Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

Overview of Lecture 17

▪ Module B
• Assignment Problems
▪ Module D
• Nonlinear Programming Solution Techniques

Lecture 17 1
Assignment Model
▪ Supply at each source is one unit.
▪ Demand at each destination is one unit.
▪ Example
• Four teams of officials are assigned to the four
games
• to minimize the total distance traveled.
Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 210 90 180 160
B 100 70 130 200
C 175 105 140 170
D 80 65 105 120

Lecture 17 2
Opportunity Cost Table
▪ Row Reduction
• Subtract the minimum value in each row.
Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 210 90 180 160
B 100 70 130 200
C 175 105 140 170
D 80 65 105 120

Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 120 0 90 70
B 30 0 60 130
C 70 0 35 65
D 15 0 40 55

Lecture 17 3
Opportunity Cost Table
▪ Column Reduction
• Subtract the minimum value in each column
Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 120 0 90 70
B 30 0 60 130
C 70 0 35 65
D 15 0 40 55

Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 105 0 55 15
B 15 0 25 75
C 55 0 0 10
D 0 0 5 0

Lecture 17 4
Solution Test
▪ Draw the minimum # of horizontal or
vertical lines to cross out all zeros.
Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 105 0 55 15
B 15 0 25 75
C 55 0 0 10
D 0 0 5 0

▪ Three Lines → Only three assignments.


▪ But, we need four assignments.

Lecture 17 5
Next Step
▪ Subtract the minimum value from all values
not crossed out.
▪ Add the minimum value to the cells where
two lines intersect.

Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 105-15 0 55-15 15-15
B 15-15 0 25-15 75-15
C 55 0+15 0 10
D 0 0+15 5 0

Lecture 17 6
Optimal Solution
Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C)
A 90 0 40 0
B 0 0 10 60
C 55 15 0 10
D 0 15 5 0

▪ Four lines are required to cross out zeros.


▪ Solution 1 ▪ Solution 2
• A-> Atlanta • A-> Clemson
• B-> Raleigh • B-> Atlanta
• C->Durham • C->Durham
• D->Clemson • D->Raleigh
• With 450 miles • With 450 miles
▪ Multiple optimal Solutions.
Lecture 17 7
Unbalanced Assignment Problem
▪ Supply > Demand
• Official E is added.
• Dummy game is added with zero distances.

A R

B 𝑐𝑖𝑗
A

C D

D C
0
0 0 0
E 0 Dum.

Lecture 17 8
Unbalanced Assignment Problem

Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C) Dummy
A 210 90 180 160 0
B 100 70 130 200 0
C 175 105 140 170 0
D 80 65 105 120 0
E 95 115 120 100 0

Lecture 17 9
Unbalanced Assignment Problem
▪ Demand > Supply
• One more game is added.
• Dummy official is added with zero distances.

A R

B 𝑐𝑖𝑗
A

C D

D C
0
0 0
0
Dum. 0 E

Lecture 17 10
Unbalanced Assignment Problem

Game Sites
Officials Raleigh (R) Atlanta (A) Durham (D) Clemson (C) E
A 210 90 180 160 95
B 100 70 130 200 115
C 175 105 140 170 120
D 80 65 105 120 100
Dummy 0 0 0 0 0

Lecture 17 11
Nonlinear Programming
▪ Upto now
• All the objective functions and constraints were
linear.

▪ However,
• Many realistic problems have nonlinear relationships.
• A solution technique is more complex and difficult.
• Local optimal vs global optimal.

Lecture 18 12
Nonlinear Profit Analysis
▪ Break-Even Model
• Parameters
• 𝑣: sales volme (i.e., demand),
• 𝑝: price
• 𝑐𝑓 : fixed cost
• 𝑐𝑣 : variable cost
• Revenue: 𝑣𝑝
• Cost: 𝑐𝑓 + 𝑣𝑐𝑣
• Profit: 𝑍 = 𝑣𝑝 − 𝑐𝑓 − 𝑣𝑐𝑣

Lecture 18 13
Nonlinear Profit Analysis

• More realistic if demand is decreasing in price.


• This means that the demand depends on price.
• Assume that
• 𝑣 = 1,500 − 24.6𝑝

1,500

𝑝
60.98

Lecture 18 14
Nonlinear Profit Analysis
▪ Profit with the Demand function
• 𝑍 = 𝑣𝑝 − 𝑐𝑓 − 𝑣𝑐𝑣
= 1,500 − 24.6𝑝 𝑝 − 𝑐𝑓 − 1,500 − 24.6𝑝 𝑐𝑣

• Assume 𝑐𝑓 = 10,000 𝑎𝑛𝑑 𝑐𝑣 = 8, then

Nonlinear or
2
• 𝑍 = −24.6𝑝 + 1,696.8𝑝 − 22,000. Quadratic Function.

