0% found this document useful (0 votes)
41 views41 pages

10 11 FVM Convecton Diffusion

This document discusses numerical methods for solving the 1D Burgers equation, which models convection and diffusion. It introduces the equation and its exact solution. It then covers important discretization schemes like forward time centered space (FTCS) and upwind differencing. FTCS is conditionally stable depending on the grid size. Upwinding suppresses oscillations but introduces numerical diffusion. Higher order schemes like second order upwind and QUICK provide better accuracy. The finite volume method for discretizing the Burgers equation is also presented, covering approaches to estimate the convective fluxes at cell faces.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
41 views41 pages

10 11 FVM Convecton Diffusion

This document discusses numerical methods for solving the 1D Burgers equation, which models convection and diffusion. It introduces the equation and its exact solution. It then covers important discretization schemes like forward time centered space (FTCS) and upwind differencing. FTCS is conditionally stable depending on the grid size. Upwinding suppresses oscillations but introduces numerical diffusion. Higher order schemes like second order upwind and QUICK provide better accuracy. The finite volume method for discretizing the Burgers equation is also presented, covering approaches to estimate the convective fluxes at cell faces.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 41

1-D Convection-Diffusion Equation

Now
Focus on a prototype equation: Burgers Equation
Representative of Navier-Stokes equations
Can understand several key numerical issues
which are directly applicable to Navier-Stokes
equations

Numerical Issues
Different discretization schemes
Space
Time

Accuracy
Stability
Solution of discrete equations

Burgers Equation
Burgers Equation

u
u u 2u
t

1-D

u
u
2u
u
2
t
x
x

Linearized

u
u
2u
c
2
t
x
x

1-D Eqn. and Exact Solution


Problem statement

ut cux uxx ;

0 x 1

B.C.: u 0, t 0, u 1, t 1
I.C.: u x, 0 0

Exact Solution

Pe cL
( Peclet Number )

e xPe 1
u x, t Pe
e 1

Pe

u
SS

time

(boundary layer)
5

Boundary Layer Behavior


Boundary layer inversely proportional to Pe

L
L2
,
T

1
Pe
c

Pe , BL

L ,
T

fluid mechanics:

Pe
L

Re

Discretization Scheme: FTCS


FTCS: Forward in time & Central in space
u
u
2u
c
2
t
x
x
n 1
i

uin 1 uin
uin1 uin1
uin1 2uin uin1

2
Note:
t
2x

t n
ct 1 n
n
n
n
u
u

2
u

i 1
i 1
i 1
i
i 1
2
x 2
x

n
i

C c
n 1
i

t
x

CFL

using finite
difference
method

C n
C n

n
ui 1 1 2 ui ui 1
2
2

FTCS: : Preliminary Look at Stability


Useful parameter: cell Peclet number or

cell Reynolds number

Pe x

cx

Similar to Re for fluid mechanics


equations, but Re is related to

momentum diffusion: Pe is more


general

FTCS: Preliminary Look at Stability


1
Consider the FDE:
C n
C n

n 1
n
ui ui 1 1 2 ui ui 1
2
2

1
n
n
2 Pe x ui 1 1 2 ui 2 Pe x uin1
2
2

1st timestep: u14 2 Pex u50 1 2 u40 2 Pex u30

1
@ t 0

oscillations/wiggles/ 2- wave

t1

1
For Pe x 3 : u
2
8
Pe x 2 : u14 0
1
4

2 Pex

1 (I.C.)

t2

FTCS: Stability Analysis

Pex 2

will yield oscillations which increase in time.

Diffusion tries
to push left

Convection tries
to push to right

10

FTCS: Stability Analysis


Grid resolution for stability:

Pe x c

x 2

or

x 2

"Resolution Condition"

grid size or resolution


requirement for BL type
of solutions (along
streamwise direction)

11

First-Order Upwind Scheme


Central difference not suitable for high cell Peclet No.
Upwinding: to suppress the oscillation for high Pe
One-sided operator along the wind direction

ui ui 1
c x for c 0
u

x ui 1 ui
c
for c 0
x

c>0

c<0

12

First-Order Upwind Scheme


Finite difference scheme:

uin1

uin 1

2 Pex

1 Pex 2 uin 1 Pe x uin1


0
no oscillations

for CD

Scheme is first-order accurate: First-Order Upwind


Good since suppresses oscillations

13

First-Order Upwind Scheme


Bad since accuracy is low

Scheme is first-order accurate: First-Order Upwind


Effective viscosity:

eff

Pex
1
2

14

First-Order Upwind Scheme


Numerical viscosity:
If

numerical exact 1

Pex 100 , then numerical

Pex
2

exact 50

for large values of Pex, the numerical


solution is independent of exact!

15

First-Order Upwind Scheme

FOU

exact

CD

Any alternative scheme which suppresses


oscillations but not too dissipative ?
16

Alternative: Higher-Order Upwind


Alternative: higher-order upwind schemes:
Second-order upwind
QUICK

17

Second-Order Upwind Scheme


Second-order upwind (SOU), or
Linear upwind difference (LUD)
c

u
u 4ui 1 3ui
c i 2
,c 0

x i
2x

u 4ui 1 3ui
c i2
,c 0

2x

c>0

18

QUICK Scheme
Quadratic Upwind Interpolation for Convective
Kinematics (Leonard, 1979)

2uin1 3uin 6ui 1 ui 2


u
c
c
,c 0
x i
6x

u 6ui 1 3ui 2ui 1


c i2
,c 0

6x

nu N
ai ui
n
x
i 1

a
i 1

0 for any scheme

c>0

19

QUICK Scheme
QUICK = CD + dissipation

ui 1 ui 1 uin1 3uin 3ui 1 ui 2


u
c
c

,c 0
x i
6x
2x

ui 1 ui 1 ui 2 3ui 1 3ui ui 1
c

,c 0

6x
2x

20

Finite Volume Method for the


Burgers Equation

21

Finite Volume Method for Burgers Equation


Equation


u
x
x x

( for steady flow)

