MECH3780

Fluid Mechanics 2 and CFD

Computation Fluid
Dynamics (CFD)
Lecture 4 - Finite Difference
Reference: CFD Lecture Notes by Assoc Prof Dr Kamarul Arifin Ahmad
Objectives

• To introduce the finite difference in CFD

• To give overview about the finite difference
method in CFD
 Three Paradigms
 Finite Difference Method (FDM)
 Related to the differential form of the conservation laws
 Easily implemented on uniform, Cartesian meshes
 Requires transformations when implemented on conformal meshes
 Finite Volume Method (FVM)
 Related to the integral form of the conservation laws
 Most intuitive method
 Easily applied on conformal meshes
 Finite Element Method (FEM)
 Mathematically rigorous method
 Implementation more difficult
 Three Paradigms
 Finite Difference Method (FDM)
 Related to the differential form of the conservation laws
 Easily implemented on uniform, Cartesian meshes
 Requires transformations when implemented on conformal meshes
 Finite Volume Method (FVM)
 Related to the integral form of the conservation laws
 Most intuitive method
 Easily applied on conformal meshes
 Finite Element Method (FEM)
 Mathematically rigorous method
 Implementation more difficult
 Three Paradigms
 Finite Difference Method (FDM)
 Related to the differential form of the conservation laws
 Easily implemented on uniform, Cartesian meshes
 Requires transformations when implemented on conformal meshes
 Finite Volume Method (FVM)
 Related to the integral form of the conservation laws
 Most intuitive method
 Easily applied on conformal meshes
 Finite Element Method (FEM)
 Mathematically rigorous method
 Implementation more difficult
Finite Difference Approximations
for First Order Derivatives

u du ui − ui −1
Backward diff. ≈
dx ∆x

du ui +1 − ui
Forward diff. ≈
dx ∆x

du ui +1 − ui −1
i−1 i i+1 x Central diff. ≈
dx 2∆x

du ∆u
= Lim
dx ∆x → 0 ∆x
Finite Difference Approximations
for First Order Derivatives

Consider a Taylor’s series expansion for xi+1

du ∆x 2 d 2u ∆x 3 d 3u ∆x n d n u
u i +1 = u i + ∆x + 2
+ 3
++ n
+
dx i 2! dx i 3! dx i n! dx i

⇒
du u i +1 − u i ∆x d 2u ∆x 2 d 3u
= − − −
dx i ∆x 2
2! dx i 3! dx i3

du u − ui
⇒ = i +1 + err (∆x )
dx i ∆x
Finite Difference Approximations
for First Order Derivatives

If we say that du u − ui
≈ i +1
dx i ∆x

there will be a truncation error in the approximation, where:

∆x d 2u ∆x 2 d 3u
err (∆x ) = − 2
− 3
−
2! dx i 3! dx i

The truncation error is said to be order ∆x, written O(∆x), because

the leading term in the error function ∝ ∆x.
Hence we may write,
du u i +1 − u i
= + O (∆x )
dx i ∆x
This is known as a first order forward difference.
Finite Difference Approximations
for First Order Derivatives

First order approximation of the temporal derivative

u in +1 − u in
n
du
= + O (∆t )
dx i ∆t

n+1

n
i−1 i i+1
Finite Difference Approximations
for First Order Derivatives

A backward difference can be obtained by expanding for xi−1

u i −1 = u i − ∆x
du
+
∆x 2 d 2u
−
∆x 3 d 3u
++
(− ∆x ) d n u
n
+
dx i 2! dx 2 i 3! dx 3 i n! dx n i

⇒
du u i − u i −1 ∆x d 2u ∆x 2 d 3u
= + − −
dx i ∆x 2
2! dx i 3! dx i3

du u − u i −1
⇒ = i + O (∆x )
dx i ∆x
Finite Difference Approximations
for First Order Derivatives
A central difference approximation may be obtained by a
(weighted) average of the forward and backward differences:

Backward
du u − u ∆x d 2
u ∆x 2
d 3
u
= i i −1
+ − −
dx i ∆x 2! dx i2
3! dx i 3

Forward du u i +1 − u i ∆x d 2u ∆x 2 d 3u
= − − −
dx i ∆x 2
2! dx i 3! dx i3

Central du u i +1 − u i −1 ∆x 2 d 3u ∆x 4 d 5u
= − − −
dx i 2∆x 3! dx i3
5! dx i5

u − u i −1
+ O (∆x 2 )
du
⇒ = i +1
dx i 2∆x
Finite Difference Approximations
for Second Order Derivatives
An approximation to d 2u dx 2 may be obtained from a
(weighted) difference between the forward and backward Taylor’s
series:
du ∆x 2 d 2u ∆x 3 d 3u ∆x n d n u
Eqn.1 u i +1 = u i + ∆x + 2
+ 3
++ n
+
dx i 2! dx i 3! dx i n! dx i

Eqn.2 u i −1 = u i − ∆x
du
+
∆x 2 d 2u
−
∆x 3 d 3u
++
(− ∆x ) d n u
n
+
2 3 n
dx i 2! dx i 3! dx i n! dx i

Eqn.1 + Eqn.2
d 2u ∆x 2n d 2n u
u i + 1 + u i −1 = 2u i + ∆x
2
++ +
2
dx i (2n )! 2n
dx i

d 2u u i +1 − 2u i + u i −1 2∆x 2 d 4u 2∆x 2(n −1) d 2n u

⇒ = − −− −
2
dx i ∆x 2
4! dx i 4
(2n )! dx i 2n
Finite Difference Approximations
for Second Order Derivatives

d 2u u i +1 − 2u i + u i −1 2∆x 2 d 4u 2∆x 2(n −1) d 2n u

= − −− −
2
dx i ∆x 2
4! dx i 4
(2n )! dx i 2n

u i +1 − 2u i + u i −1
+ O (∆x 2 )
d 2u
⇒ =
2
dx i ∆x 2