0% found this document useful (0 votes)
56 views4 pages

Consistency, Stability and Convergence

consistency, stability and convergence

Uploaded by

saugatpandey
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
56 views4 pages

Consistency, Stability and Convergence

consistency, stability and convergence

Uploaded by

saugatpandey
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Computational Fluid Dynamics or CFD is the analysis of physical systems

like the fluid flow and heat transfer by performing computer-based simulations. We take
up a system, find the underlying governing differential equations, and determine suitable
numerical methods to approximate them. A numerical approximation is necessary
because a problem in differential equations can rarely be solved analytically. By using
suitable approximation methods, these differential equations can be converted into an
algebraic form for which a number of solution methods exist. These solution methods
can be coded into a computer to solve the algebraic approximation of the governing
differential equations over a predefined domain to simulate the concerned physics.

Thus, we can understand how essential numerical schemes are in making a problem
solvable. However, not all numerical schemes ensure a physical solution (or a solution
at all) to any given problem. The choice of a numerical method is characterized by three
important criteria - Consistency, Stability and Convergence. A numerical
discretization scheme, when applied to a governing equation, needs to meet all these
three criteria in order to provide a viable solution. This article aims to provide an
understanding of what these terms mean and the nuances in their definition.

Consistency

Consistency is the property that relates the discretized equation and the original partial
differential equation. We know that we use Taylor’s series to discretize the partial
differential equations and we neglect the higher order terms depending on the desired
accuracy. These neglected terms introduce an error into the solution called the
truncation error or the discretization error. Consistency is the requirement that the
truncation error of a discretized scheme should vanish as grid size and time step tend to
zero.

Let u_bar represent the true solution of a partial differential equation for some
continuous function and u_bar represent the exact value of its discretized form, we say
that the discretization scheme is consistent with the original partial differential equation
if
Stability

Stability is the property that expresses the relationship between the numerical solution
of a discretized equation with its exact solution. There will always be a discrepancy
between the exact value of the discretized equation and its value computed by using a
computer. This difference arises due to the development of round-off errors while
computing the numerical solution.

Computers can operate only on a limited machine precision i.e., there are only a certain
number of bits available to represent the decimal number system in the binary format.
Most of the times these many bits are sufficient to capture a decimal number exactly but
not always. Decimal numbers like 0.4 whose binary conversion is 0.000110011 needs
infinite bits for exact representation which is impossible. Consequently, this results in
chopping off the extra digits which cannot be accommodated or in other words rounding
off. These round-off errors, even though initially small, may grow exponentially with
each iteration and will cause the solution to blow up (give unphysical results) over a
larger number of iterations.

Therefore, a discretization scheme is considered stable if the round-off error ε

, reduces with each iteration.

The round-off error can be defined as the difference between the numerical value u

, and the exact value u~ of the discretized equation.


Now we can write as the previous equation as,

We know that the exact solution of the discretized equation will not change because that
is what we are approaching with each iteration. Therefore, finally the stability of a
system is ensured if

Therefore, we can conclude that a discretization scheme is stable if there is positive


dissipation/or no dissipation in the numerical solution when marching from one step to
another. If there is any negative dissipation, that means that the above condition would
not be satisfied and that would lead to increasing errors at each iteration and hence the
solution would be unstable.

Convergence

Convergence is the property that relates the numerical solution of the discretized
equation with the exact solution of the differential equation. It is said to be achieved
when the numerical solution approaches the true solution of the underlying partial
differential equation as grid size and time step approach to zero.
The convergence of a numerical scheme actually depends on the two properties
mentioned above. According to the Lax-Richtmyer Equivalence theorem, if a
numerical scheme is consistent and stable, then it is automatically convergent. But it is
important to remember that this theorem only applies to well-posed linear PDEs.

Hence, any numerical scheme has to satisfy the three mentioned criteria in order to
provide reliable simulation results.

You might also like