Numerical Methods in Thermo-Fluid Dynamics I

Finite-Difference-Method 1

Dr.-Ing. Manuel Münsch

Friedrich-Alexander Universität Erlangen-Nürnberg
Lehrstuhl für Strömungsmechanik

Erlangen
 Erlangen

Schedule
No. Week(s) Practical Lecture Exercise
Monday: 10:00-11:30 Monday: 14:15-15:45 Wednesday: 10:15-11:45
Room: online Room: online Room: online
1. 02.11.20-06.11.20 Matlab Tutorial I Introduction and Navier-Stokes Eq. No Exercise
2. 09.11.20-13.11.20 Matlab Tutorial II; Computer Exercise 1 Simplification of NSE up to Boundary Layer Eq. Hand out Exercise 1
3. 16.11.20-20.11.20 Work on Computer Ex. 1 Initial Value Problems: Time Stepping (ODEs) Work on Exercise 1
4. 23.11.20-27.11.20 Work on Computer Ex. 1 Finite Difference Method 1 Presentation Ex. 1 &
Hand out Ex. 2
5. 30.11.20-04.12.20 Work on Computer Ex. 1 Finite Difference Method 2 Work on Ex. 2
6. 07.12.20-11.12.20 Work on Computer Ex. 1 Consistency, Stability and Convergence Presentation Ex. 2 &
Hand out Ex. 3 + 4
7. 14.12.20-18.12.20 Submit & Present Computer Ex. 1 Finite Volume Method 1 Work on Ex. 3 + 4
Computer Ex. 2
8. 21.12.20-23.12.20 Work on Computer Ex. 2 Finite Volume Method 2 Work on Ex. 3 + 4
24.12.20-06.01.21 Holiday Holiday Holiday
07.01.21-08.01.21 Presentation Ex. 3 + 4 &
Hand out Ex. 5
9. 11.01.21-15.01.21 Submit & Present Computer Ex. 2 Numerical solution of the Navier-Stokes-Equations 1 Work on Ex. 5
Computer Ex. 3
10. 18.01.21-22.01.21 Work on Computer Ex. 3 Numerical solution of the Navier-Stokes-Equations 2 Work on Ex. 5
11. 25.01.21-29.01.21 Work on Computer Ex. 3 Numerical solution of the Navier-Stokes-Equations 3 Presentation Ex. 5 &
Hand out Ex. 6
12. 01.02.21-05.02.21 Work on Computer Ex. 3 Numerical solution of the Navier-Stokes-Equations 4 Work on Ex. 6
Hand out Ex.6
13. 08.02.21-12.02.21 Submit & Present Computer Ex. 3 Boundary Conditions Presentation Ex. 6

Dr.-Ing. Manuel Münsch | FAU | 2

 Erlangen

Outline

• PART 1: Repetition
• Initial value problems: Time-Stepping

• PART 2: Finite-Difference-Method 1
• Introduction
• FDM via Taylor Series
• FDM via Polynomials

Dr.-Ing. Manuel Münsch | FAU | 3

PART 1: Repetition

Erlangen
 Erlangen

Repetition
Time Marching
• Time Marching Approaches are very similar to general solution strategies
of initial values problems (like Blasius equation) which becomes more
obvious by a reformulation of the equation:

∂(ρφ) ∂(ρuj φ) ∂ ∂φ
=− + (Γ ) + Sφ = Res(φ(t ), t ) (1)
∂t ∂ xj ∂ xj ∂ xj
with residual Res(φ(t ), t ) containing all convective and diffusive fluxes

• for comparison: ordinary differential equation / initial value problem (IVP)

d φ(t )
= f (φ(t ), t ) with φ(t0) = φ0 (2)
dt

Dr.-Ing. Manuel Münsch | FAU | 5

 Erlangen

Repetition
Time Marching
• Initial Value Problems (IVP):
• initial condition -> time step to approximate the state of the system after some
time evolution
• Accuracy: first, second and higher order methods
• Explicit Methods: comp. cheap, conditionally stable
• Implicit Methods: comp. expensive, unconditionally stable
• Single Step versus Multi Step Methods
• Stiff problems: implicit methods

