BHARTER

Introduction to
Numerical Computin1g

INTRODUCTION
Numerical computations play an indispensable role in solving real
mathematical, physical and engineering problems. They have beer.
use for centuries even before digital
computers appeared on the sce
Great mathematicians like Gauss, Newton,
Lagrange, Fourier and mau
others in the eighteenth and nineteenth centuries
developed numer
techniques which are still widely used. The advent of digital compu
has, however, enhanced the speed and accuracy
tations.
of numerical cor
What is numerical computing? It is
to this fundamental question before we
important to understand the anse
proceed further. Numerivai
computing is an approach for soiving complex mathematical problems
using only simple arithmetic operations. The approach
involve
formulation of mathematical models of physical situations that can be
solved with arithmetic operations. It requires development, analysis and
use of algorithms.
Numerical computations invariably involve a large numbe:
arithmetic calculations and, therefore, require fast and efficient
computing
devices. The microelectronics revolution and the
of high power, low cost
subsequent developrment
personal computers have had a profound impact
on
the application of numerical computing methods to solve scierif
problems.
The traditional numerical
computing methods usually deal with
foiowing topics:
1. finding roots of
equations
2 Numerical Methods

equations
algebraic
of linear
systems analysis
2. solving and
regression
interpolation
3.
4. n u m e r i c a l i n t e g r a t i o n
5 . n u m e r i c a l d i f f e r e n t i a t i o n

equations
differential

solution of
6. value problems
boundary
7. matrix
problems methods availabi.

solution of some of the popular

8. will discuss
book we
In this areas.
each of
these

2 NUMERIC DATA

involve two types

data,
of da namely, crete
computing may obtained counting are alled by
Numerical
Data that a r e
De
continuous
data. total number of items in
data and of discrete
data a r e the
diserete data. Examples participating in a race.
number of people
the total a r e called continuono
measurement us
a box, or
obtained through a vehicle as given by
that are ot
Data data a r e the speed
Examples of continuous a s m e a s u r e d hu
data, of a patient
temperature
or
a
speedometer,
thermometer.

L3 ANALOG COMPUTING

problem by using a tool which a

refers to the principle of solving
Analog For example, the electronic
to the problem.
in a way analogous the problem to be
operates act analogously to
analog computer
circuits in an
that vary continuously,
is based on inputs
solved. Analog computing were earliest computers
temperature. The
such as current, voltage or Calculations
basis of electrical voltages.
analog and functioned on the voltages.
multiplying and dividing
were perfprmed by adding, subtracting, is limited by the precision
computers are fast, but their accuracy
Analog can be read.
with which the physical quantities
are analog in nature: time, tem-
Many real life measurable quantitiesinstance. Analog methods
are
pre
for
perature, pressure, and speed, calculation. An
have to be represented in a
ferred when these quantities
is a machine used in a postal
example of application of analog computers
to convert the weight of a package
into the cost of postage
department
needed for mailing. is the
The basic requirement in the application
of analog computers
the physical system ot
writing down of differential equations describing
interest. Given the differential equations, the analog result may be
electrical circuts
obtained either by direct method, in which equivalent
are directly used to simulate the time variations of the dependent
functional method, in wnien
variables of the physical system, or the indicated oy
electronie cireuits perform the mathematical operations
terms of the differential equation.
 Introduction to Nurmerical Computing 3

14 DIGITAL COMPUTING
which
A digital computer is a computing device that operates on inputs
are discrete in nature. The input data are numbers (or digits) that may
symbols. Just as a digital
represent numerals, letters, or other special an hour, a digital
clock counts the seconds and minutes in
directly the results.
computer counts discrete data values to compute but
the analog information,
Today's digital computers can cope with the
have to convert it into digital form. They do this by measuring
they
value of analog quantity at regular
intervals and converting that
to that
measurement number of electrical pulses corresponding
into a the
time and hands on
measurement. analog watch, for example,
In an
converts the
watch face change continuously; a digital
watch, however,
on
intervals, marked by the numbers changing
passage of time into tiny
the dial.
are more accurate than analog computers. Analog
Digital computers correct value,
within 0.1 per cent of the
accurate to
computers may be 1s
whereas digital computers can
obtain whatever degree of accuracy
correct number of decimal places.
They are

required by choosing the like numbers,

and output specific units
designed to read, store, manage, used for
computers are widely
letters, or punctuation marks. Digital
different applications and a r e
often called general purpose
many
computers.

