Engineering Programming

in MATLAB : A Primer
February, 2000

Mark A. Austin
Institute for Systems Research,
University of Maryland,
College Park,
Maryland 20742, U.S.A.

c
Copyright 2000
Mark A. Austin. All rights reserved. These notes may not be reproduced
without expressed written permission of Mark Austin.

Engineering Programming in Matlab

Contents

Introduction to Engineering Computations

1 Introduction to Engineering Computations

1.1 Applications of Computers in Engineering
1.2 Recent Advances in Computing
Advances in Computing Since 1970
1.3 Computer Hardware Concepts
Hardware Components in a Simple Computer
1.4 Operating System Concepts
1.5 Computer Networking Concepts
Client-Server Network Architectures
The Internet
Internet Access
Protocols for Internet Communication
Internet Domain Names and Addresses
Internet Services
The World Wide Web
1.6 Hardware-Software Life Cycle
1.7 Principles of Engineering Software Design
Models of Software Systems Development
Components of Software Systems Development
Modular Program Development
Abstraction
Top-down and Bottom-up Software Design
1.8 Computer Programming Language Concepts
High- and Low- Level Computer Languages
Compiled and Interpreted Programming Languages
Procedural and Object-Oriented Programming Languages
0

3
6
6
7
7
9
9
12
12
13
14
15
15
16
17
18
22
24
25
27
30
31
32
33
33
34
35

The Syllabus

1.9 When to Program in MATLAB?

1.10 Review Questions
1.11 Review Exercises

II

MATLAB Programming Tutorial

2 Introduction to MATLAB
2.1 Getting Started
2.2 Professional and Student Versions of MATLAB
Entering and Leaving MATLAB
Online help
2.3 Variables and Variable Arithmetic
Defining Variables
Arithmetic Expressions
Numerical Precision of MATLAB Output
Built-In Mathematical Functions
Program Input and Output
2.4 Matrices and Matrix Arithmetic
Definition and Properties of Small Matrices
Reading and Saving Datasets
Application of Mathematical Functions to Matrices
Colon Notation
Submatrices
Matrix Arithmetic
Matrix Element-Level Operations
2.5 Control Structures
Logical Expressions
Selection Constructs
Looping Constructs
2.6 General-Purpose Matrix Functions
Sorting the Contents of a Matrix
Summation of Matrix Contents
Maximum/Minimum Matrix Contents
Random Numbers
2.7 Program Development with M-Files
User-Defined Code and Software Libraries
Program Development Cycle
Script M-Files

1
35
36
37

38
41
41
42
42
43
44
44
46
49
50
52
54
55
60
61
62
63
64
70
72
72
75
76
78
78
79
79
80
82
82
83
86

Engineering Programming in Matlab

Function M-Files
Handling Name Conflicts
2.8 Engineering Applications
2.9 Review Questions
2.10 Programming Exercises

86
92
93
112
115

3 MATLAB Graphics
3.1 Simple Two-Dimensional Plotting
Histograms, Bar Charts, and Stem Diagrams
Multiple Plots
3.2 Three-Dimensional Plots
3.3 Mesh and Surface Plotting
3.4 Contour Plots
3.5 Subplots
3.6 Hard Copies of MATLAB Graphics
3.7 Preparing MATLAB Graphics for the World Wide Web
3.8 Review Questions
3.9 Programming Exercises

124
124
130
132
134
135
138
139
142
143
143
144

4 Solution of Linear Matrix Equations

4.1 Definition of Linear Matrix Equations
Geometry of Two- and Three-dimensional Systems
4.2 Hand Calculation Procedures
4.3 Types of Solutions for Systems of Linear Matrix Equations
4.4 Case Study Problem : Three Linear Matrix Equations
4.5 Singular Systems of Matrix Equations
4.6 Engineering Applications
Structural Analysis of a Cantilever Truss
Analysis of an Electrical Circuit
Least Squares Analysis of Experimental Data
Distribution of Temperature in Chimney Cross-Section
4.7 Review Questions
4.8 Programming Exercises

149
149
150
151
152
154
156
157
157
163
166
172
181
182

Part I

Introduction to Engineering
Computations

Chapter 0

Introduction
This text begins with a tutorial describing the concepts on which modern engineering computations are built. In our experience, students are much better prepared to learn a new programming
language if they are already familiar with these basic concepts.
After briefly explaining the range of application programs that are found in engineering organizations, Chapter 1 quickly reviews the major contributions of computer technology over the past
thirty years. This historical review helps us to see where and how technology has evolved, and provides perspective for where computing and computer technologies are likely to head in the next five
to ten years. We then examine the hardware components in a simple computer, the components and
purposes of a simple operating system, and the role computer networks are playing in modern-day
applications of engineering computing. The latter includes introductions to client/server computing,
the Internet, and the World Wide Web (WWW).
Chapter 2 introduces the principles upon which modern engineering software systems are
built. Topics include the hardware-software life cycle, the economics of software development,
top-down and bottom-up development strategies, software modularity, and information hiding. We
conclude this chapter with an introduction to programming language concepts, including high- and
low-level programming languages, compiled versus interpreted languages, scripting and markup
languages, and so forth.

Chapter 1

Introduction to Engineering
Computations

1.1

Applications of Computers in Engineering

During that past three decades, remarkable advances have occurred in the processing speed
of computers, the capacity of computers to store, manipulate and present large quantities of data
and information, and the ability of computers to communicate with other computers over networks.
Evidence of these advances can be found in present-day engineering offices where computers are
used in at least four broad capacities:
1. For storage and manipulation of data and information. Modern databases can store and
manipulate a variety of data and information, including commercial off-the-shelf products,
materials, and services; experimental data; the results of a numerical computations; models
of designs, design documents and drawings; Geographic Information Systems (GIS) imagery;
and so forth.
2. For communication over computer networks. Networking tools and technologies allow for
the exchange of data and information over networks, and for computers to jointly contribute
to the solution of large engineering analyses. Perhaps the greatest use of computer networks
is for communication via E-mail.
3. For desktop publishing. Word processing packages such as LaTeX and Microsoft Word, and
picture editors such as Corel Draw and Photoshop enhance an engineers ability to write and
edit publications.
4. For numerical and symbolic computations. Engineering analysis programs (e.g., programs
for control systems and finite element analysis; MATLAB and Mathematica) are needed for
the solution of engineering problems. The majority of engineers use commercial software for
numerical and symbolic calculations, requiring preparation/programming of input files while
some engineers will write their own software.
6

Chapter 1

From a business point of view, the most useful application programs will directly improve the performance and reliability, productivity, and economic competitiveness of engineering systems development. The participating application programs should be fast and accurate, flexible, reliable, and
of course, easy to use. And they should work together. A good example of the last requirement
can be found in modern-day computer-aided design (CAD) systems where engineering analysis
programs are integrated with project management tools, databases of project requirements, organizational resources, and commercial off-the-shelf products, materials, and services.
An unfortunate problem caused by these advances is the gap many engineers are finding
between their knowledge of these technologies and the opportunities they afford. Solutions to this
problem are complicated by the large number of activities in which engineers participate and the
inability of many present-day engineering application programs to operate across a variety of hardware platforms and operating systems. Keeping up-to-date with computational technologies is really
a lifelong endeavor because some of the application tools and computer programming languages we
will use in five to ten years are only just being invented.

1.2

Recent Advances in Computing

A good way of beginning to understand where computers and programming languages

might be headed in the near future, is to take a look at where they have come from in the recent past.
We therefore begin this section with a little history.

Advances in Computing Since 1970

For more than a decade now, computers have been providing approximately 25% more
power per dollar per year. Together with the aforementioned advances in technology and market
driven forces, these changes have stimulated the exploration of many new ideas and paradigms.
Figure 1.1 summarizes, for example, the major modes of operation and key technologies for
computing versus decade for the past 30 to 35 years (this diagram has been adapted from an article
in Scientific American [16]). The highlights are:
1970s : In the early to mid 1970s, mainframe computers were commonplace. They had a computational speed of 1 to 2 MIPS (millions of instructions per second) and were largely viewed
as machines for research engineers and scientists. Compared to todays standards, computer
memory was very expensive, and human-computer interaction was primitive. In fact, scientists and engineers interacted with a computer by sitting at a terminal and typing commands
on a keyboard. The computer would respond by sending text to the terminal screen.

Engineering Programming in Matlab

BATCH

TIME-SHARING

DESKTOP

NETWORKS

Decade
1960s
1970s
1980s
1990s -==============================================================================
Technology

Medium-Scale
Integration

Large-Scale
Integration

Very-Large
Scale
Integration

Ultra-Scale
Integration

Users

Experts

Specialists

Individuals

Groups.

Objective

Calculate

Access

Present

Communicate

Location

Computer
Room

Terminal
Room

Desktop

Mobile.

Connectivity

Peripherals

Terminals

Desktops

Laptops
Palmtops

User
Activity

Punch and Try

(computer cards)

Remember and
Type
(interact)

See and Point

(drive)

Ask and Tell

(delegate)

Data

Alpha-numeric

Text
Vector

Fonts
Graphs

Sound
Video

Languages

Cobol, FORTRAN

PL/1, Basic

Pascal, C
SQL

C++, Java
Perl, Tcl/Tk
HTML, VRML
==============================================================================

Figure 1.1. Paradigms of computing versus decade

Most software developers wrote computer programs dedicated to a specific task (e.g., finite element analysis; control systems package; an accounting or stock control system). Many of
these packages were written in FORTRAN and run in batch mode. The ease with which
FORTRAN could be used to evaluate mathematical formulae hence the name For-mula
Tran-slation was adequate for most engineering applications, and it can reasonably be argued that the choice of language was suitable for its time (unfortunately, some engineers and
educators still think it is adequate).
1980s : In the 1980s desktop publishing systems were developed for individuals at work and home.
Computers such as the Macintosh presented users with a screen containing window, scroll bar,
and button/icon interface components, that could be defined and manipulated with a mouse

Chapter 1

device by simply pointing and clicking icons on the graphical interface.

1990s : The use of computers in engineering is now at a point where mainframe computers are being replaced by high speed engineering workstations and modern Personal Computers (PCs)
having bit-mapped graphics (a bit is short for binary digit, either a 1 or 0), global network connectivity (e.g., cellular communications; fiber optic; the World Wide Web), and multimedia
(i.e., two or more of the following; graphics, voice, digital sound, video) [13]. Laptop computers now provide mobility, without compromising too many of the computational features
available on PCs.
It is important to note that computers once viewed as a tool for computation alone, are now seen
as an indispensable tool for computations and mobile communications. Access to this information
has expanded from experts in the 1970s to groups of individuals today. A whole host of new
programming languages, operating systems, and application programs have been written to support
the new modes of functionality and day-to-day operation enabled by these technology advances.
Consider, for example, an engineer who has access to a high speed PC with multimedia interfaces
and global network connectivity, and who happens to be part of a geographically dispersed development team. The team members can use the Internet/E-mail for day-to-day communications, to share
project information among team members, to conduct engineering analyses at remote sites, and to
access information from databases on project components, materials, and services. Online assembly
of joint ventures, online bidding of projects, and online verification of project performance against
design standards may become commonplace in the near future.

1.3

Computer Hardware Concepts

The three main components in a computer are the hardware (including the computer networks connecting individual computers), the operating system, and the application programs that
operate on individual computers and across computer networks. In our opinion, programmers
should have a basic understanding of the hardware and operating system components in a simple
computer because many computer programs are written to interact with a computers input/ouput
(I/O) devices and its operating system.

Hardware Components in a Simple Computer

Figure 1.2 is a schematic of the main hardware components in a typical (simplified) personal computer. Viewed from a high-level of abstraction, a computer is an assembly of processor,
memory, and input/output (I/O) modules. A particular computer may have one or more modules of
each type, with these modules being connected in some way to produce the main function of the
computer.

10

Engineering Programming in Matlab

VIDEO CIRUITRY AND MEMORY.

BUSSES.

CENTRAL PROCESSING UNIT (C.P.U.)

DATA

INSTRUCTIONS

DATA

INSTRUCTIONS

DATA

DATA

BUSSES.
RANDOM ACCESS
MEMORY (R.A.M.)

INPUT AND OUTPUT

MASS STORAGE

Serial Ports....

Hard Disk..

DATA

DATA

CD-ROM.

DATA

INSTRUCTIONS

Parallel Ports...
Keyboard

Floppy Disk.

Sound....

DVD

DATA

BUSSES.

READ ONLY MEMORY

(R.O.M.)

Figure 1.2. Hardware model of a personal computer

In Chapter 2 we will see that a computer program is nothing other than a list of instructions
that can be followed mechanically by the computer. Machine instructions are expressed as binary
numbers that is, information represented as a sequence of zeros and ones. A computer can execute
a program only if it is expressed in a machine language that can drive the mechanical operations of
the computer.
The central processing unit (CPU) is the engine that controls the operations of the computer by executing instructions. In a conventional CPU, instructions are fetched from main memory,
decoded, and executed one at a time. This process is entirely mechanical, and so if a computer program is to execute without error, the machine code instructions must be complete and unambiguous
in their intent. Computers are designed so that they can be easily connected to external devices such
as printers and keyboards. A second purpose of the CPU is to handle interrupts. When a device
sends an interrupt signal to the CPU, it will halt what it is doing, take care of the device request,
and then resume the original computation. The ability of a CPU to work on the solution of multiple
tasks is called multitasking.
Within the CPU the control unit controls the fetch, decode, and execute cycles for instructions stored in memory. The arithmetic control unit carries out arithmetic and logical operations
on words of data. A word of data is as long as the length of the hardware register in bits. The word
length of a computer refers to the size of the unit of data in bits that a CPU can process at a time.
Computers with a large word length process data faster than computers with a small word length.

11

Chapter 1
Term
Abbreviation
Number of Bytes/Bits
=============================================================
Byte
B
1 Byte = 8 Bits
Kilobyte
KB
1024 = 210 Bytes
Megabyte
MB
1,048,576 = 220 Bytes
Gigabyte
GB
1,073,741,824 = 230 Bytes
Terabyte
TB
240 Bytes
=============================================================

Table 1.1. Terms used to quantify storage

The first processor had 4-bit word lengths. Currently, 32-bit PCs and engineering workstations are
commonplace. 64-bit computers will dominate the marketplace in a few years.
The internal storage unit (not to be confused with the computers primary memory) is fast
internal memory that temporarily stores and manipulates data. It also contains busses (i.e., wires)
for communication of the CPU with I/O devices, and mass storage known as random access memory
(RAM) and read only memory (ROM).
The speed of a CPU is closely linked to the size of computer chips from which it is constructed. Broadly speaking, the more transistors a chip has, the more information it can process.
State of the art manufacturing processes in 1997 allowed for chips having a miniaturization of
1/290-th of a human hair-width. The result, Pentium processor technology, has a maximum speed
of 200 to 230 millions of cycles per second (i.e., 200 to 230 MHz). Pentium Pro chips (manufactured at miniaturization of 1/400-th of a human hair-width) will be able to run at speeds of up to
500 MHz. Predictions are that by the year 2001, next-generation Intel chips will contain 64 bit
processing and operate at 1000 MHz.
The primary memory in a computer, called RAM (an acronym for random access memory, stores data and low-level program instructions as sequences of binary digits. Present day PCs
and engineering workstations have 16 to 128 Megabytes (MB) of RAM (see Table 1.1 for a definition of terms) plus some read only memory (ROM). Mass storage is where programs, data, images
and so forth, are permanently stored. For example, standard CD-ROMs store up to 650 MB, which
is enough memory for approximately 70 minutes of audio/music. The new digital video disc (DVD)
format, also sometimes called digital versatile disc, will be able to store up to 12 times as much data
(8.5 GB using both sides of the disk). This is enough capacity to hold a full-length feature film with
Dolby multichannel digital audio.
A bus is an electronic pathway in a digital computer that provides a communication path
for data to flow between the CPU and its memory, and between the CPU and peripheral devices
connected to the computer (e.g., monitor, printer, keyboard and mouse, network interface). Com-

12

Engineering Programming in Matlab

puter systems are designed so that they can be easily expanded by adding new devices. When a new
device is added to a computer, a software package known as a device driver must be installed so
that the CPU can communicate with the new device. In Figure 1.2, busses provide multiline paths
for rapid data transfer between different sections of the main computer board.

1.4

Operating System Concepts

An operating system is the set of programs that provides an interface between the computer hardware and the computer users. The operating system manages the sharing of a computers
resources and its memory contents, the low-level details of loading and executing programs, file
storage and retrieval, assignment of processes to the screen and keyboard I/O devices, and communication with other computers. The term operating system kernel describes the set of programs in
an operating system that implements the most primitive of that systems functions, including those
for process management, memory management, basic I/O control, and security. Together these operating system features give the computer much of its functionality, including an environment for
writing and running programs. The operating system and its components are also the first programs
to run when the computer is turned on.
Two of the most popular operating systems are UNIX and WINDOWS 95 (tm), and fortunately, there are many similarities in their basic design and functionality. When you are just starting
to learn how a new operating system works, understanding how the file system works is the most important (and potentially confusing) first step. In UNIX and WINDOWS 95, for example, you need
to know how hierarchies of files are handled by the operating system and the operations that can be
used to assemble, manipulate, and navigate a file system hierarchy. You may also need to learn how
to remote login to another computer over a computer network. Interested readers should refer to
Appendix 1 for a detailed discussion of these concepts for UNIX.

1.5

Computer Networking Concepts

A computer network is simply two or more computers connected together. Computer

networks enable humans and computers to communicate by sending messages, and to share data
and information resources by exchanging files. Two types of computer network are as follows:
1. Local Area Network (LAN) : A LAN is a network where computers are connected together
directly. Usually the connection will be some type of cable.
2. Wide Area Network (WAN) : A WAN is simply a network of LANs connected together. The
connections in present-day WANs are becoming a mixture of cable, fiber optic, and satellite
communications.

13

Chapter 1

Communication among LANs is handled by special-purpose computers called routers. Routers

connect LANs to form a WAN. WANs can then be connected to form even larger WANs.

Client-Server Network Architectures

The sharing of information across computer networks is often implemented as two (or
more) programs running on separate computers. One program, called the server, provides a particular resource. A second program, called the client, makes use of that resource. The server and
client programs may be running on different machines located in separate rooms or even separate
countries. Computer networks, where one server provides information to many clients, are said to
have client/server architectures.
Network

Kernel

Kernel

Kernel

Kernel

Client

Client

Server

Client

SCHEMATIC OF SIMPLIFIED CLIENT - SERVER NETWORK

Client sends message to Server
Kernel

Kernel

Client

Server
Server sends data / information to client
machine.

HOW CLIENT S RECEIVE DATA AND INFORMATION FROM A SERVER

Figure 1.3. Model of communication in client-server system

Figure 1.3 shows, for example, a simplified network where one server machine is connected to
three client machines. Any one of the client machines can send a message to the server machine
requesting data, information, or even, access to certain operating system or application-package
processes. The server machine will respond by sending the requested information/service to the
client.
Client/server network architectures are increasing in popularity because of the advantages
they afford. By localizing data, information, and operating system/application package processes

14

Engineering Programming in Matlab

on a single server machine, and providing access to client machines on an as-needed basis, maintenance of operating system software and application program software is simplified considerably.
Moreover, by moving much of the processing power from stand-alone client machines to powerful
server machines, an opportunity exists to design client machines having minimal operating system functionality. These so-called network computers are expected to retail for considerably less
than current PC computers.

The Internet
Millions of computers are now connected together in a massive worldwide network of
computers called the Internet. The word Internet literally means network of networks, and on
any given day it connects roughly 20 million users in more than 50 countries see, for example, the
countries shaded in black in Figure 1.4. The Internet is rather unique in the sense that nobody owns,
is in charge of, or pays for the entire cost of running the Internet.
Development of the Internet dates back to 1969 when the Defense Advanced Research
Projects Agency arm of the Department of the Defense commissioned the construction of an experimental computer network. Work on the Advanced Research Projects Agency Network (ARPANET)
centered around a perceived problem in the Department of Defense - how to keep U.S. military sites
in communication in the event of a nuclear war. If just a few metropolitan areas were wiped out,
communications could be severely disrupted. Two years later the network was connecting 15 nodes,
all of them research centers. The Internet has since passed through the watchful eyes of many universities and research organizations, and grown in size to include millions of computer hosts.
=======================================================
Number of
Number of
Date
Host Computers
Date
Host Computers
=======================================================
Aug. 81
213
Jan. 89
80,000
May 82
235
Oct. 90
313,000
Aug. 83
562
Oct. 91
617,000
Oct. 84
1,024
Oct. 92
1,136,000
Oct. 85
1,961
Oct. 93
2,056,000
Nov. 86
5,089
Oct. 94
3,864,000
Dec. 87
28,174
Jul. 95
6,642,000
Jul. 88
33,000
Jul. 96
12,881,000
=======================================================

Table 1.2. Number of hosts on the Internet 8/81 to 7/96

Table 1.2 shows the number of hosts on the Internet versus time for August 1981 through

15

Chapter 1

Figure 1.4. Countries having access to the Internet

July 1996. The estimate for the year 2000 is that 40 million people will be connected to the Internet.

Internet Access
The majority of present day home users access the Internet using a modem and phone
lines the upper speed, or bandwidth, at which information can be transmitted is 33.6 kilobits
per second. A modest increase in bandwidth to 64 or 128 kilobits per second is possible with
Integrated Services Digital Network (ISDN) technology. With 64 kilobits/sec bandwidth connection
to the Internet, users can reserve part of the bandwidth channel for voice or fax calls. Most large
research organizations and universities are connected to the Internet with one or more T1 lines
having bandwidth 1.46 megabits/sec. In the near-future, significant increases in bandwidth (i.e., 10
megabits/sec). may be possible with cable modems.

Protocols for Internet Communication

Before two computers can exchange data and information over a network, they must agree
on a specification, or protocol, for communication. Protocols cover issues such as how data will
be formatted, conventions for control and coordination of information exchange, and handling of
errors.
At the network level, the Internet Protocol (IP) specifies how data are to be physically

16

Engineering Programming in Matlab

transmitted from one computer to another, and the Transmission Control Protocol (TCP) ensures
that all the data sent using IP are received without error. Together these protocols are known as
TCP/IP, and they form the foundation of many other high-level application-oriented protocols for
sending packets of information (e.g., files, E-mail, web pages) across networks.

Internet Domain Names and Addresses

The Internet sends packets of information across a network using a model that is simply an
electronic counterpart to the way letters are posted inside an envelope containing a delivery address
and a return address. Perhaps the delivery address is the location of your Internet provider and the
return address is the location of your home.
Every computer on the Internet has its own unique Internet address. You cannot send
an E-mail message to someone, transfer a file via FTP, or access web pages located on a specific
computer, unless you know his or her Internet address. In our letter-and-envelope analogy, there
is the numerical form of an Internet address (it looks like xxx.xxx.xxx.xxx), and its vernacular
counterpart. Although the numerical form is somewhat akin to a postal zip code, and an important
part of the address delivery, people naturally prefer to work with addresses written in textual terms.
Computer domain names follow a three-part format:
persons userid
@ domain name of computer(s)

|
|
|
Part 1
the "at" sign
Part 3

For example, the internet address austinisr.umd.edu has the person-id austin and the domain name isr.umd.edu. The numerical counterpart of isr.umd.edu is 128.8.111.4. As
these examples show, there is never blank space between components of the Internet address.
One important point to note is that user ids need not be unique. For instance, two people
with the family name Austin can have the login name austin as long as they operate on separate
domains.
austin@isr.umd.edu
austin@kiwi.berkeley.edu

However, the domain name of the computer(s) must be unique.

Computer domain names are composed of subdomain names that follow the nomenclature
outlined in Tables 1.3 and 1.4. For example, in the computer address isr.umd.edu, the subdomain name edu tells us that the computer is located at an educational institution. The subdomain

17

Chapter 1
=====================================================================
Domain
Meaning
Domain
Meaning
=====================================================================
com
Commercial
mil
Military
edu
Educational
net
Networking
gov
US Government
org
Nonprofit
=====================================================================

Table 1.3. Organizational domain names

=====================================================================
Domain
Meaning
Domain
Meaning
=====================================================================
at
Austria
dk
Denmark
au
Australia
fr
France
br
Brazil
jp
Japan
ca
Canada
nz
New Zealand
de
Germany (Deutschland)
uk
United Kingdom
=====================================================================

Table 1.4. Some geographical top-level domains

name umd stands for the University of Maryland. Finally, the subdomain name isr represents the
collection of computers at the Institute for Systems Research at the University of Maryland.

Internet Services
Using the Internet means sitting at a computer and having access to a number of basic
services, including:
1. E-mail : You can send or receive electronic messages from anyone on the Internet. Anything
that can be stored in text file can be mailed. Facilities also exist for converting binary files
(e.g., an executable computer program) into a format suitable for transmission via E-mail.
2. Telnet : The Telecommunications Network (Telnet) program allows users to remotely login to
computers over a network. The computer may be in the next room, or perhaps, on another
continent.
3. Gopher : Gopher allows a user to request information from an extensive list of gopher servers
on the Internet.

18

Engineering Programming in Matlab

4. File Transfer : The File Transfer Protocol (FTP) enables the copying of files from one computer to another. Anonymous FTP is a system where an organization makes certain files
available for public distribution - you can access such a computer by logging in with the
user-id anonymous. Although no special password is required, it is customary to enter your
E-mail address.
5. Usenet : An abbreviation for users network, Usenet is a collection of more than 5000 discussion groups centered around particular topics. Newsgroups exist for every topic imaginable,
including of course, those dedicated to the Internet and its development.
More recently these basic services have been supplemented by the World Wide Web (WWW), allowing easy access to a wide array of media.

The World Wide Web

By providing millions of users in homes, schools, and industry with access to a wide array
of media via an easy-to-use graphical user interface, the Web has captured the honorable distinction
of being the most exciting part of the Internet.
Development of the Web began in March 1989, when Tim Berners-Lee of the European
Particle Physics Laboratory (part of a larger organization known as CERN) proposed the project
as a means of transporting research and ideas effectively throughout the CERN organization. The
initial project proposal outlined a simple system of using networked hypertext (see the next section)
to transmit documents and communicate among members in the high-energy physics community.
The first graphical user interface to the Web was written by the Software Design Group at the
National Center for Supercomputing Applications (NCSA). They named their graphical interface
Mosaic. Mosaic, and the now famous Microsoft Internet Explorer and Netscape browsers, provide
mouse-driven graphical user interfaces for displaying hypertext and hypermedia documents containing forms, paragraphs, lists, and tables in a variety of fonts and font styles. Graphics images
can be linked to text and vice versa. Sound and video files can be pointed to by documents, and
down-loaded over the Internet by simply clicking on the appropriate link.
Todays browsers are simply the beginning of where web development is headed. To help
ensure that web browsers will be accepted as common business tools, software developers are busy
integrating browser technology with office productivity suites (e.g., E-mail, word processors, graphics, spreadsheets, databases) and in the case of engineering firms, design and analysis software.
Work is also currently underway to incorporate encryption technology and client authentication
abilities into web servers, thereby allowing users to send and receive secure data. These features
will ensure that sensitive data is kept private, a prerequisite for commercial transactions over the
Internet.

19

Chapter 1

MOVIE : Technologies for

Electronic Commerce
Applications of
Encryption Technology

...... Work is also currently

underway to incorporate
encryption technology and ......

WEBSTERS
DICTIONARY
Encryption is ........

Selecting a piece of hypertext can take you to .....

.... more hypertext, images, sounds, and movies.

Figure 1.5. How hypertext works

Hypertext and Hypermedia
Hypertext is basically the same as regular text with one important exception hypertext
contains connections within the text to other documents. (See Figure 1.5.) Suppose, for instance,
that you were able to somehow select (with a mouse or with your finger) the phrase encryption
technology in the previous paragraph. In a hypertext system, you would then have one or more
documents related to encryption technology appear before you - a textual description of the applications of encryption technology, or perhaps Websters definition of encryption. These new texts
would themselves have links and connections to other documents. Continually selecting text would
take you on a free-associative tour of information. In this way, hypertext links, called hyperlinks,
can create a complex virtual web of connections.
Hypermedia is hypertext with a difference - hypermedia documents contain links not only
to other pieces of text, but also to other forms of media such as sounds, images, and movies. Images
themselves can be selected to link to sounds or documents. Hypermedia simply combines hypertext
and multimedia.
Suppose that you are a project manager in a large multinational engineering organization.
Here are some examples of how web-based hypermedia might be used:
1. By clicking a mouse button on a part in an engineering drawing, you are able to see the pathway
of project requirements leading to that part being incorporated in the design.

20

Engineering Programming in Matlab

2. A web-based system might provide up-to-the-minute information on the status of a particular

project. Clicking on the textual description of a project requirement could start an audio track
describing an engineers rational for the requirement.
3. By looking at an organizations floor plan, you are able to retrieve information about an office
by simply touching a room. An inventory of office equipment, detailed office drawings, and
information on the office occupant and their current projects might appear by clicking on the
right, center, and left mouse buttons.
4. You are reading a research paper to understand why a new technology is needed by your organization. By selecting text in a research paper, you are able to view a movie of the technology
being tested in a laboratory setting. Then by clicking on a mouse button you are able to download and execute an engineering analysis package demonstrating how the technology works
for simple case-study problems.
In the implementation of these applications, we expect that the use of hypermedia will elevate a
readers ability to understand and navigate the concepts presented beyond what is likely to
occur with a serial presentation (e.g., a regular book). Realizing this goal requires at the very least,
good judgment, attention to detail in document design, and user testing.
How Does the Web Work?
Web software is designed around a distributed client/server architecture (see Figure 1.3).
A web client, called a web browser if it is intended for interactive use, is a program that can send
requests for documents to any web server. A web server is a program that, upon receipt of a request,
sends the document requested (or an error message if appropriate) back to the requesting client.
Because the task of document storage is left to the server and the task of document presentation is
left to the client, each program can concentrate on those duties and progress independently of each
other.
Send me "Company Floor Plan"

WEB BROWSER

Hypermedia HTML document containing

the "Company Floor Plan"
WEB SERVER

Figure 1.6. Web server and web client transactions

Figure 1.6 shows the sequence of transactions that would take place when the Company Floor

21

Chapter 1

Plan is downloaded from a web server and displayed on a web browser. The sequence of events is
as follows:
1. A user working at the web client selects a hyperlink in a piece of hypertext connecting to another
document such as Company Floor Plan.
2. The web client uses the address associated with that hyperlink to connect to the web server at a
specified network address and asks for the document associated with Company Floor Plan.
(See the upper half of Figure 1.6.)
3. The web server responds by sending the text and any other media within that text (pictures,
sounds, or movies) to the client. The web client renders the information for presentation on
the users web browser window. (This step is shown along the lower half of Figure 1.6.)
The language that web clients and servers use to communicate with each other is called the Hypertext Transfer Protocol (HTTP). All web clients and servers must be able to speak HTTP to send
and receive hypermedia documents. For this reason, web servers are often called HTTP servers,
and World Wide Web is often used to refer to the collective network of servers speaking HTTP as
well as the global body of information available using the protocol.
Uniform Resource Locators
The Web employs what are called Uniform Resource Locators (URLs) to represent hypermedia links and links to network services within HTML documents. Almost any file or service on
the Internet, including FTP, Gopher, and Telnet, can be represented with a URL. Table 1.5 contains
some examples.
==============================================================================
Uniform Resource Locator
Description
==============================================================================
http://www.ence.umd.edu/

Connects to an HTTP server and retrieves

an HTML file.

ftp://rtfm.mit.edu/pub/usenet/

Opens an FTP connection to the usenet

frequently-asked questions stored at
rtfm.mit.edu

gopher://gopher.tc.umn.edu/

Connects to a gopher menu at the University

Minnesota home site, inventor of the Gopher.
==============================================================================

Table 1.5. Some examples of URLs

22

Engineering Programming in Matlab

A URL is always a single unbroken line of letters and numbers with no spaces. The first
part of a URL (before the two slashes) specifies the method of access; http is perhaps the most
common. Typically, the second part of the URL is the address of the computer where the data or
service is located. Further parts may specify the names of files, the port to connect to, or the text to
search for in a database.
Sites that run web servers are often named with a www at the beginning of the network
address. For example, the Department of Civil Engineering web server at the University of Maryland
has the URL:
http://www.ence.umd.edu/welcome.html

welcome.html is the name of the HTML file for our department home page. You can ask your
web browser to connect to this web server by simply specifying the URL in the location window.