Dr.-Ing. Manuel Münsch | FAU | 6

PART 2: Finite-Difference-Method 1

Erlangen
 Erlangen

Motivation
Spreading of a contaminant

Dr.-Ing. Manuel Münsch | FAU | 8

 Erlangen

Introduction

• a general analytical solution for the Navier-Stokes equation is not available

• the prediction of fluid flow is possible by making use of numerical solution
strategies
• this requires a discretization in time & time stepping algorithms (see last
lecture)
• and in addition a discretization in space via the common approaches:
• Finite-Differences
• Finite-Volumes
• Transport Equation:

∂(ρφ) ∂(ρuj φ) ∂ ∂φ
+ = (Γ ) + Sφ (3)
∂t ∂ xj ∂ xj ∂ xj

Dr.-Ing. Manuel Münsch | FAU | 9

 Erlangen

Introduction
Finite-Difference-Method
• starting point is the differential notation of the conservation equations
(here casted into a general transport eq.)
∂(ρφ) ∂(ρuj φ) ∂ ∂φ
+ = (Γ ) + Sφ (4)
∂t ∂ xj ∂ xj ∂ xj
• this is a coupled, non-linear differential equation containing:
2 2
• partial derivatives of first ( ∂∂x ) and second order ( ∂∂x 2 ), ( ∂ x∂∂ x ) in space
i i i j

• partial derivatives of first order ( ∂∂t ) in time

• for the numerical prediction the area of integration, i.e., the domain of
interest is subdivided via a computational grid
• at each grid point the geometrical coordinates xi and the field variables like
(ui , ρ, p, T ) are defined

Dr.-Ing. Manuel Münsch | FAU | 10

 Erlangen

Introduction
1-D Grid: 2-D Grid:

Nj

j+4

j+3 (i,j)

1 i+1 i+2 i+3 i+4 i+5 i+6 i+7 i+8 i+9 Ni j+2

j+1
Boundary Points Inner Points
1
1 i+1 i+2 i+3 i+4 i+5 i+6 i+7 i+8 i+9 Ni

Boundary Points Inner Points

• local structured grids for one and two dimensional flow problems
• grid is Cartesian and in general not equidistant (∆xi 6= const .)
• boundary conditions have to be defined at the boundary points / Boundary
Value Problem (BVP)
• solutions are predicted for inner points
Dr.-Ing. Manuel Münsch | FAU | 11
 Erlangen

Introduction
Finite-Difference-Method
• in the FD approach the partial differential equation is replaced by finite
differences
• at each grid point the partial derivatives of first and second order are
replaced by appropriate difference approximations → to be discussed
today
• the previous steps are known as discretization: here, FD-discretization
• in combination with boundary and initial conditions a coupled, algebraic
equation system is obtained
Aφ = b (5)
• the equation can be solved via a digital computer
• the exact solution of the partial differential equation equals the solution of
the finite difference equation plus an Error
• the Error is a function of the number of grid points and the disrcetization
scheme
Dr.-Ing. Manuel Münsch | FAU | 12
 Erlangen

Introduction
Finite-Difference-Method
• Target: approximate the exact solution of the partial differential equation as
good as possible
-> the numerical solution has to converge towards the exact solution with
decreasing step sizes ∆xi and ∆t: converged solution
• Requirements: the numerical scheme has to be consistent and stable
(See particular Lecture)

Dr.-Ing. Manuel Münsch | FAU | 13

 Erlangen

Finite-Difference-Method
Derivation of consistent differential formulations
• Starting Point: at the particular grid points (i , j ) the flow quantities are
available
• Task: approximate the partial derivative ∂ f /∂ x at the grid points via
difference approximations:
∂f ∆f
→ + T = Finite Difference + Truncation Error (6)
∂x ∆x

