APPLIED NUMERICAL METHODS

Dr. Khaled Ahmida Ashouri

 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 The philosophy of Finite Difference Method (FDM) is to replace the derivatives of the
governing equations with algebraic difference quotients
 This results in a system of algebraic equations, which is solved at the discrete grid points in
the flow field
 So, FDM converts ODE or PDE, which may be non-linear, into system of linear equations

Derivation from Taylor’s polynomial: for fluid flow parameters u(i,j),

Let’s apply Taylor series expansion of ui+1,j about point (i,j) as follows,
𝜕𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
𝑢 , 𝑢, ∆𝑥 ⋯  (1)
𝜕𝑥 ,
𝜕𝑥 ,
2! 𝜕𝑥 ,
3!
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 Solving for 1st order partial derivative, Truncation error

𝜕𝑢 𝑢 , 𝑢, 1 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
⋯
𝜕𝑥 ,
∆𝑥 ∆𝑥 𝜕𝑥 ,
2! 𝜕𝑥 ,
3!

Finite difference representation

of partial derivative
Called: discrete equation.
First order Forward Difference

 This is normally expressed as:

𝜕𝑢 𝑢 , 𝑢,
𝑂 ∆𝑥 Truncation error
𝜕𝑥 ,
∆𝑥
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 If we now apply Taylor series expansion of ui-1,j about point (i,j) we get,

𝜕𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
𝑢 , 𝑢, ∆𝑥 ⋯  (2)
𝜕𝑥 ,
𝜕𝑥 ,
2! 𝜕𝑥 ,
3!

 Similarly,
𝜕𝑢 𝑢, 𝑢 ,
𝑂 ∆𝑥 Truncation error
𝜕𝑥 ,
∆𝑥

Finite difference representation

of partial derivative

First order Backward Difference

 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 Subtracting eqn.(2) from (1), we get, Truncation error

𝜕𝑢 𝑢 , 𝑢 ,
𝑂 ∆𝑥
𝜕𝑥 ,
2∆𝑥

Second order Central Difference

𝜕 𝑢 𝜕 𝑢 ∆𝑥
 Adding up eqns.(1) & (2), we get, 𝑢 , 𝑢 , 2𝑢 , ∆𝑥 ⋯
𝜕𝑥 ,
𝜕𝑥 ,
12

 Solving for the 2nd derivative,

𝜕 𝑢 𝑢 , 2𝑢 , 𝑢 ,
𝑂 ∆𝑥
𝜕𝑥 ,
∆𝑥
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 Summarizing:
𝜕𝑢 𝑢 , 𝑢,
≅ Forward Finite Difference
𝜕𝑥 ,
∆𝑥

𝜕𝑢 𝑢, 𝑢 ,
≅ Backward Finite Difference
𝜕𝑥 ,
∆𝑥

𝜕𝑢 𝑢 , 𝑢 ,
≅ Central Finite Difference
𝜕𝑥 ,
2∆𝑥

𝜕 𝑢 𝑢 , 2𝑢 , 𝑢 ,
≅ Second Derivative FD
𝜕𝑥 ,
∆𝑥
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

𝜕𝑢 𝑢 , 𝑢,
≅
𝜕𝑥 ,
∆𝑥
Forward difference

𝜕𝑢 𝑢 , 𝑢,
≅
𝜕𝑦 ,
∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

𝜕𝑢 𝑢, 𝑢 ,
≅
𝜕𝑥 ,
∆𝑥
Backward difference

𝜕𝑢 𝑢, 𝑢 ,
≅
𝜕𝑦 ,
∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

𝜕𝑢 𝑢 , 𝑢 ,
≅
𝜕𝑥 ,
2∆𝑥
Central difference

𝜕𝑢 𝑢 , 𝑢 ,
≅
𝜕𝑦 ,
2∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

𝜕 𝑢 𝑢 , 2𝑢 , 𝑢 ,
≅
𝜕𝑥 ,
∆𝑥

Second Derivative

𝜕 𝑢 𝑢 , 2𝑢 , 𝑢 ,
≅
𝜕𝑦 ,
∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 If we have mixed derivative, i.e, the parameter u depends on x- and y-directions,
 Let’s differentiate eqn.(1) and (2) with respect to y,
𝜕𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
𝑢 , 𝑢, ∆𝑥 ⋯  (1)
𝜕𝑥 ,
𝜕𝑥 ,
2! 𝜕𝑥 ,
3!

𝜕𝑢 𝜕𝑢 𝜕 𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
∆𝑥 ⋯  (3)
𝜕𝑦 ,
𝜕𝑦 ,
𝜕𝑥𝜕𝑦 ,
𝜕𝑥 𝜕𝑦 ,
2 𝜕𝑥 𝜕𝑦 ,
6

𝜕𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
𝑢 , 𝑢, ∆𝑥 ⋯  (2)
𝜕𝑥 ,
𝜕𝑥 ,
2! 𝜕𝑥 ,
3!

𝜕𝑢 𝜕𝑢 𝜕 𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
∆𝑥 ⋯  (4)
𝜕𝑦 ,
𝜕𝑦 ,
𝜕𝑥𝜕𝑦 ,
𝜕𝑥 𝜕𝑦 ,
2 𝜕𝑥 𝜕𝑦 ,
6
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

𝜕𝑢 𝜕𝑢 𝜕 𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥  (3)
∆𝑥 ⋯
𝜕𝑦 ,
𝜕𝑦 ,
𝜕𝑥𝜕𝑦 ,
𝜕𝑥 𝜕𝑦 ,
2 𝜕𝑥 𝜕𝑦 ,
6

𝜕𝑢 𝜕𝑢 𝜕 𝑢 𝜕 𝑢 ∆𝑥 𝜕 𝑢 ∆𝑥
∆𝑥 ⋯  (4)
𝜕𝑦 ,
𝜕𝑦 ,
𝜕𝑥𝜕𝑦 ,
𝜕𝑥 𝜕𝑦 ,
2 𝜕𝑥 𝜕𝑦 ,
6

𝜕𝑢 𝜕𝑢
 subtracting (3) from (4), and using the central FD for 𝜕𝑦
and we end up
𝜕𝑦
with, , ,

𝜕 𝑢 𝑢 , 𝑢 , 𝑢 , 𝑢 ,
𝑂 ∆𝑥, ∆𝑦
𝜕𝑥𝜕𝑦 ,
4∆𝑥∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

Mixed Derivative

𝜕 𝑢 𝑢 , 𝑢 , 𝑢 , 𝑢 ,
≅
𝜕𝑥𝜕𝑦 ,
4∆𝑥∆𝑦
 Numerical Methods Course
Dr. Khaled Ahmida

FINITE DIFFERENCE METHOD

 If we have a time variable in our equation, the Backward FD is stable and called “implicit”
FD. The forward FD is then called “explicit” and is some times unstable.