1.6

Hardware-Software Life Cycle

In engineering circles, advances in computer hardware and applications programs are

driven by market competition and the need to design, analyze, manufacture, and control complex
engineering systems. The complexity of an engineering system can be due to a number of factors
including its size (i.e., a large number of interacting parts), nonlinear relationships between the input/output (I/O) parameters, incomplete information, enhanced performance specifications, and so
forth. In any case, without the assistance of modern-day computer hardware and engineering applications programs, development of these systems would simply be intractable, if not impossible.

ENGINEERS / CUSTOMERS

DEVELOPMENT OF

DEVELOPMENT OF

HARDWARE

SOFTWARE

DEVELOPERS
STAKEHOLDERS

Figure 1.7. Hardware-software lifecycle

Figure 1.7 is a high-level view of components in the hardware-software lifecycle. Typically, it begins with a group of engineers/customers, developers and stakeholders (e.g., owners of a company)
reaching consensus on the need for a new product and its pathway of development. Sometimes
the results can be really impressive. Since the mid-1980s, for example, manufacturers of computer
hardware have doubled the computational speed of their products every 12 to 18 months.

23

Chapter 1
ECONOMICS OF COMPUTING / SYSTEMS DEVELOPMENT
S = Software

Cost of Development

H = Hardware

H
H

Systems Integration

TASK - ORIENTED PROGRAMS / MODULES

INTEGRATED SYSTEMS / SERVICES

CENTRALIZED OPERATIONS

DISTRIBUTED OPERATIONS

1970s and early 1980s

1990s

Figure 1.8. Economics of software development and integration.

The economic challenges and difficulty in following up on hardware advances with appropriate software developments are reflected in Figure 1.8. In the early 1970s, software consumed
approximately 25% of total costs, and hardware 75% of total costs for development of data intensive systems. For the most part, computer systems were stand-alone, and software developers
wrote computer programs dedicated to a specific task (e.g., finite element analysis, control systems
package).
Currently, the development and maintenance of software typically consumes more than
80% of total project costs. This change in economics is the combined result of falling hardware
costs, increased software development budgets, and a need to solve more difficult problems than in
the past. Whereas one or two programmers might have written a complete program twenty years

24

Engineering Programming in Matlab

ago, teams of programmers are now needed to write todays large software programs. Moreover, for
organizations that have made large investments in software, there is great reluctance to discard software just because a new technology has come along. What management would like to see instead are
the benefits of improved communications without having to reinvest in the basic application-specific
software. The objectives of systems integration are to try to bring this situation under control, to
ensure that the pathway forward maximizes return for the organization, and protects a companys
past investments in software and hardware [7]. Systems integration is not a technical discipline in
itself, but rather an approach to the management of organizations that recognizes the different ways
its parts interact. Understanding an organizations structure and management practices is critically
important because it is a prerequisite to computer automation.
If software developments for engineering/business applications are to have any chance of
keeping up with advances in computer hardware and networking, future developments will need to
pay close attention to software design and reuse of functions, libraries, modules, program architectures, and programming experience. A key component of the solution lies in the judicious choice
of programming language(s). In addition to having the ability to compute and evaluate formulae,
programming languages now need to handle and manipulate large quantities of data and information, and lend themselves to rapid development of interactive graphics applications in parallel and
networked computing environments. Recently developed languages such as Perl, Tcl/Tk, and Java
are expected to play a central role in making this happen [1, 11, 17].

1.7

Principles of Engineering Software Design

The discipline of software engineering is concerned with the design and implementation
of computer programs that work, are correct, and are well written. Software engineering principles
can be applied to small and large computer programs alike:
Small Computer Programs. Small computer programs are characterized by the data structures and
algorithms they employ, design of the program architecture, and the computational efficiency of the
program implementation.
Novice computer programmers tend to write programs that are small, composed of perhaps only a few hundred lines of source code. When you are learning to program in a new language,
becoming familiar with its syntax, data types, and control structures are the most important things
because without them you cannot transform a small-scale task into step-by-step programming instructions.
Large Computer Programs. In real-world engineering environments it is now commonplace for
computer programs to be hundreds of thousands (even millions) of lines long. Many of them are
so complex that even the best human minds cannot simultaneously comprehend all the details. It is

Chapter 1

25

therefore vitally important that the design and implementation of large computer programs be based
on established procedures for software development, including attention to program specification
and design, organization, coding, testing, and maintenance of software [3, 10]. Careful planning of
these activities is needed because:
1. Large programs are most often written by programming teams. Team members must be able
to understand one anothers work. When the planning, design, and coding of a project takes
several years, communication of work among employees is of utmost importance.
2. Large programs are often developed within the constraints of short time-to-delivery contract
schedules. Programming teams may not have the luxury of starting again, even after a key
design flaw is identified.
3. Well-organized programs are easier to debug. A little extra time spent planning the layout
of a program may be saved many times over when it comes time to debug and upgrade the
software. Indeed, some estimates place maintenance at 60 to 80% of the overall cost of a
software project.
4. Large programs often evolve through a series of versions or updates. The programmer making
updates is likely to be a person other than the original developer.

Models of Software Systems Development

Models of software systems development are an important component of large computer
program development because they provide the project participants with a framework for knowing
what is expected, and when. In some cases, the model may also form the basis of legal agreements,
with contract payments being tied to successful completion of a task identified in the model.
Figure 1.9 is a schematic of two models of software system development that are currently
in use. The waterfall model [14] views the software development process as a sequence of stages
that includes requirements specifications, design and testing, integration, and maintenance (we get
to the details of each step in a moment). Each phase of sequential development is completed, via
formal review, before the next phase begins. The waterfall model of development is appropriate
when the problem and solution method are well understood.
The spiral model of development, shown in Figure 1.9b, is simply a sequence of waterfall
models with risk analysis and control incorporated at regular stages in the project [2]. This riskoriented approach to iterative enhancement recognizes that implementation options are not always
clear at the beginning of a project. For example, a software project that is scheduled for development
over a number of years may need technology that has not yet come to market. Maybe the technology
will evolve as expected, and maybe it will not. The radial direction of Figure 1.9b corresponds to

26

Engineering Programming in Matlab

Requirements
Definition

System Design

Detailed Design

Integration and
Test

Operations
and Maintenance

ANALYSIS [ WHAT ]

DESIGN [ HOW ]

BUILD [ WHO ]

(a). Waterfall model of systems development

Risk Analysis.
Operational
Prototype.
Determine Objectives and Alternatives.

Risk Analysis.
Prototype 2.

Objectives and Alternatives.

Risk Analysis.
Prototype 1.

REVIEW
Requirements Plan.

Preliminary

Detailed

Life Cycle Plan.

Design.

Design.

Plan Next Phase.

Plan Next Phase.

Requirements
Validation.
Testing of

Plan Next Phase.

Components.

Service

Integration and Test.

(b). Spiral model of systems development

Figure 1.9. Models of development for software systems: (a) Waterfall, (b) Spiral

Chapter 1

27

the cumulative cost incurred in the project, and the angular direction corresponds to progress made
in completing each cycle of the spiral. Each cycle of development has the following phases:
1. Identify the design and development objectives for the cycle, as well as the alternatives that are
possible to achieve the goals.
2. Evaluate different alternatives based on objectives and constraints. Where appropriate, identify
uncertainties and risks. Risk means something that can go wrong as the consequence of
incomplete information, or perhaps, as the result of human errors.
3. Develop strategies such as simulation, prototyping, and benchmarking for resolving uncertainties and risks.
4. Plan the next stage, allowing for any of the possible lifecycle models to be used.
Of course many variations on the waterfall and spiral models are possible. For example, some
companies develop software incrementally, using a model that is essentially a chain of waterfall
models with each link in the chain corresponding to development of a software release. Subsequent
releases add capability to previous releases and fix bugs found in previous versions. The second
observation tends to be a sore point with many involved in the software industry and leads them to
make comments such as never buy software that is younger than Version 3.0. Clearly they think
that it takes at least two software versions just to find and eliminate insidious startup bugs.

Components of Software Systems Development

The key steps in the development of a large software project are as follows:
Requirements Specification. The objective of the requirements specification is to state the goals
of the program as carefully as possible. The goals could include, for example, expected input
(keyboard/files/IO board) and output. The requirements should also identify any known limitations,
possible errors, and accuracy issues.
The importance of the requirements specification is partly due to the economics of project
development at the beginning of the software lifecycle. As indicated in Figure 1.10, decisions made
at the beginning of the software lifecycle are inexpensive to make, yet have the greatest commitment
of funds. Without a clearly defined requirements specification there is a greatly increased chance of
project failure.
Table 1.6 shows the results of a study by Hewlett-Packard to quantify the relative costs, in
terms of money and/or time, of correcting design errors. From an economic point of view, the last
thing a manufacturer wants is discovery of a fatal error in the engineering system by a customer !

28
CUMMULATIVE PERCENTAGE

Engineering Programming in Matlab

100

Define Concept

Commence Production
FUNDS COMMITTED

75

50

25
FUNDS EXPENDED

PRODUCT LIFE - CYCLE

Figure 1.10. Funding commitments in product lifecycle.

Project Phase

Bug Description

Relative Cost

Design

Design team

Implementation

Programmers

10-20

Quality assurance

QA personnel

70-100

Shipment to customer

Customer

Very expensive

Table 1.6. Cost of correcting design errors

Analysis. Carefully analyzing a problem and its possible solutions is of utmost importance. Either
you have to develop an algorithm yourself, or find one that is already available. Do not hesitate to
look for established techniques; for example, an engineering problem might involve solving a set
of linear equations using a standard numerical algorithm such as Gaussian Elimination. To develop
algorithms from scratch when an already well-thought out and documented technique is available is
a waste of time.
Design. The next most important phase of program development is to develop a viable design that
you believe will work. Studies indicate that up to 50% of all errors in the software lifecycle are

29

Chapter 1

made during the design phase of the software; the remaining errors tend to be programming and
syntax errors (as described previously). Unfortunately, design errors are also the most expensive to
repair since many appear only at run-time, and often after the software has been delivered to the
customer.
Implementation. The implementation phase focuses on the specification, writing, and documentation of modules of code. When the details of specification are particularly complicated, pseudocode
is often inserted as an intermediate step. Pseudocode is simply a high-level description of the
code (or algorithm) to be implemented, where the semantics are less strict than the programming
language itself.
Always document program source code so that it can be understood by you and others at
a later date. Documentation includes recording details of your specific solution how the analysis phase will transfer to the program, what language-specific constructs were used, what method
of problem decomposition was used, and any important restrictions in your software design. An
executable program is the result of the implementation phase.
Test and Verification. In many engineering industries, such as airplane flight control, airport traffic
control, nuclear power plant control, medical instruments, and communications network control,
reliability of software is the most sought after attribute. The software must execute as expected
and without errors. Obtaining software that performs with a high degree of confidence requires a
solid program design, a careful implementation, and a carefully designed suite of test problems that
may illuminate software bugs and weaknesses. The identification of bugs and weaknesses is highly
desirable, because this is the point at which fixing them is easy.
Maintenance. It is frequently stated that up to three quarters of a programmers time is spent on
maintaining existing software. Maintenance activities include (1) repair of coding (and design)
errors, (2) adapting the software to changes in the computing environment (e.g., an update in the
operating system), and (3) adapting the software to changes in the customers requirements. Ask
yourself how much code you have written in the past that is still in use (by you or others)? Can you
easily understand programs that you wrote years ago? Would it be easy to change or modify your
programs for a new purpose?
Japan
USA

Reuse rate(%)
80
30

Average lines per year

12,447
7,290

Average errors/1000 lines

1.96
4.44

Table 1.7. Reuse of code in the US and Japan

Table 1.7 shows how important it is to write code that can be reused and easily modified

30

Engineering Programming in Matlab

for a new purpose. If you are busy reinventing the wheel while other developers are reusing their
software, eventually you will not be able to compete. These issues have to be taken into account
before and during development of new programs. Once you have written a program without considering these issues, it is difficult and erroneouss to shoehorn or force the code into a new application.
We must learn to take into account the need for future changes in programs and to design programs
accordingly.

Modular Program Development

The goal of modular programming is to break a complex task or program into an ensemble
of weakly coupled (independent) modules. Each module should be viewed as an independently
managed resource, with access highly restrained. Programs consisting of well-designed modules are
much simpler to design, write, and debug than equivalent programs that are poorly designed [5, 12].
Guidelines for the design of program modules are:
1. Psychological studies indicate that the average human can only simultaneously comprehend
seven (plus or minus two) pieces of information. Hence, one module should have no more
than seven subordinate modules.
2. There should be separation between the controller and worker modules.
3. Every module must perform a task appropriate to its place in the hierarchy.
4. Every module should only receive as much information as it needs to perform its function, and
ideally, they should exchange as little information as possible.
Modules should enforce the principle of information hiding, namely, that all information about a
module should be private to the module unless it is explicitly declared public [12]. Public data
should be avoided because it exposes implementation details that make reuse unlikely and maintenance difficult. A key benefit of information hiding is the opportunity it affords for updating the
internal details of a module (here we assume that the details of the public interface will not change)
without affecting other modules in the system.
Coupling. Coupling measures the extent to which different modules in a system are interconnected
with one other. See Figure 1.11. In design we should keep the interfaces as minimal and as simple
as possible. Designing for minimal coupling among modules helps to ensure that errors occurring
in one module will not propagate across the whole system.
Cohesion. Cohesion is a concept that describes how strongly the attributes and functions of a
module are connected together. Ideally, a module should just perform one task. In design we should
keep related functions together and unrelated functions apart.

31

Chapter 1

HIGHLY - COUPLED SYSTEM

LOOSELY - COUPLED SYSTEM

Figure 1.11. (a) highly coupled and (b) loosely coupled systems

The attributes of system coupling and cohesion work together. Generally speaking, modules with components that are well related will have the capability of plugging into loosely coupled
systems. Modules should be designed within the framework of a well-defined language or grammar. If this principle is followed, then large programs may be divided into smaller modules (which
may be compiled separately), and independent modules may be assembled together (i.e., bottom-up
design). Modules designed within the framework of a language are easier to maintain and extend.

Abstraction
Programmers often apply techniques of abstraction to the development process, meaning
that they concentrate on the essential features of one part/module of the computer program, and abstract from details of the computer program that are not immediately relevant to the current module.
In the development of computer software, two types of abstraction are common:

1. Procedural. Software development is based on a stepwise refinement of the systems abstract

function. This type of abstraction is most common in programming languages such as C,
MATLAB, and FORTRAN.
2. Data. Software development is based on the system data and the operations that can be applied
to the data. Implementations of data abstraction correspond to objects and the operations
that can be applied to the objects. This type of abstraction is common in object-oriented
programming languages such as C++ and Java.

32

Engineering Programming in Matlab

Top-down and Bottom-up Software Design

Techniques of abstraction are often used in conjunction with a combination of top-down
and bottom-up development strategies, as shown in Figure 1.12.

DECOMPOSITION

NEW PROBLEM

SUB-PROBLEMS

(a). Top-down design

COMPOSITION

INDEPENDENT MODULES

COUPLED MODULES

(b). Bottom-up design

Figure 1.12. Schematics of top-down and bottom-up design
Top-down software development begins at the conceptual level and passes through three stages:
1. The high-level design establishes the important subsystems of the design, the purpose and
intended behavior of each subsystem, and the relationship among subsystems. As already
mentioned, a good design corresponds to subsystems that are as uncoupled as possible.
2. The intermediate-level design breaks subsystems into modules. Each module should have
one well-defined purpose, hide its data from other modules, and have a minimal number of
connections to other modules in the program [9].
3. The low-level design involves detailed specification of algorithms and data structures.

Chapter 1

33

You should notice that top-down design delays detailed decisions about the program flow and data
structures until they absolutely have to be made.
The strategy of bottom-up design starts with low-level procedures, modules, and subprogram library routines, and tries to combine them into higher-level entities. A key benefit of bottomup design is its use of already implemented code. For example, numerical linear algebra packages
are one area where libraries are routinely linked to C programs for finite element and numerical
analysis, solution of differential equations, and solution of engineering control problems.
Top-down and bottom-up designs are extreme strategies for generating a hierarchical program structure and are often at odds with each other. Often the result of a top-down design is
software modules that are of a one-time-only form. They are not amenable to reuse because they
were designed as components without a preconceived vision of their future use. Similarly, generic
modules suitable for bottom-up design may contain many features not needed to solve a specific
task.
Software designers never embark on the construction of a new system without first considering available libraries/modules. Conversely, they never build software modules without a preconceived vision of their future use. A balance of the above-mentioned criteria is usually needed and
desirable.

1.8

Computer Programming Language Concepts

Computer programs are composed of data and software instructions. The software instructions in a computer program tell the computer hardware how to execute a particular task by
manipulating various types of data in a precise manner. Programming languages facilitate the development of these software instructions by providing constructs for organizing computations. The
best programming languages enable the writing of good computer programs by being easy to read,
understand, and modify.

High- and Low- Level Computer Languages

In the earliest days of computing computers were programmed by entering instructions
and data into the computer in binary form. These so-called machine languages correspond to the
instruction set of a particular computer hardware. They are entirely machine dependent (i.e., very
low-level), and can be very tedious and error prone to program.
Programming in machine languages was quickly replaced by programming in symbolic
assembly languages, which make the instructions easy to write and understand by using mnemonics
for the machine instructions. For example, the sequence of symbols STO M R1 would store the

34

Engineering Programming in Matlab

contents of register 1 in memory location M. Symbolic assembly languages are still low-level, and
tend to be machine dependent. An assembler converts the assembly language into corresponding
binary code and generates a binary program for execution.
Easy-to-read high-level languages have been around since the early 1950s. The term
high-level means that many of the details in a programs development are handled automatically,
thereby providing a pathway for programmers to write less code to get the same job done. For
example, a number of high-level programming languages have been designed with the keywords
if and for for the construction of simple branching and looping control structures. Once the program source code has been written, a compiler (see details below) will automatically generate the
low-level machine instructions to implement the control structures. Whereas one line of assembly
language code must be manually written for each low-level program instruction, high-level programming languages generate (on average) about five machine instructions for each line of code
written. Because programmers write approximately the same number of lines of source code per
day, irrespective of the development language, application programs can be implemented in a highlevel language much faster (and more cheaply) than in an assembly language [4]. Examples of
high-level programming languages are FORTRAN, C, C++, and Java. These languages are good
for building software components from data structures and algorithms from scratch.

Compiled and Interpreted Programming Languages

Because computer hardware can only follow very low-level machine code instructions,
which are difficult for humans to understand and manipulate, engineers usually write computer
programs in a high-level programming language.
HIGH LEVEL
LANGUAGE

COMPILER

LOW LEVEL
MACHINE LANGUAGE

A compiler translates the computer program source code into machine code instructions. The compiler can be thought of as a type of machine; it only understands a limited range of commands. If
the compiler does not receive proper input, it can only try to interpret the error, report what is wrong
(i.e., error messages), and exit. C, C++, and FORTRAN are among the programming languages that
are compiled. Good programmers have some idea into what the code they write compiles (assembly
language, bytecode, or machine code), because it is the efficiency of compiled code that matters,
not the original number of lines of source code.
In an interpreted computer program, high-level statements are read one by one, and translated and executed on the fly (i.e., as the program is running). Scripting languages such as Tcl/Tk
and Perl are interpreted, as are application programs written in the MATLAB programming lan-

Chapter 1

35

guage.
The Java programming language is both compiled and interpreted. High-level Java source
code is compiled into a low-level bytecode, and then interpreted by a Java Virtual Machine. Do
not worry if some of the details seem a little murky at this point we fill in the details of how this
happens in the Java tutorial.

Procedural and Object-Oriented Programming Languages

Programming languages such as FORTRAN, C, and MATLAB are procedural because
they enable a complex problem to be decomposed into a hierarchy of functions. Each function has
a well-defined set of input parameters and will return a well-defined set of output values.
Programming languages such as C++ and Java are object-oriented because they view complex problems as an assembly of data items together with sets of methods (i.e., functions) that can
manipulate the data.

1.9

When to Program in MATLAB?

As indicated in Figure 1.1, the number of programming languages engineers are using is
steadily increasing. This trend is due in part to the expanding range of tasks for which engineers are
now using computers, and in part by the limited range of tasks current programming languages can
handle well. Since computer languages are usually designed with the solution of a certain range of
problems in mind, the selection of the right language for the job at hand is of utmost importance.
Use MATLAB if you need to solve a problem that can be conveniently represented by
matrices, solved using operations from linear matrix algebra, and presented using relatively simple
two- and three-dimensional graphics. Computing the solution to a family of linear equations, and
representing, manipulating, and displaying engineering data are perhaps the two best examples of
problems for which MATLAB is ideally suited.
Not only is the MATLAB programming language exceptionally straightforward to use (every data object is assumed to be an array), but the MATLAB program code will be much shorter
and simpler than an equivalent implementation in C or FORTRAN or Java. MATLAB is therefore
an ideal language for creating prototypes of software solutions to engineering problems, and using
them to validate ideas and refine project specifications. Once these issues have been worked out,
the MATLAB implementation can be replaced by a C or Java implementation that enhances performance, and allows for extra functionality for example, a fully functional graphical user interface
that perhaps communicates with other software package over the Internet.

36

Engineering Programming in Matlab

Since the early 1990s the functionality of MATLAB has been expanded with the development of toolboxes containing functions dedicated to a specific area of mathematics or engineering.
Toolboxes are now provided for statistics, signal processing, image processing, neural nets, various
aspects of nonlinear and model predictive control, optimization, system identification, and partial
differential equation computations. MATLAB 5.0 also comes with an Application Program Interface that allows MATLAB programs to communicate with C and FORTRAN programs, and vice
versa, and to establish client/server relationships between MATLAB and other software program.

1.10

Review Questions

1. Have you been able to acquire an account and logon to a computer system at your school or
company? Otherwise, do you have access to a PC that you can use to compile and run C,
MATLAB, and Java programs while reading this book?
2. Do you understand the basic purpose of the computer components, such as the floppy disk, hard
disk, memory (RAM and ROM), CPU, video display, keyboard, mouse, and so forth?
3. What are the main purposes of an operating system?
4. What is a computer network?
5. Briefly describe how a client/server system works. Can a computer act as both a client and a
server at the same time?
6. Why were the Internet and the World Wide Web originally developed?
7. What technology has enabled web search engines, such as Yahoo and AltaVista, to index tens
of millions of web pages in just a few years of time?
8. What is hypermedia?
9. What are the basic steps involved in the development of software? Why is the process broken
down into several steps? Why would you use one language rather than another? Are they all
the same?
10. Briefly explain the terms abstraction, modularity, coupling, cohesion, and information
hiding?
11. What are the goals of top-down and bottom-up design?
12. What is the difference between procedural abstraction and data abstraction?
13. Why are present-day applications programs written in high-level programming languages?
14. Why does it typically take much longer to write an application program in assembly language
than in a high-level programming language?
15. What do the acronyms HTML and VRML stand for? What is the relationship between Java
and VRML?

Chapter 1

37

16. For what types of problems is the MATLAB programming language suited? How has the
functionality of MATLAB recently been expanded?
17. For what types of problems is the C programming language suited?
18. For what types of problems is the Java programming language suited?
19. Why does it sometimes make sense to develop a software package using more than one programming language?

1.11

Review Exercises

1.1 Acquire an account on a UNIX machine and login to your account. When you have logged in
successfully, try to navigate the file system. For example, start by typing the pwd command.
This command prints the working directory, or your current location in the file system. Now
try cd .. which moves your current directory up to the parent directory. Now try ls to
list the files in your directory. Now that you are in your parent directory, you can get back
to your personal directory by typing cd (for change directory). By using cd .. and cd
followed by a directory name, try to navigate and show the directory structure around your
home directory on the UNIX system. Make a tree-type drawing of the structure.
1.2 Login to a UNIX machine and try the commands rm, ls, cd, mkdir, rmdir and cat.
Experiment with the online UNIX manual that is, type man (for manual) command, as in
man mkdir.
Try specifying options to the commands on your computer system (e.g., ls -l in UNIX).
Almost all operating system commands have options so that you can specialize the operation
of the command. You will eventually learn how to do this with your C programs (it is convenient when you include command line options because it lets the user specify particular
features of the program without having to ask every time the program is run). You can find
the available options for an operating system command via man program on UNIX systems.
1.3 Use the operating system commands explained in Appendix 1 to build a simple directory
structure under your UNIX home directory or under the root (C:) on your WINDOWS/NT
system. Do not hesitate to create many directories; they are, after all, free!
1.4 Can you get some experience with a text editor such as vi or emacs on a computer? You
will need to have experience with this because you will use an editor to enter your programs
before they can be compiled and run. Most PCs come with a basic text editor, and there are
many public domain editors you can get free of charge.

Part II

MATLAB Programming Tutorial

38

40

Engineering Programming in Matlab

Introduction
MATLAB is a great programming language for solving problems that can be conveniently represented by matrices, that lend themselves to a solution with operations from linear matrix algebra and
that can be presented using relatively simple two- and three-dimensional graphics. Computing the
solution to a family of linear equations and representing, manipulating, and displaying engineering
data are perhaps the two best examples of problems for which MATLAB is ideally suited.
As we soon see, not only is the MATLAB programming language exceptionally straightforward (almost every data object is assumed to be an array), but also MATLAB program code will
be much shorter and simpler than an equivalent implementation in C, FORTRAN, or Java. MATLAB is, therefore, an ideal language for creating prototypes of software solutions to engineering
problems, and using them to validate ideas and refine project specifications. Once these issues have
been worked out, the MATLAB implementation can be replaced by a C or Java implementation
that enhances performance and allows for extra functionality (e.g., a fully functional graphical user
interface that perhaps communicates with other software package over the Internet).
Chapter 13 is an introduction to programming in MATLAB Version 5.0 and includes all
the basic concepts you need to know. Topics include variables and variable arithmetic, matrices
and matrix arithmetic, control structures, built-in matrix functions, M-files, and so forth. Four
engineering applications are solved at the end of Chapter 13. Where appropriate, pointers are given
to equivalent or similar implementations of these problems in C and Java. A brief introduction to
MATLAB graphics is contained in Chapter 14. Many other texts weave MATLAB graphics into the
introductory sections of MATLAB programming. However, in an effort to keep the tutorial length
short, we separate graphics from programming. Chapter 15 covers the solution of linear matrix
equations in MATLAB. We demonstrate the power of MATLAB by working step by step through
the formulation and solution of a variety of engineering applications involving families of matrix
equations and, where applicable, MATLAB graphics.

Chapter 2

Introduction to MATLAB

2.1

Getting Started

The MATrix LABoratory program (MATLAB) was

initially written with the objective of providing scientists and
engineers with interactive access to the numerical computation
libraries LINPACK and EISPACK. These libraries are carefully
tested, high-quality programming packages for solving linear
equations and eigenvalue problems [6, 15]. MATLAB enables
scientists and engineers to use matrix-based techniques to solve
problems without having to write programs in traditional languages such as C and FORTRAN. Currently, MATLAB is a
commercial matrix laboratory package that operates as an interactive programming environment with graphical output. The
MATLAB programming language is exceptionally straightforward to use since almost every data object is assumed to be an array. MATLAB is available for
many different computer systems, including Macintosh, PC, and UNIX platforms.
The purposes of this tutorial are two-fold. In addition to helping you get started with MATLAB, we want you to see how MATLAB can be used in the solution of engineering problems. The
latter objective is achieved through the presentation of a series of engineering application problems.
Throughout this tutorial, we assume that you
1. Will read a few sections and then go to a UNIX workstation or PC/Macintosh to experiment
with MATLAB
2. Are familiar with the operating system on your computer
3. Have access to supplementary material on matrices, matrix arithmetic/operations, and linear
algebra

41

42

2.2

Engineering Programming in Matlab

Professional and Student Versions of MATLAB

Professional and student editions of MATLAB Version 5 are available. The functionality
of the student edition of MATLAB is limited in the following way; each matrix is limited to 16,384
elements, large enough to study problems having 128-by-128 matrices. From this point on, we
assume that you have installed (follow the instructions accompanying the software disks) the student
version of MATLAB on your computer.

Entering and Leaving MATLAB

The procedure for entering and leaving MATLAB on UNIX and Mac/PC platforms is as
follows.
UNIX Platform. A MATLAB session may be entered by simply typing
prompt >> matlab

Here prompt >> is the operating system prompt. A window should open and start MATLAB.
When you run MATLAB under the window system, whether you start from the menu or a system
prompt, a small MATLAB logo window will appear while the program is loading and disappear
when MATLAB is ready to use. MATLAB will present the (double arrow) prompt
>>

You are now in MATLAB. From this point on, individual MATLAB commands may be given at the
program prompt. They will be processed when you hit the return key. A MATLAB session may
be terminated by simply typing
>> quit

or by typing exit at the MATLAB prompt.

Macintosh and PC Platforms. Click on the icon for MATLAB or STUDENT-MATLAB. Most of
your interaction with MATLAB will take place in a Command window, where you will be presented
with the prompt
EDU>>

The procedure for leaving MATLAB is the same as any other program operating on these platforms.

Chapter 2

43

Both PC and Macintosh platforms come with a variety of programming and application
development tools. For example, the PC platform supports an integrated M-file editor, a visual Mfile editor, and performance profiler. The Macintosh platform comes with an M-file debugger and
Workspace browser.

Online help
Online help is available from the MATLAB prompt, both generally (listing all available
commands)
>> help
[a long list of help topics follows]

and specifically
>> help demo
[a help message on MATLABs demo program follows].

The version command will tell you which version of MATLAB you are using. The helpdesk
command
>> helpdesk

will initiate the execution of a World Wide Web (WWW) browser for MATLAB Online Reference
Documentation. Check it out.

44

2.3

Engineering Programming in Matlab

Variables and Variable Arithmetic

MATLAB supports a variety of data types for the desktop solution of engineering computations. Some problems will require the representation of scalar numbers or variables whose
values are scalar numbers. Other types of computations require the use of complex numbers containing real and imaginary components. Moreover, solutions to a range of engineering problems
may be expediently computed through the use of one-dimensional arrays and/or multidimensional
matrices of scalars and complex numbers. This section explains how variables work in MATLAB.
The names and (data) types of MATLAB variables do not need to be declared because
MATLAB does not distinguish between integer, real, and complex values. In fact, any variable can
take integer, real, or complex values. Like most programming languages, variable names in MATLAB should be chosen so that they do not conflict with function or subroutine names, command
names, or the names of certain values. However, name conflicts are bound to occur and we deal
with this topic below in the subsection entitled handling name conflicts.

