KKEK2142 Numerical Methods for

Engineers 2
Department of Chemical Engineering
University of Malaya
Semester 2, Session 2016/17

Instructor: Dr Mahar Diana Hamid

Lecture 4
PDE: Elliptic Equations
Introduction on PDE

Given that a function u that Partial differential equation

depends on both x and y, the (PDE) is an equation involving
partial derivatives of u with partial derivatives of
respect to x at an arbitrary unknown function of two or
point more independent
u (x,y)u (is
x defined
x, y ) uas
( x, y )
lim variables.
x x 0 x
The order of a PDE is the
Similarly, the partial highest derivative
derivative of u with respect to appearing in the equation.
y is defined as
A PDE is linear if it is linear in
u u ( x, y y ) u ( x, y )
lim the unknown function and all
y y 0 y its derivatives, with
coefficients depending only
on the independent
MDH-KKEK2142 (Sem 2 2016/2017) 2
variables.
Examples of PDE

2u 2u
2 xy 2 u 1
x 2
y
3u 2u What is the order
x 2 8u 5 y and linearity of each
x y
2
y
3
of the PDE?
u
2
3u
2
6 x
x xy 2

2u u
xu x
x 2
y

MDH-KKEK2142 (Sem 2 2016/2017) 3

 Engineering application: linear, 2nd-order PDE.
2u 2u 2u
A 2 B C 2 D 0
x xy y (1)

Depending on the values of A,B and C, PDE represented in (1)

can be classified into elliptic (B2-4AC<0), parabolic (B2-4AC=0)
and hyperbolic (B2-4AC>0).
The classification is based on the method of characteristics
developed by Vichnevetsky, 1981.
Each category relates to specific and distinct engineering
problem context that demand special solution techniques.

MDH-KKEK2142 (Sem 2 2016/2017) 4

Categories of PDE

B2-4AC Category Example

<0 Elliptic Laplace equation (steady state with
two spatial dimensions)
2T 2T
2 0
x 2
y
=0 Parabolic Heat conduction equation (time
variable with one spatial dimension)
T 2T
k' 2
t x
>0 Hyperbolic Wave equation (time variable with
one spatial dimension)
2 y 1 2 y
2 2
x 2
c t

MDH-KKEK2142 (Sem 2 2016/2017) 5

 Engineering Practice

Elliptic equation
Characterise a steady state system
Can be identified by the absence of a time
derivative
Typically employed to determine the steady
state distribution of unknown in two spatial
dimensions
E.g.: Heated plate with the boundaries of
the plate are held at different
temperatures. Hyperbolic equation
Deals with
Parabolic equation (Propagation problem) propagation however
Determine how an unknown varies with the unknown is
space and time characterised by a
Manifested by spatial and temporal 2nd order derivative
derivatives in the heat equation. with respect to time.
The problem focusses on how the spatial As a result, the
distribution changes as a function of time. solution oscillates.
E.g.: wave, sound
MDH-KKEK2142 (Sem 2 2016/2017) 6
etc
Finite Difference: Elliptic Equations

Characterised by steady state,

boundary value problems
Can be solved by
1. Laplace equation classic Laplace method
2. Solution techniques
The Laplacian Difference Equation
(Dirichlet boundary conditions)
The Liebmann Method ( Gauss-Seidel
method applied to PDEs, overrelaxation
and iterative method)

MDH-KKEK2142 (Sem 2 2016/2017) 7

Laplace equation

Consider an
element on the
face of a thin
rectangular plate
with the thickness
of z. The plate is
insulated but at its
edges,
temperatures can
be set at a
prescribed level.
The insulation and
the thinness of the
plate mean that the
MDH-KKEK2142 (Sem 2 2016/2017) 8
 This space is intentionally left blank. Please derive
it yourself from text book.

2T 2T
2 0 Laplace Equation (2)
x 2
y

MDH-KKEK2142 (Sem 2 2016/2017) 9

Solution Techniques

For numerical solution of elliptic PDEs, the PDE is

transformed into an algebraic difference equation.
Because of its simplicity and general relevance to most areas
of engineering, we will use a heated plate as an example for
solving elliptic PDEs.
Finite difference: representations based on treating the plate as
a grid of discrete points are substituted for the partial
derivatives derived previously using Laplace equation.

MDH-KKEK2142 (Sem 2 2016/2017) 10

 The Laplacian Difference Equations
Consider a point in the middle of the plate as (i,j)

MDH-KKEK2142 (Sem 2 2016/2017) 11

 The Laplacian Difference Equations
Central differences based on the grid scheme
2T Ti 1, j 2Ti , j Ti 1, j (3)

x 2 x 2
2T Ti , j 1 2Ti , j Ti , j 1 (4)

y 2 y 2
Substitute (3) and (4) into (2) to obtain
Ti 1, j 2Ti , j Ti 1, j Ti , j 1 2Ti , j Ti , j 1
0 (5)
x 2 y 2
Assuming that x y
Ti 1, j Ti 1, j Ti , j 1 Ti , j 1 4Ti , j 0 (6)

The Laplacian Difference Equations

MDH-KKEK2142 (Sem 2 2016/2017) 12

 In addition, boundary conditions along the
edges must be specified to obtain a unique
solution.
The simplest case is where the temperature at
the boundary is set at a fixed value, Dirichlet
boundary condition.

