ANZIAM J. 47 (EMAC2005) pp.

C216C230, 2006

C216

Multi-disciplinary approach to teaching

numerical methods to engineers using Matlab.
S. I. Barry

T. Webb

(received 20 October 2005; revised 4 August 2006)

Abstract
Teaching Numerical Methods to first year engineering students is
improved by a collaboration between engineers, computer scientists
and mathematicians. Our approach has been to teach a problem
based course with six, fortnightly assignments, each revolving around
an engineering problem solved by a specific numerical method. The
collaborative delivery, engineering problem focus, and intensive tutorials all contribute to make this a very successful course that gives
the students a sound foundation to take into later years and hopefully
their careers.

School Physical, Environmental & Mathematical Sciences, University of New South

Wales at ADFA, Canberra, Australia. mailto:s.barry@adfa.edu.au

School of Aeronautical, Civil and Mechanical Engineering, University of New South

Wales at ADFA, Canberra, Australia.
c AusSee http://anziamj.austms.org.au/V47EMAC2005/Barry for this article,
tral. Mathematical Soc. 2006. Published August 31, 2006. ISSN 1446-8735

ANZIAM J. 47 (EMAC2005) pp.C216C230, 2006

C217

Contents
1 Introduction

C217

2 Background history

C218

3 How the course is run

C220

4 Example problems

C222

5 Results and impressions

C226

6 Conclusion

C229

References

C229

Introduction

A course in Numerical Methods studies the way we use computers to solve

mathematically based problems, a particularly important skill for engineering
students. Traditionally this means covering the theory and practice of numerically calculating typical mathematical and engineering problems such as
integration, interpolation, regression, root finding, solving systems of linear
equations (matrices), and solving differential equations.
There are numerous textbooks which have evolved over the last forty
years, most of which are still grounded in the original computational methods and languages. These books spend a great deal of time covering error
propagation, the benefits of one integration scheme over another, the derivation of these schemes and so on. While this is useful, especially to computer
scientists and expert practitioners in the field, it is less relevant to a practicing engineer and engineering students in their first years of study, where

1 Introduction

C218

application of in-built routines is more useful. Students are also more motivated when they can see a direct application to their chosen profession.
An example of the book/technology mismatch is simple integration. It
takes one page to describe the fundamental principles of numerical integration
via Riemann sums. It takes three lines to illustrate the Matlab command
quadl which will do many integrations numerically. Yet many books with titles Numerical Methods using Matlab (for example [4]), aimed at later year
students, spend many pages describing competing methods, barely mentioning the simple Matlab command quadl. More targeted texts such as [2]
emphasise the details of the numerical methods rather than their use in solving engineering problems. However, two recent books [3, 6] deal nicely with
the learning of Matlab with chapters on numerical methods with applications.
Our aim in developing the course Engineering Computational Methods
was to give first year students some of the basic tools for solving engineering
problems, while still giving them some experience in writing more complicated programs and a few of the numerical issued involved in computational
methods. We try to teach them the in-built Matlab commands, their uses
and limitations, as well as an introduction to programming their own simple
routines. The advantage of giving first year students an exposure to Matlab solving of problems, and the underlying numerical methods, is that these
students can then use these skills for all their later courses, hence deepening
their understanding of other topics. As an underlying principle we wanted the
course to be engineering problem focused since having realistic application
both motivates [5] students and gives them a deeper learning experience [7].

Background history

Some fifteen years ago, the Civil Engineering Program at unsw@adfa was
being redesigned and one topic under the spotlight was Numerical Methods

2 Background history

C219

which was being delivered as a service to second year engineers by a computer

science school. The course was presented from a computer science perspective
with emphasis on topics important to those designing numerical methods but
of peripheral interest to Civil Engineers. The decision was taken to deliver
the course from within the school, by engineering staff with a great deal of
professional experience using computing tools in their everyday work, as well
as training in the underlying mathematics and computing. What evolved was
a project based, tutorial style course based around the Matlab computing
language.
In 2003 there was a major restructure of the unsw@adfa Schools and
with the subsequent amalgamation of Engineering Schools it was decided
to offer a similar course to all engineering students. This required some
major changes, notably delivery to a much larger class, and also bringing
the Course back from second year to the second half of first year. In the
process of gaining approval the question was asked as to whether the Course
should be taught by computer science, engineering or mathematics, each
having legitimate claims of expertiseand the valuable student credit points
on offer.
The decision was taken to run the Course jointly by the three separate
disciplines with each putting their own perspective on the course. This presented many challenges such as: modifying the content to be manageable by
first year students; catering to different engineering strands (aeronautical,
civil, electrical and mechanical); coping logistically with the larger student
numbers (130 in 2004) who would require appropriate computing facilities
and assistance; streamlining the administration of a course taught by three
schools; and, jointly writing a set of lecture notes and problem sheets.
In this article we first discuss how we ran the course before discussing
the types of problems we assign to the students. We then give some more
detailed descriptions of two of these problems, to illustrate our approach,
before discussing the course outcomes.