Defining Variables
The equality sign is used to assign values to variables:
>> x = 3
x =
3
>>

As we soon see, variable names can be assigned to scalars, vectors (i.e., one-dimensional matrices),
and matrices. Generally speaking, variable names will be a mixture of letters, digits, and the underscore character. The first character in a variable name must be a letter. Although variable names can
be of any length, MATLAB Version 5. requires that they be unique within the first 31 characters. It
is also important to bear in mind that variable names in MATLAB are case sensitive, meaning that
variables x and X are distinct. Indeed, at this point, X is not even defined.
Table 2.1 shows some of the special built-in variable names and numbers used in MATLAB.
The following script shows how you can learn their values from the MATLAB interpreter.
>> eps
eps =

45

Chapter 2

================================================================================
Variable Name
Meaning
Value
================================================================================
ans
Represents a value computed in an expression
but not stored in a variable name.
eps
Represents the floating point precision for
2.2204 e-16
the computer being used. It is the smallest
amount with which two values can differ in
the computer.
"i" and "j"
Imaginary unit in a complex number.
sqrt(-1)
pi
Pi
3.1415926 ..
NaN
Represents "not a number." NaNs crop up in
undefined expressions (e.g., division by zero)
and in matrix elements where data is missing.
inf
Infinity typically results from a division by
( |><| )
zero or an arithmetic overflow.
clock
The current time is represented in a six element
row vector containing year, month, day, hour,
minute, and seconds.
date
The current date is represented in a character
string format.
flops
Floating point operations count.
================================================================================

Table 2.1. Special Variable Names and Numbers in MATLAB

2.2204e-16
>> pi
ans =
3.1416
>> help pi
PI

3.1415926535897....
PI = 4*atan(1) = imag(log(-1)) = 3.1415926535897....

>>

Points to note are:

1. The machines round-off, the smallest distinguishable difference between two numbers as represented in MATLAB, is denoted eps.
2. The variable ans keeps track of the last output that was not assigned to another variable. You
can use this property in a sequence of calculations where the output of one computation is

46

Engineering Programming in Matlab

used in the following calculation.

Program output can be suppressed by simply appending a semicolon (;) to command lines, as in
>> x = 3;
>> x
x =
3
>>

More than one command may be entered on a single line if the commands are separated by commas
or semicolons. To see how MATLAB handles these cases, try typing the command sequence
>> x = 3, y = 4; z = 5

% define and initialize variables x, y, and z.

In MATLAB, the % symbol indicates the beginning of a comment and, as such, the MATLAB
interpreter will disregard the rest of the command line.

Arithmetic Expressions
Table 2.2 summarizes the meaning and order of evaluation of MATLAB operators in arithmetic expressions involving scalars and variables. With the exception of the left division operator,
arithmetic expressions in MATLAB follow a fairly standard notation. The MATLAB commands
>> 2+3;
>> 3*4;
>> 42;

% Compute the sum "2" plus "3"

% Compute the product "3" times "4"
% Compute "4" raised to the power of "2"

are examples of basic arithmetic operations. The lower half of Table 2.2 shows the precedence and
order of operator evaluation for arithmetic expressions. Operators having the highest precedence
(i.e., a low precedence number) are evaluated first. So, for example, power operations are performed
before division and multiplication, which are done before subtraction and addition. For example,
>> 2+3*42;

generates ans = 50. That is,

47

Chapter 2
Meaning Of Arithmetic Operators
============================================================================
Operator
Meaning
Example
============================================================================

Exponentiation of "a" raised to the

23 = 2*2*2 = 8
power of "b".
*
Multiply "a" times "b".
2*3 = 6
/
Right division (a/b) of "a" and "b".
2/3 = 0.6667
\
Left division (a\b) of "a" and "b".
2\3 = 3/2 = 1.5
+
Addition of "a" and "b"
2 + 3 = 5
Subtraction of "a" and "b"
2 - 3 = -1
============================================================================
Precedence Of Arithmetic Expressions
============================================================================
Operators
Precedence
Comment
============================================================================
()
1
Innermost parentheses are evaluated first.

2
Exponentiation operations are evaluated
right to left.
* /
3
Multiplication and right division operations
are evaluated left to right.
\
3
Left division operations are evaluated right
to left.
+ 4
Addition and subtraction operations are
evaluated left to right.
============================================================================

Table 2.2. Meaning and Precedence of Arithmetic Operators

2 + 3*42
==> 2 + 3*16
==> 2 + 48
==> 50

<== exponent has the highest precedence.

<== then multiplication operator.
<== then addition operator.

Arithmetic expressions involving operators of equal precedence are evaluated left to right. Of
course, parentheses may be used to group terms or to make them more readable. For example,
>> (2 + 3*42)/2

generates ans = 25. That is,

(2 + 3*42)/2

<== evaluate expression within

parentheses. Exponent has

48

Engineering Programming in Matlab

==> (2 + 3*16)/2
==> (2 + 48)/2
==> (50)/2
==> 25

highest precedence.
<== then multiplication operator.
<== then addition operator inside
parentheses.
<== then division operator

Even though the addition operator has lower precedence than the divide operation, the order of
evaluation can be easily altered with the use of parentheses. The second key use for parentheses
is for function calls, that is, a function name followed by parentheses containing zero or more
arguments. In this case, the function calls associated with a set of parentheses will be evaluated;
innermost levels of parentheses first. For example, the step by step order of evaluation for
>> 4.0*sin( pi/4 + pi/4 )

is
4*sin( pi/4 + pi/4 )
==> 4*sin( pi/4 +

==> 4*sin( 0.7854

==> 4*sin( 0.7854
==> 4*sin( 1.5708
==> 4*1.0

<== begin evaluation of left-hand

side multiplication.
pi/4 )
<== evaluate expression within
function parentheses, starting
with leftmost division.
+ pi/4 )
<== evaluate right-hand side
division.
+ 0.7854 ) <== evaluate sum.
)
<== sin(pi) function call.
<== finish evaluation of left-hand
side multiplication.

==> 4.0

In this example, sin(x) is a function call to compute the sine of angle x measured in radians.
MATLAB has a flops function for counting the number of floating point operations
needed to complete a MATLAB command or block of MATLAB commands. We can easily verify,
for example, that five floating point operations are needed to evaluate the arithmetic expression
>> 4*sin( pi/4 + pi/4 );
>> flops
ans =
5
>>

49

Chapter 2

Numerical Precision of MATLAB Output

All arithmetic is done to double precision, which for 32-bit machines means to about 16
decimal digits of accuracy. MATLAB automatically prints integer values as integers and floating
point numbers to four decimal digits of accuracy, with blank lines being inserted between textual
lines of output. Exponential format is automatically used when the value of a number falls outside
the range of numbers that can be printed using the default format.
=========================================================================
MATLAB Command
Meaning
Example
=========================================================================
format short
Default -- 4 decimal places
20.1235
format long
Output printed to
20.12345678901234
14 decimal places
format short e
Exponential format with
2.0123e+01
4 decimal places
format long e
Exponential format with
2.012345678901234e+01
14 decimal places
=========================================================================

Table 2.3. Formatting Options for MATLAB Output

A summary of formatting options for the printing of the variable
>> a = 20.12345678901234

is shown in Table 2.3. Two useful commands are format long, which instructs MATLAB to
display floating point numbers to 16 digits of accuracy, and format compact, which instructs
MATLAB to abbreviate its output by removing all blank lines. Consider, for example, the script of
code
>> x = 2.345

% define variable "x".

x =
2.3450
>> format compact
% "compact" version of output.
>> format long
% "double precision" output.
>> x30
ans =
1.271409381050112e+11
>>
>> format short
% switch back to "default" output.

50

Engineering Programming in Matlab

To save space in this tutorial, all our calculations from this point on are conducted with the format
compact in place.

Built-In Mathematical Functions

MATLAB has a platter of built-in functions for mathematical and scientific computations (see Table
2.3). You should remember that the arguments to trigonometric functions are given in radians. Also
bear in mind that MATLAB has functions to round floating point numbers to integers: They are
round, fix, ceil, and floor.
Example. Verify that
sin(x)2 + cos(x)2 = 1.0

for some arbitrary values of x. The MATLAB code is

>> x = pi/3;
>> sin(x)2 + cos(x)2 - 1.0
ans =
0
>>

Alternatively, we can write the arithmetic expression in terms of the ans variable storing the angle.
>> pi/4;
ans =
0.7854
>> sin(ans)2 + cos(ans)2 - 1.0
ans =
0
>>

Active Variables. When you want to know the active variables, you can use who. For example,
>> who
Your variables are:
ans
x
>>

The command whos gives a detailed listing of the active variables, together with information on
the number of elements, and their size in bytes.
The command clear removes an item from the active variable list. For example, try
typing clear x.

51

Chapter 2

Trigonometric Functions
===============================================================================
Function
Meaning
Example
===============================================================================
sin (x)
cos (x)
tan (x)
asin (x)
acos (x)
atan (x)
atan2 (x,y)

Compute the sine of x (x is in radians).

Compute the cosine of x (x is in radians).
Compute the tangent of x (x is in radians).
Compute the arcsine of x. Complex results
are obtained if abs(x) > 1 for some elements.
Compute the arccosine of x. Complex results
are obtained if abs(x) > 1 for some elements.
Compute the arctangent of the elements of x.

sin(pi)
cos(pi)
tan(pi/4)
asin(pi/2)

=
=
=
=

0.0
1.0
1.0
1.0

acos(pi/2) = 0.0
atan( [ 0.5 1.0 ] ) =
0.4636
0.7854
atan2 (1,2) =
0.4636

Compute the four-quadrant inverse tangent of

the real parts of x and y. This function
returns an angle between -pi and pi.
exp (x)
Compute ex where e is the base for natural
exp(1.0) = 2.7183
logarithms.
log (x)
Compute the natural logarithm of x to the
log(2.7183) = 1.0
base e.
log10 (x)
Compute the logarithm of x to the base 10.
log10(100.0) = 2.0
===============================================================================
Mathematical Functions
===============================================================================
Function
Meaning
Example
===============================================================================
abs(x)
ceil(x)
fix(x)

Return absolute value of x

abs( -3.5 ) = 3.0
Round x toward positive infinity
ceil (-3.8) = -3
Round x toward the nearest integer towards
fix (-3.8) = -3
zero.
floor(x)
Round x toward minus infinity.
floor (-3.8) = -4
rem(x,y)
Return the remainder of x/y. In C, this function is implemented as the modulo operator.
round(x)
Round x to the nearest integer.
round(3.8) = 4
sign(x)
Return -1 if x is less than 0, 0 if x equals sign(3) = 1
zero, and 1 if x is greater than zero.
sqrt(x)
Compute the square root of x.
sqrt(3) = 1.7320...
===============================================================================

Table 2.4. Common Trigonometric and Mathematical Functions

52

Engineering Programming in Matlab

Program Input and Output

Throughout this tutorial, we employ a variety of techniques for input and output (I/O) of
MATLAB variables and matrices. We see, for example, that variables and matrices can be defined
explicitly, with built-in functions, and using data that are loaded into MATLAB from an external
file. In this section we are concerned only with I/O of variables and introduce other features of I/O
as needed.
MATLAB has two functions for the basic input of variables from the keyboard and for
formatted output of variables.
Input of Variables from the Keyboard. MATLAB has a built-in function input that enables the
value of variables to be specified at the keyboard. For example, the command
A = input(Please type the value of Coefficient A :);

will print the message

Please type the value of Coefficient A :

enclosed between single quotes on the computer screen. MATLAB will then wait for a numerical
value to be typed at the keyboard, followed by the return key. In this particular case, the result of
the numerical input will be assigned to variable A.
Formatted Output of Variables. MATLAB uses the function fprintf for formatted output of
messages and numbers. The general syntax is as follows:
fprintf( format , matrices or variables )

The first argument contains the text and format specifications to be printed, and it is followed by
zero or more matrices and variables. MATLABs format argument operates in an almost identical
manner to formatted output in the C programming language. For example, the specification %f is
used for floating point numbers, and %e is used for exponential notation. The command
fprintf( Volume of sphere = %f\n , 3.4 )

generates the output

>> fprintf( Volume of sphere = %f\n , 3.4 )
Volume of sphere = 3.400000
>>

53

Chapter 2

The contents of the function argument between single quotes .... is called a format string. The
sequence of characters %f is an example of a conversion specification. Conversion specifications
begin with a percent sign (%) and are followed by one or more formatting options for output left/right
justification, numeric signs, size specification, precision specification, decimal points, and padded
zeros. In this particular case, the %f simply tells printf() to output the contents of 3.4 as a
floating point number. By default, the floating point number will be printed with six decimal places
of precision.
We use the \n escape character to print a carriage return and advance the screen output
(or cursor) to the beginning of a new-line. Omitting a newline character in printf statements will
result in long stream of output on the same line the output will either run off the right-hand side
of the computer screen, or possibly jump to the next line automatically (actual behavior will depend
on the computer system).
One of the neat features of printf() is its ability to accept a variable number of arguments. In practical terms one argument value must be added to the argument list of printf() for
every field with a % specifier in the string constant. For example, the script of code
>> myPi = pi;
>> fprintf(myPi = %5.3f 2*myPi = %5.3f\n, myPi, 2*myPi )
myPi = 3.142 2*myPi = 6.283
>>

shows how a single call to fprintf() can print two floating point numbers.
Alternative Conversion Specifications. Our approximation of could have been printed using
a number of user-defined formatting options; for example, the left- and right-hand columns of the
following table
Modified printf() statement
======================================
fprintf(myPi = %f \n, myPi );
fprintf(myPi = %14.7f \n, myPi );
fprintf(myPi = %14.7e \n, myPi );
fprintf(myPi = %14.7E \n, myPi );
fprintf(myPi = %14.7g \n, myPi );
fprintf(myPi = %14.7G \n, myPi );
fprintf(myPi = %-14.7f \n, myPi );
fprintf(myPi = %-14.7e \n, myPi );
======================================

Output
=====================
myPi = 3.141593
myPi =
3.1415927
myPi = 3.1415927e+00
myPi = 3.1415927E+00
myPi =
3.141593
myPi =
3.141593
myPi = 3.1415927
myPi = 3.1415927e+00
=====================

show modifications to the print statement shown above. The format specification %14.7f tells
printf() to output the contents of myPi as a floating point number in a field 14 digits wide

54

Engineering Programming in Matlab

(including the decimal point), with 7 digits of precision to the right-hand side of the decimal point.
To ensure that the specification has enough width for the decimal point and the minus sign (if
needed), the total number of digits in the output field should be at least 3 larger than the number of
digits appearing after the decimal point.
The floating point specifications e and E tell printf to output the floating point number
in exponential format. Unless the floating point number to be printed equals zero, the number
before the letter e will represent a value between 1.0.. and 9.99... The two-digit part after the e
represents an exponent value expressed as a signed decimal integer. The floating point number will
be approximately equal to the first component multiplied by 10 raised to value of the exponent.
The total number of digits in the output field should be at least 7 larger than the number of digits
appearing after the decimal point. The floating point specifications g and G tell printf to select
the better of the f or e formats. Although the rules for selecting the format are implementation
dependent, as a general guideline, if the number to be printed falls within the range-of-conversion
specification, use of the f format is likely. Otherwise, the floating point number will be printed in
exponential format. In the seventh and eighth print statements, the output is left justified by inserting
a minus (-) at the front of the conversion specification.
Saving and Restoring Variables. To save the value of the variable x to a plain text file named
x.value use
>> save x.value x -ascii

To save all variables in a file named mysession.mat in reloadable format, use

>> save mysession

To restore the session, use load mysession. PC and Macintosh versions also come with a File
menu for saving and retrieving data.

2.4

Matrices and Matrix Arithmetic

A matrix (or array) of order m by n is simply a set of numbers arranged in a rectangular

block of m horizontal rows and n vertical columns. We say

A =

a11
a21
..
.
am1

a12 a1n
a22 a2n
.. . .
..
.
.
.
am2 amn

(2.1)

55

Chapter 2

is a matrix of size (m n). Sometimes we say matrix A has dimension (m n). The numbers that
make up the array are called the elements of the matrix and, in MATLAB, no distinction is made
between elements that are real numbers and complex numbers. In the double subscript notation aij
for matrix element a(i,j), the first subscript i denotes the row number, and the second subscript
j denotes the column number.
By definition, a row vector is simply a (1n) matrix and a column vector is a (m 1)
matrix. The ith element of a vector
h

V =

v1 v2 v3 v4 vn

(2.2)

is simply denoted vi . The MATLAB language has been designed to make the definition and manipulation of matrices and vectors as simple as possible.

Definition and Properties of Small Matrices

Matrices can be introduced into MATLAB by explicitly listing the elements through the
use of built-in matrix functions and by M-files and external data files. We explain how matrices are
defined in M-files and external files in a moment. For now, consider the statements
>> A = [1 2 3 4; 5 6 7 8; 9 10 11 12; 13 14 15 16 ];

and
>> A = [ 1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 16 ]

Both statements create a four-by-four matrix and assign its contents to a variable A. In the explicit
declaration of matrices, a matrix is entered in row-major order, meaning all the first row, then all the
second row, and so forth. Matrix rows are separated by a semicolon or a new line, and the elements
within a row of a matrix may be separated by commas as well as a blank. The elements of a matrix
are enclosed by brackets. Matrix elements that are floating point numbers are specified in the usual
way (e.g., 3.1415926). Blank spaces must be avoided when listing a number in exponential form
(e.g., 2.34e-9).
Row and column vectors are declared as matrices having either one row or column, respectively.

56

Engineering Programming in Matlab

Accessing Matrix Elements. The matrix element located in the ith row and jth column of A is
referred to in the usual way:
>> A(1,2), A(2,3)
ans =
2
ans =
7
>>

Similarly, the elements in a row or column vector may be accessed by simply typing the vector
name followed by the index number inserted between brackets. For example,
>> V = [ 5 4 3 2 1 ];
>> V (2)
ans =
4
>>

% declare a row vector having 5 elements

% access the second element in vector V

The elements of vectors and matrices may be easily modified with statements of the form
>> A(2,3) = 10;
>> V(2)
= 10;

% modify matrix element A (2,3)

% modify vector element V (2)

Size of a Matrix. The MATLAB function size returns a one-by-two matrix containing the number
of rows and columns in a matrix. The script of code
>> size ( [ 1 2 3 4 5; 6 7 8 9 10; 11 12 13 14 15 ] )
ans =
3
5
>>

shows how the size function returns the number of rows and columns in a 3-by-5 matrix.
Equal Matrices. Two matrices A and B are equal if they have the same number of rows and
columns, and all the corresponding elements are equal (i.e., aij = bij for i = 1 m and j =
1 n). In MATLAB, a copy of the matrix
>> A = [ 1 2 3; 4 5 6 ];

can be made by simply writing

57

Chapter 2
>> B = A
B =
1
4

2
5

3
6

>>

Square Matrices. If a matrix has the same number of rows as columns (i.e., m = n), then we say
that it is square and that the matrix is of order n. The group of elements a11 , a22 ann are
called the principal diagonal elements.
Empty Matrices. MATLAB allows for the definition of matrices where one or more dimensions of
the matrix may be empty, that is, contain no matrix elements. For example, the statement
>> A = [];

defines an empty matrix A. The MATLAB function isempty() can be used to test whether a
matrix is empty (see Table 2.5 for more details on calling this function).
Defining Matrices with Built-In MATLAB Functions. Table 2.5 shows that MATLAB has a
variety of built-in functions for the definition of small matrices. The script of code
>> eye (3,4)
ans =
1
0
0
1
0
0
>>

0
0
1

0
0
0

shows, for example, how a matrix of 3 rows and 4 columns with 1s along the matrix diagonal is
generated by the function call eye(3,4).
Diagonal Matrix. A square matrix A whose elements aij = 0 for all i 6= j is called a diagonal
matrix. We write A = diag( a11 , a22 , a33 , , ann ). A diagonal matrix whose elements on the
principal diagonal are all 1 is called the unit matrix or identity matrix. This special
matrix has notation I.
Let X be a row or column vector containing n elements. diag(X) is the (nn) diagonal
matrix with the elements of X placed along the diagonal. Consider, for example
>> X = [ 4 3 2 1 ];
>> diag(X)
ans =

58

Engineering Programming in Matlab

=========================================================================
Function
Meaning
Example
=========================================================================
eye (n)
Returns a n-by-n identity matrix.
eye (3)
eye(m,n)
Returns a m-by-n matrix of ones along the
eye (3,4)
matrix diagonal and zeros elsewhere.
zeros(n)
Returns a n-by-n matrix of zero elements.
zeros (3)
zeros(m,n)
Returns a m-by-n matrix of zero elements.
zeros (3,4)
=========================================================================

Table 2.5. Defining Small Matrices with MATLAB Functions

4
0
0
0

0
3
0
0

0
0
2
0

0
0
0
1

>>

Conversely, if A is a square matrix then diag(A) is a vector containing the diagonal elements of
A. Consider, for example
>> A = [ 1 2 3; 4 5 6; 7 8 9 ];
>> diag (A)
ans =
1
5
9
>>

Lower and Upper Triangular Matrices. A lower triangular matrix L is one where aij = 0 for all
entries above the diagonal. An upper triangular matrix U is one where aij = 0 for all entries below
the diagonal. That is,

L =

a11
a21
..
.

0
a22
.. . .
.
.

0
0
..
.

am1 am2 amn

U =

a11 a12 a1n

0 a22 a2n
..
.. . .
..
.
.
.
.
0
0 amn

(2.3)

59

Chapter 2

The MATLAB functions triu and tril extract the upper and lower sections of a matrix, respectively. For example, the script
>> U = triu(A)
U =
1
2
0
5
0
0
>> L = tril(A)
L =
1
0
4
5
7
8
>>

3
6
9

0
0
9

extracts the upper and lower triangular sections of matrix A defined in the previous section. The
function calls triu(A,k) and tril(A,k) generate square matrices of values from A with zeros
below/above the k-th diagonal. For example, the script of code
>> U1 = triu(A, 1);
>> U2 = triu(A,-1);

generates the matrices

0 2 3
1 2 3

U1 = 0 0 6 and U2 = 4 5 6
0 0 0
0 8 9

(2.4)

Building Matrices from Blocks. Large matrices can be assembled from smaller matrix blocks. For
example, with matrix A in hand, we can enter the following commands:
>> C = [A; 10 11 12];
>> [A; A; A];
>> [A, A, A];

<== generates a (4x3) matrix

<== generates a (9x3) matrix
<== generates a (3x9) matrix

As with variables, use of a semicolon with matrices suppresses output. This feature can be especially
useful when large matrices are being generated. If A is a 3-by-3 matrix, then
>> B = [ A, zeros(3,2); zeros(2,3), eye(2) ];

will build a certain 5-by-5 matrix. Try it.

60

Engineering Programming in Matlab

Reading and Saving Datasets

Suppose that the array of numbers
1
2
3

10.0
15.0
14.0

.... data items removed from file ...

28
29
30

9.0
4.0
0.0

is stored in a file rainfall.dat and that each row of the file corresponds to the date and daily
rainfall measurement (in mm) for one calendar month. The command
>> load rainfall.dat

will read the contents of rainfall.dat into an array called rainfall, having 30 rows and 2
columns. The command
>> save rainfall.dat rainfall -ascii

will save the contents of array rainfall in the data file rainfall.dat. The array elements
will be written in ASCII using 8 digits of accuracy. The command
>> save

will save array rainfall, and all the other matrices and variables in the MATLAB workspace, in
a nonreadable mat file called matlab.mat. The contents of matlab.mat can be reloaded into
MATLAB at a later date by simply reissuing the load command. For an extensive list of save
and load command options, see the online documentation.

61

Chapter 2

Application of Mathematical Functions to Matrices

In your high school mathematics classes you probably learned that sin(x) is the sine of
an angle x measured in radians. A typical MATLAB command could be
>> x = pi/4;
>> sin ( x )
ans =
0.7071
>>

MATLAB makes an important departure from traditional mathematics in the way it deals with
matrices and the mathematical and trigonometric functions listed in Table 2.3. Instead of only
allowing for the computation of mathematical formulae on a single variable (as demonstrated in the
previous example), MATLAB enables mathematical formulae to be computed on the entire contents
of a matrix by writing only one line of code. So how does this work? When we write the statement
>> x = pi/4;

we naturally think of x as being a variable. However, MATLAB, treats x as if it were a matrix

containing only one element. The same numerical result can be obtained by writing
>> sin ( [ pi/4 ] )
ans =
0.7071
>>

If the dimensions of the input matrix are expanded to something like

>> x = [ 1 2 3 4 5; 6 7 8 9 10]
x =
1
2
3
4
5
6
7
8
9
10
>>

then the command

>> sin(x)
ans =
0.8415
-0.2794
>>

0.9093
0.6570

0.1411
0.9894

-0.7568
0.4121

-0.9589
-0.5440

62

Engineering Programming in Matlab

generates a two-by-five matrix containing the sine computations for each of the matrix elements in
x. In other words, MATLAB systematically walks through all the elements in x and computes the
sine of the corresponding matrix element.
We will soon see that this feature of MATLAB applies to nearly all the functions listed
in Table 2.3. The true benefit of this feature lies in the development of MATLAB software for the
solution of problems requiring repetitive calculations. Because the need for looping constructs can
be eliminated in many cases, the complexity of user-written code needed to implement a numerical
algorithm in MATLAB can be significantly simpler than counterpart implementations in C and Java,
for example.

Colon Notation
A central part of the MATLAB language syntax is the colon operator, which produces a
list. For example
>> -3:3
ans =
-3
>>

-2

-1

The default increment is by 1 but that can be changed. For example

>> x = -3 : .3 : 3
x =
Columns 1 through 7
-3.0000
-2.7000
-2.4000
Columns 8 through 14
-0.9000
-0.6000
-0.3000
Columns 15 through 21
1.2000
1.5000
1.8000
>>

-2.1000

-1.8000

-1.5000

-1.2000

0.3000

0.6000

0.9000

2.1000

2.4000

2.7000

3.0000

In this particular case you may think of x as a list, a vector, or a matrix, which begins at -3 and
whose entries increase by .3, until 3 is surpassed. Generally speaking, the colon operator can be
used anywhere in MATLAB code where a generated list is appropriate. Consider, for example, the
block of statements for generating a table of sines
>> x = [0.0:0.1:2.0] ;
>> y = sin(x);
>> [x y]

63

Chapter 2

The first command generates a column vector of elements ranging from zero to two in increments of
0.1. You should note that because sin operates entrywise, the second command generates a column
vector y from the vector x. The third command takes the x and y column vectors and places them
in a 21-by-2 matrix table. Go ahead, try it.
Colon notation can also be combined with the earlier method of constructing matrices. For
example, the command
>>

A = [1:6 ; 2:7 ; 4:9];

generates a 3-by-6 matrix A.

Submatrices
Any matrix obtained by omitting some rows and columns from a given matrix A is called a
submatrix of A. A very common use of the colon notation is to extract rows, or columns, as a sort
of wildcard operator, which produces a default list. For example,
>> A(1:4,3)

is the column vector consisting of the first four entries of the third column of A. A colon by itself
denotes an entire row or column. So, for example,
>> A(:,3)

is the third column of A, and A(1:4,:) is the first four rows of A. Arbitrary integral vectors can
be used as subscripts. For example, the statement
>> A(:,[2 4])

generates a two column matrix containing columns 2 and 4 of matrix A. This subscripting scheme
can be used on both sides of an assignment statement. The command
>> A(:,[2 4 5]) = B(:,1:3)

replaces columns 2,4, and 5 of matrix A with the first three columns of matrix B. Note that the entire
altered matrix A is printed and assigned. Try it.

64

Engineering Programming in Matlab

Matrix Arithmetic
The following matrix operations are available in MATLAB:
Operator
Description
Operator
Description
============================================================
+
addition

transpose
subtraction
\
left division
*
multiplication
/
right division

power
============================================================

These matrix operations apply, of course, to scalars (1-by-1 matrices) as well. If the sizes of the
matrices are incompatible for the matrix operation, an error message will result, except in the case
of scalar-matrix operations (for addition, subtraction, and division as well as for multiplication) in
which case each entry of the matrix is operated on by the scalar.
Matrix Transpose. The transpose of a mn matrix A is the nm matrix obtained by interchanging
the rows and columns of A. For example, the matrix transpose of
"

A=

1 2 3 4
5 6 7 8

is AT =

1
2
3
4

5
6
7
8

(2.5)

The matrix transpose is denoted AT . In MATLAB, the transpose of a matrix is computed by following the matrix name with the single quote [apostrophe]. For example,
>> A = [ 1 2 3 4; 5 6 7 8 ];
>> A
ans =
1
5
2
6
3
7
4
8
>>

We say that a square matrix A is symmetric if A = AT , and it is skew-symmetric if A = -AT .

Matrix Addition and Subtraction. If A is a (m n) matrix and B is a (r p) matrix, then the
matrix sum C = A + B is defined only when m = r and n = p (see Figure 2.1a). The matrix sum is a
(m n) matrix C whose elements are

65

Chapter 2

cij = aij + bij

(2.6)

for i = 1, 2, m and j = 1, 2, n. For example,

"

if
A =

"

and
B =

2 1
4 6
4 2
0 1

,
then
"

C=A+B=

2 1
4 6

"

4 2
0 1

"

6 3
4 7

(2.7)

The MATLAB commands for this computation are

>> A = [ 2
>> B = [ 4
>> C = A +
C =
6
4
>>

1; 4 6 ];
2; 0 1 ];
B
3
7

Matrix subtraction is identical to matrix addition, except that cij = aij bij for i = 1, 2, m and
j = 1, 2, n. The matrix addition and matrix subtraction operations require mn floating point
operations (flops).
Dot Product. Let
h

X=

x1 x2 x3 xn

be a row vector containing n elements and

(2.8)

66

Engineering Programming in Matlab

C(i,j)

A(i,j)

B(i,j)

Row i of A

C(i,j)

Column j of B

Figure 2.1. Schematics of (a) matrix addition and (b) matrix multiplication

Y =

y1
y2
y3
..
.

