DSME 6620

Decision Models and Applications

Lecture Six

Nonlinear Programming

1
DSME 6620
Decision Models and Applications

Agenda

• Definition of nonlinear programming (NLP)

– Local vs. global optimal solutions
– Solving NLP by Excel Solver
– Textile manufacturing problem revisited

• NPL examples
– Location problem
– Production and pricing problem
– Portfolio selection problem

• Summary of nonlinear programming

2
DSME 6620
Decision Models and Applications

Definition of NLP

• A linear function is one that contains only two types of

operations: 1) addition and 2) multiplication between a
constant and a variable.
– All the functions we used in previous lectures are linear
functions
– Example of linear functions:

• A formulation belongs to nonlinear programming as long

as one function in the formulation (either the objective
function or one of the constraint function) is nonlinear.
– Example of nonlinear functions:
DSME 6620
Decision Models and Applications

Local vs. Global Optimal Solutions

• NLP is different from LP and IP (with linear functions) in

the sense that, given a starting point, there is no
guarantee that the true optimal solution can be found.

• Why? Well, there are two types of optimal solutions

defined for NLP:
– Local optimal solution: a solution better than any other
feasible solution in its neighborhood,
– Global optimal solution: the best solution in the entire
feasible region (the true optimal solution).

1800
1580.05
1600

1400

1200

1000

800

600

400

200

0
2 4 6 8 10 12 14 16 18

Maximize f(x) = - 0.25 x4.5 + 10 x3.5 - 137.5 x2.5 + 750 x1.5 - 1000 x0.5
Subject to 2 ≤ x ≤ 18 4
DSME 6620
Decision Models and Applications

Solving NLP by Excel Solver

In general, in Solver, the default setting for NLP works

well . Please ensure that Solver be informed that the
problem under study is a nonlinear programming one:

5
DSME 6620
Decision Models and Applications

Solving NLP by Excel Solver

• Due to the possibility of ending up with a local optimal

solution (instead of a global optimal solution), it is
recommended that we try several starting points when
using Solver for a NLP problem

• A number of Excel functions are discontinous in nature

and thus are nonlinear in essence
– Logical functions, such as IF( ) or AND( )
– Mathematical functions that are discontinuous, such as
ROUND( ) or CEILING( )
– Statistical functions, such as RAND( )

6
DSME 6620
Decision Models and Applications

Textile Manufacturing Problem

Revisited
The Great Threads Company is capable of manufacturing
shirts, shorts, pants, skirts, and jackets. Each type of clothing
requires Great Threads to acquire the appropriate type of
machinery. The machinery needed to manufacture each type of
clothing must be rented at the weekly rates shown below. This
table also lists the amounts of cloth and labor required per unit
of clothing, as well as the sales price and the unit variable cost
for each type of clothing. There are 4000 labor hours and 4500
square yards (sq yd) of cloth available in a given week. The
company wants to find a solution that maximizes its weekly
profit.

Rental Labor Cloth Selling Variable

Cost Hours (sq yd) Price Cost/Unit
1) Shirts $1500 2 3.0 $35 $20
2) Shorts $1200 1 2.5 $40 $10
3) Pants $1600 6 4.0 $65 $25
4) Skirts $1500 4 4.5 $70 $30
5) Jackets $1600 8 5.5 $110 $35
7
DSME 6620
Decision Models and Applications

Textile Manufacturing Problem

Revisited
Analysis & Formulation
Can we formulate the problem by making use of the
Excel function IF?

8
DSME 6620
Decision Models and Applications

Location Problem

The Rappaport Communications Company provides cellular

telephone services in several mid-western states. The
company is planning to expand its customer base by offering
cellular service in northeastern Ohio to the cities of
Cleveland, Akron, Canton, and Youngstown. The company
will install the hardware necessary to serve customers in
each city on pre-existing communications towers in each
city. Locations of these towers are summarized in next slide.

However, the company also needs to construct a new

communication tower somewhere between these cities to
handle inter-city calls. This tower will also allow cellular
calls to be routed onto the satellite system for world-wide
calling service. The tower the company is planning to build
can cover areas within a 40-mile radius. Thus, the tower
needs to be located within 40 miles of each city.

The company aims to minimize the total distance

from the tower to each of the four cities.

9
DSME 6620
Decision Models and Applications

Location Problem

y
50 Cleveland
(5,45)
40

Akron (?, ?) Youngstown

30
(12,21) (52,21)
20

10 Canton
(17,5)
0
0 10 20 30 40 50 60
x

From Geometry: the Euclidean distance

between (x1,y1) and (x2,y2) can be calculated by

 x1  x2    y1  y2 
2 2
D ist a n ce=

10
DSME 6620
Decision Models and Applications

Location Problem

Analysis & Formulation

11
DSME 6620
Decision Models and Applications

Production & Pricing Problem

A company makes two products, product one and product

two, from two resources, materials and labor. At the
beginning of each month, the company sets the production
quantities of the two products and determines their selling
prices.