Integrate over control volume

uA e uA w


A A
x e
x w

E
22

Finite Volume Method for Burgers Equation


Define

F u convection coefficient mass flux per unit area


D

diffusion coefficient conductance


x

Convective and diffusive fluxes on east/west faces

Fw u w , Fe u e
Dw

, De e
xWP
xPE

23

Finite Volume Method for Burgers Equation


Central difference for diffusion terms

Fee Fww De E P Dw P W
Convective fluxes assumed known provisionally

e and w are unknowns to be determined

Central difference
First-order upwind
Second-order upwind
QUICK

24

Convective Flux: Central difference


Estimate e/w by averaging nodal values

e 12 P E
w 12 W P

uniform grid

25

Central difference
Discretized equation:

Fe
Fw

P E
W P De E P Dw P W
2
2

F
F
F
F

Dw w De e P Dw w W De e E
2
2
2
2

Nodal equation:

aPP aW W aEE

26

Central difference
Like to maintain diagonal dominance, i.e., aP=aW+aE:
add and subtract Fe Fw P

F
F
F
F

Dw w De e Fe Fw P Dw w W De e E
2
2
2
2

Discrete algebraic equation

aPP aW W aEE
Fw
2
F
aE De e
2
aP aW aE Fe Fw
aW Dw

27

Central difference
Consider grid with 5 nodes: 4 control volumes
CV1

CV2

x=0

CV3

CV5

CV4

CV6

x=1

Interior control volumes: CV3 and CV4

Fe
F
P E w W P De E P Dw P W
2
2

28

Central difference
CV1

CV2

CV3

CV5

CV4

x=0

CV6

x=1

Boundary control volumes: CV2 and CV5

Fe
P E F11 De E P D1 P 1
2
F
CV5: F66 w P W D6 6 P Dw P W
2
CV2:

D1 D6

2
2D ,
x

F1 F6 F
29

Central difference: discrete equation set


CV1

aPP aW W aEE Su

CV2

CV3

CV5

CV4

CV6

0
x=0

x=1

aP aW aE Fe Fw SP
AW

AE

CV 2

DF 2

CV 3, CV 4

DF 2

DF 2

CV 5

DF 2

SP

2D F

0
2D F

Su

2D F 1
0

2D F 6
30

First-Order Upwind (FOU)


Estimate e/w by upwinding

P , Fe 0
e
E , Fe 0

Fee p max Fe , 0 E max Fe , 0


Fww W max Fw , 0 P max Fw , 0
aPP aW W aEE

aW Dw max Fw , 0
aE De max Fe , 0
aP aW aE Fe Fw

Note : max Fe , 0 Fe max Fe , 0

31

Hybrid Scheme
Hybridize FOU and central difference schemes based on
face Peclet number

u w
Fw
Pe w

Dw w xWP

Pe w 2 : Use CD

Pe w 2 : Use FOU and neglect diffusion flux



Fw
aW max Fw , Dw
, 0

aE max Fe , De e , 0
2

32

Second-Order Upwind (SOU)


Estimate e/w by linearly extrapolating
2 upwind neighbors

3
1

2 P 2 W , if Fe 0
e 3
1

2 E 2 EE , if Fe 0

33

Second-Order Upwind (SOU)


Convective flux:

Fee 32 P 12 W Fe ,0 23 E 12 EE Fe ,0
Deferred correction approach:

Fee P 12 P W Fe ,0 E 12 E EE Fe ,0
FOU

Higher-order components

34

Second-Order Upwind (SOU)


Discrete equation:

aPP aW W aEE Su

same as FOU
aE De Fe , 0

Su 12 P W Fe , 0 12 E EE Fe , 0
aW Dw Fw , 0

35

QUICK Scheme
Estimate e/w by quadratic interpolation
using 2 upwind neighbors and 1
downwind neighbor

3
3
1

8 E 4 P 8 W , if Fe 0
e 3
3
1

E
P
4
8
8 EE , if Fe 0

36

QUICK Scheme
Convective flux:

Fee 83 E 34 P 81 W Fe , 0 34 E 83 P 81 EE Fe , 0
P 18 W 14 P 83 E Fe , 0 E 83 P 14 E 81 EE Fe , 0

37

SOU or QUICK: Boundary Treatment


Treatment of boundary control volume:
w
1

CV1

e
2

West face: flux from boundary values


East face: two options
FOU
Ghost cell to estimate value at ghostpoint W

38

Matrix Equation: A.=b


aP2 2 aE2 3 aW 2 1 b2
aP33 aE34 aW32 b3
aP4 4 aE4 5 aW4 3 b4
aP55 aE56 aW54 b5
aP2

aW3
0

aE2

aP3

aE3

aW4

aP4

aW5

u2
u1

0 b Su2
2 2
0 3 b3
.

aE4 4 b4

aP5 5 b5 Su5

u3

u4

aW 2 1

Su3

S u4

aE56

u5
u6

Matrix Equation
aP2 2 aE2 3 aW 2 1 b2
aP33 aE34 aW32 b3
aP4 4 aE4 5 aW4 3 b4
aP55 aE56 aW54 b5
aP2

aW3

aE2
aP3

aE3

aW4

aP4
aW5

0 Su2
2
3
.

aE4 4

aP5 5 Su5

Tri-diagonal matrix

aW 2 1

Su3

S u4

aE56

Closure
Spatial Discretization of 1-D
convection-diffusion equation using
FVM

41

You might also like