• the difference approximation is called consistent if:

 
∂f ∆f
lim T = lim − =0 (7)
∆x →0 ∆x →0 ∂ x ∆x
• the finite difference can be of different type, accuracy and characteristics
• the finite difference can be approximated via Taylor Series or Polynomials

Dr.-Ing. Manuel Münsch | FAU | 14

Finite-Difference-Approximations via Taylor Series

Erlangen
 Erlangen

Finite-Difference-Method
Taylor Series
Consider a continuous, infinitely differentiable function f (x ) at point x. Then,
the value of the function f at position x + ∆x can be calculated from a Taylor
Series expanded about point x:
∂ f ∆x ∂ 2f (∆x )2 ∂ 3f (∆x )3 ∂ n f (∆x )n
f (x + ∆x ) = f (x ) + + 2 + 3 + ... + n (8)
∂ x 1! ∂ x 2! ∂ x 3! | ∂ x{z n! }
HOT

• or in index notation on a 1-D Grid with grid spacing h1 = xi −1 − xi and

h2 = xi +1 − xi
∂ f h2 ∂ 2f (h2)2 ∂ 3f (h2)3
fi +1 = fi + + 2 + 3 + HOT (9)
∂ x 1! ∂ x 2! ∂ x 3!
2 2
∂ f h1 ∂ f (h1) ∂ 3f (h1)3
fi −1 = fi − + 2 − 3 + HOT (10)
∂ x 1! ∂ x 2! ∂ x 3!
• with higher order terms represented by term HOT
Dr.-Ing. Manuel Münsch | FAU | 16
 Erlangen

Finite-Difference-Method
Taylor Series Approximations
Illustration of function f and underlying base points:
f

fi
fi −1 fi +1

fi +2
fi −2
h1 h2

xi −2 xi −1 xi xi +1 xi +2 x

Dr.-Ing. Manuel Münsch | FAU | 17

 Erlangen

Finite-Difference-Method
Taylor Series: Forward Difference Approximation of ∂ f /∂ x
• the derivation of the FDS approximation for ∂∂xf at point i starts from:
∂ f h2 ∂ 2f (h2)2 ∂ 3f (h2)3
fi +1 = fi + + + 3 + HOT (11)
∂ x 1! ∂ x 2 2! ∂ x 3!

• solve for ∂ f /∂ x:
∂f fi +1 − fi ∂ 2f h2 ∂ 3f (h2)2
= − 2 − 3 − HOT (12)
∂x h2 |∂ x 2! ∂ x {z 3! }
T

• with truncation error of first order accuracy:

T = O((h2)1) (13)

Dr.-Ing. Manuel Münsch | FAU | 18

 Erlangen

Finite-Difference-Method
Taylor Series: Backward Difference Approximation of ∂ f /∂ x
• the derivation of the BDS approximation for ∂∂xf at point i starts from:
∂ f h1 ∂ 2f (h1)2 ∂ 3f (h1)3
fi −1 = fi − + − 3 + HOT (14)
∂ x 1! ∂ x 2 2! ∂ x 3!

• solve for ∂ f /∂ x:
∂f fi − fi −1 ∂ 2f h1 ∂ 3f (h1)2
= + 2 − 3 + HOT (15)
∂x h1 |∂ x 2! ∂ x{z 3! }
T

• with truncation error of first order accuracy:

T = O((h1)1) (16)

Dr.-Ing. Manuel Münsch | FAU | 19

 Erlangen

Finite-Difference-Method
Taylor Series: CDS Approximations of ∂ f /∂ x (A)
• the derivation starts from substracting eq. (14) from eq. (11):
∂f ∂ 2f h12 − h22 ∂ 3f h13 + h23
fi +1 − fi −1 = (h1 + h2) + 2 + 3 + HOT (17)
∂x ∂x 2! ∂x 3!

