RSM 270 Operations Management

Linear Programming

RSM 270 / Linear Programming I 1

 Previous Class (# 9): Inventory Management Policies
» Combine EOQ and Newsvendor models to obtain good (R,Q) policies for fixed order quantity (Q)
models.
» Use Newsvendor model to obtain good order up to policies for fixed review period (P) models.
» Key Concepts: Fixed-time versus fixed-quantity, service-level, inventory position, pipeline inventory,
cycle stock, lead time, review period, exposure period, order-up-to level.

Today’s Class (# 10): Linear Programming I

» Linear programming (LP) terminology and formulation.
» Solving an LP graphically when there are only 2 variables by identifying the feasible regions and
finding corner points
» Use “Solver” from Excel to solve LP’s with more than 2 variables
» Key Concepts: Objective function, decision variables, constraints, feasibility, corner points, optimal
solution, shadow price, sensitivity analysis, binding constraint.

RSM 270 / Linear Programming I 2

 Logistics
 Assignments
– Assignment 6 has been posted. It is due on Sunday, March 31st, 11:59 pm.
– If you have any questions regarding how to submit your assignments, please contact Hamid at
[email protected].

 Case 2 has been released and will be due on April 7th, 11:59pm via Quercus.
– Working in groups is optional.
– All groups from Case 1 have been dissolved. They do not carry forward
automatically.
– You must actively reach out/confirm your group members for Case 2.
– If you work in a group, you may work in groups of up to five.
– Eligible group members are anyone in the Thursday sections of RSM270
(Saunders).
RSM 270 / Linear Programming I 3
 Steps in Developing a LP Model

1) Problem Formulation

2) Solution Techniques

3) Interpretation of Results

RSM 270 / Linear Programming I 4

 Steps in Developing a LP Model

1) Problem Formulation

2) Solution Techniques

3) Interpretation of Results

RSM 270 / Linear Programming I 5

 Step 1: Problem Formulation

There are three components to a LP model:

1. Decision variables

2. Objective Function
» Profit or Cost Parameters
» Maximization/Minimization

3. Constraints
» Constraint Parameters
» Right-Hand-Side Constants

RSM 270 / Linear Programming I 6

 Step 1: Problem Formulation

There are three components to an LP model:

1. Decision variables

2. Objective Function
» Profit or Cost Parameters
» Maximization/Minimization

3. Constraints
» Constraint Parameters
» Right-Hand-Side Constants

RSM 270 / Linear Programming I 7

 Step 1: Problem Formulation

There are three components to an LP model:

1. Decision variables

2. Objective Function
» Profit or Cost Parameters
» Maximization/Minimization

3. Constraints
» Constraint Parameters
» Right-Hand-Side Constants
RSM 270 / Linear Programming I 8
 Step 1: Problem Formulation

There are three components to an LP model:

1. Decision variables

2. Objective Function
» Profit or Cost Parameters
» Maximization/Minimization

3. Constraints
» Constraint Parameters
» Right-Hand-Side Constants
RSM 270 / Linear Programming I 9
 Step 1: Problem Formulation

The objective function and each constraint are linear functions of

the decision variables

Objective function: Maximize or Minimize

Constraints: “³” or “£” or “=”

At least At most Exactly equal

RSM 270 / Linear Programming I 10

 Example LP Formulation Flair Furniture

 Two products: Tables and Chairs

 Decision: How many units of each product to produce each month?

 Objective: Maximize profit

 Constraints: Limited resources

– Labor hours available
– Raw material available
– Management/Market restrictions

RSM 270 / Linear Programming I 11

 Flair Furniture Co. (FF) Data

Tables Chairs
(per table) (per chair)
Unit Profit $7 $5 Hours Available
Carpentry 3 hrs 4 hrs 2400
Painting 2 hrs 1 hr 1000

Other Management/Market Limitations:

• Make no more than (at most) 450 chairs
• Make no less than (at least) 100 tables

RSM 270 / Linear Programming I 12

 Example LP Formulation: Flair Furniture Continued

Decision Variables:
Let T = Number of tables to make each month
C = Number of chairs to make each month

Objective Function (OF): Maximize total profit

Maximize 7T + 5C

RSM 270 / Linear Programming I 13

 Example LP Formulation: Flair Furniture Continued

Constraints:

 FF has 2400 hours of carpentry time available

3 T + 4 C < 2400 (hours)

 FF has 1000 hours of painting time available

2 T + 1 C < 1000 (hours)

RSM 270 / Linear Programming I 14

 Example LP Formulation: Flair Furniture Continued

Constraints:

 Maximum production of 450 Chairs per month

C < 450

 Minimum production of 100 Tables per month

T > 100

 Non-negativity: no negative production

T > 0, C > 0

RSM 270 / Linear Programming I 15

 Example LP Formulation: Flair Furniture Summary

Objective: Maximize Z = 7T + 5C (profit)

Subject to: 3T + 4C < 2400 (carpentry hrs.)

