10.000 5 2 3

Toledo 300.000 20.000 4 3 4
Denver 375.000 30.000 9 7 5
Kansas city 500.000 40.000 10 4 2
Current plant
ST. Louis 30.000 8 4 3
Annual
30.000 20.000 20.000
Demand
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – DISTRIBUTION SYSTEM DESIGN

0-1 variables can be used in this distribution system

design problem to develop a model for choosing the best
plant locations and for determining how much to ship
from each plant to each distribution center.
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – DISTRIBUTION SYSTEM DESIGN

The annual fixed cost of operating the new plants (k$)

175 y1 + 300 y2 + 375 y3 + 500 y4

x ij = the units shipped from Plant i to Distribution center j

(i = 1, 2, 3, 4, 5 and j = 1, 2, 3) *unit in thousand

The annual transportation cost (k$)

(5x11 + 2x12 + 3x13) + (4x21 + 3x22 + 4x23) +

(9x31 + 7x32 + 5x33 ) + (10x41 + 4x42 + 2x43)+

(8x51 + 4x52 + 3x53)

DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – DISTRIBUTION SYSTEM DESIGN

Capacity constraints:
x11 + x12 + x13 ≤ 10y1 Detroit capacity

x21 + x22 + x23 ≤ 20y2 Toledo capacity

x31 + x32 + x33 ≤ 30y3 Denver capacity

x41 + x42 + x43 ≤ 40y4 Kansas city capacity

x51 + x52 + x53 ≤ 30 St. Louis capacity

Demand constraints:
x11 + x21 + x31 + x41 + x51 = 30 Boston demand

x12 + x22 + x32 + x42 + x52 = 20 Toledo demand

x13 + x23 + x33 + x43 + x53 = 20 Houston demand

Variable constraints:
xij ≥ 0 for all i, j;

y1, y2, y3, y4 =0; 1

DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – DISTRIBUTION SYSTEM DESIGN

 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

If the company locates its principal

place of business (PPB) in a county, then
its branches can be established in that
county and any adjacent counties.

Objective: Determine the minimum

number of PPBs necessary to establish
branches throughout the 20-county region.

xi =
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

xi =

To minimize the number of PPBs Min

The company may locate a branch in a county if:
• That county contains a PPB
• That county is adjacent to another county with a PPB

Constraints:
• Ashtabula county: x1 + x2 + x12 + x16
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

xi =

To minimize the number of PPBs Min

The company may locate a branch in a county if:
• That county contains a PPB
• That county is adjacent to another county with a PPB

Constraints:
• Ashtabula county: x1 + x2 + x12 + x16
• Lake county: x1 + x2 + x3 + x12
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

xi =

To minimize the number of PPBs Min

The company may locate a branch in a county if:
• That county contains a PPB
• That county is adjacent to another county with a PPB

Constraints:
• Ashtabula county: x1 + x2 + x12 + x16
• Lake county: x1 + x2 + x3 + x12
• Cuyahoga county: x2 + x3 + x4 + x9 + x10 + x12+ x13
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

xi =

To minimize the number of PPBs Min

The company may locate a branch in a county if:
• That county contains a PPB
• That county is adjacent to another county with a PPB

Constraints:
• Ashtabula county: x1 + x2 + x12 + x16
• Lake county: x1 + x2 + x3 + x12
• Cuyahoga county: x2 + x3 + x4 + x9 + x10 + x12+ x13
• ……
• Carroll county: x11 + x14 + x19 + x20
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – BUSINESS LOCATION

 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – CAPITAL BUDGETING

Faced with limited capital for the next 04 years, company needs to select the most profitable projects.
Present value*: The estimated net present value is the net cash flow discounted back to the beginning of year 1

The four 0-1 decision variables are:

P = 1 if the Plant expansion project is accepted; 0 if rejected
W = 1 if the Warehouse expansion project is accepted; 0 if rejected
M = 1 if the New machinery project is accepted; 0 if rejected
R = 1 if the New product research project is accepted; 0 if rejected
DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – CAPITAL BUDGETING

DISTRIBUTION AND NETWORK MODELS

APPLICATIONS INVOLVING 0-1 VARIABLES – CAPITAL BUDGETING

 DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – MULTIPLE-CHOICE

Faced with limited capital for the next 04 years, company needs to select the most profitable projects.
Present value*: The estimated net present value is the net cash flow discounted back to the beginning of year 1

The four 0-1 decision variables are:

P = 1 if the Plant expansion project is accepted; 0 if rejected
W = 1 if the Warehouse expansion project is accepted; 0 if rejected
M = 1 if the New machinery project is accepted; 0 if rejected
R = 1 if the New product research project is accepted; 0 if rejected
 DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – MULTIPLE-CHOICE CONSTRAINT

If the company actually has 3 warehouses and it just

wants to expand only one warehouse.
Newly defined variables:
W1 = 1 if the 1st warehouse is chosen; 0 if rejected;

W2 = 1 if the 2nd warehouse is chosen; 0 if rejected;

W3 = 1 if the 3rd warehouse is chosen; 0 if rejected.

Multiple-choice constraint reflects the

requirement that exactly 01 of these
warehouses be selected: W1 + W2 + W3 = 1
 DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – MUTUALLY EXCLUSIVE CONSTRAINT

If the company actually has 3 warehouses and it just

wants to expand only one warehouse.
Newly defined variables:
W1 = 1 if the 1st warehouse is chosen; 0 if rejected;

W2 = 1 if the 2nd warehouse is chosen; 0 if rejected;

W3 = 1 if the 3rd warehouse is chosen; 0 if rejected.

