Professional Documents
Culture Documents
Routines may not be available for the 1. Numerical methods are extremely
language or the type of computer we are powerful problem-solving tools. They
using. are capable of handling large systems of
An affordable commercial package may equations. As such, they greatly enhance
not contain all of the methods we need. your problem-solving skills.
Adaptation of methods to fit special 2. Many problems cannot be approached
situations may be impossible to using canned programs. If you are
accomplish with pre-written tools. conversant with numerical methods and
Programming our own routines helps us are adept at computer programming, you
will have the capability of designing
to understand the steps of a process
more fully; it also allows us to explore a your own programs to solve problems
without having to buy expensive
problem beyond the mere determination
of a solution. Such exploration may software.
3. Numerical methods are an efficient
provide insights that can expand our
view of a particular problem or help us vehicle for learning to use computers. It
is well known that an effective way to
in a different problem.
learn programming is to actually write
Even if pre-written codes are available,
computer programs.
an understanding of the numerical
4. Numerical methods provide a vehicle
processes allows us to choose the best
for you to reinforce your understanding
method among several for performing a
of mathematics. One function of
task.
numerical methods is to reduce higher
We become better educated in the tools
mathematics to basic arithmetic
of our profession. operations.
Characteristics of
Numerical Errors
Numerical Methods
An error in estimating or determining a We can deal with a complicated
quantity of interest can be defined as a problem more easily if we can break it down
deviation from its unknown true value. into relatively small, manageable pieces.
Errors can be classified based on their
A code segment of the task that
sources as non-numerical and numerical
must be performed at different stages of the
errors.
problem has to be written only once.
Non-numerical errors include (1)
Modules may be tested
modeling errors, (2) blunders and mistakes,
independently of other parts of the code; in
and (3) uncertainty in information and data.
this way, debugging (or error detection and
Numerical errors include the following: correction) is simplified.
1. Truncation Errors – results from the Well-chosen modules for production
termination of an infinite series codes simplify program maintenance and
expansion after a finite number of terms. upgrading because modification of one
2. Round-off Errors – results when module either leaves other modules
numbers are shortened to fit into unaffected or minimizes any necessary
computer words of limited length. changes in other modules.
3. Mathematical-approximation Errors –
result from using approximate
mathematical expressions for a Mathematical
relationship or solution. For example,
numerical methods sometimes call for Background
the use of a linear model for
representing nonlinear mathematical The following are the types of mathematical
expressions. subject areas that are covered in this course:
4. Error (e) – is the difference between the
computed (xc) and true (xt) values of a 1. Roots of Equations. The solution of
number x. many scientific and engineering
e=x c −x t problems requires finding the roots of
equations that can be nonlinear. These
5. Relative True Error (er) – is defined as problems are concerned with the value
the error e relative to the true value. of a variable or a parameter that satisfies
a single equation.
x c −x t 2. Simultaneous Linear Equations. Many
e=
xt engineering and scientific problems can
be formulated in terms of systems of
simultaneous linear equations. When
Modularity in these systems consist of only a few
equations, a solution can be found
Programming analytically using the standard methods
from algebra, such as substitution.
Modules are code segments for However, complex problems may
performing specific tasks; they are variously involve a large number of equations that
known as procedures, functions, cannot realistically be solved using
subroutines, etc., depending on the analytical methods. In these cases, we
language. Some advantages of modularity in will need to find the solution
programming are as follows. numerically using computers.
3. Numerical Integration. A physical
interpretation of numerical integration is
the determination of a close
approximation to the area under a curve domains. The methods that will be
of a function that has been determined considered are finite-difference methods
either from experimental data or from a in which derivatives of a function are
mathematical expression. Numerical replaced by finite-difference equations.
integration formulas play an important 6. Curve Fitting. In engineering, curve
role in the solution of differential fitting plays an important role in the
equations. analysis and interpretation of
4. Ordinary Differential Equations. experimental data and in the correlation
ODE’s are of great significance in of mathematical models formulated
engineering practice. This is because from fundamental engineering
many physical laws are expressed in principles. It involves the determination
terms of the rate of change of a quantity of a continuous function which results in
rather than the magnitude of the quantity the most “reasonable” or “best” fit of
itself. Two types of problems are experimentally measured values of
addresses: initial-value and boundary- (x1,y1) through (xn,yn). The least-
value problems. squares approximation is one form of
5. Partial Differential Equations. PDE’s curve fitting in which we seek the
are used to characterize engineering parameters of a given model that best fit
systems where the behavior of a the data
physical quantity is expressed in terms
of its rate of change with respect to two
or more independent variables. The
course is intended to provide an
introduction to methods of PDE and will
focus on the classical problems of
parabolic, elliptic, and hyperbolic
equations with relatively simple