You are on page 1of 14

Computational Techniques

Module 9: Partial Differential Equations


Dr. Niket Kaisare
Department of Chemical Engineering
Indian Institute of Technology - Madras

Introduction

Ordinary Differential Equations


States vary in either time or one spatial dimension

Partial Differential Equations


States vary in:
Time and space; or
More than one spatial dimensions

In other words, PDEs represent systems with


two or more independent variables

An Example

A conducting rod

A conductingg block

An Example

A conducting rod
d 2T

dx

T Ta

T ( 0)
T ( L)

A conductingg block
2T

2T

2 0
2
x
y

T (0, y )
T ( L, y )

T ( x,0) T Ta
T ( x, d ) T Ta

Plug Flow Reactor

Steady State Model


Model:

FA dCdxA r C A

Initial Condition:

C A x 0 C0

Unsteady Model
Model:
Initial
I ii lC
Conditions:
di i

C A
C A
F
A
r C A
t
x
C A (t 0), x C0 ; C A t ,( x 0) Cin

Typical PDEs of Interest

First order PDEs

D E F G 0
x
y

Second order PDEs


2

2
2

A 2 B
C 2 D E F G 0
xy
x
y
x
y

Linear if (A to G) are functions of x and y only


Homogeneous if G = 0

Overview

Classification of PDEs

Method of Lines
Convert PDEs to ODEs

Finite Difference
Convert to (non-)linear
(non )linear equations

Parabolic PDEs

When B2 4AC = 0
Example: Transient Axial Dispersion Reactor

2
C A

A D C A r C

FA
A
2
t
x
x

T
Typically
ll observed
b
d when
h diffusive/conductive
d ff
/ d
terms
are present or dominant in the system

Parabolic PDEs: Solution Methods

Forward in Time, Central in Space (FTCS)


Apply forward difference in time:
O(t)
Apply central difference in space:
O(x2)
PDE Algebraic: Solve using NR / Gauss-Siedel

Crank-Nicholson Method (semi-implicit)


(
p )
Apply Midpoint Method in time
O(t2) accurate and more stable

Method of Lines
Central difference in space: PDE ODEs in time

Hyperbolic PDEs

When B2 4AC > 0


Wave-like solution which evolves in one (or
more) characteristic directions
Example: Transient PFR

C A
C A
FA
r C A
t
x

T
Typically
i ll observed
b
d when
h convective
ti tterms are
dominant in the system

Hyperbolic PDEs: Solution Methods

Forward in Time, Central in Space (FTCS)


Unstable and hence not used

Upwind
p
Difference
Apply forward difference in time:
O(t)
Apply backward difference in space: O(x)
PDE Algebraic: Solve using NR / Gauss-Siedel

Crank-Nicholson Method
Semi-implicit, higher accuracy and better stability

Method of Lines
Upwind difference in space: PDE ODEs in time

Elliptic PDEs

When B2 4AC < 0


Solution depends on both boundaries
Example: Heat conduction in a slab

2T
x

2T
y

Solution depends on all boundary conditions and the


PDEs need to be solved simultaneously.
simultaneously

Elliptic PDEs: Solution Methods

Central difference in all the directions


Accuracy:
O(x2), O(y2)
PDE Algebraic: Solve using NR / Gauss-Siedel
Conceptually simple, but implementation is tedious
and difficult because all equations
q
are solved
simultaneously.

Overview of this Module

Classification of PDEs

Method of Lines
Convert PDEs to ODEs

Finite Difference
Convert to (non-)linear
(non )linear equations

You might also like