Taylor Series Method for forming finite difference equation (DEM)
 

Basic Mathmatical Requirements:

f(x) is a single valued function and if in the interval a<x<b

Its first (N-1) derivatives are continuous

Its Nth derivative exists, f^N(x)

Taylor Series in Cartesian coordinates:

where

Example:

Heat conduction two dimensional - steady state with internal heat generation

Approximate each differential using Taylor Series:


Add and truncate

thus

A similar expression is obtained for 

We've been using the obvious subscript abbreviation T_o for T(x_o.y_o). We'll simplify notation further with the notation:

Then the finite difference equation at a given point becomes

with an error

The error is abbreviated as:

Improvement of the approximation

Simple option is to decrease the mesh size (increase number of mesh points)
Another option is to Obtain a higher order approximation

High order approximation

We must consider more points on the mesh when working with the equations. Introduce extended notation:

We want an approximation to the second derivative that only involves T_o, T_E, T_W, T_EE, and T_WW. Look at the Taylor expansions giving the values of T at these 4 points.




Now multiply the equation for T_E by the unknown coefficient "A", multiply the equation for T_W by the unknown "B", multiply the equation for T_EE by the unknown "C", and multiply the equation for T_WW by the unknown "D". Sum these modified equations. We want the summed equations to provide an approximation to the second derivative of T, so we require that the coefficient of the second derivative in the summed equation be equal to 1. We further require that the coefficients of the first, third, and fourth derivatives be zero. This gives us four equations that can be solved for values of A, B, C, and D. They are:


Replacing these values in the summed equation and rearranging gives:


What is the expression for the error, ?

Cylindrical coordinate system (DEM)
 

Example: Steady State Heat Conduction
 

Assume and, are uniform

In the r, direction, define

Write the Taylor Series expansion for each of these variables




In this instance we are free to either deal with all four expansions as a single sum, or group the radial and theta equations separately. I will follow the pattern in the previous higher order example. Multiply the equation for T_E by the unknown coefficient "A", multiply the equation for T_W by the unknown "B", multiply the equation for T_N by the unknown "C", and multiply the equation for T_S by the unknown "D". Sum these modified equations. We want the summed equations to provide an approximation to the radial and azimuthal derivatives in the conduction equations, so we require that the coefficient of the first derivative of T with respect to r in the summed equation be equal to 1/r_o. The coefficient of the second radial derivative must be one. We require that the coefficient of the first derivative with respect to theta be zero. The coefficient of the second derivative must be 1/r_o². This again gives us four equations that can be solved for values of A, B, C, and D. The solution is:

Thus the finite difference equation is

What is the error?




-

Or

Special treatment near origin

When symmetry exists at the origin (Geometrical & Thermal)

    Indeterminate

Thus

When no symmetry at r = 0, use Cartesian coordinates for the node at the origin:

Everywhere else use cylindrical coordinate system

Application for Higher Order Boundary Conditions

Other Lectures / Home