INTRODUCTION - WHY NUMERICAL METHODS?

The ready availability of high-speed computers and easy-to-use powerful software packages has had a
major impact on engineering education and practice in recent years. Engineers in the past had to rely on
analytical skills to solve significant engineering problems, and thus they had to undergo a rigorous
training in mathematics. Today’s engineers, on the other hand, have access to tremendous amount of
computation power under their fingertips, and they mostly need to understand the physical nature of the
problem and interpret the results. But they also need to understand how calculations are performed by the
computers to develop an awareness of the processes involved and the limitations, while avoiding any
possible pitfalls.

1. Limitations

Analytical solution methods are limited to highly simplified

problems in simple geometries (Fig. 1). The geometry must be such
that its entire surface can be described mathematically in a coordinate
system by setting the variables equal to constants. That is, it must
fit into a coordinate system perfectly with nothing sticking out or in.
In the case of one-dimensional heat conduction in a solid sphere of
radius ro, for example, the entire outer surface can be described by
r = ro. Likewise, the surfaces of a finite solid cylinder of radius ro
and height H can be described by r = ro for the side surface and FIG 1
z = 0 and z = H for the bottom and top surfaces, respectively. Even minor complications in geometry can
make an analytical solution impossible. For example, a spherical object with an extrusion like a handle at
some location is impossible to handle analytically since the boundary conditions in this case cannot be
expressed in any familiar coordinate system. Even in simple geometries, heat transfer problems cannot
be solved analytically if the thermal conditions are not sufficiently simple. For example, the consideration
of the variation of thermal conductivity with temperature, the variation of the heat transfer coefficient
over the surface, or the radiation heat transfer on the surfaces can make it impossible to obtain an
analytical solution. Therefore, analytical solutions are limited to problems that are simple or can be
simplified with reasonable approximations.
 2. Better Modelling

We mentioned earlier that analytical solutions are exact solutions since they do not involve any
approximations. But this statement needs some clarification. Distinction should be made between an
actual real-world problem and the mathematical model that is an
idealized representation of it. The solutions we get are the
solutions of mathematical models, and the degree of applicability
of these solutions to the actual physical problems depends on the
accuracy of the model. An “approximate” solution of a realistic
model of a physical problem is usually more accurate than the
“exact” solution of a crude mathematical model (Fig. 2). When
attempting to get an analytical solution to a physical problem,
there is always the tendency to oversimplify the problem to make
the mathematical model sufficiently simple to warrant an
FIG 2
analytical solution. Therefore, it is common practice to ignore any
effects that cause mathematical complications such as nonlinearities in the differential equation or the
boundary conditions. So it comes as no surprise that nonlinearities such as temperature dependence of
thermal conductivity and the radiation boundary conditions are seldom considered in analytical solutions.
A mathematical model intended for a numerical solution is likely to represent the actual problem better.
Therefore, the numerical solution of engineering problems has now become the norm rather than the
exception even when analytical solutions are available.

3. Flexibility

Engineering problems often require extensive parametric studies to understand the influence of some
variables on the solution in order to choose the right set of variables and to answer some “what - if ”
questions. This is an iterative process that is extremely tedious and time-consuming if done by hand.
Computers and numerical methods are ideally suited for such calculations, and a wide range of related
problems can be solved by minor modifications in the code or input variables. Today it is almost
unthinkable to perform any significant optimization studies in engineering without the power and
flexibility of computers and numerical methods.

4. Complications

Some problems can be solved analytically, but the solution procedure is so complex and the resulting
solution expressions so complicated that it is not worth all that effort. With the exception of steady one-
dimensional or transient lumped system problems, all heat conduction problems result in partial
differential equations. Solving such equations usually requires mathematical sophistication beyond that
acquired at the undergraduate level, such as orthogonality, eigenvalues, Fourier and Laplace transforms,
Bessel and Legendre functions, and infinite series. In such cases, the evaluation of the solution, which
often involves double or triple summations of infinite series at a specified point, is a challenge in itself.
Therefore, even when the solutions are available in some handbooks, they are intimidating enough to
scare prospective users away.
Accuracy, Consistency, Stability and Convergence

Accuracy

Accuracy refers to the correctness of a numerical solution when compared to an exact solution. In most
cases, we do not know the exact solution. It is therefore more useful to talk of the truncation error of a
discretization method. The truncation error associated with the diffusion term using the finite difference
method is O Δx 2 , as shown by Equation 2.5.This simply says that if d 2φ dx2 is represented by the first
term in Equation 2.5, the terms that are neglected are of O Δx 2 . Thus, if we refine the mesh, we expect
the truncation error to decrease as Δx 2. If we double the x-direction mesh, we expect the truncation error
to decrease by a factor of four. The truncation error of a discretization scheme is the largest truncation
error of each of the individual terms in the equation being discretized. The order of a discretization method
is n if its truncation error is O Δx n . It is important to understand that the truncation error tells us how
fast the error will decrease with mesh refinement, but is not an indicator of how high the error is on the
current mesh. Thus, even methods of very high order may yield inaccurate results on a given mesh.
However, we are guaranteed that the error will decrease more rapidly with mesh refinement than with a
discretization method of lower order.

Consistency

A consistent numerical method is one for which the truncation error tends to vanish as the mesh becomes
finer and finer. (For unsteady problems, both spatial and temporal truncation errors must be considered).
We are guaranteed this if the truncation error is some power of the mesh spacing Δx (or Δt). Sometimes
we may come across schemes where the truncation error of the method is O Δx Δt . Here, consistency is
not guaranteed unless Δx is decreased faster than Δt. Consistency is a very important property. Without
it, we have no guarantee that mesh refinement will improve our solution.

Stability

The previous two properties refer to the behavior of the discretization method. Stability is a property of
the path to solution. For steady state problems, for example, we obtain a discretized set of algebraic
equations which must be solved. We may choose to solve this set using an iterative method. Depending
on the properties of the method,
solution errors may be amplified or damped. An iterative solution method is unstable or divergent if it
fails to yield a solution to the discrete set. It is also possible to speak of the stability of time-marching
schemes. When solving unsteady problems, we will use numerical methods which compute the solution
at discrete instants of time, using the solution at one or more previous time steps as initial conditions.
Stability analysis allow us to determine whether errors in the solution remain bounded as time marching
proceeds. An unstable time-marching scheme would not be able to reach steady state in an unsteady heat
conduction problem, for example (assuming that a steady state exists).
It is possible to analyze iterative and time marching methods using stability analysis. However, this is
most convenient for linear problems, and is usually too difficult for most realistic problems. Here, non-
linearities in the governing equations, boundary conditions, and properties, as well as coupling between
multiple governing equations, make a formal analysis difficult. In reality the practitioner of CFD must
rely on experience and intuition in devising stable solution methods.

Convergence

We distinguish between two popular usages of the term convergence. We may say that an iterative method
has converged to a solution, or that we have obtained convergence using a particular method. By this we
mean that our iterative method has successfully obtained a solution to our discrete algebraic equation set.
We may also speak of convergence to mesh independence. By this, we mean the process of mesh
refinement, and its use in obtaining solutions that are essentially invariant with further refinement.