(2.9)

yn
be a column vector containing the same number of elements. The dot product, also sometimes called
the scalar product or inner product, is a special case of matrix multiplication (see the next section)
and is defined by

X Y =

n
X

xi yi

(2.10)

i=1

For example, the dot product of row vector X = [ 1, 2, 3, 4, 5 ], with its matrix transpose, can be computed in MATLAB by simply typing

67

Chapter 2
>> X = [ 1, 2, 3, 4, 5 ];
>> Y = X;
>> X*Y
ans =
55
>>

The results of this computation can be easily verified by hand

X*Y =
=
=
=

[ 1 2 3 4 5 ]*[ 1 2 3 4 5 ]
1*1 + 2*2 + 3*3 + 4*4 + 5*5
1 + 4 + 9 + 16 + 25
55

The dot product operation on two vectors containing n elements requires n floating point operations
(flops).
Matrix Multiplication. Let A and B be (m n) and (r p) matrices, respectively. The matrix
product A B is defined only when interior matrix dimensions are the same (i.e., n = r). The
matrix product C = A B is a (m p) matrix whose elements are

cij =

n
X

aik bkj

(2.11)

k=1

for i = 1, 2, m and j = 1, 2, n. From a geometric point of view, cij is the dot product of
the ith row of A with the jth column of B (see Figure 2.1b.). Assuming that matrices A and B are as
defined in the previous section:
"

C = AB =

2 1
4 6

# "

4 2
0 1

"

The MATLAB commands for this operation are

>> A = [ 2 1; 4 6 ];
>> B = [ 4 2; 0 1 ];
>> C = A*B
C =
8
5
16
14
>>

24+10 22+11
44+60 42+61

"

8 5
.
16 14
(2.12)

68

Engineering Programming in Matlab

Computation of the matrix product requires 2mnp flops. When the dimensions of the matrices are
approximately equal, then computational work is 2n3 flops.
Example. Because the number of rows equals the number of columns in square matrices, square
matrices can always be multiplied by themselves. The triple matrix product of
>> A = [ 2 1; 4 6 ];

can be evaluated by writing

>> A*A*A
ans =
48
56
224
272
>>

or by simply writing
>> A3
ans =
48
224
>>

56
272

Example. A magic square is a square matrix that has equal sums along all its rows and columns.
For example,
>> magic(4)
ans =
16
2
5
11
9
7
4
14
>> magic(4)

3
10
6
15

13
8
12
1

In this particular case, the elements of each row and column sum to 34. Now we can use matrix
multiplication to check the magic property of magic squares:
>>
>>
>>
>>
>>

A = magic(4);
b = ones(4,1);
A*b;
v = ones(1,4);
v*A;

<== (4x1) matrix containing row sums.

<== (1x4) matrix containing column sums.

69

Chapter 2

Scalar Multiplication of Matrices. Scalars multiply matrices on an element-by-element basis. For

example, with matrix A in place
>> 2*A
ans =
4
2
8
12
>> A/4
ans =
0.5000
1.0000
>> 4\A
ans =
0.5000
1.0000
>>

% multiply elements of matrix "A" by "2".

% right division of "A" by "4".

0.2500
1.5000
% left division of "A" by "4".
0.2500
1.5000

Matrix Division. Left and right division of matrices is a generalization of left and right division in
variable arithmetic. Let A and B be two matrices. Right division of two matrices, written
>> A/B

corresponds to the solution of the linear matrix equations A.X = B. Similarly, left division of two
matrices is written
>> A\B

and corresponds to the solution of the linear matrix equations X.A = B. We discuss the nature of
these solutions in more detail in Chapter 15.
Matrix Inverse. The inverse of a square matrix A, denoted A1 , is given by the solution to the
matrix equations
A A1 = A1 A = I.
In MATLAB, A1 can be computed with inv(A).
Arithmetic on Submatrices. MATLAB provides for the computation of arithmetic operations on
submatrices. Suppose that matrix A is
>> A = [

>>

1
7
13
19

2
8
14
20

3
9
15
21

4
10
16
22

5
11
17
23

6;
12;
18;
24 ];

70

Engineering Programming in Matlab

Columns 2 and 4 of A can be multiplied on the right by the 2-by-2 matrix [1 2;3 4];
>> A(:,[2,4]) = A(:,[2,4])*[1 2;3
A =
1
14
3
20
5
7
38
9
56
11
13
62
15
92
17
19
86
21
128
23
>>

4]
6
12
18
24

Matrix A is altered and printed.

Matrix Element-Level Operations

MATLAB has a special convention whereby a dot positioned in front of the arithmetic
operations listed in the upper half of Table 2.2 will force entry-by-entry computation of the matrix
operation. By definition, the matrix addition and matrix subtraction operations are already computed
on an element-by-element basis. Matrix operations such as multiplication and left and right matrix
division are not. The best way of seeing how this works is to simply walk through a couple of
examples.
Example. We start with
>> [1,2,3,4].*[1,2,3,4]
ans =
1
4
9
16
>>

Here the expression [1,2,3,4].*[1,2,3,4] evaluates to

[1,2,3,4].*[1,2,3,4] ==> [ 1*1 ,2*2 ,3*3 ,4*4 ]
==> [
1 , 4 , 9 , 16 ].

The same result can be obtained with [1,2,3,4].2. Similarly, the expression
>> [1,2,3,4]./[1,2,3,4]
ans =
1
1
1
1
>>

evaluates to a vector of 1s.

Example. Suppose that we wish to verify sin(x)2 + cos(x)2 = 1 for a variety of argument values
x. The block of code

71

Chapter 2
>> x = [ -pi : pi/4 : pi ];

creates a one-by-nine array x containing values that range from to in increments of /4. The
command
>>

sin(x).2 + cos(x).2

generates the output

ans =
1

>>

The order of evaluation is as follows. First, one-by-nine matrices are computed for sin(x) and
cos(x). Element-level operations square each element in sin(x) and then each element in cos(x).
The output is simply the sum of the latter two matrices.

72

2.5

Engineering Programming in Matlab

Control Structures

Control of flow in the MATLAB programming language is achieved with logical expressions, branching constructs for the selection of a solution procedure pathway, and a variety of looping constructs for the efficient computation of repetitious operations.

Logical Expressions
In MATLAB, a logical expression involves the use of relational and logical operands for
the comparison of variables and matrices of the same size. Table 2.5 summarizes the six relational
operators and three logical operators in the MATLAB programming language.
Relational Operators
=========================================
Operator
Description
=========================================
<
Less than
>
Greater than
<=
Less than or equal
>=
Greater than or equal
==
Equal
=
Not equal
=========================================
Logical Operators
=========================================
Operator
Description
Precedence
=========================================

Not
1
&
And
2
|
Or
3
=========================================

Table 2.6. Summary of Relational and Logical Operators

Relational operators allow for the comparison of variables, and matrices of the same size. In the
case of variable comparisons, the outcome will be 1 if the comparison is true, and 0 otherwise.
Indeed in MATLAB, as in other languages such as C and Java, true is represented by a nonzero
integer (usually one) and false is represented by zero. To see how this works in practice, we look at
the script of code:
>> 3 < 5

73

Chapter 2
=========================================================================
Function
Description
=========================================================================
any(x)
For vector arguments x, any(x) returns 1 if any of the
elements of x are nonzero. Otherwise, any(x) returns 0.
When x is a matrix argument, any(x) operates on the columns
of x, returning a row vectors on 1s and 0s.
all(x)

For vector arguments x, all(x) returns 1 if all the

elements of the vector are nonzero. Otherwise it returns 0.
For matrix arguments x, all(x) operates on the columns
of x, returning a row vector of 1s and 0s.

find (x)

This function finds the indices of the nonzero elements

of x. The function argument can be combined with logical
and relational expressions.

isnan (x)

isnan(x) returns 1s where the elements of x are NaNs and

0s where they are not. Recall that NaN means "Not a Number".

finite (x)

finite(x) returns 1s where the elements of x are finite

and 0s where they are not.

isempty(x)

In MATLAB, an empty matrix has a zero size in at least

one dimension. isempty(x) returns 1 if x is an empty
matrix and 0 otherwise.
=========================================================================

Table 2.7. Logical Functions in MATLAB

ans =
1
>> a = 3 == 5
a =
0
>>

In the first example, 3 is less than 5, so the result of the relational expression is true (i.e., 1).
The second example contains both an assignment = operator, and an equality == operator.
Because the precedence of evaluation for the == operator is higher than the assignment operator,
the relational comparison is made first, and then the result is assigned to variable a. In this particular
case, 3 is not equal to 5, and so variable a assumes a value of 0 (i.e., false).
When relational operands are applied to matrices of the same size, as in

74

Engineering Programming in Matlab

>> A = [ 1 2; 3 4 ];
>> B = [ 6 7; 8 9 ];
>> A == B
ans =
0
0
0
0
>> A < B
ans =
1
1
1
1
>>

the result will be a matrix of 0s and 1s giving the value of the relationship between corresponding
entries.
Logical expressions can be combined by using the logical operands, and, or, and not (see
the lower half of Table 2.5). The following table shows how the logical operators can be combined.
=======================================================
A
B
A & B
A | B
A
=======================================================
True
True
True
True
False
True
False
False
True
False
False
True
False
True
True
False
False
False
False
True
=======================================================

Logical operands also apply to matrices of the same size. In the script
>> A = [ 1 2; 3 4 ];
>> B = [ 6 7; 8 9 ];
>> A & B
ans =
1
1
1
1
>>

all the matrix elements in ans evaluate to 1 because matrices A and B contain only nonzero matrix
elements.
MATLAB also has an ensemble of built-in functions for the evaluation of logical expressions involving matrices. A summary of their capabilities is given in Table 2.5.

75

Chapter 2

Selection Constructs
Generally speaking, a selection construct enables the details of a programs problemsolving procedure to be tied to the evaluation of one or more logical expressions. The following
diagram shows the syntax for three commonly used selection constructs in the MATLAB programming language:
If-end construct
================

If-else-end construct
=====================

If-elseif-end construct
=======================

if < condition1 >,

< program1 >
end;

if < condition1 >,

< program1 >
else
< program2 >
end;

if < condition1 >,

< program1 >
elseif < condition2 >,
< program2 >
elseif < condition3 >
< program3 >
.....
elseif < conditionN >
< programN >
end;

In each of these constructs, the block of statements <program1> will be executed when the
logical expression <condition1> evaluates to true. Otherwise, the program control moves to
the next program construction. For the leftmost construct, this means the end of the selection
construct. For the if-else-end construct, the block of statements <program2> will be executed
when <condition1> evaluates to false. For example, in the block of statements
>> a = 2;
>> if a <
c =
else
c =
end;
>> c
c =
4
>>

b = 1;
b,
3;
4;

the logical expression a<b evaluates to false, thereby causing the second block of program statements to be executed.
Another variation is the if-elseif-end construct. MATLAB will systematically evaluate
the sequence of logical expressions <condition1>, <condition2> <conditionN>

76

Engineering Programming in Matlab

until one evaluates to true. After the corresponding block of <program> statements has been
executed, the program control will jump to the end of the if-elseif-end construct.

Looping Constructs
MATLAB provides a number of looping constructs for the efficient computation of similar
calculations. The syntax for the while and for looping constructs is
While-loop construct
=====================

For-loop construct
=====================

while < condition1 >,

< program1 >
end

for i = <array of values>

< program1 >
end

In the while looping construct, the block of statements <program1> will be executed while the
logical expression <condition1> evaluates to true. For example, the following script
A "while looping" construct
===========================
>> i = 2;
>> while ( i < 4 ),
y = 4*i,
i = i + 1,
end
>>

Iteration No
i
i < 4
y
==================================
1
2
3

2
3
4

true
true
false

8
12

shows a simple while loop and an analysis of the values involved in the looping construct.
In the for looping construct, the block of statements <program1> will be executed for
each of the vectors i defined by the column elements in <array>. For example, the looping
construct
A "for looping" construct
=========================
>> for i = [2,4,5,6,10],
c = 2*i
end
>>

Iteration No
i
c
=========================
1
2
3
4
5

2
4
5
6
10

4
8
10
12
20

77

Chapter 2

executes five times. The values for the variable i during execution are successively 2, 4, 5, 6, and
10. When <array> is a two-dimensional matrix, the values of i will be vectors corresponding
to the matrix columns of <array>. For example, an extension of our previous example is
A "for looping" construct
=========================

Iteration No
i
c
==================================

>> for i = [ 2 4 5 6 10;

1 2 3 4 5 ],
c = 2*i
end
>>

1
2
3
4
5

[ 2
[ 4
[ 5
[ 6
[ 10

1]
2]
3]
4]
5]

[
[
[
[
[

4
2
8
4
10 6
12 8
20 10

]
]
]
]
]

Looping constructs may be nested of course. Here is an example of the contents of a matrix being
initialized inside a nested for loop:
MATLAB source code
=======================
>> for i=1:2,
for j=1:2,
A(i,j) = i/j;
end
end
>>

i
j
A(i,j)
=====================================
1
1
2
2

1
2
1
2

A(1,1)
A(1,2)
A(2,1)
A(2,2)

=
=
=
=

1/1
1/2
2/1
2/2

=
=
=
=

1.0
0.5
2.0
1.0

There are actually two loops here, with one nested inside the other; they define A(1,1), A(1,2),
A(2,1), and A(2,2) in that order.
Programming Tip. Looping constructs in the MATLAB programming language need to be used
with care. The first potential problem is with while loops and the implicit danger of the program
control becoming trapped inside a loop because <condition1> never evaluates to false.
A second important issue is performance. Generally speaking, MATLAB code written
with the for and while looping constructs will execute much slower than if MATLABs implied
looping constructs are employed. For example, the block of MATLAB code
>> x = [ 0 : 0.1 : 2.0 ];
>> y = zeros(1,21);
>> for i = 1:21,
y(i) = sin(x(i));
end
>>

78

Engineering Programming in Matlab

will have faster execution if it is simply written

>> x = [ 0 : 0.1 : 2.0 ];
>> y = sin(x);

This is because the former looping constructs are interpreted, and MATLABs implied looping
constructs are executed as a low-level compiled code.

2.6

General-Purpose Matrix Functions

MATLAB has an ensemble of built-in matrix functions that are useful for general-purpose
engineering and scientific computations.

Sorting the Contents of a Matrix

When x is a matrix, sort(x) sorts each column of x in ascending order. For example
>> x = [

1
3
4;
5 -4 16;
-4 -8 -10 ];
>> sort(x)
ans =
-4
-8
-10
1
-4
4
5
3
16
>>

sort(x) also returns an index matrix i, containing a mapping between elements in the x matrix
and the y (sorted) matrix. For example
>> [ y, i ] = sort(x)
y =
-4
-8
-10
1
-4
4
5
3
16
i =
3
3
3
1
2
1
2
1
2
>>

In colon notation y(:,j) = x(i(:,j),j). Hence, by walking down the first column of the
index matrix i, we see that matrix elements x(3,1), x(1,1), and x(2,1) have increasing
values.

79

Chapter 2

Summation of Matrix Contents

When x is a vector, sum(x) returns the sum of the elements of x. When x is a matrix,
sum(x) returns a row vector with the sum over each column. For example,
>> sum ( [

1
5
-4

3
4;
-4 16;
-8 -10 ] )

ans =
2

-9

10

>>

Watch out for this function being used in statistical analysis of experimental data and least squares
analysis.

Maximum/Minimum Matrix Contents

Maximum/Minimum Matrix Contents
The MATLAB function max(A) returns the maximum element in vector A, and min(A)
returns the minimum element in vector A. When A is a matrix, max(A) will return a row vector
containing the maximum value in each column of A. Similarly, min(A) will return a row vector
containing the minimum value in each column of A. For example, the script of code
>> A = [ 1 2 3 4; 5 6 7 8 ];
>> max (A)
ans =
5
6
7
8
>> min (A)
ans =
1
2
3
4
>>

shows how the maximum and minimum values of each column of A are computed.
Some MATLAB functions can return more than one value. In the case of max, the interpreter returns the maximum value and also the column index where the maximum value occurs. For
example, the script of code
>> [m, i] = max(A)
m =
5
6
7
i =
2
2
2
>>

8
2

80

Engineering Programming in Matlab

shows how the maximum value in each column of A is located at index 2.

Random Numbers
The MATLAB function rand returns numbers and matrices of numbers containing elements that are uniformly distributed between zero and one. Theoretical considerations indicate that
the average value of these elements should be close to 0.5.
For example, the function call rand(3) will return a three-by-three matrix with random
entries. The function call
>> A = rand(10,30);

generates a 10-by-30 matrix of elements uniformly distributed between zero and one, and assigns
the result to matrix A. The average value of the matrix elements can be computed by simply writing
>> average = sum(sum(A))/300
average =
0.5059
>>

In the nested function call sum(sum(A)), the sum of the elements in each matrix column is
computed, and then the sum of the matrix column sums is evaluated. The average value of the
matrix elements corresponds to the sum of matrix element values in A divided by the total number
of matrix elements, 300 in this case.
The function randn generates matrices with elements chosen from a normal distribution
with mean 0.0 and variance 1.0.
The mean value and scatter of uniform and normally distributed random numbers may be
adjusted by multiplying the contents of rand and randn by suitable linear transformations. For
example, to obtain a 3-by-5 matrix of elements uniformly distributed between -1 and 1, we can
simply write:
>> B = 2*rand(3,5) - ones(3,5)
B =
-0.1650
0.8609
-0.8161
0.3735
0.6923
0.3078
0.1780
0.0539
-0.1680
>>

0.4024
0.8206
0.5244

-0.4751
-0.9051
0.4722

Sequences of random numbers will be generated when the functions rand and randn are called
repeatedly. Sequences of random numbers that are repeatable in other words, the same sequence of

Chapter 2

81

random numbers will be the same each time the MATLAB program is executed can be generated
by initializing a seed number. The syntax is
>> rand (seed, n );
>> randn (seed, n );

where n is the seed number (greater than unity). Further information on initializing sequences of
random numbers in MATLAB can be obtained by typing help rand.

82

2.7

Engineering Programming in Matlab

Program Development with M-Files

We have so far assumed in this tutorial

that all MATLAB commands will be typed in
at the keyboard. Practical considerations dictate that this mode of operation is suitable only
for the specification of the smallest problems,
perhaps a handful of MATLAB commands or
less. A much better problem-solving approach
is to use a text editor to write the commands
in an M-file, and then ask MATLAB to read
and execute the commands listed in the M-file.
This section describes two types of M-files.
Script M-files correspond to a main program
in programming languages such as C. Function
Figure 2.2: Matlab Graphic
M-files correspond to a subprogram or userwritten function in programming languages such as C. An M-file can reference other M-files, including referencing itself recursively.

User-Defined Code and Software Libraries

Computer programs written in MATLAB are a combination of user-defined code that
is, the computer program code we write ourselves and collections of external functions located
in software libraries or MATLAB toolboxes. Software library functions are written by computer
vendors, and are automatically bundled with the compiler. Software libraries play a central role in
the development of C programs because the small number of keywords and operators in C is not
enough to solve real engineering and scientific problems in a practical way. What really makes C
useful is its ability to communicate with collections of functions that are external to the user-written
source code. This is where much of the real work in C programs takes place.
The ANSI C standard requires that certain libraries are provided with all implementations
of ANSI C. For example, the standard library contains functions for I/O, manipulation of character
strings, handling of run-time errors, dynamic allocation and deallocation of memory, and functions for C program interaction with the computers operating system. Mathematical formulae are
evaluated by linking a C program to the math library. Engineering application programs may also
communicate with graphical user interface, numerical analysis, and/or network communications
libraries.
Generally speaking, if there is a library function that meets your needs, by all means use

Chapter 2

83

it. The judicious use of library functions will simplify the writing of your C programs, shorten the
required development time, and enhance C program portability. Software is said to be portable if it
can, with reasonable effort, be made to execute on computers other than on the one on which it was
originally written.

Program Development Cycle

When you are a novice programmer, the two most important issues you are likely to be
concerned about are learning the syntax of the C language and becoming familiar with the step-bystep details of planning, writing, compiling, running, testing, and documenting small C programs.
The pathway of C program development is summarized in Figure 2.3. The key steps are as follows:
1. Develop Pseudocode. Pseudocode is a nickname for an English description of ideas that will
eventually be translated into a programming language. Like most engineering design processes, development of pseudocode descriptions is inherently iterative. Pseudocode descriptions begin with general high-level statements of the required problem solving procedures and
are successively refined until they look more or less like a program. For us, pseudocode is
a description of the program steps that can eventually be written in C.
Beginning the program development cycle with a pseudocode problem description is important because it forces us to deal with design issues at an early stage, when corrections and
changes are easy to make. For example, with a pseudocode description of a problem solving
procedure in hand, the correctness of an algorithm can be tested by stepping through it in your
head or on scratch paper. Most experienced programmers will attest to the fact that errors in
logic are the most difficult to find, especially in the latter stages of program development.
Getting programs to compile (syntax, structure, etc.) is the easy part.
2. Edit Program. The pseudocode is translated into a C program by using a text editor for
the creation of source code files containing a description of the problem to be solved. Most
computer systems support a variety of text editors. Because some of them will be much easier
to learn and use than others, it is a good idea to ask your instructor for advice on the best place
to start and for information on basic editor commands to insert and modify text.
All file names in our C programs will consist of a base name with an optional period and
suffix. The first character of the file name must be a letter, possibly followed by more letters
and numbers. C source code files are where the C language description of our engineering
problem descriptions will go and, by convention, they will be stored in files having names
with a .c suffix. Header files contain function declarations and symbolic constants, and act
as an interface between user-defined C code and one or more external software libraries. By
convention, header file names will have a .h suffix. We soon see that object files are generated

84

Engineering Programming in Matlab

Software
Libraries
Programmer uses editor to
Input

create source code file(s).

Write

Program

Pseudo-code

Source Code

Compiler

Program

Loader /

Executable

Object Code

Linker

Program

Syntax Errors
Programmer edits source

Program Design Errors

Output

code file to correct errors

Document

Test /

Program

Verification

Figure 2.3. Program Development Cycle for MATLAB (needs fixing!!!!)

by the compiler and have a .o suffix. The suffix convention is used by compilers to identify
the type of information stored in a file.
3. Compile Program. A C compiler is a computer program that translates high-level problem
descriptions, written in the C language, into equivalent low-level language descriptions that
can be understood by a particular type of computer (i.e., the computer hardware). You should
think of the C compiler as a type of machine that only understands the grammatical rules of
the C programming language. If the compiler does not receive proper input it can only try
to interpret the error, report what is wrong by printing error messages, and exit. If the rules
of the language are satisfied, however, the compiler will generate one or more object files
containing low-level machine language instructions. The linker is a software program that
resolves all cross references among object files, connects the user-written code to relevant
software library modules and, finally, generates a single executable program file. The chain
of compilation activities is summarized by the center row of blocks in Figure 2.3.
4. Performance Analysis. Performance analysis consists of an evaluation of the program performance with respect to the design specifications. As indicated on the right-hand side of Figure
2.3 most computer programs will receive some form of input from the keyboard, or perhaps
from a file, and generate output to the computer screen, or perhaps to a file. The critical
question to ask at this point is does the computer program do what it was designed to do?

Chapter 2

85

5. Debug and Test Program. The first few drafts of a C program nearly always contain errors (or
bugs). You will need to correct (or debug) your program if there are syntax, executional, or
logic errors, as determined in the performance analysis or compilation. Syntax errors are those
encountered when the program is compiled into machine code; if you have typing errors, then
the compiler will not be able to understand what you are doing. Execution errors are found
when the program is run, for example, if the bounds of an array are exceeded or, perhaps, if
division by zero occurs. Logic errors are usually discovered after the program has finished
executing or during execution. If execution does not proceed as expected, then perhaps there
is a flaw in the logic of the program. In each case, symbolic debuggers such as dbx [8] may
be used to interactively run the program and detect errors (see following comment). It may be
faster and easier to go back and reconsider the design of the program. Have you effectively
transferred the design to C, or is there a flaw in the design? Answering these questions first
may lead to a solution faster than learning a whole new bag of tricks with a debugger.
6. Document Code. You should document your source code so that other programmers know what
the source code does and how it may be used. The process of documenting your program
should be integrated into the program development cycle. Some (experienced) programmers
will tell you that they first develop the code and then document it. It is our experience,
however, that with the pressures of life being what they are (everyone is pressed for time),
this strategy of development results in code that works but is poorly or never documented.

In this solution of practical engineering problems, several iterations of Steps 1 through 6 may be
required before a fully developed, correct, working program is obtained. Indeed, the purpose of
the looping constructs in Figure 2.3 is to show that when an error occurs in the language syntax,
program design, and/or program performance, the programmer must return to Steps 1 and 2, make
the necessary adjustments to the program design/source code, and recompile and test the program.
M-File Preparation

UNIX Platforms
In a UNIX environment, a standard text editor can be used to prepare the contents of an Mfile. The M-file should be located in the same directory as the MATLAB program. After the M-file
has been created and saved, the list of commands inside the M-file can be executed by moving to
the MATLAB window and typing the M-file name without the .m extension. To be more precise,
suppose that we create a program file
myfile.m

86

Engineering Programming in Matlab

in the MATLAB language. The commands in this file can be executed by simply typing the command myfile at the MATLAB prompt. The MATLAB statements will run like any other MATLAB function. You do not need to compile the program since MATLAB is an interpretative (not
compiled) language.
UNIX commands can be invoked from inside the MATLAB environment by preceding the
command by a ! (e.g., !ls will list the files located in the current working directory).
Macintosh and PC Environments
An M-file can be prepared in a Mac/PC environment by clicking on new in the file menu.
A new window will appear. After the list of commands has been typed into the file, it can be saved
by clicking on save as in the edit menu. The file can be executed from the command window by
typing the M-file name without the .M extension. A second pathway in the Macintosh environment
is to click on save and go in the file menu.

Script M-Files
Suppose that we use our favorite text editor to create a file called sketch.m, containing
[x y] = meshgrid(-3:.5:3, -3:.5:3);
z = x.3 + y.3 + x.2 - y.2
surf(x,y,z);

Without getting into too many details on MATLAB graphics at this point, the first two lines generate
a rectangular grid of x,y,z points in a format suitable for a three-dimensional MATLAB plot. The
function call surf(x,y,z) asks MATLAB to generate a surface plot. The simple command
>> sketch

will force MATLAB to open the file sketch.m and systematically execute the sequence of commands within. The result, shown at the top of Section 13.6, has been captured from MATLAB
graphics on a UNIX workstation, and will be the same as if you had entered the three lines of the
file at the prompt.

Function M-Files
Function files provide extensibility to MATLAB by allowing you to create new problemspecific functions having the same status as other built-in MATLAB functions. Unlike some other
traditional programming languages, each function file contains one function. Functions are like

87

Chapter 2

scripts, but for the purposes of enhancing computational speed, are compiled into a low-level bytecode when called for the first time.
Square Root Calculation
In this section we develop a MATLAB function file called sqroot.m to compute the
square root of a positive number N via the recursive relationship


xn+1

1
N
= xn +
.
2
xn

(2.13)

A good initial estimate of the square root is xo = N/2. The program uses the criterion
|

xn+1 xn
|
xn

(2.14)

for a test on convergence, where is a very small number. In our implementation, is taken as eps,
the floating point precision for the computer being used (see Table 2.1).

Computer Program 2.1 : Square Root Calculation

function sqroot(x)
% SQROOT Compute square root by Newtons method
% Check that value of function argument is positive.
if x <= 0,
error(In sqroot() : argument x must be positive);
end;
% Initial guess.
xstart = x/2;
% Iteration loop to compute square root.
for i = 1:100
xnew = ( xstart + x/xstart)/2;
disp(xnew);
if abs(xnew - xstart)/xnew < eps,
break,

%
%
%
%

new estimate of square root.

print xnew.
check convergence of iterations.
iterations.

88

Engineering Programming in Matlab

end;
xstart = xnew;

% update estimate of square root.

end

When MATLAB executes a function M-file for the first time, it will open the appropriate text M-file
and compile the function into a low-level representation that will be stored within MATLAB. For
those cases where an M-file function references other M-file functions, they will also be compiled
and placed in MATLABs memory.
The command sequence:
>> format long
>> format compact
>> sqroot(20)

loads the contents of sqroot.m into MATLAB memory and then computes the square root of 20
by iteration
>> sqroot(20)
6
4.66666666666667
4.47619047619048
4.47213779128673
4.47213595499996
4.47213595499958
4.47213595499958
>>

A function M-file will terminate its execution when either a return statement is encountered or,
as is the case in this example, an end-of-file (EOF) is reached.
The first line of Program 2.1 contains declarations for the function name, sqroot, and its
input argument(s). Without this line the file would simply be a script file. MATLAB requires that
M-file functions be stored in files having the same name (i.e., the function sqroot must be stored
in an M-file called sqroot.m).
Your MATLAB programs should contain lots of comments telling the reader in plain English what is occurring. Some day that reader will be you, and you will wonder what you did.
Comment lines up to the first noncomment line in a function M-file are the help text returned when
you request online help. For Program 2.1, this means
>>

help sqroot

89

Chapter 2

SQROOT Compute square root by Newtons method

>>

We also urge you to use the indented style that you have seen in the previous programs. It makes
the programs easier to read, the program syntax easier to check, and forces you to think in terms of
building your programs in blocks.
Matrices and variables in a function file have scope that is local by default. The sqroot
function is simple enough that all the relevant details can be handled with x, xstart, and xnew.
The statement disp( xnew ) displays the matrix xnew as text, without printing the matrix
name.
Error messages play an important role in nearly all computer programs since they help you
debug future programs. Because the algorithm in Program 2.1 can only compute the square root of
positive floating point numbers, we need to check that sqroot() is called with positive argument.
The block of MATLAB code
if x <= 0,
error(In sqroot() : argument x must be positive);
end;

prints an error message when argument x is negative. For example, the script of code
>> sqroot(-4)
Error using ==> sqroot
In sqroot() : argument x must be positive
>>

tries to computer the square root of -4. The function error() displays the message enclosed
within single quotes and causes an exit from the M-file to the keyboard.
Variations on Standard Use. It is important to bear in mind that our implementation of sqroot()
is very minimal. As already demonstrated, function calls of the type sqroot(20) are possible.
However, we want the ability to integrate function calls into expressions involving matrices. This
can be done by changing the function declaration to
function [y] = sqroot(x)

Now expressions of the type

90

Engineering Programming in Matlab

>> x = 3.5 * sqroot ( 10000 )
x =
350
>>

are possible.
Statistics of Experimental Data
In this example, we show how the mean and standard deviation of an array of experimental
data can be computed inside a single MATLAB function. Suppose that
x = [x1 , x2 , x3 , , xN ]T

(2.15)

is a column vector of data points collected in an experiment. The mean value of the data points is
given by
"

N
1 X
xi
x =
N i=1

(2.16)

and the standard deviation by

"

x =

N
X
[xi x ]2
i=1

#1/2

"

N h i
1 X
x2 x 2
N i=1 i

#1/2

(2.17)

For the case where an experiment generates a two-dimensional array of experimental data, Equations (2.16) and (2.17) can be applied to each column in the two-dimensional matrix.

Computer Program 2.2 : Statistics of Experimental Data

function

[mean, stdev] = stat(x)

% STAT Mean and standard deviation

%
For a vector x, stat(x) returns the
%
mean and standard deviation of x.
%
For a matrix x, stat(x) returns two row vectors containing,

91

Chapter 2
%

respectively, the mean and standard deviation of each column.

[m

n] = size(x);

if m == 1
m = n;
end

% handle case of a row vector

mean = sum(x)/m;
stdev = sqrt(sum(x. 2)/m - mean.2);

The first line of Program 2.2 tells MATLAB that function stat will accept an array argument, and
return a one-by-two matrix containing the mean and standard deviation of the experimental data
points. Notice that mean and stdev are both computed inside the function body.
As indicated in the documentation for Program 2.2, when x is a row or column vector, stat(x)
returns the mean and standard deviation of x. For a general matrix x, stat(x) returns two row
vectors containing, respectively, the mean and standard deviation of each column. With Program
2.2 placed in the M-file stat.m, the MATLAB commands
>> y = [1:10];
>> [ym, yd] = stat(y)
ym =
5.5000
yd =
2.8723
>>

compute the mean and standard deviation of the integers 1 through 10. The mean and standard
deviation of the entries in the vector y are assigned to ym and yd, respectively.
Variations on Standard Use. Two variations on the standard use of stat.m are possible. First,
single assignments can also be made with a function having multiple output arguments. For example,
>> xm = stat(x)

(no brackets needed around xm) will assign the mean of x to xm. In this case, standard deviation
of x will be lost. Second, when x is a two-dimensional matrix, stat(x) will return the matrices
containing mean and standard deviation of each column in x. For example, when

92

Engineering Programming in Matlab

>> x = [ 1 2; 3 4; 5 6 ];

stat(x) gives
>> [ xm, xstd ] = stat(x);
>> xm
xm =
3
4
>> xstd
xstd =
1.6330
1.6330
>>

Programming Tip. Generally speaking, function M-files are more difficult to debug than script
M-files because you cannot use MATLAB to print the value of variables inside the function. We
therefore suggest that you develop function files first as script files and then once the script file
works properly, encapsulate the script inside a function declaration with appropriate arguments and
comment statements.

