MEE 514 – Computational Fluid Dynamics

1 Lecture 8
1D Convection-Diffusion Equation
Dr Gilbert Accary
LECTURE PLAN
 1D convection-diffusion problem
 Governing equation
 Discretized equation
 Central differencing scheme
 Properties of a discretization scheme
 Upwind scheme, false diffusion
 Hybrid scheme
 Power law scheme

2
1D STEADY CONVECTION-DIFFUSION PROBLEM

T1 T = T1 T = T2 T2

T (x)

 Simplified problem:

3
 GOVERNING EQUATION
 Generalized convection-diffusion equation:
𝜕 
𝛻. 𝑉 𝛻. Γ 𝛻 𝑄
𝜕𝑡
 The flow must verify the continuity equation:
𝜕
𝛻. 𝑉 0
𝜕𝑡
 1D steady convection-diffusion problem without source:
𝑑 𝑑 𝑑 𝑑 𝑢
𝑢  0 (continuity)
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥

4
 Analytical solution:
DISCRETIZED EQUATION (CONT.)
 Integrating the convection-diffusion equation:
𝑑 𝑑 𝑑
𝑢 . 𝑑𝑣  . 𝑑𝑣 with 𝑑𝑣 𝐴. 𝛿𝑥
𝑑𝑥 𝑑𝑥 𝑑𝑥
𝑑
 Using Gauss theorem: 𝑢 . 𝑑𝑠  . 𝑑𝑠
𝑑𝑥

yields 𝑑 𝑑
𝑢𝐴 𝑢𝐴 𝐴 𝐴
𝑑𝑥 𝑑𝑥

5
DISCRETIZED EQUATION (CONT.)

 Using the convection and diffusion fluxes concept:

𝐹 𝐹  𝐷   𝐷  
 
where: 𝐹 𝑢 𝐹 𝑢 𝐷 𝐷
𝛿𝑥 𝛿𝑥
Assuming: 𝐴 𝐴 𝐴

 Integrating the continuity equation:

𝑑
𝑢 . 𝑑𝑣 𝑢 𝑑𝑠 0 → 𝑢𝐴 𝑢𝐴 →𝐹 𝐹
𝑑𝑥

6
CENTRAL DIFFERENCING SCHEME
 It is logical to try a linear interpolation to compute
the cell face values for the convective terms:
   
 
2 2
 The discretized equation becomes:
   
𝐹 𝐹 𝐷   𝐷  
2 2
 This can be re-arranged as:
0 (continuity)
𝐹 𝐹 𝐹 𝐹
𝐷 𝐷 𝐹 𝐹  𝐷  𝐷 
2 2 2 2

 Which can be written: 𝑎  𝑎  𝑎 

𝐹 𝐹 7
𝑎 𝐷 𝑎 𝐷 𝑎 𝑎 𝑎
2 2
CENTRAL DIFFERENCING SCHEME (CONT.)
 Consider the problem of transport of property :

with L = 1 m,  = 1 kg/m3, and  = 0.1 kg/m.s

 Analytical solution:

1. u = 0.1 m/s using 5 equally spaced cells

2. u = 2.5 m/s using 5 equally spaced cells
3. u = 2.5 m/s using 20 equally spaced cells
8
CENTRAL DIFFERENCING SCHEME (CONT.)
1. u = 0.1 m/s, x = 0.2 m  F = 0.1 kg/m2.s, D = 0.5 kg/m2.s
2. u = 2.5 m/s, x = 0.2 m  F = 2.5 kg/m2.s, D = 0.5 kg/m2.s

3. u = 2.5 m/s, x = 0.05 m  F = 2.5 kg/m2.s, D = 2 kg/m2.s

CD scheme
Good result for
Low F/D ratio

1. F/D = 0.2
2. F/D = 5
3. F/D = 1.25

9
 PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

𝐹𝑙𝑢𝑥 𝐹𝑙𝑢𝑥
Linear Approximation: Always consistent

𝐹𝑙𝑢𝑥

𝜙 𝜙 𝜙 𝜙 yields 10
Γ Γ 𝐹𝑙𝑢𝑥𝑒𝑠 𝑞 𝑞
𝛿𝑥 𝛿𝑥
 PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

Quadratic Interpolation: Not necessarily consistent

11
Quadratic Interp. is not always bad  QUICK scheme
PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

 In the absence of a source term,  must be bounded by its

boundary values (Ex: Steady state conduction).
 Diagonal dominance is sufficient condition for a convergent
iterative method (Scarborough - 1958)
∑𝑎 1 at all nodes
𝑎 1 at one node at least

 In finite volume method:

𝑎 ∑𝑎 𝑆 ∑𝑎 since SP < 0
12
𝑎 ∑ 𝑎 at Dirichlet boundary conditions
PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

 Boundedness  All anb should have same sign (all > 0)