A unit of product one costs $50 to make, while a unit of

product two costs $20. The utilization of materials and labor
and the available quantities of resources are shown in the
table I.

The company believes that either product’s demand depends

on its selling price and is estimated to be linearly decreasing
in that price, plus some noise. The company will use the
latest ten months’ information on pricing and demand
(shown in Table II) to estimate that relationship.

Assume produced units in a month can be used to meet the

demand of that month. Leftovers at the end of a month will
be wasted, hence the company won’t want to produce more
than the estimated demand for either product.

12
DSME 6620
Decision Models and Applications

Production & Pricing Problem

Table I: Utilization and availability of resources
Materials (lbs) Labor (hour)
Product One (per unit) 8 4
Product Two (per unit) 2 2
Availability 20,000 12,000

Table II: Price and demand during latest ten months

Product One Product Two
Month Price Demand Price Demand
1 284 2339 130 1172
2 287 2301 143 1052
3 277 2385 124 1302
4 285 2342 132 1116
5 293 2203 138 1105
6 298 2214 133 1097
7 303 2215 118 1353
8 272 2475 127 1271
9 308 2148 125 1199
10 277 2384 123 1283
13
DSME 6620
Decision Models and Applications

Production & Pricing Problem

Analysis & Formulation

14
DSME 6620
Decision Models and Applications

Portfolio Selection Problem

Harry Markowitz
Financial economist
Nobel Prize recipient

• Portfolio selection problem - Given a set of

investments (such as stocks, gold, and T-bills), how
do financial analysts determine the portfolios that
have the lowest risk with a high return (or the
highest return with a low risk)?

• This question was answered by Harry Markowitz in

1950s. For his work on this and other investment
portfolio topics, he received Nobel Prize in
Economics in 1991.

• Portfolio selection is the basis of most methods of

asset allocation used by the Wall Street.

15
DSME 6620
Decision Models and Applications

Portifolio Selection Problem

Ray Dodson is an independent financial advisor. He recently

met with a new client, Paula Ribben, who wanted Ray’s
advice on how best to diversify her investments. Paula has
invested a good portion of her retirement savings in the stock
of IBC. During the past 12 years, this stock has produced an
average annual return of 7.64% with a variance of
approximately 0.0028. Paula would like to earn more on her
investment, but is very cautious and doesn’t like to take risks.
She has asked Ray to recommend a portfolio of investments
that would provide at least 12% average return with as little
additional risk as possible. After some research Ray identified
two additional stocks, from NMC and NBS, that he believes
could help meet Paula’s investment objectives. The
performance of the stocks are given in the table of next slide.

Ray wants to determine what percentage of Paula’s funds

should be allocated to each of the stocks in order to achieve
an overall expected return of 12% while minimizing the
variance of the total return on the portfolio.

16
DSME 6620
Decision Models and Applications

Portifolio Selection Problem

Year IBC NMC NBS

1 11.2% 8.0% 10.9%
2 10.8% 9.2% 22.0%
3 11.6% 6.6% 37.9%
4 -1.6% 18.5% -11.8%
5 -4.1% 7.4% 12.9%
6 8.6% 13.0% -7.5%
7 6.8% 22.0% 9.3%
8 11.9% 14.0% 48.7%
9 12.0% 20.5% -1.9%
10 8.3% 14.0% 19.1%
11 6.0% 19.0% -3.4%
12 10.2% 9.0% 43.0%

17
DSME 6620
Decision Models and Applications

Portifolio Selection Problem

IBC

Average return = 7.64%

NMC

Average return = 13.43%

NBS

Average return = 14.93%

18
DSME 6620
Decision Models and Applications

Portifolio Selection Problem

Analysis & Formulation

19
DSME 6620
Decision Models and Applications

Summary of NLP

• In a NLP problem, nonlinearity exists either in the

objective function, or some of the constraint
functions.
– There exist two types of optimal solutions in a NLP
problem, local optimal solutions and global optimal
solutions.

• The outcome of solving a NLP may depend on the

starting point that we choose.
– It is recommended that we try several starting points
when using Solver for a NLP problem.
– On the other hand, for LP or IP (with linear
functions), regardless of the starting point, we can
always get the optimal solution.

• We may read Chapter 6 of the textbook to reinforce

our understanding of NLP.

20