2T + 1C < 1000 (painting hrs.)
C < 450 (max no. of chairs)
T > 100 (min no. of tables)
T> 0, C > 0 (non-negativity)

RSM 270 / Linear Programming I 16

 Steps in Developing a LP Model

1) Problem Formulation

2) Solution Techniques

3) Interpretation of Results

RSM 270 / Linear Programming I 17

 Step 2: Solution Techniques
There are many different ways to solve a linear program:

- Graphical Solutions
- Microsoft Excel (Solver)
- Simplex Method
- Karamarkar’s Algorithm
- IBM CPLEX
- Gurobi
- MATLAB
- Mathematica

RSM 270 / Linear Programming I 18

 Step 2: Solution Techniques (Graphical)

 Graphing an LP model helps provide insight into LP models and their

solutions

 Graphical methods are typically restricted to only two dimensions

(i.e., two decision variables)

 Note that the same properties with two decision variables apply to
all LP models and solutions with n decision variables!

RSM 270 / Linear Programming I 19

 Example Graphic Solution: Flair Furniture Continued

The x-y axis represents each decision variable!

C

0 T
RSM 270 / Linear Programming I 20
 Example Graphic Solution: Flair Furniture Continued

Carpentry
Constraint Line: C

3T + 4C = 2400 6𝟎𝟎 =
𝟐𝟒𝟎𝟎
𝟒 Infeasible
600
(i.e., 3T + 4C < 2400) 3T
> 2400 hrs

+4
C=
24
00
Intercepts:
Feasible
(T = 0, C = 600) Region
(T = 800, C = 0) 0
0 800 T
𝟐𝟒𝟎𝟎
𝟖𝟎𝟎 =
𝟑
RSM 270 / Linear Programming I 21
 Example Graphic Solution: Flair Furniture Continued
C
Painting 1000

Constraint Line:
2T + 1C = 1000

2T
+1
(i.e., 2T + 1C < 1000) 600

C=
100
0
Intercepts:
(T = 0, C = 1000) Feasible
Region
(T = 500, C = 0)
0
0 500 800 T
RSM 270 / Linear Programming I 22
 Example Graphic Solution: Flair Furniture Continued
C
1000
Max Chair Line:
C = 450

(i.e., C < 450) 600

450

Min Table Line:

T = 100 Feasible
Region
(i.e., T > 100) 0
0 100 500 800 T
RSM 270 / Linear Programming I 23
 Example Graphic Solution: Flair Furniture Continued
C

Objective Function Line: 500

Optimal Point
(T = 320, C = 360)
Profit
400
=

7T
+5
C=
300
7T + 5C

$2
7T

7T
,8 0

+5
+5

0
Optimal point satisfies all constraints of the model.

C=
C=
200

$4
$2
To see whether this is an accurate statement, you

,0 4
,1 0
can go ahead and plug in the optimal point in each

0
0
constraint of the model.
100 Feasible
Region

RSM 270 / Linear Programming I

0 0 100 200 300 400 500 T 24
 Example Graphic Solution: Flair Furniture Continued
C
Additional Constraint: 500 New optimal point
T = 300, C = 375
Need at least 75 more
chairs than tables 400 Feasible
T = 320
Region
C > T + 75 C = 360
300 No longer
or feasible

200
C – T > 75

100
Left Hand Side (LHS) Right Hand Side (RHS)

RSM 270 / Linear Programming I

0 0 100 200 300 400 500 T 25
 Main Insights and Definitions

 Feasible solution: A point that satisfies all the constraints in the

problem.
 The set of all feasible solutions is called the feasible set or
feasible region. It is always polygonal.
 Corner Point Property: An optimal solution to an LP model must
lie at one or more corner points.
 Optimal point (region): The corner point or region with the best
objective function value
– (i.e., the largest for a maximization problem; the smallest for a
minimization problem)

RSM 270 / Linear Programming I 26

 Main Insights and Definitions

 A constraint is said to be binding or active if it is satisfied with

equality at the optimal solution. That is LHS = RHS.
– Geometrically, an active constraint is one that passes through the
optimal solution.

 A constraint is said to be inactive or nonbinding if it is satisfied

with strict inequality at the optimal solution.
– Geometrically, an inactive constraint is one that does not pass through
the optimal solution.

RSM 270 / Linear Programming I 27

 Special Situations (next class)

 Redundant Constraints

 Alternate Optimal Solutions

 Infeasible Solution

 Unbounded Optimal Directions

RSM 270 / Linear Programming I 28
 Example Graphic Solution: Flair Furniture Continued
C
500

Corner
400 Feasible Points
Region
300

200

Optimal
100 Point

0
0 100 200 300 400 500 T
RSM 270 / Linear Programming I 29
 Example Excel Solution: Flair Furniture Continued

Use Solver from Microsoft Excel

RSM 270 / Linear Programming I 30

 Screenshot of Flair Furniture Model

RSM 270 / Linear Programming I 31

 Screenshot of Solver Options
Solver Entries

RSM 270 / Linear Programming I 32

 Screenshot of Flair Furniture Solution

Which constraints are binding?