2 Background history

C220

How the course is run

The Course comprises six engineering problems taken from real world experience or related to more advanced engineering courses and was developed by
the second author with assistance from the other two lecturers and engineering colleagues. Development of course notes, framing the mathematical and
numerical basis and overall management of the course fell to the first author.
A third lecturer (Darcy Brooker from School of Information Technology and
Electrical Engineering) provided further numerical and computational insight.
Each topic is presented over a two week period, with two lectures, and four
hours of supervised computer laboratory sessions. The lectures are jointly
delivered by the three lecturers who also conduct the laboratory tutorials
with the assistance of tutors.
A typical lecture delivery would start with a description of the engineering problem to be solved by the engineer, followed by descriptions of the
mathematical background, the relevant Matlab functions, programming
techniques and numerical issues. Most of this would be delivered over two
hours at the start of each topic. The deep learning happens in the computer
laboratories where the students tackle the assignments under supervision.
Through each topic the student has to come to an understanding of the
engineering problem, be able to convert it into an equivalent mathematical
formulation, design Matlab code to solve the problem and finally reinterpret
it as the solution to the engineering problem. The students submit a written
report with attached Matlab code together with its output.
The assignments account for about 50% of the assessment with an exam
accounting for the remainder. The three hour, closed book exam consists
of small problems designed to reflect the knowledge gained by doing the
assignments.

3 How the course is run

C221

Designing the problems

Some examples of the problems are:
finding the current in an electrical circuit with many components (linear
systems);
finding the forces in elements of a truss frame (linear systems);
replacing missing wave height data (interpolation);
estimating the transmission of solar energy through the atmosphere
(integration);
modelling the pressure trace following an explosion (regression);
finding the displacement of a pole due to the force of a wave (differential
equations);
finding the wave length of a tsunami for a given depth (solving nonlinear
equations).
Each problem was very carefully designed so that it reflected a real life
engineering problem but had definable outcomes in the form of a report to a
fictitious supervisor (thus emphasising that this was an engineering problem,
the outcome of which had to be communicated clearly to someone). The
work had to be doable within the four hour tutorial, with perhaps one extra
hour or so for preparing the report. The students were also prompted to
explore some mathematical and computing issues so the assignment was not
just a simple application of a one-line formula, but an exploration of uses
and failings of black-box computing tools.

3 How the course is run

C222

Designing the course notes

As part of the course, detailed notes were produced, given the absence of our
idea of a suitable textbook. These notes are available on the Course web site
and were hypertext pdf, so that index and table of contents link directly to
the appropriate page. All graphs in the text were produced by Matlab, and
when a student clicks on these graphs the appropriate Matlab code appears
and can be copied by the student. Hence students can easily use the code as
templates or examples for their own work. In addition, a detailed Matlab
help page was produced which incorporates our own Matlab guide as well
as several on-line guides taken from other universities (with permission) [1].

Example problems

To illustrate our approach we outline two problems: finding the zero of a

nonlinear function for wavelength; and, solving differential equations for a
bending beam.

Zeroing nonlinear functions

This assignment involved finding the wavelength, L, of a tsunami for a given
water depthan apparently trivial task but with enough aspects to make it
interesting for the student. The governing equation is


g T2
2d
f (L) = L
tanh
= 0,
(1)
2
L
where g is gravity, T = 2 880 s is the period (calculated from an earlier
assignment), and d 4 000 m is the water depth typical for the Indian Ocean.
Typical questions for the students were briefly as follows:

4 Example problems

C223

1. Plot f (L) and graphically estimate the zero (L).

2. Use fzero(@f,106 ), fzero(@f,1), fzero(@f,[1,106 ]) to find the zero.
3. Approximate the function using tanh x x to get
f1 (L) L

gT 2
L

gT 2 d .