A concave function.

Lecture 18 15
Nonlinear Profit Analysis
▪ Derivative of the Curve = Slope
𝜕𝑍
• 𝜕𝑝
= −49.2𝑝 + 1,696.8.
• At the highest point, the slope=0.
𝜕𝑍
• 𝜕𝑝
=0 → 𝑝 = 34.49.
𝑣 = −49.2 ∗ 34.49 + 1,696.8 = 651.6.
𝑍 = 7,259.45
▪ Classical Optimization
• The use of calculus to determine the optimal
solution.

Lecture 18 16
Constrained Optimization
▪ Recall that
• 𝑍 = −24.6𝑝2 + 1,696.8𝑝 − 22,000.
▪ Unconstrained Optimization Model
• A Nonlinear Objective Function with no Constraints.
• Take a derivative, Set it equal to Zero, Solve!

▪ Constrained Optimization Model


• One or More Constraints
• Nonlinear Programming Model

Lecture 18 17
Graph

C
7,259.45 B
A

20 34.49 40

▪ Constraint
• 𝑝 ≤ 20
• Point A is optimal (on the boundary)
▪ Constraint
• 𝑝 ≤ 40
• Solution is not on the boundary. Point C is optimal.

Lecture 18 18
Multiple Constraints
▪ Western Clothing Company
• It produces two products: product 1 (designer jeans)
and product 2 (straight-leg jeans)
• Decision variables:
𝑝1 𝑝𝑟𝑖𝑐𝑒 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 1 𝑎𝑛𝑑 𝑝2(𝑝𝑟𝑖𝑐𝑒 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 2)
• Demands are the functions of prices
• 𝑥1 = 1,500 − 24.6𝑝1 𝑎𝑛𝑑 𝑥2 = 2,700 − 63.8𝑝2
• Maximize 𝑍 = (𝑝1 − 12)𝑥1+ 𝑝2 − 9 𝑥2
Cloth
subject to 2𝑥1 + 2.7𝑥2 ≤ 6,000 𝑦𝑑.
Cutting time
3.6𝑥1 + 2.9𝑥2 ≤ 8,500 𝑚𝑖𝑛
Sewing time
7.2𝑥1 + 8.5𝑥2 ≤ 15,000 𝑚𝑖𝑛

Lecture 18 19
Solution Methods
▪ Two Popular Solution Methods
• Substitution Method
• Method of Lagrange Multipliers

Lecture 18 20
Substitution Method
▪ This is applied to
• Cases with equality constraints.
• By substitution, eliminate the constraint effectively.
• Constrainted model → Unconstrainted model.
▪ Break-Even Model
• 𝑍 = 𝑣𝑝 − 𝑐𝑓 − 𝑣𝑐𝑣
• 𝑣: sales volme (i.e., demand),
• 𝑝: price
• More realistic if demand is decreasing in price.
• 𝑣 = 1,500 − 24.6𝑝

Lecture 18 21
Substitution Method – Break Even Model
▪ New Profit with New Demand
• 𝑍 = 𝑣𝑝 − 𝑐𝑓 − 𝑣𝑐𝑣
= 1,500 − 24.6𝑝 𝑝 − 𝑐𝑓 − 1,500 − 24.6𝑝 𝑐𝑣
• If 𝑐𝑓 = 10,000 𝑎𝑛𝑑 𝑐𝑣 = 8,
• 𝑍 = −24.6𝑝2 + 1,696.8𝑝 − 22,000.
▪ Slope of a curve = Derivative of the Curve
𝜕𝑍
• = −49.2𝑝 + 1,696.8.
𝜕𝑝
• At the highest point, the slope=0.
𝜕𝑍
• =0 → 𝑝 = 34.49.
𝜕𝑝
𝑣 = −49.2 ∗ 34.49 + 1,696.8 = 651.6.
𝑍 = 7,259.45

Lecture 18 22
Substitution Method – Pottery Company
▪ Beaver Creek Pottery Company
• 𝑥1 : 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑜𝑤𝑙𝑠,
• 𝑥2 : 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑚𝑢𝑔𝑠.
▪ Unit profit:
• Cost is increased as production amount is increased.
• Bowl: 4 − 0.1𝑥1
• Mug: 5 − 0.2𝑥2
▪ Total Profit
• 𝑍 = 4 − 0.1𝑥1 𝑥1 + 5 − 0.2𝑥2 𝑥2
= −0.1𝑥12 + 4𝑥1 − 0.2𝑥22 + 5𝑥2