MDH-KKEK2142 (Sem 2 2016/2017) 13

Consider a heated plate where boundary temperatures are held
at constant levels

MDH-KKEK2142 (Sem 2 2016/2017) 14

 A balance for node (1,1) using equation (6) is:
T2,1 T0,1 T1,2 T1,0 4T1,1 0
T01 75
T10 0
T2,1 T1,2 4T1,1 75
Similar equations can be developed for other
interior points to result a set of simultaneous
equations.

MDH-KKEK2142 (Sem 2 2016/2017) 15

 The result is a set of nine simultaneous equations with
nine unknowns:

4T1,1 T2,1 T1,2 75

T1,1 4T2,1 T3,1 T2,2 0
T2,1 4T3,1 T3,2 50
T1,1 4T1,2 T2,2 T1,3 75
T2,1 T1,2 4T2,2 T3,2 T2,3 0
T3,1 T2,2 4T3,2 T3,3 50
T1,2 4T1,3 T2,3 175
T2,2 T1,3 4T2,3 T3,3 100
T3,2 T2,3 4T3,3 150

MDH-KKEK2142 (Sem 2 2016/2017) 16

Boundary Conditions

We will address problems that involve boundaries

at which the derivative is specified and
boundaries that are irregularly shaped.
Derivative Boundary Conditions
Known as a Neumann boundary condition.
For the heated plate problem, heat flux is
specified at the boundary, rather than the
temperature.
If the edge is insulated, this derivative becomes
zero. i.e. from Fouriers Law.

MDH-KKEK2142 (Sem 2 2016/2017) 17

 The Liebmann Method
Most numerical solutions of Laplace equation involve systems that are
very large.
For larger size grids, a significant number of terms will be zero.
For such sparse systems, most commonly employed approach is
Ti 1, j Ti 1, j which
Gauss-Seidel, Ti , j 1 Twhen
i , j 1
applied to PDEs is also referred as
Ti , j
Liebmanns method.
4

It is solved iteratively for j=1 to n and i=1 to m, and it will converged

on a stable solution. new
Overrelaxation: T value of T from present iteration
employed
Ti , j Ti , j (1 )Ti , j
new new old to accelerate the rate of convergence by
i , j

applying the following formula after each Ti , j iteration:

old
value of T from previous iteration
weighting factor between 1 and 2

The iteration is repeated until the percent relative error (a)I,j fall
below a prescribed stopping criterion s

MDH-KKEK2142 (Sem 2 2016/2017) 18

Example

Use Liebmanns method to solve for

the temperature of the heated plate in
slide no. 14. Employ over relaxation
with a value of 1.5 for the weighting
factor. Stopping criterion is 2%.

MDH-KKEK2142 (Sem 2 2016/2017) 19

 Secondary Variables
For

the heated plate case, The resultant heat flux:

a 2nd variable is the rate of
heat flux across the The direction of qn is given
plates surface. This
as:
quantity can be computed
from Fouriers Law:
for qx > 0

From finite difference

approx. for 1st derivative
for qx < 0

MDH-KKEK2142 (Sem 2 2016/2017) 20

Example:

Employ the results of Example in slide

18 with no iteration to determine the
distribution of heat flux for the heated
plate in slide no. 14. Assume that the
plate is 40 x 40 cm and is made out of
aluminum (k=0.49 cal/(s.cm.C))

MDH-KKEK2142 (Sem 2 2016/2017) 21

Consider a boundary node (0,j)
on the left edge of a heated plate
T1, j T1, j T0, j 1 T0, j 1 4T0, j 0
To approximate the derivative
normal to the edge (x derivative),
an imaginary point (-1,j) is
located at a distance x beyond
the edge

Representing the first derivative

in the x direction at (0,j) by the Thus, the derivative has
finite differential equation
been incorporated into the
T T1, j T1, j balance.

x 2 x
Similar relationships can
T
T1, j T1, j 2x be developed for
x derivative boundary
T conditions at the other
2T1, j 2x T0, j 1 T0, j 1 4T0, j 0
x edges.
MDH-KKEK2142 (Sem 2 2016/2017) 22
Irregular Boundaries
Will be covered in 4th
year elective

Many engineering
problems exhibit
irregular boundaries.
Consider a plate where
the lower left boundary is
circular.
Noticed that we have
affixed parameters 1, 2,
1, 2 to each of the
lengths surrounding the
nodes.
MDH-KKEK2142 (Sem 2 2016/2017) 23
 First derivatives in the x
A similar equation can
direction can be approximated be developed in the y
T as:
Ti , j Ti 1, j dimension
T
2
2 Ti , j 1 Ti , j Ti , j 1 Ti , j
x i 1,i 1x 2
y 2
y 1 ( 1 2 ) 2 ( 1 2 )
T Ti 1, j Ti , j

x i ,i 1 2 x
T T Substitute these

2T T x i ,i 1 x i 1,i equation into the

x 2 x x 1x 2 x Laplace equation
2
Ti , j Ti 1, j Ti 1, j Ti , j

2T 1x 2 x
2
x 2 1x 2 x
2
2T 2 Ti 1, j Ti , j Ti 1, j Ti , j
2
x 2 x 1 (1 2 ) 2 (1 2 )
MDH-KKEK2142 (Sem 2 2016/2017) 24