(2)

4. Use tanh x x x3 /3 to obtain

L4 (g T 2 d)L2 + gT 2 d3 4 2 0 ,

(3)

and use the roots command in Matlab to find the zeros of this function.
5. Write a Newtons method code to find the zero.
6. If d = 4 000 50 what error does one get with L?
We expected the students to comment on what their results meant. For
example, the plot of the f (L) along with the approximate versions of f (L)
using the asymptotics of tanh are illustrated in Figure 1.
The student should note that the fzero(@f,106 ) gives the answer accurately, since L = 106 is a good first guess, but after several iterations. However, fzero(@f,1) gives the incorrect result of L = 0 since Matlab actually
finds the discontinuity of f (L) at L = 0 , while fzero(@f,[1,106 ]) (giving
Matlab the range x [1, 106 ]) gives the result accurately and quickly.
p
The student then shows that L gT 2 d gives an acceptable result with
an error of 0.003%. Doing the cubic approximation also gives an accurate
answer, but that one has to know enough to discard the inaccurate root at
L = 14 500 . Doing a little programming to do Newtons method, the student
then finds that an accurate answer is obtained with fewer steps than using
fzero even with a starting value of L = 1 .

4 Example problems

C224

F(L)

1.5

x 10

f(L)
Linear approx
Cubic approx
zero
false zeros

f(L) [m]

0.5
0
0.5
1
1.5
5

L [m]

10
5

x 10

Figure 1: Objective function f (L) with various approximations.

4 Example problems

C225

By showing that the error in d gives an error of L = 0.006% the student

should then realise that super-accurate and efficient methods are of little use
if the error in the data swamps the method used.
p
The student finally concludes that L gT 2 d gives an acceptable solution given the errors in d, but that if more accuracy is required then Newtons
method is the fastest, and Matlab in-built command fzero works almost
as well, with some guidance. Hence, perhaps the conclusion we most want
the student to realise is this: no method is ideal and gives perfect answers
every time; one must always consider many options and think carefully about
the results.

Differential equations
A second example is finding displacement y(x), for a pole bending under
wave drag force, as governed by
2

y 00 = (h x)2 (1 + y 0 )3/2 ,

y(0) = y 0 (0) = 0 ,

(4)

where is a constant, x is the height up the pole, and h is the water depth.
Here we get the student to solve this using Matlab ode45 having first solved
the linear approximation
y 00 = (h x)2
(5)
both analytically and numerically. One of the problem solving strategies addressed in this assignment is to first solve a problem with known solution (the
linear approximation) before solving more complicated nonlinear problems.
In addition we get the student to write a simple Euler solution and comment
on the various solutions.
This assignment coincided with the first year mathematics course where
converting a second order differential equation to systems of first order differential equations was being discussed. There was thus opportunity to address

4 Example problems

C226

the problem of compartmentalism by reinforcing and expanding on the material from the mathematics course.

Results and impressions

Developing this course was not easy. Apart from the administrative difficulties of a course taught jointly by three lecturers, and three schools, there
were the difficulties of reconciling our different approaches. As expected,
each of us wanted different emphasis placed on different parts of the course.
However, the resultant course was much better than we could have done
separately. For example, the engineering examples were more realistic that
would have been invented by a mathematician, but the input of mathematics
and computer science meant the students appreciated more the background
to the solution method and did more rigorous testing of their routines. An
additional benefit was that collaboration extended beyond the confines of
the course with more interaction between schools on the development and
administration of our Engineering Mathematics courses. We also found that
each of us learned a great deal, with even our Matlab programming skills
improving with the input of ideas from the other lecturers.
We have seen many benefits for the students with this Course. Students
learned how to write well structured assignments. We are very strict on
presentation and completeness with the solutions. Initial assignments were
typically rather incomplete and scrappy, whereas by the end most were producing well polished assignments with well crafted and commented computer
code as well as sound engineering interpretation. At the beginning many students were computer phobic. This disappeared by the end and most students
were happy and confident in solving problems with Matlab . While the assignments had to be done individually, we saw many pleasing aspects of
teamwork in the tutorials. Students were genuinely helping each other and
keeping the fine balance between collaboration and plagiarism. Most student