Handling Name Conflicts

Suppose that we do not know sum is a built-in function and type the MATLAB statement
>>
>>
>>
>>

x =
y =
z =
sum

1;
2;
3;
= x + y + z;

with the intent of using sum to represent the sum of values stored by the variables x, y, and z.
The name sum now represents a variable and MATLABs built-in sum function is hidden (you can
check this with the command who).
When a name is typed at the prompt or used in an arithmetic expression, the MATLAB
interpreter evaluates the name by systematically walking through four steps
1. It looks to see if the name is a variable.
2. It looks to see if the name is a built-in function.
3. It looks in the current directory to see if the name matches a script file (e.g., sum.m).
4. It looks in the MATLAB search path for a script file matching the name (e.g., sum.m).
Clearing the variable sum (i.e., by typing clear sum) reactivates the built-in function sum.

93

Chapter 2

2.8

Engineering Applications

Now that we are familiar with MATLABs matrix and M-file capabilities, we work step by
step through the design and implementation of four programs. They are
1. A MATLAB program that computes and plots the relationship between Fahrenheit and Celsius
temperature.
2. A MATLAB program that computes and plots the time-history response (i.e., displacement
versus time and velocity versus time) of an undamped single degree of freedom (SDOF)
oscillator.
3. A MATLAB program that will prompt a user for the coefficients in a quadratic equation, and
then compute and print the roots.
4. A MATLAB program that reads experimental data from a data file, and then computes and
plots a histogram of the data. The mean and standard deviation of the experimental data are
computed with stat.m (see Section 2.7).
A key issue in the design of almost every MATLAB program is the problem of finding a good balance among the use of user-defined code and built-in MATLAB library functions. The advantages
of user-defined functions are that they enable the development of customized problem-solving procedures. However, writing and testing user-defined code can be a very time-consuming process.
Library functions have the benefit of enabling reuse of code that has already been written and thoroughly tested. A judicious use of library functions can result in significant reductions in the time
and effort needed to write and test MATLAB programs.
Obtaining a good balance in the use of user-defined code and library functions requires
experience and, to some extent, is an art. This means that novice MATLAB programmers must
place a priority on becoming familiar with the availability and purposes of library functions.
Temperature Conversion Program

Problem Statement. The relationship between temperature measured in Fahrenheit (Tf ) and temperature Celsius (Tc ) is given by the equation
9
Tf = Tc + 32
5

(2.18)

Write an M-file that computes and plots the temperature conversion relationship for the range -50
through 100 degrees Celsius.

94

Engineering Programming in Matlab

Fahrenheit versus Celsius Temperature Conversion
250

200

Temperature (Fahrenheit)

150

100

50

50

100
50

50

100

Temperature (Celsius)

Figure 2.4. Fahrenheit Versus Celsius

In addition to having a suitable title, the vertical and horizontal axes of your graph should be properly labeled. See, for example, Figure 2.4.
Pseudocode. We begin our implementation with the observation that Equation (2.18) is linear. It
follows that the temperature conversion relationship can be plotted by simply drawing a straight
line between the (Tc , Tf ) coordinates evaluated at Tc = -50 and Tc = 100. The pseudocode for this
problem is
allocate two-by-one arrays to hold the Celsius and
Fahrenheit temperatures.
set the Celsius array values to -50 and 100 degrees.
compute the Fahrenheit temperatures.
plot the array values.
label the horizontal/vertical axes.

95

Chapter 2
add a plot title.

Program Source Code. Here is the program source code.

Computer Program 2.3 : Temperature Conversion Program

%
%
%
%
%
%

======================================================================
temperature.m -- Compute and plot a graph of Celsius versus Fahrenheit
for the range -50 through 100 degrees Celsius.
Written By : Mark Austin
March 1997
======================================================================

% Allocate arrays for Celsius and Fahrenheit temperatures.

tempC = zeros(2,1);
tempF = zeros(2,1);
% Compute temperatures at graph end-points.
tempC(1)
tempC(2)

= -50; tempF(1)
= 100; tempF(2)

= 9*tempC(1)/5 + 32;
= 9*tempC(2)/5 + 32;

% Plot and label the graph

plot( tempC, tempF );
grid;
xlabel(Temperature (Celsius));
ylabel(Temperature (Fahrenheit));
title(Fahrenheit versus Celsius Temperature Conversion);

Running the Program. As indicated in the block of comment statements at the top of the source
code, Program 2.3 is stored in the M-file temperature.m. The command
>> temperature

executes the script file and generates Figure 2.4.

Program Architecture. Table 2.8 shows how key tasks in the problem statement have been mapped
to user-defined MATLAB code and function calls in MATLAB.

96

Engineering Programming in Matlab

=========================================================================
Tasks
*
User-defined code
MATLAB library
=========================================================================
Define arrays tempC and tempF. *
In temperature.m
zeros()
*
Initialize tempC.
*
In temperature.m
Compute tempF.
*
In temperature.m
*
Plot tempF versus tempC.
*
In temperature.m
plot()
Add a grid to the plot.
*
grid
Label the horizontal axis.
*
xlabel()
Label the vertical axis.
*
ylabel()
Add a plot title.
*
title()
=========================================================================

Table 2.8. Mapping of Tasks to MATLAB Code

As already mentioned, all the user-defined MATLAB code is located in the script M-file
temperature.m. A variety of built-in MATLAB functions are employed for the matrix allocation
and to generate the graph.
Analysis of the MATLAB Code. The Celsius and Fahrenheit temperature values are stored in the
arrays tempC and tempF, respectively. Once array tempC has been initialized, the contents of
tempF are evaluated via Equation (2.18). The command
plot( tempC, tempF );

creates a linear plot of array tempF versus tempC. The contents of tempC are the data points along
the horizontal axis. The data points along the vertical axis are stored in tempF. The command grid
adds a grid to the plot. Labels along the horizontal and vertical axes, and a plot title, are added to
the figure with the three commands
xlabel(Temperature (Celsius));
ylabel(Temperature (Fahrenheit));
title(Fahrenheit versus Celsius Temperature Conversion);

Free Vibration Response of Undamped Single Degree of Freedom (SDOF) System

Problem Statement. The free vibration response of an undamped single degree of freedom (SDOF)
oscillator is given by

y(t) = y(0) cos(wt) +

v(0)
sin(wt)
w

(2.19)

97

Chapter 2

where t is time (seconds), and m and k and the mass and stiffness of the system, respectively.
y(t)
p
and v(t) are the displacement and velocity of the system at time t. By definition, w = (k/m) is
the circular natural frequency of the system.
Write an M-file that will compute and plot the displacement versus time (i.e., y(t) versus t) and
velocity versus time (i.e., v(t) versus t) for the time interval 0 through 10 seconds when mass, m =
1 and stiffness, k = 10. The initial displacement and velocity are y(0) = 1 and v(0) = 10, respectively.
To ensure that your plot will be reasonably smooth, choose an increment in your
p displacement and
velocity calculations that is no larger than 1/20th of the system period T = 2 (m/k).
Pseudocode. We begin our analysis by noting that v(t) is simply the derivative of the displacement
with respect to time. Hence, in mathematical terms

v(t) =

dy(t)
= y(0)w sin(wt) + v(0) cos(wt)
dt

(2.20)

Since neither y(t) nor v(t) are linear functions, we implement our solution in three stages. First, a
two-dimensional array will be allocated for the system response storage. Components of the system
response (i.e., y(t) and v(t)) will then be computed and stored in the array columns. Finally, the
array contents will be plotted. The pseudocode for this problem is
define number of points, "npoints", for plotting.
setup array response (npoints,3) for storing t, y(t), and v(t).
column 1 will store the time values
column 2 will store the system displacement
column 3 will store the system velocity
define problem parameters;
sdof mass "m"
sdof stiffness "stiff"
time increment dt for plotting purposes.
initial displacement "x0"
initial velocity
"v0"
for i = 1 to
column 1
column 2
column 3
end loop

npoints
of array = time t = (i-1)*dt;
of array = y(t)
of array = v(t)

plot y(t) versus t for t = 0 to 10 seconds.

hold the plot.
plot v(t) versus t for t = 0 to 10 seconds.

98

Engineering Programming in Matlab

add a grid to the plot.
label the horizontal/vertical axes.
add a plot title.

Notice how certain statements have been indented to show the structure of the eventual program.
Once preliminary ideas are written this way the steps can be refined until the pseudocode is very
close to the target language.
The fidelity of the system response evaluation is controlled by the variable npoints. In addition to using several arrays for storage and plotting purposes, the pseudocode employs a looping
construct for the systematic evaluation of the system response at regular intervals between t = 0 and
t = 10 seconds.
Program Source Code. Here is the program source code.

Computer Program 2.4 : Free Vibration Response of Undamped SDOF System

%
%
%
%
%

=================================================================
sdof.m -- Compute dynamic response of sdof system.
Written By : Mark Austin
March 1997
=================================================================

% Setup array for storing and plotting system response

nopoints = 501;
response = zeros(nopoints,3);
% Problem parameters and initial conditions
mass
stiff
w
dt

=
=
=
=

1;
10;
sqrt(stiff/mass);
0.02;

displ0
= 1;
velocity0 = 10;
% Compute displacement and velocity time history response
for i = 1 : nopoints
time = (i-1)*dt;
response(i,1) = time;

Chapter 2

99

response(i,2) = displ0*cos(w*time)
+ velocity0/w*sin(w*time);
response(i,3) = -displ0*w*sin(w*time) + velocity0*cos(w*time);
end
% Plot displacement versus time
plot(response(:,1), response(:,2));
hold;
% Plot velocity versus time
plot(response(:,1), response(:,3));
grid;
xlabel(Time (seconds));
ylabel(Displacement (m) and Velocity (m/sec));
title(Time-History Response for SDOF Oscillator);

Running the Program. Assume Program 2.4 is stored in the script M-file sdof.m. The command
>> sdof

generates the curves shown in Figure 2.5.

How do we know that these graphs are correct? From a mathematical viewpoint, we expect that
the natural period of this system will be
T = 2*pi*sqrt(m/k) = 6.282/sqrt(10) = 2.0 seconds,

A quick visual inspection of Figure 2.5 reveals that both y(t) and v(t) oscillate with a natural period
of 2 seconds (the time-step increment, dt = 0.02 sec easily satisfies the stated criteria for a smooth
graph). The second point to notice is that at t = 0 seconds, the displacement and velocity graphs both
match the stated initial conditions. Moreover, you should observe that because the initial velocity is
greater than zero, we expect the y(t) curve to initially increase. It does. A final point to note is the
relationship between the displacement and velocity. When the oscillator displacement is at either its
maximum or minimum value, the mass will be at rest for a short time. In mathematical terms, peak
values in the displacement curve correspond to zero values in the velocity curve.
Program Architecture. The left- and right-hand sides of Table 2.9 show how key tasks in the
SDOF problem statement have been mapped to user-defined MATLAB code and calls to MATLAB
functions. Once again, because this problem is relatively straightforward, all the user-defined source
code is located within one M-file, sdof.m.

100

Engineering Programming in Matlab

TimeHistory Response for SDOF Oscillator
15

Displacement (m) and Velocity (m/sec)

10

10

15

5
6
Time (seconds)

10

Figure 2.5. Time-history response of SDOF oscillator

Built-in MATLAB functions are used for the response array allocation, the sin() and cos()
trigonometric calculations, and the square root calculation. Of course, we could have substituted
sqrt() in the MATLAB library with the sqroot() function from Program 2.1. Both implementations give the same numerical result. However, use of the sqroot() function requires that
MATLAB automatically locate and compile sqroot.m after it has been referenced from sdof.m.
Analysis of the MATLAB Code. The variables displ0 and velocity0 store the SDOF displacement and velocity at time, t = 0. The looping construct
for i = 1 : nopoints
time = (i-1)*dt;
response(i,1) = time;
response(i,2) = displ0*cos(w*time)
+ velocity0/w*sin(w*time);
response(i,3) = -displ0*w*sin(w*time) + velocity0*cos(w*time);
end

101

Chapter 2

============================================================================
Tasks
*
User-defined code
MATLAB library
============================================================================
Setup array response (npoints,3). *
In sdof.m
zeros()
Compute circular natural freq w.
*
sqrt()
*
Compute y(t) and v(t).
*
In sdof.m
sin()
*
cos()
*
Plot y(t) and v(t) versus t.
*
In sdof.m
plot()
Hold the graphics.
*
hold
Label the horizontal axis.
*
xlabel()
Label the vertical axis.
*
ylabel()
Add a grid to the plot.
*
grid
Add a plot title.
*
title()
============================================================================

Table 2.9. Mapping of Tasks to MATLAB Code

systematically walks along the rows of the response array and evaluates the time t, displacement
y(t), and velocity v(t) for columns one through three of response. The execution speed will be
rather slow because the looping construct is interpreted. The command
plot( response(:,1), response(:,2) );

draws the contents of column two in array response versus column one. In other words, a plot of
y(t) versus t is drawn. The graph of v(t) versus t is generated with
plot( response(:,1), response(:,3) );

A far more efficient way of computing and storing the system response is with the commands
time
= 0.0:0.02:10;
displ
= displ0*cos(w*time)
+ velocity0/w*sin(w*time);
velocity = -displ0*w*sin(w*time) + velocity0*cos(w*time);

The first statement generates a (1x501) matrix called time having the element values 0, 0.02, 0.04
.... 10.0. The dimensions of matrices displ and velocity are inferred from the dimensions of
time with the values of the matrix elements given by the evaluation of formulae on the right-hand
side of the assignment statements. The required plots can be generated with
plot(time, displ);
hold;
plot(time, velocity);

102

Engineering Programming in Matlab

The first statement creates a plot of vector displ versus vector time. The hold command places
a hold on the current plot and all axis properties so that subsequent graphing commands may be
added to the existing graph. The third statement creates the plot of v(t) versus t.
Of course, the benefits of fast evaluation of the SDOF formulae and convenient storage of
the results in array response can be combined. The block of code
time = [ 0:0.02:10 ];
response(:,[1]) = time;
response(:,[2]) = displ0*cos(w*time)
+ velocity0/w*sin(w*time);
response(:,[3]) = -displ0*w*sin(w*time) + velocity0*cos(w*time);

defines a (501x1) matrix called time, and then uses the second technique to compute (501x1)
matrices of x(t) and v(t). The results of these calculations are assigned to the columns of response
with submatrix notation
response(:,[k])

for the kth column of response.

Quadratic Equation Solver

Problem Statement. It is well known that the roots of a quadratic equation are given by solutions
to
p(x) = ax2 + bx + c = 0

(2.21)

A number of solution cases exist. For example, when both a = 0 and b = 0, we consider the equation
to be extremely degenerate and leave it at that. When a = 0 and b 6= 0, we consider the case
degenerate; in this situation, the equations reduce to p(x) = bx + c = 0, which has one root.
Otherwise, we have
"

Roots =

b2 4ac
2a

(2.22)

The term b2 4ac is known as the discriminant of the quadratic equation.

Write a program that will interactively prompt a user for the equation coefficients a, b, and c at
the keyboard, and then compute and print its roots along with an appropriate message for the type

103

Chapter 2

of solution. Appropriate messages are extremely degenerate, degenerate, two real

roots, and two complex roots. For example, if a = 1, b = 2, and c = 3, the computer program
should print
Two complex roots : Root 1 = 1.0000 + 1.414214 i

(2.23)

Root 2 = 1.0000 1.414214 i

(2.24)

Pseudocode. Because this problem contains numerous details and types of solutions, we refine our
pseudocode over several iterations of development.
Iteration 1. A first draft pseudocode version for this task might look similar to the following:
announce the quadratic equation solving program.
request a, b, and c from the user.
calculate the discriminant.
if the discriminant is >= 0 then the roots are real:
compute and print real roots x1 and x2
else the roots are complex
the real part is -b/2a
the imaginary part is the square root of ( the
negative of the discriminant ) / 2a
print the results.

Iteration 2. Using successive refinement we can develop this pseudocode further into
print quadratic equation solving program
print Enter the coefficients a, b, and c
read the coefficients
discriminant = b2 - 4ac
if the discriminant is >= 0 then equation has real roots x1, x2.
x1 = -b/2a - sqrt( discriminant) / 2a
x2 = -b/2a + sqrt( discriminant) / 2a
else, the discriminant is < 0, and roots are complex numbers
(re, im) with
re = -b/2a
im = sqrt( - discriminant) / 2a
print Complex roots

Iteration 3. Iteration 2 still does not look like MATLAB code so at least one more stage of development is desirable

104

Engineering Programming in Matlab

print quadratic equation solving program
print Enter the coefficients a, b, and c
input coefficients a, b, c
if( a is equal to zero )
compute roots to determinate equations
exit
discriminant = b2 - 4ac
if (discriminant >= 0.0)
% The roots are real
root1 = -b/2a - sqrt( discriminant )/2a
root2 = -b/2a + sqrt( discriminant )/2a
print The first root is root1
print The second root is root2
else
% The roots are complex
re = -b/2a
im = sqrt( -discriminant )/2a
print The first complex root is, re, im
print The second complex root is, re, -im

This version of pseudocode is very close to MATLAB, and we are ready to begin writing the MATLAB code for our quadratic equation solver.
Program Source Code. The source code for Program 2.5 is contained in two files, a script M-file
called quadratic.m and a function M-file called discriminant.m. Here are the details.

Computer Program 2.5 : Compute Roots of Quadratic Equation

%
%
%
%
%
%
%
%
%
%

========================================================================
quadratic.m -- Coefficients are read in from keyboard
Roots of Quadratic are printed to screen.
Note : Naive implementation of quadratic equation solver. This algorithm
does not take into account possible loss of accuracy when two
floating point numbers of almost equal size are subtracted.
Written By : Mark Austin
July 1997
========================================================================

% Print Welcome Message

disp(Welcome to the Quadratic Equation Solver (Version 1));
disp(====================================================);

105

Chapter 2
% Prompt User for Coefficients of Quadratic Equation
disp(Please enter coefficients
A = input (Enter coefficient a
B = input (Enter coefficient b
C = input (Enter coefficient c

for equation a.x2 + b.x + c);

: );
: );
: );

% Print Quadratic Equation to Screen

fprintf(Equation you have entered is : %g.x2 + %g.x + %g\n, ...
A, B, C);
% Compute Roots of simplified equations : A equals zero
RootsFound = 0;
if A == 0 & B == 0,
fprintf(Cannot solve extremely degenerate equation );
fprintf(%14.8g = 0.0\n, C );
RootsFound = 1;
end;
if A == 0 & B = 0 & RootsFound == 0,
Root1 = - C/B;
fprintf(Degenerate root : Root = %14.8g\n, Root1 );
RootsFound = 1;
end;
% Compute Roots of Quadratic Equation : A not equal to zero
if RootsFound == 0,
Discriminant = discriminant(A,B,C);
if Discriminant >= 0,

% Compute discriminant of
% quadratic equation.
% Case for two real roots

Root1 = -B/2.0/A - sqrt( Discriminant )/2.0/A;

Root2 = -B/2.0/A + sqrt( Discriminant )/2.0/A;
fprintf(Two real roots : Root1 = %14.8g\n, Root1 );
fprintf(
: Root2 = %14.8g\n, Root2 );
else
% Case for complex roots
fprintf(Two complex roots : Root1 = %14.8g + %14.8g i\n, ...
-B/2.0/A, sqrt( -Discriminant )/2.0/A);
fprintf(
: Root2 = %14.8g + %14.8g i\n, ....
-B/2.0/A, -sqrt( -Discriminant )/2.0/A);
end;
end;

106

Engineering Programming in Matlab

function

[ discrim ] = discriminant( A, B, C )

% DISCRIMINANT : Compute discriminant in quadratic equation.

%
discrim = B*B - 4*A*C;

Running the Program. The following script of code

>> quadratic
Welcome to the Quadratic Equation Solver (Version 1)
====================================================
Please enter coefficients for equation a.x2 + b.x + c
Enter coefficient a : 1.2
Enter coefficient b : 3.4
Enter coefficient c : 5.6
Equation you have entered is : 1.2.x2 + 3.4.x + 5.6
Two complex roots : Root1 =
-1.4166667 +
1.6308655 i
: Root2 =
-1.4166667 +
-1.6308655 i
>>

shows a typical session of I/O for Program 2.5. You should verify that this solution is correct by
substituting a = 1.2, b = 3.4, and c = 5.6 into Equation (2.22).
Although this example does not demonstrate it, a key limitation of this program is the absence
of checking for loss of numerical accuracy that occurs when two floating point numbers of almost
equal size are subtracted. Situations of this type arise, for example, when a = 1.0, b = 1000000.0,
and c = 1.0. An algorithm for overcoming this problem is explained in Problem ?? of the C tutorial.
Program Architecture. Table 2.10 shows how key tasks in the quadratic equation solution procedure have been mapped to user-defined MATLAB code and MATLAB function calls.
Other than than the discriminant computation in discriminant.m, the heart of our quadratic
equation solver is contained in quadratic.m. When MATLAB executes quadratic.m for
the first time and encounters the reference to a function discriminant, the function M-file
discriminant.m will be located and compiled into MATLABs memory.

107

Chapter 2

==========================================================================
Tasks
* User-defined Code
MATLAB Library
==========================================================================
Announce quadratic program.
* In quadratic.m
disp()
Read equation coefficients.
* In quadratic.m
input()
*
Compute discriminant.
* In discriminant.m
Square root computation.
* In quadratic.m
sqrt()
Solve quadratic equation.
* In quadratic.m
Print roots of quadratic equation. *
fprintf()
*
==========================================================================

Table 2.10. Mapping of Tasks to MATLAB Code

The MATLAB function disp() displays messages enclosed within single quotes. fprintf()
displays messages containing formatted output. The function input() prompts a user for keyboard input (there is no point in reinventing the wheel). Finally, we employ the MATLAB function
sqrt() for the square root calculation in Equation (2.22).
Analysis of the MATLAB Code. The script M-file, quadratic.m, contains the commands
needed to compute and print the roots of the quadratic equation. The function M-file, discriminant.m,
contains the function
function

[ discrim ] = discriminant( A, B, C )

accepting three matrix arguments, and returning a matrix

discrim = B*B - 4*A*C;

containing the equation discriminant. We implicitly assume in this function declaration that A, B,
and C will be one-by-one matrices. The execution of discriminant is terminated by an EOF.
Statistical Analysis of Experimental Data

Problem Statement. Suppose that the concentration of spores of pollen per square centimeter are
measured over a 15-day period and stored in a data file expt.dat.
1
2
3
4

12
35
80
120

108

Engineering Programming in Matlab

5
6
7
8
9
10
11
12
13
14
15

280
290
360
290
315
280
270
190
90
85
66

The first and second columns of expt.dat store the day of the experiment and the measured
pollen count, respectively.
Write a MATLAB program that will read the contents of the data file into an array and create
and label a two-dimensional bar plot showing the pollen count versus day. The program should
then compute the mean and standard deviation of the pollen count, and plot and label dashed lines
for the mean pollen count and the mean pollen count one standard deviation.
Pseudocode. We begin our analysis by noting that the mean and standard deviation of the experimental data can be computed by the function stat() in Program 2.2. Hence, the pseudocode for
this problem is
read the contents of file expt.dat into the array expt.
draw and label the bar chart.
call the function stat() to compute the mean and standard
deviation of the experimental data.
construct a working array of coordinate points for plotting the
dashed lines -- horizontal lines are required for:
mean value - 1 standard deviation.
mean value alone.
mean value + 1 standard deviation.

Program Source Code. Here is the program source code.

Computer Program 2.6 : Statistical Analysis of Experimental Data

109

Chapter 2
%
%
%
%
%

====================================================
expt.m -- Statistical analysis of experimental data.
Written By : Mark Austin
July 1997
====================================================

% Store experimental results in array

load expt.dat
% Generate bar plot of experimental results
bar(expt(:,1), expt(:,2),b)
xlabel(Day of Expt);
ylabel(Pollen Count);
% Compute terms from experimental results.
[xm, xd] = stat(expt(:,2))
% Create and display mean value of pollen count
mean_minus = xm(1,1) - xd(1,1);
mean_plus = xm(1,1) + xd(1,1);
data = [

1, xm(1,1), mean_minus, mean_plus;

15, xm(1,1), mean_minus, mean_plus ];

hold;
plot (data(:,1), data(:,2), b);
plot (data(:,1), data(:,3), b:);
plot (data(:,1), data(:,4), b:);
text(1,
xm(1,1) + 10,Mean Pollen Count);
text(1,mean_minus + 10,Mean - Std);
text(1, mean_plus + 10,Mean + Std);

Running the Program. The script of code

>> format compact
>> expt
xm =
184.2000
xd =
114.2309
>>

110

Engineering Programming in Matlab

shows the command needed to run Program 2.6 and the textual output that is generated.

Figure 2.6. Pollen count versus day of experiment.

Figure 2.6 shows the bar chart of pollen count measurements and mean/standard deviation statistics. Solid- and dashed-line summaries of the mean data value and the mean value one standard
deviation are then superimposed on the bar chart.

111

Chapter 2

Program Architecture. Table 2.11 shows how tasks in the problem statement have been mapped
to user-defined MATLAB code and calls to built-in MATLAB functions.
============================================================================
Tasks
*
User-defined code
MATLAB library
============================================================================
*
File holding the experimental data. *
expt.dat
*
============================================================================
*
Read the file expt.dat into an
*
In expt.m
load
array "expt".
*
*
Draw the bar chart.
*
In expt.m
bar()
Label the horizontal axis.
*
xlabel()
Label the vertical axis.
*
ylabel()
*
Compute the mean and std of
*
In stat.m
the experimental data.
*
*
Hold the graphics.
*
In expt.m
hold
Draw the dashed horizontal lines.
*
plot()
Label the dashed lines.
*
text()
==========================================================================

Table 2.11. Mapping of Tasks to MATLAB Code

After the script M-file expt.m has loaded the contents of expt.dat into an array expt containing two columns and created a bar chart of the data, the function M-file stat.m is called to
compute the mean and standard deviation of the experimental data.
Analysis of the MATLAB Code. The command
load expt.dat;

loads the content of data file expt.dat into the array expt. Then the command
[xm, xd] = stat( expt(:,2) )

calls the user-defined function stat defined in stat.m to compute the mean and standard deviation of data values stored in the second column of expt. The mean and standard deviation

112

Engineering Programming in Matlab

are represented by xm and xd, both one-by-one matrices. A two-by-four array, data, holds the
coordinate values of the solid- and dashed-line segments for plotting purposes.
Otherwise, Program 2.6 relies on a variety of function calls to the MATLAB graphics
library to draw and label the bar chart. A detail description of these functions and their capabilities
are found in Chapter 14.
Further Information

1. The MATLAB program comes with a lot of M-file examples. To find their location on your
computer, type the MATLAB command path. This will lead you to some really nifty demos.

2.9

Review Questions

1. What does the helpdesk command do?

2. What is the maximum array size supported by the student edition of MATLAB Version 5.
3. MATLAB output can be rather lengthy. What is an easy way of shortening the length of the
output?
4. What is the purpose of the eps constant?
5. What does the input function do?
6. Briefly explain how precedence of arithmetic operators works in MATLAB.
7. Explain step by step how the arithmetic expression
>> 1/2*( 2 + 3*42 )

is evaluated in MATLAB.
8. Consider the script of code
>> ix = 1;
>> ij = 2*ix;
>> ik = 2*(ix==1) + (3*ij = 6);

What is the value of ik?

9. By default, MATLAB prints floating point numbers to four decimal places of accuracy. How
would you adjust this option?
10. What are the three ways of defining a matrix in MATLAB?

113

Chapter 2

11. What is the output generated by the MATLAB commands?

>> A = [ 1 2; 3 4 ];
>> B = [ A 2*A ]

12. What is the output generated by the sequence of commands?

>> x = [ 1, 2, 3; 4, 5, 6 ];
>> sin(2*x);

13. What does the command

>> x = -10.0: 0.2: 10.0

do?
14. What is the output generated by the sequence of commands?
>> x = [ -pi : pi/2 : pi ];
>> sin(x).2 + cos(x).2

15. If A is a (m n) matrix and B is a (r p) matrix, what restrictions must exist on m, r, n, and

p for the matrix sum C = A + B to be defined?
16. If A is a (m n) matrix and B is a (r p) matrix, what restrictions must exist on m, r, n, and
p for the matrix product C = A B to be defined?
17. Suppose that
>> y = [ 1 2 3 ; 4 5 6; 7 8 9 ];

How would you use the for looping construct to compute the sum of the elements in matrix
y?
How would you use the sum() function to sum the matrix elements in y (a one line answer
will suffice)?
18. Why does the fragment of code

114

Engineering Programming in Matlab

>> y = zeros(1:1000);
>> for i = 1:1000
y(i) = 2*sin(i);
end

execute slower than

>> y = 2*sin([1:1000]);

19. Suppose that a matrix data contains

data = [ 1.5 1.0 3.0;
6.5 -1.2 12.4;
2.5 -1.0 3.8;
2.4 8.1 5.8 ];

How would you use MATLAB to compute the maximum matrix element value in each row
and column of matrix data?

115

Chapter 2

2.10

Programming Exercises

2.1 Beginner. Figure 2.10 shows a mass m resting on a frictionless surface. The mass is connected to two walls by springs having stiffnesses k1 and k2 .

k_1

k_2
m

Figure 2.7. Mass-spring system

The natural period of the mass-spring system is:
r

T = 2

m
k1 + k2

(2.25)

Write a MATLAB program that will prompt a user for m, k1 , and k2 , check that the supplied
values are all greater than zero, and then compute and print the natural period of the massspring system.
2.2 Beginner. Suppose that during squally conditions, regular one second wind gusts produce a
forward thrust on a yacht sail corresponding to
(

F (t) =

4 + 15 t 135 t3
(731 171t) /140

0.0 t 0.3,
0.3 < t 1.0.

