FVM – Diffusion problems | Dr Mohd Faizal Mohamad

FINITE VOLUME METHOD FOR DIFFUSION PROBLEMS

The governing equation of steady diffusion is derived from the general transport
equation for property φ as shown below;
𝜕(𝜌𝜑)
⃗ ) = 𝑑𝑖𝑣(𝛤 𝑔𝑟𝑎𝑑 𝜑) + 𝑆𝜑
+ 𝑑𝑖𝑣(𝜌𝜑𝑢
𝜕𝑡
(1)
By deleting the transient and convective terms, this gives;
𝑑𝑖𝑣(𝛤 𝑔𝑟𝑎𝑑 𝜑) + 𝑆𝜑 = 0

(2)
The control volume integration of the equation (2) yields;

∫ 𝑑𝑖𝑣(𝛤 𝑔𝑟𝑎𝑑 𝜑)𝑑𝑉 + ∫ 𝑆𝜑 𝑑𝑉 = ∫ 𝑛⃗ . (𝛤 𝑔𝑟𝑎𝑑 𝜑)𝑑𝐴 + ∫ 𝑆𝜑 𝑑𝑉 = 0

𝐶𝑉 𝐶𝑉 𝐴 𝐶𝑉

(3)
1.0 One dimensional steady state diffusion
The steady state diffusion of a property φ in 1D domain as shown in Figure 1 is
governed by
𝑑 𝑑𝜑
(𝛤 ) + 𝑆 = 0
𝑑𝑥 𝑑𝑥
Where Γ is the diffusion coefficient and S is the source term.

Control volume boundaries

φA=constant

φB=constant

D
C E
A B

Control volume Nodal points

Figure 1 One-dimensional steady state diffusion

There are three steps to solve the problem using finite volume method (FVM);
1. Grid generation
2. Discretisation
3. Solution of equations
 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

Step 1: Grid generation

The first step in the FVM is to divide the domain into discrete control volumes.

∆𝑥 = 𝛿𝑥𝐶′𝐸′

C’ D E’
C E

𝛿𝑥𝐶′𝐷 𝛿𝑥𝐷𝐸′

𝛿𝑥𝐶𝐷 𝛿𝑥𝐷𝐸

Figure 2 Control volume between nodal points C and E

Figure 2 shows the developed control volume between two nodal points C and E with
point D is the general nodal point. C’ and E’ are the control volume faces located on
the left and right sides of the general nodal point D, respectively.

Step 2: Discretisation
For the control volume defined in the Figure 2, the discretised equation at nodal point
D is given by;

𝑑 𝑑𝜑 𝑑𝜑 𝑑𝜑
∫ (𝛤 ) 𝑑𝑉 + ∫ 𝑆𝑑𝑉 = (𝛤𝐴 ) − (𝛤𝐴 ) + 𝑆̅∆𝑉 = 0
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝐸′ 𝑑𝑥 𝐶′
∆𝑉 ∆𝑉

(4)
Here,
A = cross-sectional area of control volume face
ΔV = volume
𝑆̅ = average value of source S over the control volume
In order to derive useful forms of the discretised equations, the interface diffusion
coefficient 𝛤 and the gradient 𝑑𝜑⁄𝑑𝑥 are required. Linear approximations so called
 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