• solve for ∂ f /∂ x:
∂f fi +1 − fi −1 ∂ 2f h1 − h2 ∂ 3f h13 + h23
= − 2 − 3 + HOT (18)
∂x h1 + h2 ∂x 2! ∂ x (h1 + h2)3!
| {z }
T

• with truncation error:

h13 + h23
T = O((h1 − h2), ) (19)
h1 + h2
→ first order for h1 6= h2
→ second order for h1 = h2
Dr.-Ing. Manuel Münsch | FAU | 20
 Erlangen

Finite-Difference-Method
Taylor Series: CDS Approximations of ∂ f /∂ x (B)
• to improve accuracy for arbitrary grids where h1 6= h2 one can eliminate
2
the leading term of the truncation error: ∂∂xf2
• for this multiply eq. (11) with h12, multiply eq. (14) with h22 and substract the
resulting equations
• afterwards solve for ∂ f /∂ x:
∂f h12(fi +1 − fi ) + h22(fi − fi −1) ∂ 3f h1h2
= − 3 + ... (20)
∂x h1h2(h1 + h2) |∂ x 3{z! }
T

• with truncation error:

T = O(h1 · h2) (21)
→ formally of second order if / since h1 ≈ h2

Dr.-Ing. Manuel Münsch | FAU | 21

 Erlangen

Finite-Difference-Method
Taylor Series: CDS Approximations of ∂ f /∂ x (B)
• for equidistant grids where h1 = h2 = h:
∂f fi +1 − fi −1 ∂ 3 f h2 ∂ 5 f h4
= − 3 − 5 (22)
∂x 2h |∂ x 3! {z ∂ x 5}!
T

• with truncation error:

T = O(h2) (23)
→ second order accuracy

Dr.-Ing. Manuel Münsch | FAU | 22

 Erlangen

Finite-Difference-Method
Taylor Series: Approximations of ∂ f /∂ x
• FDS, BDS and especially CDS are the most used FD approximations

(Figure basing on the lecture script: Numerische Strömungsmechanik, M. Breuer, FAU/LSTM, 2006)

Dr.-Ing. Manuel Münsch | FAU | 23

 Erlangen

Finite-Difference-Method
Taylor Series: One-sided Approximations of ∂ f /∂ x
• base points don’t have to be centered around point i
• this is usefull close to the boundary of a grid where only a limited number
of neighbouring points are available or no neighbouring points are
available
f
h3

h1 h2

xi xi +1 xi +2 x

Dr.-Ing. Manuel Münsch | FAU | 24

 Erlangen

Finite-Difference-Method
Taylor Series: One-sided Approximations of ∂ f /∂ x
• here, an additional point i + 2 is considered:
∂ f h3 ∂ 2f h32 ∂ 3f h33
fi +2 = fi + + − + ... (24)
∂ x 1! ∂ x 2 2! ∂ x 3 3!

• consider eq. (11)·h32− (24)·h12 and solving for ∂ f /∂ x:

∂f h12(fi +2 − fi ) − h32(fi +1 − fi ) ∂ 3f h1h3

=− + 3 + ... (25)
∂x h1h3(h3 − h1) |∂ x 3{z
! }
T =O(h1 h3 )

Dr.-Ing. Manuel Münsch | FAU | 25

 Erlangen

Finite-Difference-Method
Taylor Series: One-sided Approximations of ∂ f /∂ x
• for h1 = h2 = h3/2 = h:
∂f fi +2 + 3fi − 4fi +1 ∂ 3 f h2
=− + 3 + ... (26)
∂x 2h |∂ x 3{z! }
T =O(h2 )
• here, with truncation error:
T = O(h2) (27)
→ of second order accuracy