Mutually exclusive constraint reflects the

requirement that no more than 01 of these
warehouses be selected: W1 + W2 + W3 1
DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – k OUT OF n ALTERNATIVES CONSTRAINT

Suppose that W1, W2, W3, W4, and W5 represent 05 potential warehouse expansion projects

• If 2 of the five projects must be selected => W1 + W2 + W3 + W4 + W5 2

• If no more than 2 of the five projects must be selected => W1 + W2 + W3 + W4 + W52

DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – CONDITIONAL CONSTRAINT

Conditional constraint: The acceptance of one option is conditional on the acceptance of another.

For instance: the Warehouse expansion project was conditional on the Plant expansion project
FEASIBILITY TABLE

The conditional constraint that enforces the restriction is => W P

Ràng buộc: cái này được chọn hay không là phụ thuộc vào cái kia
DISTRIBUTION AND NETWORK MODELS

MODELING FLEXIBILITY – COREQUISITE CONSTRAINT

Corequisite constraint: Two options are dependent on each other.

For instance: the warehouse expansion project had to be accepted whenever the plant expansion

project was accepted, and vice versa

FEASIBILITY TABLE

The corequisite constraint that enforces the restriction is => W P

 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

CONJOINT ANALYSIS is a market research technique that can be used to learn how prospective
buyers of a product value the product’s attributes.
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

CONJOINT ANALYSIS

In a typical Conjoint Analysis, a sample of consumers are asked to

express their preference for specially prepared pizzas with chosen levels
for the attributes.

PIZZA’S 4 FOUR MOST IMPORTANT ATTRIBUTES LEVEL

Crust (Vỏ) Thin / Thick
Cheese (Phô mai) Mozzarella / Blend
Sauce (Nước sốt) Smooth / Chunky
Sausage flavor (Hương vị xúc xích) Mild / Medium / Hot

Then regression analysis is used to determine the part-worth for each of

the attribute levels. In essence, the part-worth is the utility value that a
consumer attaches to each level of each attribute.
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

CONJOINT ANALYSIS

Salem Foods is planning to enter the pizza market, where 02 existing brands, Antonio and King,
have the major share of the market.

Salem Foods
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

CONJOINT ANALYSIS

Salem Foods is planning to enter the pizza market, where 02 existing brands, Antonio and King,
have the major share of the market.

Salem Foods

Consumer 1’s current favorite pizza is the Antonio’s brand,

which has a thick crust, mozzarella cheese, chunky sauce, and medium-flavored sausage
=> Consumer 1’s utility for the Antonio’s brand pizza is: 2 + 6 + 17 + 27 = 52
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

CONJOINT ANALYSIS

Salem Foods is planning to enter the pizza market, where 02 existing brands, Antonio and King,
have the major share of the market.

Salem Foods

King’s pizza
which has a thin crust, a cheese blend, smooth sauce, and mild-flavored sausage
=> Consumer 1’s utility for the King’s brand pizza is: 11 + 7 + 3 + 26 = 47
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

Assuming the 8 consumers in the current study is representative of the marketplace for pizza,
we create an integer programming model that helps Salem design a pizza,
which have the highest utility for enough people.
* In Marketing literature, the problem being solved is called the share of choice problem.

The decision variables are defined as follows:

lij = 1 if Salem Foods chooses level i for attribute j; 0 otherwise

yk = 1 if consumer k chooses the Salem Foods pizza; 0 otherwise

The number of customers preferring the Salem brand pizza is just the sum of the y k variables,

=> The objective function is: Max (y1 + y2 + . . . + y8)

 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

lij = 1 if Salem Foods chooses level i for attribute j; 0 otherwise

yk = 1 if consumer k chooses the Salem Foods pizza; 0 otherwise

The objective function is: Max (y1 + y2 + . . . + y8)

The 1st consumer’s utility of a particular type of pizza:

Utility for 1st consumer = (11.l11 + 2.l21) + (6.l12 + 7.l22) + (3.l13 + 17.l23) + (26.l14 + 27.l24 + 8.l34)

To succeed with its brand, Salem Foods realizes that it must entice consumers in the marketplace to
switch from their current favourite brand of pizza to the Salem Foods product.
Consumer 1 only purchases the Salem instead of Antonio’s brand pizza if the levels of the attributes
for the Salem are chosen such that:
Utility for 1st consumer = (11.l11 + 2.l21) + (6.l12 + 7.l22) + (3.l13 + 17.l23) + (26.l14 + 27.l24 + 8.l34) > 52
(Consumer 1’s utility for his current favourite Antonio’s brand pizza is: 2 + 6 + 17 + 27 = 52)
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

lij = 1 if Salem Foods chooses level i for attribute j; 0 otherwise

yk = 1 if consumer k chooses the Salem Foods pizza; 0 otherwise

The objective function is: Max (y1 + y2 + . . . + y8)

For instance, y1 = 1 when the 1st consumer prefers the Salem pizza and y1 = 0 otherwise.
Thus, the constraint for 1st consumer:

(11.l11 + 2.l21) + (6.l12 + 7.l22) + (3.l13 + 17.l23) + (26.l14 + 27.l24 + 8.l34) ≥ 1 + 52.y1

Four more constraints must be added, one for each attribute.

l11 + l21 = 1 l12 + l22 = 1 l13 + l23 = 1 l14 + l24 + l34 = 1
 DISTRIBUTION AND NETWORK MODELS

APPLICATIONS 0-1 VARIABLES – PRODUCT DESIGN & MARKET SHARE OPTIMIZATION

The Optimal solution to this ILP:

l11 = 1 l22 = 1 l23 = 1 l14 = 1

y1 = 1 y2 = 1 y6 = 1 y7 = 1

Salem Foods

Salem Food’s pizza

which has a thin crust, a cheese blend, chunky sauce, and mild-flavored sausage