5 Results and impressions

C227

quickly became experts in the basics of setting up a Matlab program, using a range of plotting features, writing function subroutines, and sorting out
how to use new commands. They also gained skills in the overall approach to
solving engineering problemssomething that will continue into their later
years.
Positive comments were received from the students in staff-student review
meetings, and elsewhere, and also from our colleagues. The results of a
student survey showed that the majority of the class saw the value in what
we were doing although in the first year (2004) they commented that the
assignments were too long.
We also had relevant comment from students who had not done the
course. In one instance, one of us (Steve Barry) found several second year students working on an engineering assignment that required Matlab . These
students had not done the course because 2004 was the first year it was offered. In helping them Steve showed them what was available in the Course.
They then enthusiastically used the notes to learn what they needed and were
very vocal in their disappointment at not being taught this course when they
were in first year. They felt it was a crucial course for them and one they
wished they had done.
While the Course appears to be successful even in its first year of offering,
it has not been without its problems:
1. We had some instances of plagiarism in 2004. We designed the assignments to be doable within the laboratory time frame; however, only
about half of the students were completing their work in the laboratory times. In retrospect, we made the assignments in this first year
too long, and hence many students felt undue pressure to finish and
resorted to plagiarism. In 2005 the assignments were shorter and more
focused, and there were no cases of plagiarism. In addition we had a
flexible marking scheme whereby a student may count either four, five
or six of the assignments towards their final mark. This meant that

5 Results and impressions

C228

the pressure to perform well on each assignment was reduced, and a

student could do badly on an individual assignment and still do well
overall.
2. Administratively, the timetabling of computer laboratories was very
difficult with the constraints of the number of students, the heavy engineering timetable and the equally heavily booked computer labs. This
was made easier in 2005 since we advised students at the beginning of
the year to purchase their own notebook computers with Matlab
something that about 80% of students did and which the students found
very useful, being able to bring their work easily to us for help.
3. There were numerous administrative problems with a joint course of
this nature, from mis-scheduling of exams to the simple task of three
people editing the one set of notes.
4. Given the problem-based computer nature of the course, it would have
been nice to have assignments worth 100% of the mark, or an exam
using computers. However, given the large student numbers, and the
possibility of plagiarism, a written exam was necessary. In addition,
unsw@adfa does not have the facilities for simultaneous computing
exams for 130 students. As students move to notebook computers we
should be able to address this issue.
While there were many areas we were not able to cover in this course, we
felt the students learned the skills to apply their programming and Matlab
knowledge to any topic they might later come across. This is being reinforced
in computational engineering courses in later years as many assignments are
being set for Matlab solution.

5 Results and impressions

C229

Conclusion

Our teaching of applicable numerical methods to Engineering students benefits greatly from being jointly taught by engineering, mathematics and computer science academics. The students are more motivated, see a more balanced course and quickly learn how to solve an engineering problem with
Matlab . While initially a difficult course to organise and develop, once set
up, it is a course whose benefits to the student and the staff go well beyond
the confines of the course itself.

Acknowledgements: We thank Darcy Brooker for computational insights

and the input of colleagues and students.

References
[1] S. Barry. Matlab guide and online tutorial help. http:
//www.pems.adfa.edu.au/~s8704008/MATLAB.dir/MATLAB.html.
C222
[2] S. J. Chapra. Applied Numerical Methods with Matlab for Engineers
and Scientists. McGraw Hill, 2005.
http://www.mhhe.com/engcs/general/chapra/ C218
[3] A. Gilat. Matlab. An introduction with applications. Wiley, 2005.
http://www.wiley.com/WileyCDA/WileyTitle/
productCd-0471694207.html C218
[4] G. Lindfield and J. Penny. Numerical Methods using Matlab. Prentice
Hall, USA, 2000. URL C218

References

C230

[5] V. E. Martinez Luaces and G. E. Guineo Cobs. Numerical Calculus and

Analytical Chemistry. Proc. 2nd Int. Conf. on Teaching Mathematics,
Crete, Wiley, 2002.
http://www.math.uoc.gr/~ictm2/Proceedings/pap289.pdf C218
[6] H. Moore. Matlab for Engineers. Pearson, 2007. http://www.
pearsoned.co.uk/Bookshop/detail.asp?item=100000000114129
C218
[7] D. Smith. How people learn ... Mathematics. Proc. 2nd Int. Conf. on
Teaching Mathematics, Crete, Wiley, 2002.
http://www.math.uoc.gr/~ictm2/Proceedings/invSmi.pdf C218