Dr.-Ing. Manuel Münsch | FAU | 26

 Erlangen

Finite-Difference-Method
Taylor Series: Approximations of ∂ f /∂ x
Of course, higher order approximations basing on additional grid points are
also possible.
In general:
• First Order:
∂f fi +1 − fi
= + T1 (28)
∂x ∆x
• Second Order:
∂f fi +1 − fi −1
= + T2 (29)
∂x 2∆x
• Third Order:
∂f −fi +2 + 6fi +1 − 3fi − 2fi −1
= + T3 (30)
∂x 6∆x
• Fourth Order:
∂f −fi +2 + 8fi +1 − 8fi −1 + fi −2
= + T4 (31)
∂x 12∆x
Dr.-Ing. Manuel Münsch | FAU | 27
 Erlangen

Finite-Difference-Method
Taylor Series: Truncation Error
Here Tm is the truncation error:
∂ 2f ∆x ∂ 3f ∆x 2
T1 =− 2 − 3 + HOT1 = O(∆x 1) (32)
∂ x 2! ∂ x 3!
3 2
∂ f ∆x ∂ 5f ∆x 4
T2 =− 3 − 5 + HOT2 = O(∆x 2) (33)
∂ x 3! ∂ x 5!
∂ 4f ∆x 3 ∂ 5f ∆x 4
T3 = 4 + 5 + HOT3 = O(∆x 3) (34)
∂ x 4! ∂ x 5!
5 4
∂ f ∆x ∂ 7f ∆x 6
T4 = 5 + 7 + HOT4 = O(∆x 4) (35)
∂ x 5! ∂ x 7!

-> the exponent of the leading term is specifying the truncation error of the
scheme:
Tr = O(∆x r ) with order of accuracy: r (36)
-> thus, the power of decay of the discretization error when ∆x − > 0

Dr.-Ing. Manuel Münsch | FAU | 28

 Erlangen

Finite-Difference-Method
Taylor Series: Truncation Error
When the grid is refined by ∆x /n the truncation error is:
 r
∆x 1
Tr ( )= Tr (∆x ) (37)
n n

• first order scheme and n = 2:

∆x 1
Tr ( ) = Tr (∆x ) (38)
2 2
• second order scheme and n = 2:
∆x 1
Tr ( ) = Tr (∆x ) (39)
2 4
• third order scheme and n = 2:
∆x 1
Tr ( ) = Tr (∆x ) (40)
2 8
• fourth order scheme and n = 2:
∆x 1
Tr ( ) = Tr (∆x ) (41)
Dr.-Ing. Manuel Münsch | FAU | 2 16 29
 Erlangen

Finite-Difference-Method
Taylor Series: Approximation of ∂ 2f /∂ x 2
∂ 2f
• according to the previous derivations also approximations for ∂x 2 can be
derived

• 2 possible procedures:
∂f
A here, eq. (11) and eq. (14) have to be combined such that the leading term ∂x is
skipped
2
B express ∂∂xf2 as ∂∂x ∂∂xf

Dr.-Ing. Manuel Münsch | FAU | 30

 Erlangen

Finite-Difference-Method
Taylor Series: Approximation of ∂ 2f /∂ x 2
∂f
A combine eq. (11) and eq. (14) such that the leading term ∂x is skipped

• here, eq. (11) ·h1+ eq. (14) ·h2:

∂ 2f 2 (h1(fi +1 − fi ) − h2(fi − fi −1)) ∂ 3f h1 − h2
= + 3 + HOT (42)
∂x 2 h1h2(h1 + h2) ∂
| x 3!{z }
T =O((h1 −h2 )

• for h1 = h2 = h:
∂ 2f fi +1 − 2fi + fi −1 ∂ 4 f h2
2
= 2
− 4 + HOT (43)
∂x h |∂ x 12{z }
T =O(h2 )
• here, with truncation error:
T = O(h2) (44)
→ of second order accuracy
Dr.-Ing. Manuel Münsch | FAU | 31
 Erlangen

Finite-Difference-Method
Taylor Series: Approximation of ∂ 2f /∂ x 2
∂ 2f ∂ ∂f
B express ∂x 2 as ∂x ∂x

• with CDS:
∂f
− ∂∂xf |i −1/2
∂ 2f ∂ x |i +1/2
= (45)
xi +1/2 − xi −1/2
∂x 2
• here, ghost points are used at the positions i ± 1/2