RSM 270 / Linear Programming I 33
 Steps in Developing a LP Model

1) Problem Formulation

2) Solution Techniques

3) Interpretation of Results

RSM 270 / Linear Programming I 34

 Step 3: Interpretation of Results

 What is the model telling you to do? What actions should you
take to:
– Minimize your costs? Maximize your profits?
– Read and interpret the MS Excel Answer Report

 Why does this make sense?

 Sensitivity and Economic Analysis
– How is the optimal solution affected by changes and estimation errors
in the problem data?
– Read and interpret the MS Excel Sensitivity Report

RSM 270 / Linear Programming I 35

 Answer and Sensitivity Reports

RSM 270 / Linear Programming I 36

 Flair Furniture Continued: Answer Report
Objective Cell (Max)
Cell Name Original Value Final Value
$D$6 Profit Total Profit = 0 4040

Variable Cells
Cell Name Original Value Final Value Integer
$B$5 Number of units Tables 0 320 Contin
$C$5 Number of units Chairs 0 360 Contin

Slack value for each constraint is the

difference between the LHS and RHS.
Constraints
Cell Name Cell Value Formula Status Slack
$D$11 Minimum chairs LHS 320 $D$11>=$F$11 Not Binding 220
$D$8 Carpentary hours LHS 2400 $D$8<=$F$8 Binding 0
$D$9 Painting hours LHS 1000 $D$9<=$F$9 Binding 0
$D$10 Maximum chairs LHS 360 $D$10<=$F$10 Not Binding 90
RSM 270 / Linear Programming I 37
 Shadow Prices

 Each constraint has an associated shadow price.

 The shadow price is the change in the optimal objective value
per unit increase in the right-hand side of the constraint, given
that all other data remain the same.

 Shadow prices worthy of consideration are the ones whose slack

variables are zero at optimality. That is,
– the constraint is binding because RHS – LHS = 0

RSM 270 / Linear Programming I 38

 Flair Furniture Continued: Sensitivity Report
Microsoft Excel 15.0 Sensitivity Report
Worksheet: [Solver Example - Flair Furniture.xlsx]Sheet1

Range Information
Shadow Price Information
Variable Cells
Final Reduced Objective Allowable Allowable
Cell Name Value Cost Coefficient Increase Decrease
$B$5 Number of Units Tables 320 0 7 3 3.25
$C$5 Number of Units Chairs 360 0 5 4.333333333 1.5

Constraints
Final Shadow Constraint Allowable Allowable
Cell Name Value Price R.H. Side Increase Decrease
$D$11 Minimum tables 320 0 100 220 1E+30
$D$8 Carpentry hours 2400 0.6 2400 225 900
$D$9 Painting hours 1000 2.6 1000 600 150
$D$10 Maximum chairs 360 0 450 1E+30 90
RSM 270 / Linear Programming I 39
 Flair Furniture Continued: Sensitivity Analysis Questions
 Flair Furniture is offered 100 more painting hours at a cost of $250. Should they take
the deal?
– Yes. The objective is maximization and 100 * 2.6 (profit) > 250 (cost)

 Flair Furniture is offered 100 more carpentry hours at a cost of $250. Should they
take the deal?
– No. The objective is maximization and 100 * 0.6 (profit) < 250 (cost).

 What would be the impact of decreasing the minimum number of tables in 50 units?
– No impact on the optimal solution

 What would be the impact of increasing the maximum number of chairs in 50 units?
– No impact on the optimal solution
RSM 270 / Linear Programming I 40
 General Principles on Shadow Prices

 The unit of the shadow price is the unit of the objective function divided by the
unit of the constraint:
D (optimal objective function value)
Shadow Price =
D (RHS value)

 In terms of Microeconomic Theory, the Shadow Price of a given constraint is the

Marginal Value of the resource whose units are expressed in the constraint.
 Shadow prices values are valid in a range. Outside of that range, it is necessary to
resolve the LP.
 The shadow price for any non-binding constraint will be zero, i.e., LHS – RHS = 0

RSM 270 / Linear Programming I 41

 We Can Extend this to More Variables!

42
RSM 270 / Linear Programming I 42
 Linear Programming Applications

 Product mix
 Make-buy
 Media selection
 Marketing research
 Portfolio allocation
 Market microstructure
 Shipping & transportation
 Multi-period scheduling

RSM 270 / Linear Programming I 43

 Summary: Linear Programming

1) Define the decision variables

2) Formulate LP using the decision variables
• Write the objective function equation
• Write each of the constraints equations

3) Implement the model (e.g., in Excel)

4) Solve (e.g., graphically or Excel’s Solver)

5) Interpret your results and do sensitivity analysis
RSM 270 / Linear Programming I 44
 Readings & More…

 For a better grasp of the key concepts, the following readings are
only recommended:
– Appendix A of Jacobs and Chase.

 Next Class: Linear Programming II

RSM 270 / Linear Programming I 45