COMPUTING
PROCESS OF NUMERICAL
1.5
involves formulation of
As stated earlier, numerical computing basic
that c a n be solved using
mathematical models of physical problems
arithmetic operations. The process
of numerical computing can be roughly
1.1:
which are illustrated in Fig.
divided into the following four phases
mathematical model
1. formulation of a
numerical method
2. construction of a n appropriate
method to obtain a solution
3. implementation of the
4. validation of the solution critical to the
suitable mathematical model is
The formulation of a defined as
mathematical model c a n be broadly
solution of the problem. A
that expressés the
a formulation
of certain mathematical equation
or process. Models may range
essential features of a physical system
from a simple algebraic equation
to a complex set of differential equations.
mathematical equations that might
Figure 1.2 shows various types of
mathematical models of physical processes.
result while forinulating siatement of
mathematical model begins with a
The formulation of a
considered. The factors may
associated factors to be
the problem and the
other laws of conservation in physics.
concern tie baiance of forces and
4 Numerical M e t h o d s

Introduction to Numerical Computing 5

(
 Numerical Computing Z
Introduction to

Numerical Methods
At 0, we get,
Taking the limits
>

have many uncertainties and unknowns. It might

Real life problems for approximating
therefore, be necessary
to make certain assumptions
that a r e considered C P
those features of the problem
and to include only
solution. An over-simplilied model may have only this differential equation is
critical to the final be enhanced later, if necessary. The
Solution of
The model may P ) = Po e
limited usefulness. m o r e difficult. We
make the solution procedure 0.
model refinement may e n h a n c e m e n t of the model and at time t
=

The
maintain the balance of where Po is the population C =
C1- C2
must always o n the growth
constant

solution required. depends

accuracy of the would be to try The population growth
model is available, o u r first step stable if C2. C =

Once a mathematical population will be

analytical solution. In most cases, the mathematical process
is t h e
to obtain a n explicit the n u m e r i c a l
computing
to analytical solutions or they may not be
with
models may not be amenable The third phase of This phase is
concerned

techniques. In such cases, we have to

selected.
using analytical of the method
solved efficiently mathematical models.
implementation
numerical methods to solve tasks
construct appropriate the following three
numerical method is a computational technique algorithm
As mentioned earlier, a 1. design of a n
number of basic arithmetic operations. 2. writing of a program
which involves only a finite to obtain the
results
of
be several alternative numerical o n a computer the validation
For a given problem, there might 3. executing it the next step is
different factors or trade-ofs before selecting able to obtain the results, to s e e that it
methods. We must consider Once w e a r e
verification of the
results
of equation, type of computer available, Validation m e a n s the w e m u s t go
back
a particular method-such as type the process. If it is not, then
and programming and maintenance efforts within the desired limits of a c c u r a c y .
accuracy, speed of execution, is
the following:
required. and check each of
Modelling is the process of translating
a physical problem into a
1. mathematical model itself
numerical method
selected
mathematical problem. The process involves 2.
used to implement
the method
1. making a number of simplifying assumptions 3. computational algorithm selection of a n
alternate
m e a n modification
of the model,
2. identification of important variables This may a combination
(or
of them).
3. postulation of relationships between the
variables improving the algorithm
numerical method o r
again. Figure 1.3
mathematical models is introduced, the cycle begins
This book is mainly concerned with the solution of Once modification real world
m o v e s from the
a

illustrates how the numerical computing cycle

using numerical techniques.
and back.
to mathematical world

Example1,1 Real world

Mathematical worid
of
Formulate a mathematical model for predicting the population growth Mathematical
Physical problem
a city. problem or model

Assumptions: and time interval.

Birth and death rates are proportional to population
Parameters:
Pit) - population at time t Validation of model Solution
At
AP-increase in population in time interval
Then, dosy bsporibro
AP = births in At - deaths in At

= C, P t ) at - C2P(t) A
baugsa
Application Solution techniques
= (C - C2) PU) At
and toolss
Fig. 1.3 Another way of looking at the computing process
AP
Growth rate a=
At
CP(t)
 merical Methods Introduction to Numerical Computing 2

I4CHARACTERISTICS OF NUMERICAL COMPUTING Efficiency

solution of a
Numerical methods exhibit certain computational characteristics during One moreconsideration in choosing a numerical method for
heir implementation. It is important to consider these characteristics mathematical model is efficiency. It means the amount of effort required
method that
while choosing a particular method for implementation. The human and computer to implement the method. A
by both effort and yet
characteristics that are critical to the success of implementation are: requires less of computing time and less of programming
ccuracy, rate of convergence, numerical stability, and efficiency. achieves the desired accuracy is always preferred.