• in detail:
∂f fi +1 − fi 1
|i +1/2 = with xi +1/2 = (xi + xi +1) (46)
∂x xi +1 − xi 2
∂f fi − fi −1 1
|i −1/2 = with xi −1/2 = (xi + xi −1) (47)
∂x xi − xi −1 2

Dr.-Ing. Manuel Münsch | FAU | 32

 Erlangen

Finite-Difference-Method
Taylor Series: Approximation of ∂ 2f /∂ x 2
• inserting the upper relations into eq. (45):

∂ 2f 2 [(xi − xi −1)(fi +1 − fi ) − (xi +1 − xi )(fi − fi −1)]

= (48)
∂x 2 (xi − xi −1)(xi +1 − xi )(xi +1 − xi −1)

→ same result as for approx. A with h1 = xi − xi −1 and h2 = xi +1 − xi

Dr.-Ing. Manuel Münsch | FAU | 33

Finite-Difference-Approximations via Polynomials

Erlangen
 Erlangen

Finite-Difference-Method
Polynomial Approach
Express the function f (x ) in the vicinity of xi as polynomial of n-th order,
where n stands for the order of accuracy, and fit the polynomial through n+1
points.
Example:

f (x ) = a0 + a1(x − xi ) + a2(x − xi )2 with n = 2 (49)

• here: a0, a1 and a2 are unknown

• these parameters have to be estimated by 3 equations basing on the grid
points xi , xi +1, xi −1
∂f ∂ 2f
• once f (x ) is estimated also the derivatives ∂x or ∂x 2 can be derived

Dr.-Ing. Manuel Münsch | FAU | 35

 Erlangen

Finite-Difference-Method
Polynomial Approach
Example:
f (x ) = a0 + a1(x − xi ) + a2(x − xi )2 with n = 2 (50)

• here: a0, a1 and a2 are unknown

• these parameters have to be estimated by 3 equations basing on the grid
points xi , xi +1, xi −1
• Point i:
fi = a0 + a1(xi − xi ) + a2(xi − xi )2 = a0 (51)
• Point i+1:
fi +1 = a0 + a1(xi +1 − xi ) + a2(xi +1 − xi )2 (52)
• Point i-1:
fi −1 = a0 + a1(xi −1 − xi ) + a2(xi −1 − xi )2 (53)

Dr.-Ing. Manuel Münsch | FAU | 36

 Erlangen

Finite-Difference-Method
Polynomial Approach
• solving for a0, a1 and a2 is leading to
a0 = fi (54)
fi +1(xi − xi −1)2 − fi (xi +1 − xi −1)[(xi − xi −1) − (xi +1 − xi )]
a1 =
(xi +1 − xi −1)(xi +1 − xi )(xi − xi −1)
fi −1(xi +1 − xi )2
− (55)
(xi +1 − xi −1)(xi +1 − xi )(xi − xi −1)
fi +1(xi − xi −1) − fi (xi +1 − xi −1) + fi −1(xi +1 − xi )
a2 = (56)
(xi +1 − xi −1)(xi +1 − xi )(xi − xi −1)
• for an equidistant grid with (xi − xi −1) = (xi +1 − xi ) = ∆x
a0 = fi (57)
fi +1 − fi −1
a1 = (58)
2∆x
fi +1 − 2fi + fi −1
a2 = (59)
Dr.-Ing. Manuel Münsch | FAU | 2∆x 2 37
 Erlangen

Finite-Difference-Method
Polynomial Approach
• Polynome:
f (x ) = a0 + a1(x − xi ) + a2(x − xi )2 (60)
• first derivative
∂f
= ao + 2a2(x − xi ) (61)
∂x
• second derivative
∂ 2f
2
= 2a2 (62)
∂x
• Thus:
∂f fi +1 − fi −1
= a1 = (63)
∂x 2∆x
2
∂ f fi +1 − 2fi + fi −1
= 2a2 = (64)
∂x 2 ∆x 2
-> corresponds to CDS approximations

