Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
141 views16 pages

Finite Difference Method Guide

The document discusses the finite difference method for numerically solving partial differential equations. It begins by explaining that finite difference replaces continuous fields with infinite degrees of freedom by a discretized field with finite nodes. It then provides examples of discretization equations for Poisson problems in 2D and cylindrical/polar coordinates. Various methods for solving the resulting difference equations are presented, including successive over-relaxation to accelerate convergence. Formulations for curved boundaries and interfaces between different materials are also covered.

Uploaded by

Martua Mario
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
141 views16 pages

Finite Difference Method Guide

The document discusses the finite difference method for numerically solving partial differential equations. It begins by explaining that finite difference replaces continuous fields with infinite degrees of freedom by a discretized field with finite nodes. It then provides examples of discretization equations for Poisson problems in 2D and cylindrical/polar coordinates. Various methods for solving the resulting difference equations are presented, including successive over-relaxation to accelerate convergence. Formulations for curved boundaries and interfaces between different materials are also covered.

Uploaded by

Martua Mario
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 16

Finite difference method

Chapter 3
Similar to other numerical methods, the aim of finite difference
is to replace a continuous field problem with infinite degrees of
freedom by a discretized field with finite regular nodes. The
partial derivatives of the unknown function are approximated
by the difference quotients at a set of finite discretization
points.
In this chapter discretization equations of Poissonian problems
both in translational symmetrical and axi-symmetrical
coordinates are discussed.
Discretization mode for two-dimensional problem
 Assume that a two-dimensional area Q is bounded by
the contour T, the potential function u within the
domain Q satisfies Poisson's equation and is subject to
the Dirichlet boundary condition shown as:

 the distances between grid lines are


called steps or mesh lengths while each
intersection of grid lines is called a node
Taylor's series

functions u(x) and u(y) at any node 0

 Indicate that the first order derivatives of the function are approximated by
expressions of the central difference quotients of the first order. The function u
at any point '0’ is dependent on the function value of its neighbouring nodes
and the step length. The accuracy is higher if the length of h is smaller.
Taylor's series

functions u(x) and u(y) at any node 0

 Second order for five-pointed star


 From following equation, the discretization form of Poisson's equation is :

 F0 is the source density of point O.


 Poisson's equation corresponding to the discretization form of an
asymmetric five-pointed star.

 For symmetrical five-pointed star we use Laplacian equation


Discretization mode for triangular and 3-D problem

triangular problem

3D problem
Discretization equation in polar coordinates

Laplacian equation

If 𝜃 is small enough,

The error depends on the


value of 𝜃. If 𝜃 = 150, then
the error is less than 0.41 %;
if 𝜃 = 7.50, then the error is
less than 0.039%.
Discretization formula of axisymmetric fields

Laplacian equation for cylindrical

The potential distribution is independent of the


coordinate 𝜃

r-z plane similar to x-y plane, hr = hz = h, r =ih


Discretization formula of the non-linear magnetic field

Reluctivity v=1Τ𝜇 is a function of the magnetic flux density. Poisson’s


equation for the magnetic vector potential in 2-D case is

Assuming that the current density and the reluctivity of the medium are
constant in the area of each mesh, the discretization equation can be

In non-linear problem an under-relaxation factor 𝛽 of v may be


introduced in some cases to accelerate the iteration as in
Difference equations for time-dependent problems

Explicid method

Implicid method
Solution Method for Difference Equation
Successive over-relaxation (SOR) method

𝑛+1 𝑛+1
In this equation the new approximate values 𝑢𝑖−1,𝑗 and 𝑢𝑖,𝑗−1 are used in the (n + 1) th
iteration as soon as they are available. This method is called the Gauss-Seidel iteration

Unfortunately, if the total number of the nodes is large, the convergence speed of the
Gauss-Seidel iteration is still too slow. The over-relaxation method is one of the most
generally effective, stable and successful methods to accelerate the convergence speed.

(𝑛)
The convergence criterion gives that the residuals 𝑅𝑖,𝑗 at every interior node
become less than a predetermined error e
Difference formulations of arbitrary boundaries and
interfacial boundaries between different materials
Difference formulations of a curved boundary
Difference formulations for interface of different materials

The difference equation at node 0 is

At boundary conditions

You might also like