(2.26)

F(t) has units kN. Write a MATLAB program that computes and prints F (t) for 0 t 3
seconds. The program output should look something like the following:
Time
Thrust
(seconds)
(kN)
=========================

116

Engineering Programming in Matlab

0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
3.00

4.00
5.64
4.61
4.31
4.00
5.64
4.61
4.31
4.00
5.64
4.61
4.31
4.00

2.3 Beginner. The adjacent figure shows a triangle defined by the vertex coordinates (x1 , y1 ),
(x2 , y2 ), and (x3 , y3 ).
y

( x_2 , y_2 )

( x_3 , y_3 )
( x_1 , y_1 )
x

Write a MATLAB program that will

1. Interactively prompt a user for the x and y coordinates at each of the triangle vertices.
2. Print the triangle vertices in a tidy table.
3. Compute and print the area of the triangle (make sure that it is positive) and its perimeter.
2.4 Beginner. Figure 2.8 is a schematic of an irregular polygon having seven sides. Suppose
that the x and y vertex coordinates are stored as two columns of information in the file
polygon.dat.
1.0
1.0

1.0
5.0

117

Chapter 2

1
0
0
1
0
1

11
00
00
11
000
111
00
11
000
111
000
111
000
111
000
111
000
111
000
111
000
111
000
111
0
1
000
111
0
1
000
111
0
1

00
11
00
11
111
000
00
11
000
111
000
111
000
111
000
111
0
1
000
111
0
1

3
2

1
0
0
1
0
1

1
0
0
1
0
1
2

Figure 2.8. Seven-sided irregular polygon

6.0
7.0
4.0
3.0
3.0

5.0
3.0
3.0
2.0
1.0

Write a MATLAB script file that will read the contents of polygon.dat into an array
polygon, and then compute and print
1. The minimum and maximum polygon coordinates in both the x and y directions
2. The minimum and maximum distance of the polygon vertices from the coordinate system
origin
3. The perimeter and area of the polygon
Note. You should use the max() and min() functions provided by MATLAB for Part 2. In
Part 3, you should use the fact that the vertices have been specified in a clockwise manner.
2.5 Beginner. The single-payment compound-interest factor,

118

Engineering Programming in Matlab

F = P [1 + i]n

(2.27)

defines the future value of a present day investment P that earns an interest rate i for n years.
During the first year P dollars grows to P.(1+i) dollars. In the second year, P.(1+i)
dollars grows to P.(1+i).(1+i) dollars, and so forth. The term (1 + i)n is called the
compound-interest factor.
Write a MATLAB program that will prompt a user for P, i and n. Use your program
to demonstrate that when $1,000 is invested at 12% for 4 years, the future sum is $1,573.50.
2.6 Beginner. Suppose that a savings bank offers a tiered rate of interest that increases with the
account balance as follows:

5%

Interest Rate( balance ) =

8%

10%

$0.0 balance $5, 000,

$5, 000 < balance $10, 000,
$10, 000 < balance,

(2.28)

Write a MATLAB program to:

1. Draw a plot of rate of interest versus balance for account balances up to $25,000.
2. Suppose that a customer deposits $6,000 for 20 years. Compute and print the compound
balance for years 1 through 20.
Note. When the interest rate is not constant over n years, the term (1 + i)n in Equation 2.27
is replaced by (1 + i1 ) (1 + i2 ) (1 + in ).
2.7 Beginner. Leibnezs series is given by:

1 1 1 1

= 1 + + +
4
3 5 7 9

(2.29)

Write a MATLAB script file to compute Leibnizs series summation for 1000 terms. First,
use a for-loop construct to compute the series summation. Repeat the experiment using an

119

Chapter 2

array for the series coefficients, appropriate matrix element-level operations for the alternating
signs. and sum() for the summation of series terms.
Use the MATLAB facilities tic and toc to monitor the time needed to compute
each implementation. How does the relative speed of the two methods vary as a function of
the number of terms in the series?
2.8 Intermediate. Using no more than three lines of MATLAB code, and no looping constructs,
demonstrate that the series summation:

S=

1
1
1
1
+
+
+
+
12 23 34 45

(2.30)

approaches 1.
2.9 Beginner. An efficient way of computing the cube root of a number N is to compute the root
of
f (x) = x3 N = 0

(2.31)

with the method of Newton Raphson, namely:



xn+1 = xn

f (xn )
f 0 (xn )

(2.32)

Substituting Equation (2.31) into Equation (2.32) and rearranging terms gives the recursive
relationship
"

xn+1

1 2x3n + N
=
3
x2n

(2.33)

Write a MATLAB program that will

1. Prompt the user for a number N, and compute the cube root of N via Equation (2.33). The
details of the cube root calculation should be contained within a function M-file called
cuberoot.m. A suitable function declaration is

120

Engineering Programming in Matlab

[ answer ] = cuberoot ( N )

Your M-file function should use Equation (2.14) for a test on convergence.
2. Print the number N, its cube root, and the number of iterations needed to compute the
result.
Hint: Your solution to this problem should be similar to Program 2.1.
2.10 Intermediate. Write and test a function M-file matadd.m for the element-by-element addition of matrices A and B using Equation (2.6). An appropriate function declaration is
[ matrixsum ] = matadd ( A, B )

After your M-file function has checked that matrices A and B have compatible sizes, Equation
(2.6) should be evaluated inside a set of 2 nested for loops.
Write a test program to allocate and initialize matrices A and B, and compute their
sum using the matadd function and MATLABs built-in library for matrix addition. What is
the relative speed of these two approaches?
2.11 Intermediate. Repeat the experiment described in Problem ?? but for the multiplication of
two matrices A and B. An appropriate function declaration is
[ matrixproduct ] = matmult ( A, B )

Again, the matrix multiplication should be computed with a set of 2 nested for loops.
2.12 Intermediate. Monte Carlo methods solve problems by experiments with random numbers
on a computer. They have been around since about the mid-1940s. A relatively straightforward way of estimating is to conduct an experiment where darts are randomly thrown at a
square board of side length D, as shown in Figure 2.10.


Given that the area of the circle Acircle = D2 /4 , and the area of the square Asquare = D2 ,
then
"

Acircle
=4
Asquare

(2.34)

In the Monte Carlo experiment, N darts are thrown at the board. Let Pi = (xi , yi ) be the
coordinate point of the ith dart (i = 1,2,3 ... N). Point Pi is inside the circle if

121

Chapter 2

Length = D.

Length = D.

Figure 2.9. Estimating via integration

"

D2
x2i + yi2
4

(2.35)

If at the conclusion of the experiment X darts have landed inside the circle (X N), then a
Monte Carlo estimate of is


X
4
N

(2.36)

Write a MATLAB program that will

1. Simulate this experiment by generating pairs of uniformly distributed random numbers,
one for the x-coordinate and a second for the y-coordinate. The random numbers should

122

Engineering Programming in Matlab

be scaled so that they cover the interval [ -D/2, D/2]. Calculate whether each data
point Pi lies inside the circle. At the conclusion of the experiment, sum the number of
darts that lie inside the circle and evaluate Equation (2.36).
2. Print the total number of trials in the experiment (N), the total number of trial points
inside the circle (X), and the Monte Carlo estimate of .
3. Draw Figure 2.10 together with the (x, y) coordinates of the dart throwing experiment.
2.13 Intermediate. Extend the functionality of Program 2.1 so that it will compute the square root
of matrix elements in matrices containing only positive matrix elements. For example, the
function call
>> sqroot( [ 1 2 3; 4 5 6; 7 8 9 ] )

should generate the output:

ans =
1.0000
2.0000
2.6458

1.4142
2.2361
2.8284

1.7321
2.4495
3.0000

When one or more of the matrix elements is not positive, sqroot.m should print an appropriate error message and terminate its execution.
2.14 Intermediate. A key limitation of the quadratic equation solver in Program 2.5 is the absence
of checking for loss of numerical accuracy due to subtraction cancelation. We note from
Equation (2.22) that this will occur whenever the quadratic equation has two real roots, and
either coefficient a or coefficient c (or both) is very small compared to coefficient b.
Rather than naively applying Equation (2.22) directly, numerical accuracy can be
improved by computing the quantity
h

Q = 0.5 b + sign(b)

b2 4ac

(2.37)

where sign(b) is a function that gives 1 for positive b and -1 for negative b. Equation
(2.37) follows from multiplying Equation (2.22) by the unity fraction
"

b b2 4ac

b b2 4ac

(2.38)

Chapter 2

123

The roots of the quadratic are c/Q and Q/a, respectively. Extend Program 2.5 so that it
avoids numerical errors due to subtractive cancellation. Develop some testcase problems to
show that your improved implementation works even when Program 2.5 provides inferior
solutions.

Chapter 3

MATLAB Graphics

Graphics are an indispensable part of engineering education and professional practice because they provide insight into the complicated relationships that exist between multi-dimensional
scientific and engineering phenomena. One of the neat features of MATLAB is its graphics capabilities. Using only a few simple MATLAB commands, two- and three-dimensional plots and subplots
can be created, with labels and titles, axes, and grids.

3.1

Simple Two-Dimensional Plotting

MATLAB has an ensemble of functions for the simple two-dimensional plotting of functions and data that take the form (xi , yi ), i = 1, 2, 3, , n. They are
plot(x,y) Draws a linear plot of vector y versus vector x.
semilogx(x,y) Draws a plot of vector y versus vector x, using a logarithmic scale for x and a
linear scale for y.
semilogy(x,y) Draws a plot of vector y versus vector x, using a logarithmic scale for y and a
linear scale for x.
loglog(x,y) Draws a plot of vector y versus vector x, using logarithmic scales for both x and y
axes.
In each of these function calls, the contents of x are the data points along the horizontal axis of
the plot. The data along the vertical axis of the plot is stored in a second vector y. Vectors x and
y should have the same length. If x or y is a matrix, then the vector is plotted versus the rows or
columns of the matrix, whichever line up.
By default, MATLAB will plot each data point pair on a set of axes scaled to cover the
range of values in x and y, and connect the marked data points with straight line segments. The
latter helps to highlight trends implied by the sequence of data point pairs.
124

125

Chapter 3

For function calls that involve logarithmic transformations, it is important to remember that
the logarithm of x is undefined for negative x. MATLAB will handle these accidental oversights by
printing an error message indicating that data points have been omitted from a plot.
Coordinate labels, plot titles, grid, and textual messages can be added to simple twodimensional plots with
grid on Add grid lines to a plot. The command grid off takes them off. The command grid
by itself toggles the grid state.
xlabel(x axis label) Add text beside the x axis on the current axis.
ylabel(y axis label) Add text beside the y axis on the current axis.
title(title of plot) Add text at the top of the current axis.
text( x, y, text) Add text string to a plot where (x, y) is the coordinate of the center leftedge
of the character string taken from the plot axes.
gtext(text) Activate the use of the mouse to position a cross-hair on the graph at which point
the text will be placed when any key is pressed.
Example. Draw the graph of


sin(2x)
y(x) =
2x

for 10 x 10

(3.1)

The sequence of MATLAB commands

>> x = -10 : 0.2 : 10;
>> y = 1/2*sin(2.*x)./x;
Warning: Divide by zero
>> plot(x,y)
>> grid
>> xlabel(x)
>> ylabel(sin(2x)/2x)

generates Figure 3.1.

Vector x is a (1 101) matrix holding the coordinates of domain [-10,10] partitioned
into intervals of 0.2, and y is a vector of the same length giving the values of y = sin(2x)/2x within

126

Engineering Programming in Matlab

the partition. Equation (3.1) is evaluated successfully for each of the elements in matrix x except
x(51) = 0, where a divide by zero occurs. Does this make sense? By writing a Taylor series
expansion for sin(2x) and dividing through by 2x, it is relatively easy to show that y(0) = sin(0)/0
= 1. MATLAB does not pick up on this point. Instead, the element of vector y corresponding to x
= 0 is assigned NaN, and the plot() function treats the array element y(51) as missing data (see
the upper sections of Figure 3.1).
1

0.8

sin(2x)/2x

0.6

0.4

0.2

0.2

0.4
10

0
x

10

Figure 3.1. y = sin(2x)/2x for -10 x 10

Note. One way of mitigating this problem is to replace the zero element(s) in matrix x with eps,
that is
>> x = -10 : 0.2 : 10;
>> x = x + (x==0)*eps

In the second MATLAB command, a component of matrix x will be incremented by eps when the
logical expression (x==0) evaluates to true. Otherwise, the components of x remain unchanged.

127

Chapter 3

Now the y matrix is generated without error, and sin(0)/0 evaluates to 1.

Axes of Plots. In Figure 3.1, MATLAB has automatically fixed the range of x and y values so that
the plotted function fills the space that is available (i.e., xmin = min(x), xmax = max(x)). In doing
so, the missing data item at y(0) has been unintentionally hidden by the border of Figure 3.1.
Perhaps it would be more evident if the range of y values in Figure 3.1 were rescaled to cover the
interval [ -0.4 , 1.2]?
MATLAB provides the function axis to control the scaling and appearance of axes in a
plot. In this tutorial, we make frequent use of the function call
axis( [xmin, xmax, ymin, ymax ] ) Set the range of x and y values in the current plot to [
xmin, xmax] and [ ymin, ymax], respectively.
In addition, the command axis(equal) changes the current axis box size so that equal tick
mark increments on the x- and y-axis are equal in size. Similarly, the command axis(off)
turns off all axis labeling and tick marks. Axis labeling and tick marks is turned back on again with
axis(on).
Thus, the sequence of MATLAB commands
>>
>>
>>
>>
>>
>>
>>
>>

x = -10 : 0.2 : 10;

x = x + (x==0)*eps;
y = 1/2*sin(2.*x)./x;
plot(x,y)
axis([ -10 10 -0.4 1.2 ])
grid;
xlabel(x)
ylabel(sin(2x)/2x)

generates a graph nearly identical to Figure 3.1, in this case with the range of x and y values covering
[-10,10] and [-0.4,1.2], respectively, and with the correct evaluation of sin(2x)/2x at x =
0.
Parametrically Defined Curves. Plots of parametrically defined curves can also be made. For
example, a plot of the parametric curve
[x(t), y(t)] = [cos(3t), sin(2t)] for 0 t 2
can be created with the MATLAB commands

(3.2)

128

Engineering Programming in Matlab

======================================================================
Line type
solid (-),
dashed (-),
dotted (:),
dashdot (-.)
Mark type

point (.),
x-mark (x).

plus (+),

star (*),

circle

(o),

Color

yellow (y), magenta (m),

cyan (c),
red (r),
green (g),
blue (b),
white (w),
black (k).
======================================================================

Table 3.1. Summary of Line Types and Line Colors

>>
>>
>>
>>

t = 0: 0.02: 2*pi;
plot( cos(3*t), sin(2*t) )
grid
axis(equal)

The result is shown in Figure 3.2. Notice how we have used the axis(equal) command to
equalize the scales in the x- and y-axis directions. This makes the parametric plot take its true shape
instead of an oval.
Setting Line and Mark Types, and Colors. As already demonstrated in Figures 3.1 and 3.2,
MATLAB will automatically connect the data points in a plot by solid lines. When the data points
are closely spaced, this gives the appearance of a smoothly drawn curve.
Table 3.1 contains the line and mark types, and color options, that can be used in MATLAB plots.
For example, the script
>> t = 0: 0.04: 2*pi;
>> plot( t.*cos(3*t), 2.*t.*sin(2*t), o )
>> grid

generates the pathway of o points shown in Figure 3.3. In this case, we have deliberately omitted
the axis(equal) command so that the horizontal and vertical axes can be independently
scaled to fill the rectangular space available on a page.
Combinations of line and mark types, and color settings, can be specified by simply
bundling the combined settings into a text string enclosed by single quotes. For example, the command
>> plot( t.*cos(3*t), 2.*t.*sin(2*t), og )

tells MATLAB to plot Figure 3.3 with green circles.

129

Chapter 3
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1

0.5

0.5

Figure 3.2. Parametric contour (x,y) = [ cos(3t), sin(2t) ]

8
6
4
2
0
2
4
6
8
10
12
6

Figure 3.3. Parametric contour (x,y) = [ t*cos(3t), 2t*sin(2t) ]

130

Engineering Programming in Matlab

Clearing Plots. The command clf deletes all objects from the current figure. The command cla
deletes all objects (e.g., lines, text) from the current axes.

Histograms, Bar Charts, and Stem Diagrams

The function specifications that follow are a subset of MATLAB functions available for
plotting histograms, bar charts, stair-step plots, and stem diagrams:
bar(x,y) Draws a bar graph of the elements of vector y at locations specified in vector x. The x
values must be ascending order.
hist(x,y) Plot a histogram of the values in vector y using the bins specified in vector x.
stairs(x,y) A stair-step graph is a bar graph without internal lines. This function call draws a
stair-step graph of the elements of vector y at locations specified in vector x.
stem(x,y) Plots the data sequence y as stems from locations specified in x. Each stem is terminated by a circle positioned at the data value.
For a complete list of options for each function, type
>> help function_name

(e.g., help bar) within MATLAB.

Example. The block of MATLAB code
>> x = 1:6;
>> y = zeros(1,6);
>> for i = 1:600
ii
= ceil( 6*rand(1) );
y(1,ii) = y(1,ii) + 1;
end,
>> bar(x,y)
>> axis([0, 8, 0, 125 ]);
>> grid
>> xlabel(Number on Dice (1 through 6));
>> ylabel(Number of scores in Experiment);

simulates an experiment where a regular die is thrown 600 times and the total number of scores, 1
through 6, is counted and plotted as a bar chart (see Figure 3.4). You should observe that the sum
of column heights 1 through 6 is 600 and that the average column height is 100.
Figure 3.5 is a graphical representation of the same die-throwing experiment, but in this
case, the grid has been removed and the frequency of die scores is displayed as a stem chart.

131

Chapter 3

120

Number of scores in Experiment

100

80

60

40

20

0
0

3
4
5
Number on Dice (1 through 6)

Figure 3.4. Bar chart of scores in die-throwing experiment

120

Number of scores in Experiment

100

80

60

40

20

0
0

3
4
5
Number on Dice (1 through 6)

Figure 3.5. Stem chart of scores in die-throwing experiment.

132

Engineering Programming in Matlab

Multiple Plots
A straightforward way of plotting multiple curves on the same graph is with the command
plot(x,y,w,z, ... ) Opens a graphics window and draws a linear plot of vector y versus vector x,
and vector w versus vector z on the same graph. Again, if x, y, w or z is a matrix, then the
vectors are plotted versus the rows or columns of the matrix, whichever line up.
MATLAB will plot the data set y versus x first, then w versus z and so forth. This strategy of
implementation means that the length of vectors x and y need not be the same as w and z.
The following examples illustrate three ways of making multiple plots on a single graph,
as shown in Figure 3.6.
Example. Consider the script of code
>>
>>
>>
>>
>>
>>
>>

x = 0 : .01 : 2*pi;
y1 = sin(x);
y2 = sin(2*x);
y3 = sin(4*x);
plot( x, y1 ,--, x, y2, :, x, y3,+ )
grid
title (Dashed, line, and dotted line graph)

Three plots are drawn with one call to plot(). Here we use a dashed line and dotted line for the
first two graphs, while a + symbol is placed at each node for the third.
Example. When one of the arguments to plot is a vector and the other is a matrix, plot will
graph each column of the matrix versus the vector. For example
>> x = 0 : .01 : 2*pi;
>> y = [ sin(x), sin(2*x), sin(4*x) ];
>> plot(x,y)

x is an array having one row and 629 columns. y is a matrix having 629 rows and 3 columns.
Holding Figures
So far in this tutorial we have generated all our figures, including those with multiple plots,
with one call to plot. Another way of generating figures with multiple plots is with the command
hold, which freezes the current graphics screen so that subsequent plots can be superimposed on
it.
Example. Figure 3.6 can also be generated with the sequence of commands

133

Chapter 3

Figure 3.6. Overriding default line- and point- types

>>
>>
>>
>>
>>
>>
>>

x = 0 : .01 : 2*pi;
plot( x,
sin(x), -- )
hold
plot( x, sin(2*x), : )
plot( x, sin(4*x), + )
grid
title (Dashed, line, and dotted line graph)

The hold can be released by entering hold again. The command hold on holds the current plot
and all axis properties so that subsequent graphing commands add to the existing graph. Conversely,
the command hold off returns to the default mode whereby plot commands erase the previous
plots and reset all axis properties before drawing new plots.

134

3.2

Engineering Programming in Matlab

Three-Dimensional Plots

MATLAB uses the function plot3, the three-dimensional counterpart of plot, to display
arrays of data points (xi , yi , zi ), i = 1, 2, 3, , n in three-dimensional space. The syntax is
plot3(x,y,z) When x, y, and z are vectors of the same length, this function plots a line in threedimensional space through the points whose coordinates are the elements of x, y, and z.
There are many variations on this functions use, and we simply recommend you type help plot3
for a list of options.
Example. Suppose that 8 experimental measurements (x,y,z) are stored in the rows of a matrix
called data:
>> data = [ 2.5 1.3 0.0;
0.0 2.0 0.0;
1.0 3.0 0.0;
2.5 3.5 4.0;
3.0 1.0 -2.0;
2.0 -1.0 -2.0;
3.5 4.0 -2.5;
0.0 1.0 0.0 ];
>>

The block of MATLAB commands

>>
>>
>>
>>

plot3 ( data(:,1), data(:,2), data(:,3), o )

axis([ -1 4 0 5 -5 5 ])
grid
xlabel(x), ylabel(y), zlabel(z)

generates the data points, grid, and axes in Figure 3.7. The first thing that you should notice about
this script is how we have used colon notation (e.g., data(:,1)) to extract the vectors x, y, and
z from data. This script also demonstrates that many of the optional features we have used with
plot, such as line and mark types, grid, and axis settings, can also be used with plot3.
Now we add the labels Start and End to the first and last points in array data. The
block of MATLAB code
>> a = size ( data )
>> text( data(
1,1) + 0.2 , data(
1, 2), Start );
>> text( data( a(1,1),1) + 0.2 , data( a(1,1), 2), End );

135

Chapter 3

0
Start
End
5
5
4

4
3

3
2

1
y

0
0

Figure 3.7. Three-dimensional plot of experimental data points.

uses the function size to obtain the number of rows and columns in array data a(1,1) contains
the number of rows in data, and a(1,2) the number of columns. The text function then adds
the desired labels at the coordinates of the first and last data points.

3.3

Mesh and Surface Plotting

MATLAB provides a suite of functions for creating three-dimensional mesh and surface
plots of functional relationships

z = f (x, y)

(3.3)

above a rectangular region defined in the x-y plane. MATLAB represents z = f (x, y) as an array of
data points above a regular grid of points lying in the (x-y) plane (see Figure 3.8). The surface shape

136

Engineering Programming in Matlab

(or mesh shape) is highlighted by connecting the neighboring data points by straight-line segments,
with the result in many cases looking like a fishing net.
Three steps are needed to create a mesh or surface plot:
Z
Data Point ( x_i, y_i, z_i )

Matrix "yy" defines coordinates

in y - direction.
Y

Matrix "xx" defines coordinates in x - direction.

Figure 3.8. Three-dimensional mesh above regular grid of points in (x-y) plane
1. Generate Two-Dimensional Grid in (x-y) Plane. A rectangular grid of points in the (x-y)
plane is generated by defining vectors xx and yy for the coordinate positions of the nodes
along the x- and y-axes. Then, the function call
[ x, y ] = meshgrid ( xx, yy )

transforms the domain into rectangular arrays x and y for the efficient evaluation of z =
f (x, y). The rows of the output array x are copies of the vector xx, and the columns of the
output array y are copies of the vector yy.
2. Compute z Values at (x,y) Coordinates of Grid. Now that matrices x and y of coordinates
in the (x-y) domain are place, we can systematically evaluate Equation (3.3) for each (x,y)
coordinate pair. The results are stored in a rectangular array z.
We soon see that by writing f (x, y) in terms of matrix element-level operations, in many
cases, the entire evaluation of (x, y) coordinate pairs can be achieved with only one MATLAB
statement.
3. Draw Mesh and Surface Plots. Three-dimensional mesh and surface plots may be produced
with the functions:

137

Chapter 3

mesh(z) Creates a three-dimensional mesh plot of the elements of matrix z. If z is an

m-by-n matrix, then the x- and y-axes will cover the ranges 1 through m and 1 through
n, respectively.
surf(z) Draws a three-dimensional surface plot of the elements of matrix z. Otherwise, it
is the same as mesh(z).
mesh(x,y,z) Creates a three-dimensional mesh plot of the elements of matrix z, the x- and
y-axes labeled to cover the range of values in matrices xx and yy.
surf(x,y,z) Creates a three-dimensional surface plot of the elements of matrix z, the the xand y-axes labeled to cover the range of values in matrices xx and yy.
The mesh and surf functions can also be used with various color options, and we refer you
to the online MATLAB documentation for the relevant details.
Example. Suppose that we want to create a three-dimensional plot of the function
h

i sin(y)

z = f (x, y) = x2 + y 2

(3.4)

over the domain 10 x 10 and 10 y 10.

The block of MATLAB code
>>
>>
>>
>>

xx = -10: 0.4: 10;

yy = xx;
yy = yy + (yy==0)*eps
[x,y] = meshgrid(xx,yy);

generates matrices xx and yy containing coordinates of 10 x 10 and 10 y 10

divided into intervals of 0.4. The third statement moves points along the line y = 0 to y = eps so
that MATLAB will not generate a divide by zero in its evaluation of Equation (3.4). Finally, the
meshgrid function returns (51 51) matrices x and y containing the x- and y-coordinates in the
rectangular domain. The command
>> z = (x.2 +

y.2).*sin(y)./y

systematically evaluates z, a (51 51) matrix, for each of the (x-y) coordinate pairs in the rectangular domain. Finally, the commands

138

Engineering Programming in Matlab

100
80
60

40
20
0
20
40
10
10

5
5

0
0

5
y

5
10

10

Figure 3.9. Three-dimensional mesh.

>> mesh(xx,yy,z)
>> xlabel(x), ylabel(y), zlabel(z)

generate a three-dimensional mesh plot, with the x- and y-axes scaled to cover the range of points
contained in matrices xx and yy. The result is shown in Figure 3.3.

3.4

Contour Plots

A contour plot is an elevation map containing families of lines connecting regions of equal
elevation. Sometimes it is convenient to think of a contour as a slice of a region at a particular
elevation.
In MATLAB, contour plots are generated from three-dimensional elevation data:
contour(z) Generates a contour plot of matrix z where the matrix elements are treated as heights
above the (x-y) plane.

Chapter 3

139

contour(x,y,z) Generates a contour plot where x and y are vectors specifying coordinates on the
x- and y- axes. Again, z is a matrix whose elements are treated as heights above the (x-y)
plane.
contour(x,y,z,v) Matrices x, y and z are as previously defined. Vector v tells contour to draw
length (v) contour lines at the elevations specified in the elements of v.
Three-dimensional mesh and surface plots may be drawn with a contour diagram lying in the x-y
plane. The relevant function specifications are
meshc(x,y,z) This function is the same as mesh, except that a contour plot is drawn beneath the
mesh.
surfc(x,y,z) This function is the same as surf, except that a contour plot is drawn beneath the
surface.

Example. With matrices xx, yy, x, y, and z in place for Figure 3.3, a two-dimensional contour
map can be drawn and labeled with
>> contour(xx,yy,z)
>> xlabel(x), ylabel(x)

Similarly, a three-dimensional surface plot with a contour map drawn beneath can be generated with
the command
>> surfc(xx,yy,z)

The results are shown in Figures 3.10 and 3.11.

3.5

Subplots

MATLAB graphics windows will contain one plot by default. The subplot command can
be used to partition the graphics screen so that either two or four plots are displayed simultaneously.
Two subwindows can be displayed either side by side or top and bottom. When the graphics
window is partitioned into four subwindows, two are on the top and two are on the bottom.
The syntax for setting up a subplot is

140

Engineering Programming in Matlab

10
8
6
4
2
y

0
2
4
6
8
10
10

0
x

10

Figure 3.10. Contour plot.

100

50

50
10
10

5
5

0
0

5
10

10

Figure 3.11. Combined surface and contour plot.

141

Chapter 3

subplot(i,j,k) The subplot function takes three integer arguments i, j, and k. Integers i and j
specify that the graphics window should be partitioned into an i-by-j grid of smaller windows.
The subwindows are numbered from left to right, top to bottom. Integer k specifies the kth
subplot for the current graphics window.
Example. The sequence of commands that follow generates the array of subplots shown in Figure
3.12.
% Define array of temperatures in chimney cross section
temp = [ NaN
NaN
NaN
NaN
200.0
147.3
96.5
47.7
0.0

NaN
NaN
NaN
NaN
200.0
146.9
96.0
47.4
0.0

NaN
NaN
NaN
NaN
200.0
145.2
93.9
46.1
0.0

NaN
NaN
NaN
NaN
200.0
140.4
88.9
43.2
0.0

200.0
200.0
200.0
200.0
200.0
128.2
78.8
37.8
0.0

147.3
146.9
145.2
140.4
128.2
94.3
60.6
29.6
0.0

96.5
96.0
93.9
88.9
78.8
60.6
40.2
19.9
0.0

47.7
47.4
46.1
43.2
37.8
29.6
19.9
10.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

;
;
;
;
;
;
;
;
];

% Generate contour and surface subplots.

subplot(2,2,1);
contour( temp )
title(Contour Plot);
subplot(2,2,2);
mesh( temp )
title(Mesh Plot);
subplot(2,2,3);
surf( temp )
title(Surface Plot);
subplot(2,2,4);
surfc( temp )
title(Surface Plot with Contours);
% ==============================================
% The End !

In the function calls contour( temp ), mesh( temp ), surf( temp ), and surfc(
temp ), the elements of matrix temp are treated as elevations above the (x-y) plane.
You may recognize array temp from the C tutorial, where we computed the distribution
of temperatures in one fourth of a chimney cross-section. The chimney interior is represented by
the block of NaNs. The zero elements along the right-hand side and bottom of temp represent the

142

Engineering Programming in Matlab

Mesh Plot

Contour Plot
8

200

6
100
4
0
10

10

5
2

0 0

Surface Plot

Surface Plot with Contours

200

200

100

100

0
10

10

5
0 0

0
10

10

5
0 0

Figure 3.12. Temperature profiles in chimney cross-section

outside temperature. The row and column of 200 elements are the temperature along the inside wall
of the chimney.

3.6

Hard Copies of MATLAB Graphics

To create a black-and-white postscript file of a MATLAB figure, just type

>> print name-of-figure.ps

Color postscript files can be generated with

>> print -dpsc name-of-figure.ps

Similarly, to create a color jpg file (i.e., Joint Photographic Experts Group file), try the following
command

143

Chapter 3
>> print -djpeg name-of-figure.jpg

Online information can be obtained by typing help print.

3.7

Preparing MATLAB Graphics for the World Wide Web

The UNIX tool xv can then be used to convert the postscript file into a gif file format,
suitable for reading by WWW browsers.

3.8

Review Questions

1. How can the axes in a two-dimensional plot be constrained to retain the true shape of a figure?
2. How does MATLAB handle the plotting of

f (x) =

(x 2)
(x 2)

(3.5)

at x = 2? What is a good way of avoiding the divide-by-zero scenario?

3. What does the hold command do?
4. How would you draw a three-dimensional surface together with a contour plot of elevation
levels?

144

3.9

Engineering Programming in Matlab

Programming Exercises

3.1 (Beginner). Suppose that column one in the following file, RAIN.TXT, contains the daily
rainfall for one week, and the second column contains the average rainfall for the week.
0.100000
1.000000
0.000000
0.200000
3.560000
0.000000
0.000000

0.694286
0.694286
0.694286
0.694286
0.694286
0.694286
0.694286

Write a MATLAB prorgram that will

1. Read the contents of RAIN.TXT into MATLAB
2. Plot a histogram of the daily rainfall measurements and a horizontal line for the average
weekly rainfall.
3.2 Beginner. The fragment of MATLAB code
>> t = 0:0.01:2*pi;
>> plot( sin(t), cos(t) )

plots a circle of radius 1 centered at the origin.