AcCuracy 7COMPUTATIONAL ENVIRONMENT

E:ery method of numerical computing introduces errors. They may be
namely the
ouher due to using an approximation in place of an exact mathematical The last phase of the numerical computing process, hardware,
r e s o u r c e s such a s computer
rocedure (known as truncation errors) or due to inexact representation implementation phase, requires actual
language compilers,
and manipulation of numbers in the computer (known as roundoff errors). operating system and other systems software,
tools to manipulate data and
These errors affect the accuracy of the results. The results we obtain application programs and other software
must be sufficiently accurate to serve the purpose for which the provide output in a desired form.
computer
mathematical model was built. Choice of a method is, therefore, very The computer hardware may range from a small personal
on the nature and
size of the
much dependent on the particular problem. The general nature of these to a large super computer depending s a m e results on two
Serrors will be discussed in detail in Chapter 4. problem. A program may not always produce the
different types of computers due to difference in their round off
errors.

and play a n important role

Rate of Convergence Appropriate operating systems compilers
become
in developing portable programs. UNIX and MS-DOS have
Many numerical methods are based on the idea of an iterative process.
This process involves generation of a sequence of approximations with
popular operating systems for scientific computing. FORTRAN language
has dominated the scientific computing field for the last four decades
the hope that the process will converge to the required solution. Certain and it is expected to continue its predominant role for some more years.
methods converge faster than others. Some methods may not converge It has been continuously modified and extended to support the ever
at all. It is, therefore, important to test for convergence before a method changing requirements of software engineering. The likely strong
is used. Rapid convergence takes less execution time on the computer. competitor for FORTRAN in the near future will be C and C++ languages
There are several techniques for accelerating the rate of convergence of which contain some unique features and powerful control structures.
certain methods. The concepts of convergence and divergence are Portability is another strong point of these languages.
discussed in Chapter 4. They are also discussed in various places where
specific methods are analysed for convergence.
1.8 NEW TRENDSIN NUMERICAL COMPUTING
Numerical Stability
In recent years, the increasing power of computer hardware has affected
Another problem introduced by some numerical computing methods is the approach of numerical computing in several ways. It has forced
that of numerical instability. Errors introduced into a computation, from scientists and engineers to search for algorithms that are computationally
whatever source, propagate in different ways. In some cases, these errors and efficient. An important new trend is the construction of
fast
end to grow exponentially, with disastrous computational results. A
computing process that exhibits such exponential error growth is said to
algorithms to take advantage of specialised computer hardware such as
vector computers and parallel computers. Another trend is the use of
be'numerically unstable. We must choose methods that are not only fast sophisticated interactive graphics, in which the user can view the results
but also stable. graphically and advise the computer, graphically, on how to proceed
Numerical instability may also arise due to ill-conditioned problems. further.
There are many problems which are inherently sensitive to round oft One important development which is likely to have an increasing
errors and other uncertainties. Thus, we must distinguish between sen-
impact on scientific computing is symbolic computation. Symbolic
sitivity of methods and sensitivity inherent in problems.
When the problem is ill-conditioned, there is nothing we can do to computation systems would enable us to add, multiply and divide
polynomials or rational expressions the same way we would do using
uake c 2ethod to become numerically stable. pencil and paper. They can also solve certain mathematical problems
10 Numerical Methods Introduction to Numerical Computing 1I
the time rate of change
without rounding off errors. Symbolie computation is expected to play an
4. Newton's second
law of motion states that
the resultant force acting on it.
in scientific computation. to
increasing role of momentum of a body is equal to determine the
Object-oriented numerical computing is gaining importance due to a mathematical model
Using this law, formulate n e a r the earth's
surface.
the popularity oflanguageslike C++ and Java. They incorporate concepts terminal velocity of a free falling body
that the rate of heat from a
such as encapsulation, inheritance, polymorphism and operator states
5. The Newton's law of cooling between the
overloading. They support the idea that program units should interact the difference in temperatures
with one another only through clearly defined interfaces. They also enable
liquid is proportional to Formulate a mathematical model
to

without modifying it.

liquid and the surroundings.
the extension (or reuse) of the existing code govern this law.
force is proportional
water, the retarding
6. When a boat m o v e s through equation in
Formulate a differential
1.9 MATHEMATICAL BACKGROUND to the square of the velocity.
the m a s s m and the drag coefficient k.
terms of velocity given
mathematical background. characteristics of numerical computing
This book assumes that the readers have some 7. State the four of numerical
of algebra, functions, matrices, and integral How is it affected during the process
They require basic knowledge 8. What is accuracy?
and differential calculus. computing? numerical computing?
How is it important in
9. What is convergence?
numerical instability?
10. What do you m e a n by
1.10 SUMMARY 11. Distinguish between
sensitivity of methods
and sensitivity
of

In this chapter, we have introduced the concept of numerical computing problems. a numerical computing
12. Describe resources required for implementing
and discussed the steps involved in solving a physical problem using
numerical methods. We also discussed the characteristics
of numerical process.

computing and computing r e s o u r c e s required for implementing a

numerical method

Key Termns

terative process
Accuracy
Mathematical model
Algorithm
Analog computer Numerical computing
Numenical method
C
C++ Numenical stability
Continuous data Parallel computers
Rate of convergencee
Digital computer
Discrete data Round off error

Eficiency Symbolic computation

Truncation error
FORTRAN
Validation
General purpose computers
Vector computer
enlactat
M-conditioned problems

REVIEW QUESTIONS
1. What is Numerical Computing?
computing.
2. Distinguish between analog computing and digital
3. Describe, with the help of a block diagram, the process of numer
cal computing.