Lecture 18 23
Substitution Method – Pottery Company
▪ Total Profit
• 𝑍 = −0.1𝑥12 + 4𝑥1 − 0.2𝑥22 + 5𝑥2
▪ Labor Constraint
• 𝑥1 + 2𝑥2 = 40 ℎ𝑜𝑢𝑟.
▪ Substitution
• 𝑥1 = 40 − 2𝑥2
2
• 𝑍 = −0.1 40 − 2𝑥2 + 4 40 − 2𝑥2 − 0.2𝑥22 + 5𝑥2
= 13𝑥2 − 0.6𝑥22
𝜕𝑍
• = 13 − 1.2𝑥2
𝜕𝑥2
𝜕𝑍
• =0 → 𝑥2 = 10.8 𝑚𝑢𝑔𝑠 𝑎𝑛𝑑
𝜕𝑥2
𝑥1 = 40 − 2 ∗ 10.8 = 18.4 𝑏𝑜𝑤𝑙𝑠.
𝑍 = $70.42.

Lecture 18 24
Method of Lagrange Multipliers
▪ More General Solution method
• Than the substitution method.
▪ Lagrange Multiplier
• Usually denoted by 𝜆
• Mutiplication of 𝜆 and constraints are subtracted
from the objective function.
▪ Pottery Company
• Maximize 𝑍 = −0.1𝑥12 + 4𝑥1 − 0.2𝑥22 + 5𝑥2
• Subject to 𝑥1 + 2𝑥2 = 40 ℎ𝑜𝑢𝑟.

Lecture 18 25
Method of Lagrange Multipliers - Pottery

▪ Constraint Equation
• 𝑥1 + 2𝑥2 − 40 = 0
▪ Lagrangian Function
• Multiplication of Lagrangeian multiplier
• 𝜆 𝑥1 + 2𝑥2 − 40
• Subtract it from the objective function
• 𝐿 = 4𝑥1 − 0.1𝑥12 + 5𝑥2 − 0.2𝑥22 − 𝜆 𝑥1 + 2𝑥2 − 40 .
▪ Feasible Solution
• L=objective function.

Lecture 18 26
Method of Lagrange Multipliers - Pottery

▪ Lagrangian Function
• 𝐿 = 4𝑥1 − 0.1𝑥12 + 5𝑥2 − 0.2𝑥22 − 𝜆 𝑥1 + 2𝑥2 − 40 .
• Derivative of 𝐿 with respect to three variables of
𝑥1 , 𝑥2, 𝑎𝑛𝑑 𝜆.
𝜕𝐿
• = 4 − 0.2𝑥1 − 𝜆 = 0.
𝜕𝑥1
𝜕𝐿
• = 5 − 0.4𝑥2 − 2𝜆 = 0.
𝜕𝑥2
𝜕𝐿
• = −𝑥1 − 2𝑥2 + 40 = 0.
𝜕𝜆
• Three equations for three variables.
• 𝑥1 = 18.3, 𝑥2 = 10.8, 𝜆 = 0.33 𝑤𝑖𝑡ℎ 𝑍 = $70.42.

Lecture 18 27
Method of Lagrange Multipliers
▪ Meaning of 𝝀
• Analogous to the dual variables
• Change of objective function due to a unit change
of the quantity (resource).
• 40 hours → 41 hours
• 𝑥1 + 2𝑥2 − 41 = 0
• 𝐿 = 4𝑥1 − 0.1𝑥12 + 5𝑥2 − 0.2𝑥22 − 𝜆 𝑥1 + 2𝑥2 − 41 .
• Derive three equations by taking derivatives with
respect to three variables.

Lecture 18 28
Method of Lagrange Multipliers
▪ Meaning of 𝝀
• Old solution with 40 hours
• 𝑥1 = 18.3, 𝑥2 = 10.8, 𝜆 = 0.33 𝑤𝑖𝑡ℎ 𝑍 = $70.42
• New solution with 41 hours
• 𝑥1 = 18.7, 𝑥2 = 11.2, 𝜆 = 0.27 𝑤𝑖𝑡ℎ 𝑍 = $70.75
• 0.33 (70.75-70.42) more profit, which is 𝜆 = 0.33.
• A unit increase in a resource (labor) results in 𝜆 =
0.33 increase in profit
• We are willing to pay $0.33 for one additional hour
of labor.
• Same interpretation as that given for a dual variable.

Lecture 18 29
Method of Lagrange Multipliers
▪ Meaning of 𝝀
• Positive 𝜆
• Optimal objective function value is increasing in quantity
value.
• More profit if more quantity is provided
• Less profit if less quantity is provided
• Negative 𝜆
• Optimal objective function value is decreasing in quantity
value.
• More profit if less quantity is provided
• Less profit if more quantity is provided.

Lecture 18 30
Homework
▪ Module B
• 2, 3, 8, 9, 41, 47
• Submit only problem 2.

▪ Module D
• 7, 14, 15
• Submit only problem 7.

Lecture 17 31

You might also like