1. Write a function file, circle.m, that will plot a circle of radius r centered at coordinate
(x.y). A suitable function declaration is
function circle ( x, y, r )

2. With your circle function in hand, write a short M-file that will generate the rectangular
layout of circles shown in Figure 3.13. The circles are positioned along the x coordinates
x = 5 and x = 50 and the y coordinates y = 5 and y = 50. Each circle should have a radius
2.5.
3.3 Beginner. Write a MATLAB script file that will plot and label the function
h

z = f (x, y) = ex

2 y 2

over the domain 2 x 2 and 2 y 2.

sin(y) sin(y)

(3.6)

145

Chapter 3

Figure 3.13. Rectangular layout of circles

3.4 Beginner. Write a MATLAB script file that will plot and label the function
h

z = f (x, y) = int 4.0 sin3 (x) sin3 (y) + 0.5

(3.7)

for the domain 10 x 10 and 10 y 10. In Equation (3.7), int is a function that
truncates the fractional part of a floating point number.
3.5 Intermediate. Figure 3.14 shows a sphere of radius r and density floating in water.
The weight of the sphere will be 4/3r 3 . The volume of the spherical segment displacing
water is 1/3(3rd2 d3 ).
1. Show that the depth of the sphere floating in water is given by solutions to

146

Engineering Programming in Matlab

r
d

Figure 3.14. Sphere floating in water.

f (x, ) = x3 3.x2 + 4. = 0

(3.8)

where x = d/r is a dimensionless quantity.

2. Write a MATLAB program that will compute the depth to which a sphere will sink as a
function of its radius for = 0 to = 1 in increments of 0.1.
3. Use MATLAB to plot and label the results.
Note. Perhaps the most straightforward way of solving this problem is to write a numerical
procedure that computes the root of the cubic equation. This is not as hard as it might seem
since f (0, ) is always greater than zero and f (2, ) is always less than zero. Only one
solution to Equation (3.8) lies within the interval f ([0, 2], ) and so standard root finding
techniques such as bisection and Newton Raphson will work. A more elegant way of solving
this problem is with the contour function.
3.6 Intermediate. Write a block of MATLAB code that will evaluate and plot Equation (3.1)
without error and without moving the x = 0 coordinate to x = eps.
3.7 Intermediate. Write a MATLAB script file that will plot and label the family of ellipses
defined by

[x(t), y(t)] = [A sin(t), B cos(t)]

(3.9)

for 0 t 2, A = 1, and B = 1, 2, 5.
3.8 Intermediate. Write a MATLAB program that will draw a circle, and partition and label
its perimeter into n equal segments. Line segments should then be drawn to connect all the

147

Chapter 3
2

1
0
0
1
00
11
0111111111
1
0
1
00111
11
000000000 1111111111
0001111111
111
0000000000
0
1
0000000000
1111111111
0000000
0000000
1111111
01111111
1
0000000
1111111
0000000000
000000000
111111111
000
111
0000000000
1111111111
0000000000
1111111111
000
111
000000000 1111111111
111111111
0001111111
111
0000000000
0
1
0000000000
1111111111
0000000
0000000
1111111
000
111
01111111
1
0000000
1111111
0000000111
000000000
111111111
000
111
0000000000
1111111111
0000000000
1111111111
000
00
11
00
11
1010
000000000
111111111
000
111
0000000000
1111111111
0
1
0000000000
1111111111
0000000
1111111
0000000
1111111
000
111
00
11
00
11
111111111111
000000000 000000000000
111111111
0000000000
1111111111
000
111
000000000
000
111
0111111111
1
0000000
1111111
0000000
1111111
000000000
111111111
0000000000
1111111111
000
111
000000000
111111111
000
111
0
1
0000000
1111111
0000000
1111111
000000000 111
111111111
0000000000
1111111111
000
000000000
111111111
000
111
0
1
0000000
1111111
0000000
1111111
000000000
111111111
0000000000
1111111111
000
111
000000000
111111111
000
111
0
1
0000000
1111111
0000000
1111111
0
1
00
11
00
11
000000000
111111111
0000000000
1111111111
000
111
000000000
111111111
000
01111111
1
0000000
1111111
0000000
0
1
00
11
00111
11
3

PERIMETER DIVIDED INTO 3 SEGMENTS

PERIMETER DIVIDED INTO 6 SEGMENTS

Figure 3.15. Partitioned circles.

points lying on the circle perimeter. The left- and right-hand sides of Figure 3.15 demonstrate
the required partitioning and labeling for n = 3 and n = 6, respectively.
Devise an algorithm that requires no more than n(n 1)/2 function calls to draw all
the line segments. This is not as hard as it seems simply observe, for example, that once a
line segment has connected points 1 and 6, there is no need to draw a second line connecting
points 6 and 1.
3.9 Moderate. An experiment is conducted on 36 specimens to determine the tensile yield
strength of A36 steel. The experimental results are as follows:
42.3
40.5
39.9
39.2
43.1
41.9

42.1
38.7
45.3
40.7
36.2
40.4

41.8
40.8
41.6
38.5
46.2
39.1

42.4
39.6
36.8
40.1
41.5
38.6

47.7
42.4
45.4
42.8
38.3
46.3

41.4
37.5
44.8
42.5
40.2
39.5

Write a MATLAB program and appropriate M-files that will:

1. Read the experimental test results from a file experimental.data.
2. Compute and print the maximum, minimum, average, and median tensile strengths.
3. Generate a histogram of observations versus yield stress
4. Construct a bar graph or stair-step graph of cumulative frequency versus yield stress.

148

Engineering Programming in Matlab

Note. The average value of the experimental results can be computed using formulae presented in Section 8.5. After the experimental results have been numerically sorted, the median value is the midvalue (if the number of data points is odd) or the average of the two
middle values (if the number of data points is even). The cumulative frequency versus
yield stress is given by
Z

p(x)dx

Cumulative frequency(y) =
0

where p(x) is the probability distribution of tensile yield strengths.

(3.10)

Chapter 4

Solution of Linear Matrix Equations

In this chapter, we learn how MATLAB can be used to solve engineering problems involving linear systems of matrix equations. Linear systems of equations arise in many areas of
engineering including structural mechanics, circuit simulation, control of electrical and mechanical
devices, and finite element analysis. Moderate to large families of equations commonly represent
the state of a system (e.g., equations of equilibrium, energy and momentum conservation).

4.1

Definition of Linear Matrix Equations

We begin with the basics. In expanded form, a system of m linear equations with n unknowns may be written
a11 x1
a21 x1
a31 x1
..
.

+ a12 x2
+ a22 x2
+ a32 x2
..
.

am1 x1 + am2 x2

+ a13 x3 +
+ a23 x3 +
+ a33 x3 +
..
.

+ a1n xn
+ a2n xn
+ a3n xn

= b1
= b2
= b3
..
.

(4.1)

+ am3 x3 + + amn xn = bm

The constants a11 , a21 , a31 , amn and b1 , b2 , bm are called the equation coefficients. Most
often the coefficients will be real numbers, but they could also be complex numbers. The variables
x1 , x2 xn are the unknowns in the system of equations. The matrix equivalent of Equation 4.1
is [A] {X} = {B}, where

[A] =

a11 a12

a1n
..
.
a21 a22
..
..
.
.
am1 amn

x1

, {X} =

..
.
xn

, and {B} =

b1

..
.
bm

(4.2)
149

150

Engineering Programming in Matlab

Equations 4.2 are said to be homogeneous when the right-hand side vector {B} = 0. A system of
equations is said to be under-determined when there are more unknowns than equations (i.e., m <
n). Conversely, a system of equations is termed over-determined when there are more equations
than unknowns (i.e., m > n). Over-determined systems of equations arise in linear optimization
methods, and in the problem of finding the best fit of a low-order equation to experimental data. A
well-known name for the latter application is least squares analysis.

Geometry of Two- and Three-dimensional Systems

Linear algebra plays a central role in the development of numerical equation solvers because it allows for the classification solutions to Equation 4.2. Indeed, before the development of
numerical equation solvers can proceed, we need to understand under what circumstances a system
of equations will have a unique solution. For those cases where a system of equations has more
than one solution, we also need to know how many solutions there will be and how they can be
characterized.
A good way of gaining insight into these issues is to study systems of equations whose
solutions are simple enough to be graphically displayed. This will occur for two- and three- dimensional problems (i.e., when m = n = 2 and m = n = 3).
Two-Dimensional Example. When m = n = 2, the matrix equations [A] {X} = {B} can be
interpreted as a pair of straight lines in the (x1 , x2 ) plane. That is,

a11 x1

+ a12 x2 = b1

(4.3)

a21 x1

+ a22 x2 = b2

(4.4)

The pair of Equations 4.3 and 4.4 may be interpreted as a linear transformation from two-dimensional
coordinate space (x1 , x2 ) into a two-dimensional right-hand side vector space (b1 , b2 ) (see Figure
4.1).
The problem of finding solutions to Equations 4.3 and 4.4 is equivalent to finding points
X = (x1 , x2 ) in the (X1 , X2 ) plane that will be mapped via the transformation A X into the
(B1 , B2 ) plane.
As we soon see, this problem is complicated by three types of solutions, namely, (1) no
solutions to A.X = B, (2) a unique solution to A.X = B, or (3) an infinite number of solutions to
A.X = B.

151

Chapter 4

X_2

B_2
A
[ b1 , b2 ]

[ x1 , x2 ]

Inverse ( A ).

X_1
COORDINATE SPACE

B_1
RIGHT - HAND SIDE SPACE

Figure 4.1. Matrix transformations.

4.2

Hand Calculation Procedures

There are two hand calculation procedures for computing a solution(s) to the system of
equations. In the graphical method, plots of lines in two dimensions are constructed. Solutions
to the system of equations correspond to points where the lines intersect. An example of such a
solution is shown in the left-hand schematic of Figure 4.2, labeled Unique Solution.
In the second method, the equations are premultiplied by constants in such a way that when
they are combined variables will be eliminated. For example, if Equation 4.3 is multiplied by a21
and Equation 4.4 is multiplied by a11 , then we have the system of equations
a21 a11 x1

+ a21 a12 x2 = a21 b1

(4.5)

a11 a21 x1

+ a11 a22 x2 = a11 b2

(4.6)

Subtracting Equation 4.5 from Equation 4.6 gives



x2 =

a11 b2 a21 b1
.
a11 a22 a12 a21

(4.7)

Now x2 can be back-substituted into either Equation 4.3 or Equation 4.4 for the corresponding value
of x1 . You may may have noticed that this procedure is not unique. If Equation 4.3 is multiplied by
a22 and Equation 4.4 is multiplied by a12 , subtraction of equations gives

152

Engineering Programming in Matlab

y

x
Unique Solution

Inconsistent

Multiple Solutions

Figure 4.2. Types of solutions to matrix equations.

a22 b1 a12 b2
x1 =
a11 a22 a12 a21

(4.8)

Now x2 can be back-substituted into either Equation 4.3 or and x1 can be back-substituted into
either Equation 4.3 or 4.4 for the corresponding value of x2 .

4.3

Types of Solutions for Systems of Linear Matrix Equations

We observe that the denominators of Equations 4.7 and 4.8 are the same, and correspond
to the determinant of a (2 2) matrix, namely:
"

det(A) = det

a11 a12
a21 a22

= a11 a22 a12 a21 .

(4.9)

The same principles apply to families of equations containing three unknowns (i.e., m = n = 3).
When the equations are written in the form [A]{X} = {B}, A is a (33) matrix. From a geometric
point of view, each equation describes a plane in three-dimensional space, as shown in Figure 4.3.
The solution to the system of equations corresponds to those points in three-dimensional space that
lie in all three planes. There will only be one such point when det(A) 6= 0. The mathematical
condition det(A) = 0 occurs when two or more planes are parallel. As with the two-dimensional
case, there will be either an infinite number of solutions or no solutions. An infinite number of

153

Chapter 4
PLANE 1

PLANE 3

PLANE 2

UNIQUE SOLUTION

Figure 4.3. Intersection of planes in three-dimensional space.

solutions occurs when all three planes have the same equation (or a nonzero constant multiplied by
the same equations). No solutions occur when at least two of the planes are parallel but do not have
the same equation.
When problem sizes increase above three unknowns, we can no longer rely on graphical methods to find a solution(s). The nature of solutions to linear equations is given instead by
theoretical results from linear algebra. The results are


1. A unique solution {X} = A1 {B} exists when A1 exists (i.e., det [A] 6= 0).
2. The equations are inconsistent when [A] is singular and rank [A|B] 6= rank [A].
3. If rank [A|B] equals rank [A], then there are an infinite number of solutions.


For the case of a unique solution, A1 is the inverse of matrix A, and such a matrix will exist
when the determinant of A is nonzero. An equivalent way of detecting a unique solution is with
the rank of matrix A, that is, the number of linearly independent rows in matrix A. In a nutshell, a
unique solution to the matrix equations will exist when the rank (A) equals the number of rows

154

Engineering Programming in Matlab

or columns in A. In cases 2 and 3 above, the notation [A|B] stands for the matrix A juxtaposed with
matrix B.


The MATLAB function inv (A) will return the inverse A1 of matrix A when it exists.
The MATLAB function det (A) will return the determinant of a square matrix A. The MATLAB
function rank (A) will return the rank of A, otherwise known as the number of linearly independent rows (or columns) in A.

4.4

Case Study Problem : Three Linear Matrix Equations

Suppose that the following three equations describe the equilibrium of a simple structural
system as a function of external loads and computed displacements.
3x1 1x2
1x1 + 6x2
0x1 2x2

+ 0x3 = 1
2x3 = 5
+ 10x3 = 26

(4.10)

This family of equations can be written in the form A.X = B where

3 1
0
x1

[A] = 1
x2 , and {B} =
6 2 , {X} =
5

0 2 10
26
3

(4.11)

In a typical application, matrices A and B will be defined by the parameters of the engineering
problem, and the solution matrix X will need to be computed. In this particular case, the solution
matrix

{X} =

(4.12)

makes the left- and right-hand sides of matrix Equation 4.11 equal. The following script of MATLAB code
>> A = [ 3 -1 0; -1 6 -2; 0 -2 10 ];
>> B = [ 1; 5; 26 ];
>> X = A\B

155

Chapter 4

defines matrices A and B, and computes and prints the solution to A.X = B, namely,
>> X = A\B
X =
1.0000
2.0000
3.0000
>>

Of course, this solution can be verified by first computing the inverse of A, and then postmultiplying
it by matrix B. The relevant details of MATLAB code are
>> Ainv = inv(A)
Ainv =
0.3544
0.0633
0.0633
0.1899
0.0127
0.0380
>> X = Ainv*B
X =
1.0000
2.0000
3.0000
>>

0.0127
0.0380
0.1076

We can use MATLAB to check that this result is consistent with the rank and determinant of matrix
A. First, the script of code
>> rank (A)
ans =
3
>>

shows that the matrix rank is equal to 3, the number of rows and columns in matrix A. We therefore
expect that the matrix equations will have a unique solution. A second indicator of the solution type
is the matrix determinant. The script of code
>> det (A)
ans =
158
>>

shows that the matrix determinant is nonzero, again indicating the presence of a unique solution to
the matrix equations.

156

4.5

Engineering Programming in Matlab

Singular Systems of Matrix Equations

A family of matrix equations A is said to be singular when the individual equations are
dependent. That is, one or more of the matrix equations can be written as a linear combination of
the remaining equations. For example, matrix
>> A = [ 1 2 3;
2 4 6;
4 5 6 ];

is singular because the elements of the second row are simply two times those in the first row. For a
general matrix A, such a relationship may be far from evident, and so we must rely on the rank and
determinant functions to identify singular systems. In this case, the function call
>> rank (A)
ans =
2
>>

highlights the presence of a singular system because the rank(A) = 2 is less than 3, the number
of rows and columns in A. Consequently, the inverse of A will not exist. Should a situation of this
type be accidentally overlooked in an engineering computation, MATLAB will display an error
message and return a solution vector containing NaNs or , depending on the values elements in
matrix A. Consider, for example
>> Ainv = inv(A)
Warning: Matrix is singular to working precision.
Ainv =
Inf
Inf
Inf
>>

Inf
Inf
Inf

Inf
Inf
Inf

The error condition Inf will be propagated to all subsequent arithmetic computations involving
Ainv.
Generally speaking, if a matrix determinant is computed to be exactly zero, then there is no
difficulty in identifying the three types of matrix solutions mentioned in the previous section. But
what about matrices that are nearly singular? Many practical implementations are complicated by
factors such as round-off error, finite precision, and limited ranges of numbers that may be stored in

Chapter 4

157

a computer (i.e., underflow and overflow of numbers). Perhaps matrix A is singular and a nonzero
calculation is due to numerical problems, or perhaps it is not singular. Resolving these issues is far
from a trivial matter.

4.6

Engineering Applications

Now that we are familiar with MATLABs matrix and graphics capabilities, this section
works step by step through the design and implementation of four engineering applications that
require the solution of linear matrix equations. They are:
1. A MATLAB program for the structural analysis of a cantilever truss.
2. A MATLAB program for the electrical analysis of a circuit containing resistors and batteries.
3. A MATLAB program that computes a least squares analysis of experimental data.
4. A MATLAB program that computes and plots the distribution of temperature in a chimney
cross-section.
Each problem description is accompanied by a brief discussion of the theory needed to set up the
relevant matrix equations.

Structural Analysis of a Cantilever Truss

Problem Statement. In the design of highway bridge structures and crane structures, engineers are
often required to compute the member forces and support reactions in planar truss structures. The
analysis of cantilever truss structures is governed by the following principles
1. At each joint, the sum of internal and external forces in the horizontal and vertical directions
must equal zero.
2. The sum of external forces and support reactions in the horizontal and vertical directions must
equal zero.
3. For the entire structure and all possible substructures, the sum of moments must equal zero.
Figure 4.4 shows a six-bar cantilever truss carrying 10 kN loads at joints B and C. The analysis
begins with the arbitrary assignment of element numbers in the truss, numbers for the joints, and
reaction components. We assume that

158

Engineering Programming in Matlab

F6

F5

F4

5m

F3

y
F1
x

F2

10 kN
5m

10 kN
5m

Figure 4.4. Six-bar cantilever truss.

1. The truss elements can only carry axial forces, with tensile axial forces being positive and
compressive axial forces being negative.
2. All the joints are pinned (in other words, the joints cannot transfer moments; only axial forces
from the truss elements).
The forces F1 , F2 , F6 represent the axial forces in truss members 1 through 6, and Rax , Ray ,
Rex , and Rey the support reactions at joints A and E.
The cantilever truss is statically determinate in the sense that axial forces in the elements
and the support reactions can be computed without a knowledge of material properties. Equations
of equilibrium in the horizontal and vertical directions are written for each of the five joints. The
equations are then expressed in matrix form, and solved for the member axial forces and horizontal
and vertical reaction components. For example, the forces acting on joint A are
F5

F1

R_ax
R_ay

159

Chapter 4

The equations of equilibrium in the horizontal and vertical directions are

X

X
F5
F5
Fx = F1 + + Rax = 0
Fy = + Ray = 0
2
2

(4.13)

The equations of equilibrium for joint B are

X

Fx = F1 F2 = 0

Fy = F4 10 = 0

(4.14)

The equations of equilibrium for joint C are

X

X
F3
F3
Fx = F2 + = 0
Fy = 10 = 0
2
2

(4.15)

The equations of equilibrium for joint D are

X

X
F5
F5
F3
F3
Fx = F6 = 0
Fy = F4 + + = 0
2
2
2
2

(4.16)

and for joint E, the equations of equilibrium are

X

Fx = F6 + Rex = 0

Fy = Rey = 0

(4.17)

Equations 4.13 to 4.17 are solved in two steps. First, we put Equations 4.14 to 4.16 in matrix form
and solve for the member forces F1 , F2 F6 . Then we back-substitute the member forces into
equations 4.13 and 4.17 to get the support reactions. In matrix form, Equations 4.14 to 4.16 are

1 1
0 0
0
0
0
0
0
0
0 1
0
0
0
1 1/2 0
0
0
0 1/2 0
0
0
0 1/2 0 1/2 1
0
0 1/ 2 1
1/ 2
0

In step two, the support reactions are given by

F1
F2
F3
F4
F5
F6

10

10

(4.18)

160

Engineering Programming in Matlab

Rax

Ray

R
ex

Rey

1
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

1/2
0
1/ 2
0
0 1
0
0

F1
F2
F3
F4
F5
F6

(4.19)

Program Source Code. Program 4.1 allocates memory for matrix Equations 4.18 and 4.19, initializes the nonzero matrix element values, and computes a solution to Equations 4.18 and 4.19. It
prints matrices of axial forces in the truss members and components of reaction at joints A and E.

Computer Program 4.1 : Structural Analysis of Cantilever Truss

%
%
%
%
%
%
%
%
%

============================================================================
truss.m : Compute internal forces and reactions in cantilever truss structure.
Matrices : Truss
= Represents truss geometry and element connectivity.
: Load
= External forces acting on the truss nodes.
: Force
= Internal member forces acting in truss elements.
: Support
= Relationship between truss and support reactions
: Reactions = Matrix of support reaction forces.
============================================================================

% Define problem parameters

NoElmts
= 6;
NoReactions = 4;
% Setup matrix for "truss connectivity"
Truss = zeros( NoElmts, NoElmts );
Truss( 1 , 1 ) = 1;
Truss( 1 , 2 ) = -1;
Truss( 2 , 4 ) =

1;

Truss( 3 , 2 ) =
Truss( 3 , 3 ) =

1;
1/sqrt(2);

Truss( 4 , 3 ) =

1/sqrt(2);

161

Chapter 4
Truss( 5 , 3 ) =
Truss( 5 , 5 ) =
Truss( 5 , 6 ) =

1/sqrt(2);
-1/sqrt(2);
-1;

Truss( 6 , 3 ) =
Truss( 6 , 4 ) =
Truss( 6 , 5 ) =

1/sqrt(2);
1;
1/sqrt(2);

% Setup matrix for "external loads" on truss nodes

Load = zeros( NoElmts, 1 );
Load( 2 , 1 ) = 10;
Load( 4 , 1 ) = 10;
% Print matrices for "truss connectivity" and "external loads"
Truss
Load
% Solve equations and print "internal member" forces
Force = Truss\Load
% Setup matrix for "support reactions"
Support = zeros( NoReactions, NoElmts );
Support(
Support(
Support(
Support(

1
1
2
3

,
,
,
,

1
5
5
6

)
)
)
)

=
=
=
=

-1;
-1/sqrt(2);
-1/sqrt(2);
-1;

% Compute and print "support reactions"

Reactions = Support*Force
% ============================================================
% the end!

Running the Program. Assume that Program 4.1 is contained the M-file truss.m. The script of
input/output (I/O):
>> format compact
>> truss
Truss =
1.0000
-1.0000

162

Engineering Programming in Matlab

0
0
0
0
0

0
1.0000
0
0
0

0
0.7071
0.7071
0.7071
0.7071

1.0000
0
0
0
1.0000

0
0
0
-0.7071
0.7071

0
0
0
-1.0000
0

Load =
0
10
0
10
0
0
Force =
-10.0000
-10.0000
14.1421
10.0000
-28.2843
30.0000
Reactions =
30
20
-30
0
>>

shows the command needed to run the program (i.e., truss) and the output that is generated by
the analysis.
For programming convenience, we define the variables NoElmts and NoReactions for
the number of frame elements and reactions, respectively. The matrices in Equations 4.18 and 4.19
are defined by first allocating memory for zero matrices of the appropriate size and then filling in
the nonzero matrix elements.
Validating the Results. The member "Forces" matrix contains the axial forces in elements 1
through 6. A quick examination of the matrix reveals that F1 = F6 = 10 kN (i.e., compression)
and that F4 = 10 kN (i.e., tension). Element 6 carries a tensile force of 30 kN. If you take moments
about joint A, then you will see that the axial force in element 6 times a lever arm of 5 m is balanced
by the 10 kN loads at lever arms 5 m and 10 m.
The "Support" reactions matrix contains the horizontal and vertical support reactions at
joints A and E. Two points should be noted. First, because the horizontal component of externally
applied loads is zero, we expect that the sum of the horizontal reactions at joints A and E will be
zero. They are. Second, you should also note that truss element 2 transfers all the externally applied
vertical loads to support A. The vertical reaction at support A is 20 kN, which is the sum of the two

163

Chapter 4

externally applied loads.

Analysis of an Electrical Circuit

Problem Statement. The analysis of electrical networks composed of resistance and voltage supplies is governed by three basic principles:
1. Ohms Law. The drop in voltage across a resistor R in the direction of an assumed current is
proportional to the current. In other words, V = I R.
2. Kirchoffs Law. The sum of all currents entering and exiting a node must equal zero.
3. Kirchoffs Voltage Law. The sum of voltage drops around any closed loop must sum to zero.
Consider the circuit shown in Figure 4.5, consisting of three loops, nine resistors, and one battery.

I_3

5 Ohms

I_2

2 Ohms

3 Ohms

1 Ohm

4 Ohms

4 Ohms

1 Ohm

5 Ohms

I_1

10 V

6 Ohms

Figure 4.5. Three-loop voltage-resistance circuit.

The analysis begins with the arbitrary assignment of current directions in each of the three loops.
The current in each loop will be deemed to be positive when it flows in the direction shown. For
loop 1, a positive voltage change occurs between the negative and positive battery terminals. The
voltage drops across the upper and lower resistors in loop one are 2I1 and 6I1 , and 3 (I1 I2 ) in
the left-most resistor. Applying Kirchoffs Voltage law to loop 1 gives
6 I1 + 2 I1 + 3 (I1 I2 ) = 10 V.
(4.20)
For loop 2, Kirchoffs voltage law gives

164

Engineering Programming in Matlab

I2 + 5 I2 + 3 (I2 I1 ) + 4 (I2 I3 ) = 0 V.

(4.21)

and for loop 3, Kirchoffs voltage law gives

I3 + 4 I3 + 5 I3 + 4 (I3 I2 ) = 0 V.

(4.22)

Putting Equations 4.20 to 4.22 in matrix form gives

11 3
0
I1

I2
3
13
4

I3
0 4 14

10

(4.23)

Program Source Code. Program 4.2 defines and initializes matrix Equation 4.23, and solves and
prints the solution to the currents in each of the three loops.

Computer Program 4.2 : Analysis of an Electrical Circuit

%
%
%
%
%
%
%
%

==========================================================================
electrical.m -- Compute currents in an electrical circuit
Matrices : Resist = Rows represents resistors in each loop.
: Voltage = Voltage gain in each loop provided by battery.
: Current = Current in each loop.
==========================================================================

% Setup matrix for "resistances in circuit loops"

Resist = [ 11
-3
0

-3
13
-4

0;
-4;
14 ]

% Setup matrix for "voltage gains" in circuit loops

Voltage = [ 10; 0; 0 ]
% Solve equations and print currents

Chapter 4

165

Current = Resist\Voltage
% ============================================================
% the end!

Running the Program. Assume that Program 4.2 is stored in electrical.m. The MATLAB
script
>> format compact
>> electrical
Resist =
11
-3
0
-3
13
-4
0
-4
14
Voltage =
10
0
0
Current =
0.9765
0.2471
0.0706
>>

shows the command needed to execute the program and the output that is generated. A (3 3)
matrix of system resistances is explicitly defined and initialized in one MATLAB statement. A
(3 1) matrix of system voltages is defined in second statement. With these matrices in place,
solutions to Equation 4.23 are computed and printed by simply writing:
>> Current = Resist\Voltage

Of course, we could have computed the currents in each loop of the circuit by writing inv(Resist)*Voltage
Validating the Results. You should verify that the solution is correct by multiplying
>> Resist*Current

and checking that the result is equal to Voltage.

166

Engineering Programming in Matlab

Least Squares Analysis of Experimental Data

Problem Statement. Engineers are often faced with the practical problem of having to model
complex physical processes and phenomena that are not fully understood. The lack of understanding
may be due to the overwhelming size of the system, or perhaps, because information about the
system is missing. In an effort to better understand system behavior, many engineers design and
conduct laboratory experiments, and use the experimental data in the construction of simplified
empirical models. The simplified models will be based on numerous assumptions in behavior and
may contain parameters that can be adjusted or modified to provide a best fit to the experimental
data.

eN

[ x3 , y3 ]
[ xN, yN ]
y = p(x )

[ x1 , y1 ]
e1

e3

e2
[ x2 , y2 ]

Figure 4.6. Experimental data points and line of best fit.

Figure 4.6 is a plot of experimental data points (x1 , y1 ), (x2 , y2 ), (xN , yN ), and a dashed-line
polynomial y = p(x) of best fit that has been drawn by hand. The polynomial might predict, for
example, the relationship between an input signal, x, and an output signal, y = p(x). Notice that
approximately half of the data points deviate from the line in a positive direction (i.e., are above the
line) and approximately half the data points are below the line (i.e., negative deviation).
Although it is certainly possible to find a high-order polynomial that will interpolate the
data points exactly, in many engineering applications, theoretical considerations and good engineering judgment indicate that a low-order polynomial (or simpler curve) will provide a good approximation to the data. Unfortunately, there are no hard-and-fast rules for selecting a best functional
form. An engineer should look for obvious trends, such as a linear, quadratic, cubic polynomial, for
symmetries and antisymmetries in the data, and for periodic behavior, suggesting functional forms

167

Chapter 4

containing sin() and cos() terms.

Derivation of Discrete Least Squares Equations. In this section, we formulate a mathematical
procedure for fitting a polynomial curve
p(x) = ao + a1 x + a2 x2 + + an xn

(4.24)

of degree n through N data points (x1 , y1 ), (x2 , y2 ), (xN , yN ), where (N > n + 1). To
avoid mathematical difficulties with positive/negative distances in error deviations, the objective of
discrete least squares is to minimize the sum of the squares of the distance between the y values of
the data, and y = p(x). In mathematical terms, we want to minimize
N
X

e2i =

i=1

N
X

(yi p(x))2

(4.25)

i=1

or

S(ao , a1 , , an ) =

N
X

[yi ao a1 xi an xni ]2

(4.26)

i=1

The optimal parameter settings are given by the solution to the linear equations
S
S
S
= 0,
= 0,
= 0
a0
a1
an

(4.27)

The partial derivatives of Equations 4.27 are:

"

#
N
X

i=1
"N
X

ao +
#

xi ao +

"

i=1
N
X
i=1

xni

"

ao +

"N
X
i=1
"N
X

xi a1 + +
#

x2i

i=1
N
X

"N
X

xn+1
i

a1 + +

i=1
"N
X

an

N
X

xn+1
i

"

a1 + +

xni

an

xn+m
i

an =

PN

i=1 yi

PN

i=1 yi

PN

n
i=1 xi

(4.28)

xi

(4.29)

yi

(4.30)

i=1

The family of n + 1 Equations 4.28 through 4.30 is linear in the parameters ao , a1 , an . In matrix
form, the equations may be written

168

Engineering Programming in Matlab

PN

i=1 xi

P
PN

N
2

xi
i=1
i=1 xi

.. . .
..

.
.
PN
n+1
n PN
i=1 xi

i=1 xi

PN

n
i=1 xi

PN
i=1

yi

a0

n+1 a1
N

x
i=1 i
i=1 xi yi
.. =
.. .
..

.
.
PN
PN
an
2n
n

PN

i=1 xi

i=1

(4.31)

xi yi

Force F

Analysis of Experimental Data. Suppose that you have been asked to formulate an engineering
model to describe the force-displacement relationship for the simple spring shown on the left-hand
side of Figure 4.7. The purpose of the model is to describe the functional relationship between an
applied force F and a measured displacement x.