Dr.-Ing. Manuel Münsch | FAU | 38

 Erlangen

Finite-Difference-Method
Polynomial Approach
Illustration of the Polynomial Approach for ∂ f /∂ x:

(Figure basing on the lecture script: Numerische Strömungsmechanik, M. Breuer, FAU/LSTM, 2006)

Dr.-Ing. Manuel Münsch | FAU | 39

 Erlangen

Finite-Difference-Method
Polynomial Approach
• for higher order approximations polynomials of higher order have to be
chosen
-> more grid points to estimate coefficients
-> truncation error smaller
• for higher order approaches there is a risk of oscillations in the solution
-> thus, not frequently used (See Lecture on Stability)

Dr.-Ing. Manuel Münsch | FAU | 40

 Erlangen

FDM-Schemes (1)
1st-Order FDS: ∂ u /∂ x 1st-Order BDS: ∂ u /∂ x

∆x ∆x
(i,j) (i+1,j) (i-1,j) (i,j)

ui +1,j − ui ,j ui ,j − ui −1,j
   
∂u ∂u
= (65) = (66)
∂x i ,j ∆x ∂x i ,j ∆x

Dr.-Ing. Manuel Münsch | FAU | 41

 Erlangen

FDM-Schemes (2)
2nd-Order CDS: ∂ u /∂ x 2nd-Order CDS: ∂ 2u /∂ x 2

2∆x ∆x ∆x
(i-1,j) (i+1,j) (i-1,j) (i,j) (i+1,j)

ui +1,j − ui −1,j
 
∂u ∂ 2u ui +1,j − 2ui ,j + ui −1,j
 
= (67) =
∂x i ,j 2∆x ∂x 2 i ,j (∆x )2
(68)

Dr.-Ing. Manuel Münsch | FAU | 42

 Erlangen

FDM-Schemes (3)
2nd-Order CDS: ∂ u /∂ y 2nd-Order CDS: ∂ 2u /∂ y 2

(i,j+1) (i,j+1)

∆y
(i,j)

2∆y
∆y

(i,j-1) (i,j-1)

ui ,j +1 − ui ,j −1
 
∂u 2
ui ,j +1 − 2ui ,j + ui ,j −1
 
∂ u
= (69) =
∂y i ,j 2∆y ∂y 2 i ,j (∆y )2
(70)

Dr.-Ing. Manuel Münsch | FAU | 43

 Erlangen

FDM-Schemes (3)
2nd-Order CDS: ∂ 2u /∂ x ∂ y

(i-1,j+1) (i+1,j+1)
∂ 2u
 
(∂ u /∂ y )i +1,j − (∂ u /∂ y )i −1,j
= (71)
∂x ∂y 2∆x
∆y   i ,j
∂u ui +1,j +1 − ui +1,j −1
= (72)
∂y 2∆y
 i +1,j
∂u ui −1,j +1 − ui −1,j −1
= (73)
∂ y i −1,j 2∆y
(i-1,j-1) ∆x (i+1,j-1)

With Eqs. (71)-(73):

2
ui +1,j +1 + ui −1,j −1 − ui −1,j +1 − ui +1,j −1
 
∂ u
= (74)
∂x ∂y i ,j 4∆x ∆y

Dr.-Ing. Manuel Münsch | FAU | 44

 Erlangen

Finite-Difference-Method
Conclusion
• the FDM allows to approximate the solution of Boundary Value Problems BVP
• typically 2nd order methods are used, but higher order methods can be constructed
• Solution procedure:
• choose a grid (potentially non-equidistant)
• apply FD-approximation to the unknown variable
• obtain a linear system of equations for the value of the unknown variable at each
grid point
• check accuracy by solving with finer grid size
• ... to be discussed in the next lecture

Dr.-Ing. Manuel Münsch | FAU | 45