central differencing in uniform grid gives the value for interface values and gradients
as below;
𝛤𝐶 + 𝛤𝐷
𝛤𝐶′ =
2
(5a)
𝛤𝐷 + 𝛤𝐸
𝛤𝐸′ =
2
(5b)
𝑑𝜑 𝜑𝐸 − 𝜑𝐷
(𝛤𝐴 ) = 𝛤𝐸′ 𝐴𝐸′ ( )
𝑑𝑥 𝐸′ 𝛿𝑥𝐷𝐸
(6)
𝑑𝜑 𝜑𝐷 − 𝜑𝐶
(𝛤𝐴 ) = 𝛤𝐶′ 𝐴𝐶′ ( )
𝑑𝑥 𝐶′ 𝛿𝑥𝐶𝐷
(7)
𝑆̅∆𝑉 = 𝑆𝑢 + 𝑆𝐷 𝜑𝐷
(8)
Source term may be a function of the dependent variable (linear form).
Substitute of equations (6), (7) and (8) into (4) gives;
𝜑 𝐸 − 𝜑𝐷 𝜑𝐷 − 𝜑𝐶
𝛤𝐸′ 𝐴𝐸′ ( ) − 𝛤𝐶 ′ 𝐴𝐶 ′ ( ) + (𝑆𝑢 + 𝑆𝐷 𝜑𝐷 ) = 0
𝛿𝑥𝐷𝐸 𝛿𝑥𝐶𝐷
(9)
Re-written the equation (9) yields;
𝜑𝐸 𝜑𝐷 𝜑𝐷 𝜑𝐶
𝛤𝐸′ 𝐴𝐸′ − 𝛤𝐸′ 𝐴𝐸′ − 𝛤𝐶 ′ 𝐴𝐶 ′ + 𝛤𝐶 ′ 𝐴𝐶 ′ + 𝑆𝑢 + 𝑆𝐷 𝜑𝐷 = 0
𝛿𝑥𝐷𝐸 𝛿𝑥𝐷𝐸 𝛿𝑥𝐶𝐷 𝛿𝑥𝐶𝐷
𝛤𝐸′ 𝛤𝐶′ 𝛤𝐶′ 𝛤𝐸′
( 𝐴𝐸′ + 𝐴𝐶′ − 𝑆𝐷 ) 𝜑𝐷 = ( 𝐴𝐶′ ) 𝜑𝐶 + ( 𝐴 ) 𝜑 + 𝑆𝑢
𝛿𝑥𝐷𝐸 𝛿𝑥𝐶𝐷 𝛿𝑥𝐶𝐷 𝛿𝑥𝐷𝐸 𝐸′ 𝐸
(10)
Let’s coefficients of 𝜑𝐶 and 𝜑𝐸 as 𝑎𝐶 and 𝑎𝐸 and the coefficient of 𝜑𝐷 as 𝑎𝐷 . Therefore,
equation (10) becomes;
𝑎𝐷 𝜑𝐷 = 𝑎𝐶 𝜑𝐶 + 𝑎𝐸 𝜑𝐸 + 𝑆𝑢
(11)
 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

Where,
𝑎𝐶 𝑎𝐸 𝑎𝐷
𝛤𝐶′ 𝛤𝐸′
𝛿𝑥𝐶𝐷 𝛿𝑐𝐷𝐸 𝑎𝐶 + 𝑎𝐸 − 𝑆𝐷

Step 3: Solution of equations

Discretised equations of (11) must be set up at each of the nodal points in order to
solve the problem. For control volumes that are adjacent to the domain boundaries the
general discretised equation (11) is modified to incorporate boundary conditions.

Example 1
Consider the problem of source-free heat conduction in an insulated rod whose ends
are maintained at constant temperatures of 100 °C and 500 °C. The one-dimensional
problem sketched in Figure 3 is governed by
𝑑 𝑑𝑇
(𝑘 ) = 0
𝑑𝑥 𝑑𝑥
Calculate the steady state temperature distribution in the rod. Thermal conductivity k
equals 1000 W/m·K, cross-sectional area A is 10 x 10-3 m2.

0.5 m

A B

TA = 100 Area (A) TB = 500

Figure 3 Heat conduction in an insulated rod

Solution
Step 1: Grid generation
Divide the length of the rod into five equal control volumes as shown in Figure 4.

1 2 3 4 5
TA TB
C’ E’ F’ G’

1 𝛿𝑥 𝛿𝑥 1
𝛿𝑥 𝛿𝑥
2 2

Figure 4 Five equal control volumes

 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

From the above figure 𝛿𝑥 is equal to 0.1m.