[ x_4 , F_4 ]
[ x_3 , F_3 ]

Displacement x
[ x_2 , F_2 ]
k (x )
Mass M

Force F
[ x_1 , F_1 ]
Displacement x

MASS - SPRING EXPERIMENTAL SETUP

EXPERIMENTAL DATA

Figure 4.7. Force-displacement experiment for mass-spring system

The force-displacement model is calibrated by conducting a displacement-controlled experiment,
where displacement of the mass is increased to a prescribed level and the force is measured. The
experimental results are
Data
Displacement
Force
Point
(cm)
(N)
=========================================
1
5.0
0.0
2
5.5
47.5
3
6.0
90.0
4
6.5
127.5
5
7.0
160.0

169

Chapter 4
6
7

7.5
8.0

187.5
210.0

A schematic of coordinate pairs (measured displacement, applied force) is plotted on the right-hand
side of Figure 4.7. Theoretical considerations indicate that the force-displacement relationship is
mildly nonlinear and is closely approximated by the quadratic:
Force(x) = ao + a1 x + a2 x2

(4.32)

where ao , a1 and a2 are coefficients to be determined via experiment and least squares analysis.
When n = 2, Equation 4.31 takes the form:

i=1 xi

PN
2
i=1 xi

PN

i=1 xi

PN
PN

x2i
x3i

PN

2
i=1 xi

a0

PN
i=1

yi

PN
a

PN
1
i=1 xi yi
3

=
i=1 xi

N
2
a2
PN

i=1 xi yi
4
i=1 xi

(4.33)

Program Source Code. Program 4.3 stores the experimental data in a matrix data and then
assembles and solves the matrix Equation 4.33.

Computer Program 4.3 : Least Squares Analysis of Experimental Data

%
%
%
%
%
%
%
%

==========================================================================
leastsq.m -- Compute least squares polynomial fit on experimental data
Experiment : x = displacement of spring (cm).
f = force in spring (N).
Least squares fit : p(x) = a + b.x + c.x2
==========================================================================

% Store force-displacement relationship in matrix "data"

data = [ 5.0
5.5
6.0

0.0;
47.5;
90.0;

170

Engineering Programming in Matlab

6.5
7.0
7.5
8.0

127.5;
160.0;
187.5;
210.0 ];

% Compute terms in least squares matrix and right-hand vector

N = 7;
sumx
= sum(data(:,1));
sumy
= sum(data(:,2));
sumxy = sum(data(:,1).*data(:,2));
sumx2
sumx2y
sumx3
sumx4

=
=
=
=

sum(data(:,1).*data(:,1));
sum(data(:,1).*data(:,1).*data(:,2));
sum(data(:,1).*data(:,1).*data(:,1));
sum(data(:,1).*data(:,1).*data(:,1).*data(:,1));

A = [

N
sumx
sumx2

sumx
sumx2
sumx3

sumx2
sumx3
sumx4 ]

B = [ sumy; sumxy; sumx2y]

% Compute and print constants a,b,c
Coefficients = A\B
% ============================================================
% the end!

Running the Program. Assume that Program 4.3 is contained the M-file leastsq.m. The script
of MATLAB I/O
>> format compact
>> leastsq
A =
1.0e+04 *
0.0007
0.0046
0.0046
0.0303
0.0303
0.2059
B =
1.0e+04 *
0.0823
0.5836
4.1891
Coefficients =

0.0303
0.2059
1.4282

171

Chapter 4
-750.0000
200.0000
-10.0000
>>

shows the commands used to run the program and the output that is generated. Matrices A and
B represent Equation 4.33. The matrix Coefficients contains the results of the least squares
analysis. That is, the values for ao , a1 , and a2 that provide the best fit of Equation 4.32 to the
experimental data.
Assembling the Least Squares Matrix Equations. Program 4.3 stores the system displacements
and applied forces in columns 1 and 2 of matrix data. The most straightforward and, admittedly,
inefficient way of computing the matrix element terms in Equation 4.33 is with blocks of MATLAB
code that look similar to the following:
sumx2 = 0.0;
for i = 1:N
sumx2 = sumx2 + data(i,1)*data(i,1);
end,

Here sumx2 holds the sum of x2i terms needed for matrix elements A(1,3), A(2,2), and A(3,1).
Evaluation of this looping structure in MATLAB will be slow because it is interpreted. The same
numerical result can be obtained in much less time with the single statement
sumx2

= sum( data(:,1).*data(:,1) );

Now the MATLAB function sum is applied to matrix element-level multiply operations on all the
items in column one of matrix data. The .* syntax signifies matrix element-level multiply operations, and the colon (:) operation implies all of the items within a column of matrix data.
It is possible, in fact, to completely eliminate the sumx2-type terms from the calculation
and to form the (3x3) least squares matrix in one statement block involving matrix element-level
multiplies on data. Similar expressions can be written for matrix B.
Validating the Results. The polynomial coefficients ao = 750, a1 = 200, and a2 = 10 define
the force-displacement relationship
Force(x)

The sum of deviations

= 750 + 200 x 10 x2

(4.34)

= 10 (x 5) (x 15)

(4.35)

172

Engineering Programming in Matlab

N
X
i=1

e2i =

N
X

(yi p(x))2 = 0.0

(4.36)

i=1

indicating that our second-order polynomial passes through the seven data points exactly (okay, we
confess, we set it up that way).

Distribution of Temperature in Chimney Cross-Section

Problem Statement. In this example, we use a finite difference approximation of Laplaces equation to compute the distribution of temperature in a chimney cross-section. We select this problem
because it is typical of many that naturally occur in the analysis of equilibrium states of physical systems. For example, with suitable simplifying assumptions, Laplaces equation can also describe (1)
the irrotational flow of an incompressible fluid, (2) electrostatic and magnetostatic potentials, and
(3) the hydraulic head associated with the steady-state flow of ground water in a uniform porous
medium.
Figure 4.8 shows the front elevation and square-shaped cross-section A-A for the tall chimney. The chimney is constructed from a material that is homogeneous and isotropic (i.e., the material
has the same material properties in all directions). At the cross-section, the inside and outside temperatures are 200o C and 0o C, respectively, and there is neither a net flow of heat to or from the
chimney (i.e., it is in thermal equilibrium). Finally, we assume that at the chimney cross-section,
the distribution of temperature is constant along the z axis.
Because the chimney is constructed from a material that is homogeneous, thermal conductivity will not vary with position, and because the material is isotropic, thermal conductivity will
not vary with direction. The steady-state distribution of temperature T = T (x, y) throughout the
chimney cross-section is given by solutions to Laplaces equation
2 T (x, y) 2 T (x, y)
+
= 0.
x2
y 2

(4.37)

with boundary conditions T = 0o C along the exterior of the chimney, and T = 200o C along the
chimney interior.
Finite Difference Mesh for Chimney Cross-Section
The chimney cross-section is symmetric about the x- and y-axes, and the two diagonal axes.
Our computational model takes advantage of symmetries about the x- and y-axes by modeling only

173

Chapter 4

y
T = 0 C.

x
T = 200 C.

A
z

SECTION A - A

x
FRONT ELEVATION OF CHIMNEY

Figure 4.8. Front elevation and cross-section of tall chimney.

one quarter of the chimney cross section, as shown in Figure 4.9a. Two new boundary conditions
are needed for this model; along the y-axis the temperature gradient dT /dx = 0, and along x-axis
dT /dy = 0.
If dx and dy are the mesh distance in the x-axis and y-axis directions, then a suitable finite
difference approximation to Equation 4.37 is


T (x + dx, y) 2T (x, y) + T (x dx, y)

T (x, y + dy) 2T (x, y) + T (x, y dy)
+
= 0.
dx2
dy 2
(4.38)

Equation 4.38 is simply the two-dimensional counterpart of the finite difference approximation
derived for the cable profile problem. If dx = dy, then Equation 4.38 can be rearranged to give

4T (x, y) T (x dx, y) T (x + dx, y) T (x, y + dy) T (x, y dy) = 0.

(4.39)

174

Engineering Programming in Matlab

AXIS OF SYMMETRY

T = 0 C

[ dT / dx ] = 0.

21

19

20

16

17

18

13

14

15

10

11

12

T = 0 C

T = 200 C

T = 200 C

[ dT / dy ] = 0.

(a). Finite difference mesh

-1

-1

-1

-1

-2

-1

INTERIOR STENCIL [ NODES 4 - 21 ].

-1

X - AXIS OF SYMMETRY [ NODES 1 - 3 ].

(b). Finite difference stencils

Figure 4.9. Finite difference mesh and stencils for Equation 4.37

175

Chapter 4

The leftmost schematic of Figure 4.9b shows the weighting of discrete temperatures in the finite difference approximation. The nodes along the x-axis (i.e., y = 0) satisfy the finite difference equation
4 T (x, 0) T (x dx, 0) T (x + dx, 0) 2T (x, dy) = 0

(4.40)

The leftmost schematic of Figure 4.9b shows the and along the y-axis (i.e., x = 0)
4 T (0, y) 2T (dx, y) T (0, y + dy) T (x, y dy) = 0

(4.41)

The finite difference mesh has 65 nodes, 26 of them being on the interior (i.e., T = Oo C) and
exterior (i.e., T = 200o C) boundaries. This leaves 39 nodes on the chimney interior for evaluation.
Instead of evaluating the temperature stencils at all 39 interior nodes, we compute temperature only
at the 21 nodes labeled with small filled black boxes and fill in the remaining unknowns by noting
the symmetry in temperature along the line x = y. The four-node stencil is used at nodes 1 to 3,
and the five-node stencil is used at nodes 4 to 21.
Program Source Code. Our solution to this problem uses the method of iteration to compute
the steady-state temperature profile at the internal nodes. Once the temperature profile is known we
create a two-dimensional color contour plot of the temperature distribution inside the chimney wall.

Computer Program 4.4 : Distribution of Temperature in Chimney Cross-Section

%
%
%
%

============================================================
chimney.m -- Compute and displace profiles of temperature in
chimney cross section.
============================================================

%
%

Setup working array and boundary conditions along

internal/external walls.
T = zeros(9,9);
for i = 5:9;
T(i,5) = 200;
end;
for i = 1:5;
T(5,i) = 200;
end;
for i = 6:9;

176

Engineering Programming in Matlab

for j = 1:4;
T(i,j) = NaN;
end;
end;
% Loop over internal nodes and compute new temperatures
counter
= 0;
maxchange = 200;
while (maxchange > 1)
counter = counter+1;
maxchange = 0;
k=5;
l=4;
for c = 6:8;
newtemp
= 0.25*(2*T(l+1,c)+2*T(l,c+1));
tempchange = newtemp - T(l,c);
maxchange = max(maxchange,abs(tempchange));
T(l,c)=newtemp;
for r=k:8
newtemp
= 0.25*(T(r,c-1)+T(r,c+1)+T(r-1,c)+T(r+1,c));
tempchange = newtemp - T(r,c);
maxchange = max(maxchange,abs(tempchange));
T(r,c)= newtemp;
end
newtemp
= 0.25*(T(9,c-1)+T(9,c+1)+2*T(8,c));
tempchange = newtemp - T(9,c);
maxchange = max(maxchange,abs(tempchange));
T(9,c)=newtemp;
l=l-1;
k=k-1;
end
counter;
maxchange;
% to view counter or maxchange remove
end
% Compute reflected temperature
for i = 2:4
for j = 1: 11-i
T(i,j) = T(10-j,10-i);
end
end
% Print temperature array.
T

177

Chapter 4

% Plot temperature contours.

contour(T)
hold;
% Now overlay perimeter of chimney section on contours.
perim = [ 1
9
9
5
5
1
1

,
,
,
,
,
,
,

1;
1;
9;
9;
5;
5;
1 ];

plot(perim(:,1),perim(:,2),w);
text(1.1,5.3,Temp = 200.0);
text(7.0,1.3,Temp = 0.0);
%
%

============================================================
The End!

Running the Program. Assume that Program 4.4 is contained in the M-file chimney.m. The
abbreviated script of MATLAB I/O:
>> format compact
>> chimney
.... lots of program output removed ....
T =
Columns 1 through 7
0
47.9527
96.9120
147.6536
200.0000
NaN
NaN
NaN
NaN

0
47.4495
96.1208
147.0166
200.0000
NaN
NaN
NaN
NaN

Columns 8 through 9

0
45.9241
93.6699
144.9543
200.0000
NaN
NaN
NaN
NaN

0
42.8111
88.3435
139.9098
200.0000
NaN
NaN
NaN
NaN

0
37.2518
77.7632
127.2653
200.0000
200.0000
200.0000
200.0000
200.0000

0
28.7560
59.1105
92.5014
127.2653
139.9098
144.9543
147.0166
147.6536

0
19.0375
38.4899
59.1105
77.7632
88.3435
93.6699
96.1208
96.9120

178

Engineering Programming in Matlab

Figure 4.10. Temperature profile in chimney.

0
9.3215
19.0375
28.7560
37.2518
42.8111
45.9241
47.4495
47.9527

0
0
0
0
0
0
0
0
0

>>

shows the commands needed to run the chimney analysis and compute the distribution of temperature throughout the chimney cross-section. Figure 4.10 is two-dimensional contour plot of chimney
temperature.

179

Chapter 4

Computational Procedure. The first block of code sets up a (9x9) matrix for modeling one fourth
of the chimney cross-section. The temperature along the interior and exterior walls is set to 200
and 0 degrees, respectively. In MATLAB the interior region of the chimney can be represented with
NaNs. That is, a missing data item. At this point in the program execution, the contents of matrix T
are
T =
0
0
0
0
200
NaN
NaN
NaN
NaN

0
0
0
0
200
NaN
NaN
NaN
NaN

0
0
0
0
200
NaN
NaN
NaN
NaN

0
0
0
0
200
NaN
NaN
NaN
NaN

0
0
0
0
200
200
200
200
200

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0

The main block of code walks along columns 6 through 8 and evaluates the finite difference stencils
for
Column No
Row Nos
============================
6
4 through 9
7
3 through 9
8
2 through 9
============================

For example, after the algorithm has walked along column 6 for the first time, the contents of T are
T =
Columns 1 through 7
0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

Columns 8 through 9

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
0
0
0
200.0000
200.0000
200.0000
200.0000
200.0000

0
0
0
0
50.0000
62.5000
65.6250
66.4062
83.2031

0
0
0
0
0
0
0
0
0

180

Engineering Programming in Matlab

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0

Bear in mind that the temperature at stencil T(4,6) is still zero because all the neighboring stencils
are initially zero. The three-line block of code
newtemp
= 0.25*(T(9,c-1)+T(9,c+1)+2*T(8,c));
tempchange = newtemp - T(9,c);
maxchange = max(maxchange,abs(tempchange));

computes the new temperature estimate at the node, the change in node temperature from the previous iteration, and the maximum change in temperature occurring over rows 6 through 8 for the
current iteration.
The outermost loop of the algorithm will iteratively refine the temperature profile until
satisfactory convergence occurs. For this example, we stop refining the temperature profile when
the maximum change in temperature over rows 6 through 8 is less than 1 degree. At the conclusion
of the main block of code, the temperature profile is
T =
Columns 1 through 7
0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

Columns 8 through 9

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
9.3215

0
0

0
0
0
0
200.0000
NaN
NaN
NaN
NaN

0
0
0
0
200.0000
200.0000
200.0000
200.0000
200.0000

0
0
0
92.5014
127.2653
139.9098
144.9543
147.0166
147.6536

0
0
38.4899
59.1105
77.7632
88.3435
93.6699
96.1208
96.9120

181

Chapter 4
19.0375
28.7560
37.2518
42.8111
45.9241
47.4495
47.9527

0
0
0
0
0
0
0

The final temperature profile is obtained by reflecting the temperatures along the line y = x.
Of course, the temperature profile may also be computed by writing and solving the finite
difference equations in matrix form (see Problem 4.8).

4.7

Review Questions

1. Explain how a system of m linear equations containing n unknowns can be represented in

matrix form.
2. What are the three types of solutions matrix equations can have?
3. What role does the matrix determinant play in determining whether a family of matrix equations
will have a unique solution.
4. Let A be a (n n) matrix and B be a (n 1) matrix. Under what conditions will the solution to
A.X = B have an infinite number of solutions? How would you use MATLAB to detect this
situation?
5. Suppose that a family of three equations, each having three unknowns, is graphed in threedimensional space and that it is immediately apparent that one of the equations is a linear
combination of the remaining two. If the equations are written in matrix form, what can you
say about (1) the matrix rank, (2) the matrix determinant, and (3) the matrix inverse?

182

Engineering Programming in Matlab

4.8

Programming Exercises

4.1 Beginner. Suppose that the cable profile of a small suspension bridge carrying a uniformly
distributed load

w
Cable Profile

20
Cable Hangers

10
Uniform Load along
Deck of Suspension Bridge
x
Deck of Bridge
10

DIAGRAM NOT TO SCALE

corresponds to the solution of the differential equation

d2 w
= 1.0
dx2

(4.42)

with the boundary conditions w(0) = 10 and w(10) = 20. It is easy to show that the analytic
solution to the cable profile is
1
w(x) = x2 4x + 10
2

(4.43)

Now solve Equation 4.42 via the method of finite differences.

1. What is a suitable finite difference approximation to Equation 4.42?.
2. If the cable profile is divided into five regions along the x-axis, with four internal nodes,
write down the family of finite difference equations that you would solve for the cable
profile (do not try to find a solution to these equations).

183

Chapter 4

3. Write a MATLAB program to solve the family of equations by the method of iteration.
4. Write down the family of linear matrix equations corresponding to this finite difference
problem. Write a MATLAB program that computes the solution to these equations, and
then displays the numerical solution and Equation 4.43 on the same graph.
4.2 Beginner. Figure 4.8 shows a three-loop voltage-resistance circuit, containing one battery
and seven resistors.

I_2

1 Ohm

3 Ohms

4 Ohms

2 Ohms

I_1

6 Ohms

10 V

6 Ohms

I_3

5 Ohms

Figure 4.11. Three-loop voltage-resistance circuit.

Write a MATLAB program to compute and print the magnitude of current flows in each of
the three loops. For each loop, assume that anticlockwise current flow is positive.
4.3 Intermediate. In the solution of many fluid mechanics and chemical engineering problems,
conservation of mass is a central principle. Briefly stated, conservation of mass accounts for
all sources and sinks of a material that pass in and out of a control volume (see the left-hand
side of Figure 4.12).
For a specified interval of time, the accumulation of substance is simply the sum of the inputs
minus the sum of the outputs. When the sum of the inputs equals the sum of the outputs,
accumulations are zero, and the mass within the volume will be constant. Since the mass
within the volume does not change with time, we say that such a system is in steady state.

184

Engineering Programming in Matlab

Control Volume

Input

Output

Q1, C1

Q3, C3

Mixer

Q2, C2

Accumulation of Mass

MASS MALANCE IN CONTROL

VOLUME

Q1 = 2 m^3 / sec

Q2 = 1 m^3 / sec

Q3 = 3 m^3 / sec

C1 = 0.02 kg / m^3

C2 = 0.015 kg / m^3

C3 = ?????

STEADY - STATE COMPLETELY MIXED

REACTOR

Figure 4.12. Conservation of mass in fully mixed reactor.

The principle of conservation of mass can be used to determine the concentrations of
substances in system of coupled fully mixed reactors. To see how the analysis proceeds, lets
first look at the single fully mixed reactor shown on the right-hand side of Figure 4.12. The
reactor has one input pipe and two output pipes. You should observe that the concentration at
the output pipe is not shown because it can be computed via the principle of conservation of
mass.
The mass of substance passing through each pipe is simply the flow rate, Q (m3 /sec),
multiplied by the concentration of substance C (kg/m3 ). For a system in steady state (where
the mass does not increase or decrease due to chemical reactions), conservation of mass requires

C1 Q1 + Q2 C2 = Q3 C3 .

(4.44)

Hence, the concentration of mass in the output pipe is



(Q1 C1 + Q2 C2 )
C3 =
Q3

0.055
kg/m3 .
3

(4.45)

Exactly the same principles can be used to compute the concentration of substances in the
network of fully mixed reactors shown in Figure 4.13. The concentrations of mass in reactors

185

Chapter 4
Q_p1 = 2 m^3 / sec;

Q_p2 = 1 m^3 / sec;

C_p1 = 0.02 kg/m^3;

C_p2 = C_2

Q_12 = 3 m^3/sec

Q_23 = 1 m^3/sec;

C_1

C_2

C_3

Q_p3 = 0.5 m^3/sec;

Q_24 = 1 m^3/sec;

C_p3 = 0.10 kg/m^3;

Q_41 = 1 m^3/sec;

Q_34 = 1.5 m^3/sec;

C_4

Q_p4 = 1.5 m^3/sec;

C_p4 = C_4;

Figure 4.13. Network of four fully mixed reactors.

1 through 4 are denoted by the symbols C1 , C2 , C3 and C4 . Because there are four reactors,
four simultaneous mass-balance equations are needed to describe the distribution of substance
concentrations.
1. Show that the mass-balance equations may be written

Q12
0
0
Q41
Q12 Q12
0
0
0
0 Q23 Q34
0
Q24 Q34 (Qp4 + Q41 )

C1
Q12 Cp1

C2
0

C
Q
3
p3
p3

C4

(4.46)
0

2. Develop a MATLAB program to solve Equations 4.46 for the concentrations in each
reactor.
4.4 Intermediate. In the design of highway bridge structures and crane structures, engineers are
often required to compute the maximum and minimum member forces and support reactions
due to a variety of loading conditions.

186

F4

F9

F5

F6

F7

5m

Engineering Programming in Matlab

F8

y
F1
x

F2

F3

P1 kN

P2 kN

R_ay

R_dy
5m

5m

5m

Figure 4.14. Front elevation of pin-jointed bridge truss.

Figure 4.14 shows a nine bar pin-jointed bridge truss carrying vertical loads P1 kN and P2 kN
at joints B and C. The symbols F1 , F2 , F9 represent the axial forces in truss members 1
through 9, and Ray and Rdy are the support reactions at joints A and D. (Notice that because
support D is on a roller and there are no horizontal components of external loads, horizontal
reactions will be zero.)
Write down the equations of equilibrium for joints B through F and put the equations in matrix form. Now suppose that a heavy load moves across the bridge and that, for
engineering purposes, it can be represented by the sequence of external load vectors
"

P1
P2

"

10
0

# "

P1
P2

"

5
5

# "

P1
P2

"

0
10

(4.47)

Develop a MATLAB program that will solve the matrix equations for each of the external
load conditions, and compute and print the minimum and maximum axial forces in each of
the truss members.
4.5 Intermediate-Advanced. In the detailed stages of a petroleum refinery design, an experiment is conducted to determine the empirical relationship between solubility weight (%) of
n-butane in anhydrous hydrofluoric acid at high pressures and temperature. A plot of the
experimental data

187

Chapter 4
Data Point
Temperature (C)
Solubility (%)
==============================================
1
25
2.5
2
38
3.3
3
85
7.1
4
115
11.0
5
140
19.7

on semilog graph paper indicates that solubility and temperature follow the nonlinear relationship
Solubility s(t) = ao ea1 t .

(4.48)

where ao and a1 are parameters to be determined. A linear least squares problem can be
obtained by applying the transformation loge (s(t)) = loge (ao ) + a1 t.
1. Show that the least squares estimate of parameters ao and a1 is given by solutions to the
matrix equations

N
PN

i=1 ti

PN

i=1 ti

PN

2
i=1 ti

loge (ao )
a1

PN
i=1

PN
i=1

loge (si )

ti loge (si )

(4.49)

2. Write a MATLAB program to compute parameters ao and a1 by solving matrix Equation

4.49.
4.6 Intermediate. Figure 4.15 is a three-dimensional view of a 2 by 2 km site that is believed to
overlay a thick layer of mineral deposits.
To create a model of the mineral deposit profile and establish the economic viability of mining
the site, a preliminary subsurface exploration consisting of 16 bore holes is conducted. Each
bore hole is drilled to approximately 45 m, with the upper and lower boundaries of mineral
deposits being recorded. The bore hole data is as follows:
Borehole
[ x, y ] coordinate
[ upper, lower ] mineral surfaces
============================================================================
1
[
10.0 m,
10.0 m ]
[ -30.5 m, -40.5 m ]
2
[ 750.0 m,
10.0 m ]
[ -29.0 m, -39.8 m ]
3
[ 1250.0 m,
10.0 m ]
[ -28.0 m, -39.3 m ]
4
[ 1990.0 m,
10.0 m ]
[ -26.6 m, -38.5 m ]
5
[
10.0 m, 750.0 m ]
[ -34.2 m, -41.4 m ]

188

Engineering Programming in Matlab

2000 m
z

20

00

Data Point [ x, y, z ]

Upper Surface of
mineral deposits

Lower Surface of
mineral deposits

Figure 4.15. Three-dimensional view of mineral deposits.

6
7
8
9
10
11
12
13
14
15
16

[
[
[
[
[
[
[
[
[
[
[

750.0
1250.0
1990.0
10.0
750.0
1250.0
1990.0
10.0
750.0
1250.0
1990.0

m,
m,
m,
m,
m,
m,
m,
m,
m,
m,
m,

750.0
750.0
750.0
1250.0
1250.0
1250.0
1250.0
1990.0
1990.0
1990.0
1990.0

m
m
m
m
m
m
m
m
m
m
m

]
]
]
]
]
]
]
]
]
]
]

[
[
[
[
[
[
[
[
[
[
[

-32.8
-31.8
-30.3
-36.7
-35.2
-34.2
-32.8
-40.4
-39.0
-38.0
-36.5

m,
m,
m,
m ,
m,
m,
m,
m,
m,
m,
m,

-40.6
-40.1
-39.4
-42.0
-41.2
-40.7
-40.0
-42.8
-42.1
-41.6
-40.9

m
m
m
m
m
m
m
m
m
m
m

]
]
]
]
]
]
]
]
]
]
]

With the bore hole data collected, the next step is to create a simplified three-dimensional
computer model of the site and subsurface mineral deposits. The mineral deposits will be
modeled as a single six-sided object. The four vertical sides are simply defined by the boundaries of the site. The upper and lower sides are to be defined by a three-dimensional plane

z(x, y) = ao + a1 x + a2 y

(4.50)

189

Chapter 4

where coefficients ao , a1 , and a2 correspond to minimum values of

S (ao , a1 , a2 ) =

N
X

[zi z(xi , yi )]2

(4.51)

i=1

Things to do:
1. Show that minimum value of S(ao , a1 , a2 ) corresponds to the solution of the matrix
equations

i=1 xi

PN
i=1 yi

PN

i=1 xi

PN
PN

x2i

xi yi

PN

i=1 yi

ao

PN

i=1

zi

N
a

=
x

1
i
i
i=1
i=1 xi zi

PN
PN
2

PN

i=1 yi

a2

i=1

(4.52)

yi zi

2. Write an M-file that will create three-dimensional plots of the borehole data at the lower
and upper surfaces.
3. Write an M-file that will set up and solve the matrix equations derived in part 1 for the
upper and lower mineral planes.
4. Compute and print the average depth and volume of mineral deposits enclosed within the
site.
Note. The least squares solution corresponds to the minimum value of function S(ao , a1 , a2 ).
At the minimum function value, we will have

dS
dS
dS
=
=
=0
dao
da1
da2

(4.53)

Matrix Equation 4.52 is simply the three equations 4.53 written in matrix form. You should
find that the equation of the upper surface is close to z(x,y) = -30.5 + x/500 y/200 and the lower surface close is to z(x,y) = -40.5 + x/1000 - y/850.
4.7 Intermediate. Repeat the chimney temperature problem using the following problemsolving procedure:

190

Engineering Programming in Matlab

1. Write an M-file that sets up the finite difference equations in matrix form and then computes a solution by solving A.T = B, where T is the temperature at the internal nodes of
the chimney.
2. Create a three-dimensional mesh (or surface) plot of the temperature distribution in one
fourth of the chimney cross section.
4.8 Intermediate. Figure 4.16 shows the cross-section of a long conducting metal box with a
detached lid.
0 Volts

Axis of Symmetry

100 Volts

CROSS SECTION OF METAL BOX

FINITE DIFFERENCE GRID

Figure 4.16. Cross-section of tall (infinite) metal box.

The sides and bottom of the box are at 100 Volts potential, and the lid is at ground (0 V)
potential. Laplaces equation and the method of finite differences can be used to compute
the distribution of potential inside the box. The solution procedure is almost identical to the
chimney problem described in the chapter, but with temperature changed to voltage.
The box is assumed to extend to infinity in the z direction (so that there are no edge
effects to consider). This simplifies the problem to two dimensions (x and y). The static
distribution of voltage, V = V (x, y), inside the metal box is given by solutions to Laplaces
equation

2 V (x, y) 2 V (x, y)
+
=0
x2
y 2

(4.54)

Chapter 4

191

with boundary conditions V = 0 V on the top or lid of the box and V = 100 V along both sides
and the bottom of the box. Write a MATLAB program that will
1. Compute the voltage distribution inside the box via the method of finite differences described in Chapter 8 of the C tutorial.
2. Plot a contour map of the voltage potential.
3. Optional. Change the potential on the walls and lid of the box relative to each other and
show how the voltage distribution changes.

Bibliography

[1] Arnold, K., Gosling, J. The Java Programming Language. Addison-Wesley, Reading, MA
01867, 1996.
[2] Boehm, B.W. A spiral model of software development and enhancement. IEEE Computer,
21(5):6172, 1988.
[3] Booch, G. Object-Oriented Analysis and Design with Applications. Benjamin Cummings,
Redwood City, CA 94065, 2nd edition, 1994.
[4] Brooks, F. The Mythical Man-Month. Addison-Wesley, 1975.
[5] Clements, P. C., Parnas, P. L., Weiss, D. M. The modular structure of complex systems. Proc
7th International Conf. on Software Engineering, pages 408417, March 1984.
[6] Dongarra, J.J., Bunch, J.J., Moler, C.B., Stewart, G.W. LINPACK Users Guide. SIAM, 1979.
[7] East, S. Systems Integration A Management Guide for Manufacturing Engineers. McGrawHill, 1994.
[8] Linton M. dbx. Technical report, Berkeley, CA 94720, 1982.
[9] Meyer, B. Object-oriented Software Construction. Prentice-Hall International Series in Computer Science, Hertfordshire, UK, 1988.
[10] Nievergelt, J., Hinrichs, K.H. Algorithms and Data Structures : With Applications to Graphics
and Geometry. Prentice-Hall, Englewood Cliffs, NJ 07632, 1993.
[11] Osterhout, J.K. Tcl and the Tk Toolkit. Addison-Wesley Professional Computing Series,
Reading, MA 01867, 1994.
[12] Parnas, D. L. On the criteria to be used in decomposing systems into modules. Communications of the ACM, 15:330336, December 1972.
192

Chapter 4

193

[13] Press, L. Personal computing : Technetronic education : Answers on the cultural horizon.
Communications of the ACM, 36(5):1722, May 1993.
[14] Royce, W.W. Managing the development of large software systems. In Proceedings of the
IEEE WESCON, August 1970.
[15] Smith, B.T., Boyle, J.M., Ikebe, Y., Klema, V.C., Moler, C. Matrix Eigensystem Routines :
EISPACK Guide. Springer-Verlag, 2nd edition, 1970.
[16] Tesler, L.G. Networked computing in the 1990s. Scientific American, 265(3):8693, September 1991.
[17] Wall, L., Christiansen, T., Schwartz, R. Programming Perl. OReilly and Associates, Sebastopol, CA 95472, 2nd edition, 1993.