If Not → ‘Wiggles’
 Central differencing scheme: 𝑎  𝑎  𝑎 
𝐹 𝐹
𝑎 𝐷 𝑎 𝐷 𝑎 𝑎 𝑎
2 2
𝐹 𝜌𝑢𝛿
𝑎 0 → Grid Peclet number 𝑃 2
𝐷 Γ
 In the considered example: 1. F/D = 0.2
2. F/D = 5 (wiggles) 13
3. F/D = 1.25
PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

 Transportiveness property of a flow (Roache, 1976) can

be illustrated by a constant source of  at point P.
(pure diffusion)
𝑃 0
Flow (pure convection)
𝑃 →∞
W P e E

14
𝐹 𝜌𝑢𝛿 Convection
Grid Peclet number 𝑃
𝐷 Γ Diffusion
PROPERTIES OF A DISCRETIZATION SCHEME

Conservativeness Boundedness Transportiveness

 At low Pe, face e is equally influenced by P and E

 Central differencing scheme is consistent for Pe < 2
 At high Pe, face e is more influenced by P than by E
 Central differencing scheme is not consistent

Flow

W P e E
15
UPWIND DIFFERENCING SCHEME

 To respects transportiveness property:

 If uw > 0 and ue > 0 → w = W and e = P

The discretized equation becomes:
𝐹  𝐹  𝐷   𝐷  

 If uw < 0 and ue < 0 → w = P and e = E

The discretized equation becomes:
𝐹  𝐹  𝐷   𝐷  

16
UPWIND DIFFERENCING SCHEME (CONT.)

 Identifying the coefficients of P, W, and E:

𝑎  𝑎  𝑎 
aW aE aP
Fw > 0, Fe > 0 Dw + Fw De Dw+ De+ Fe
Fw < 0, Fe < 0 Dw De Fe Dw+ De Fw

Since Fe = Fw (continuity)  𝑎 𝑎 𝑎

 Both expressions of aW and aE can be combined by:

𝑎 𝐷 Max 𝐹 , 0
𝑎 𝐷 Max 0, 𝐹 17
UPWIND DIFFERENCING SCHEME (CONT.)

 Reconsider the transport of  using upwind scheme:

Consider cases 2 (Pe = 5) and 3 (Pe = 1.25)

CD scheme
2nd order accurate
on uniform grid

Upwind scheme
1st order accurate
18
FALSE DIFFUSION IN UPWIND SCHEME

100

19
FALSE DIFFUSION IN UPWIND SCHEME (CONT.)

20
At high Reynolds, false diffusion can be large enough
to give physically incorrect results (Leschziner, 1980)
HYBRID SCHEME

 Combines central and upwind differencing schemes

(Spalding, 1972).
• Pe < 2 → CD scheme (2nd order accurate)
• Pe > 2 → Upwind scheme (1st order but No ‘Wiggles’)

 Drawback: Globally 1st order accurate

𝐹 𝜌𝑢
 For example, for a west face: 𝑃𝑒
𝐷 Γ /𝛿𝑥
Flux on west face:
𝜙 𝜙
𝑞 𝐹 𝐷 𝜙 𝜙 for 2 𝑃𝑒 2
2
𝑞 𝐹 𝜙 for 𝑃𝑒 2
(No diffusion)
𝑞 𝐹 𝜙 for 𝑃𝑒 2 21
HYBRID SCHEME (CONT.)
 The general form of the discretized equation is:
𝑎  𝑎  𝑎 
 The different expressions of aW and aE are combined by:
𝐹
𝑎 Max 𝐹 , 𝐷 ,0
2
𝐹
𝑎 Max 𝐹 , 𝐷 ,0
2
Using Fe = Fw (continuity)  𝑎 𝑎 𝑎
 Reconsider the transport of  using hybrid scheme:
Consider cases 2 (Pe = 5) and 3 (Pe = 1.25)

22
POWER LAW SCHEME
 The power-law differencing scheme of Patankar-1980
is obtained from an accurate approximation of the
exact solution of the 1D convection-diffusion problem
 It produces better results than the hybrid scheme
 It results approximately in the exact solution of the
1D convection-diffusion problem without source
term, it is however 1st order accurate in general
 Flux at face west:
1 0.1𝑃𝑒
𝑞 𝐹 𝜙 𝜙 𝜙 for 0 𝑃𝑒 10
𝑃𝑒
𝑞 𝐹 𝜙 for 𝑃𝑒 10 (No diffusion)
23
Similar expressions are written for 𝑃𝑒 < 0
POWER LAW SCHEME (CONT.)
 These expression yield:
𝑎 𝐷 Max 0, 1 0.1 𝑃𝑒 Max 𝐹 , 0
𝑎 𝐷 Max 0, 1 0.1 𝑃𝑒 Max 𝐹 ,0
 Considering the transport of  using power-law scheme:

24
QUICK SCHEME
 Quadratic Upstream Interpolation for Convective Kinetics
6 3 1
If uw > 0, 𝜙 𝜙 𝜙 𝜙 for 𝛿𝑥 𝑐𝑜𝑛𝑠𝑡.
8 8 8
𝑎  𝑎  𝑎  𝑎  𝑎  𝑏

25
 Assignment #3: Assessment of the QUICK scheme