Step 2: Discretisation
Nodal points 2, 3 and 4 can be normally applied with discretised equation in (11) but
modification of discretised equation is required at points 1 and 5 as these points are
located adjacent to the boundaries.
Discretised equations for nodal points 2, 3 and 4 can be written as;
Point 2;
𝑘𝐸′ 𝑘𝐶′ 𝑘𝐶′ 𝑘𝐸′
( 𝐴𝐸′ + 𝐴𝐶′ ) 𝑇2 = ( 𝐴𝐶′ ) 𝑇1 + ( 𝐴 )𝑇
𝛿𝑥23 𝛿𝑥12 𝛿𝑥12 𝛿𝑥23 𝐸′ 3
Point 3;
𝑘𝐹′ 𝑘𝐸′ 𝑘𝐸′ 𝑘𝐹′
( 𝐴𝐹′ + 𝐴𝐸′ ) 𝑇3 = ( 𝐴𝐸′ ) 𝑇2 + ( 𝐴 )𝑇
𝛿𝑥34 𝛿𝑥23 𝛿𝑥23 𝛿𝑥34 𝐹′ 4
Point 4;
𝑘𝐺′ 𝑘𝐹′ 𝑘𝐹′ 𝑘𝐺′
( 𝐴𝐺′ + 𝐴𝐹′ ) 𝑇4 = ( 𝐴𝐹′ ) 𝑇3 + ( 𝐴 )𝑇
𝛿𝑥45 𝛿𝑥34 𝛿𝑥34 𝛿𝑥45 𝐺′ 5
Thermal conductivity 𝑘𝐶′ = 𝑘𝐸′ = 𝑘𝐹′ = 𝑘𝐺′ = 𝑘, node spacing is uniform (𝛿𝑥12 =
𝛿𝑥23 = 𝛿𝑥34 = 𝛿𝑥45 = 𝛿𝑥) and cross-sectional area 𝐴𝐶′ = 𝐴𝐸′ = 𝐴𝐹′ = 𝐴𝐺′ = 𝐴. No
source term exists in this problem (𝑆𝑢 = 𝑆𝑝 = 0).

The above three equations can be re-written as;

Point 2;
𝑘 𝑘 𝑘 𝑘
( 𝐴+ 𝐴) 𝑇2 = ( 𝐴) 𝑇1 + ( 𝐴) 𝑇3
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝛿𝑥
𝑎2 𝑇2 = 𝑎1 𝑇1 + 𝑎3 𝑇3
Point 3;
𝑘 𝑘 𝑘 𝑘
( 𝐴+ 𝐴) 𝑇3 = ( 𝐴) 𝑇2 + ( 𝐴) 𝑇4
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝛿𝑥
𝑎3 𝑇3 = 𝑎2 𝑇2 + 𝑎4 𝑇4
Point 4;
𝑘 𝑘 𝑘 𝑘
( 𝐴+ 𝐴) 𝑇4 = ( 𝐴) 𝑇3 + ( 𝐴) 𝑇5
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝛿𝑥
𝑎4 𝑇4 = 𝑎3 𝑇3 + 𝑎5 𝑇5
 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

𝑎1 = 𝑎2 = 𝑎3 = 𝑎𝐶 𝑎3 = 𝑎4 = 𝑎 5 = 𝑎𝐸 𝑎2 = 𝑎3 = 𝑎4 = 𝑎𝐷
𝑘 𝑘 𝑎𝐶 + 𝑎𝐸
𝐴 𝐴
𝛿𝑥 𝛿𝑥

Modification requires for points 1 and 5.

Point 1;

𝑑 𝑑𝑇 𝑑𝑇 𝑑𝑇
∫ (𝑘 ) 𝑑𝑉 = (𝑘𝐴 ) − (𝑘𝐴 ) =0
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝐶 ′ 𝑑𝑥 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴
∆𝑉

𝑇2 − 𝑇1 𝑇1 − 𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴
𝑘𝐴 ( ) − 𝑘𝐴 ( )=0
𝛿𝑥 𝛿𝑥⁄2
𝑘 𝑘 𝑘 𝑘
𝐴𝑇2 − 𝐴𝑇1 − 𝐴𝑇1 + 𝐴𝑇 =0
𝛿𝑥 𝛿𝑥 (𝛿𝑥⁄2) (𝛿𝑥⁄2) 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴
Re-arrange the above equation gives
𝑘 2𝑘 𝑘 2𝑘
( 𝐴+ 𝐴) 𝑇1 = 0𝑇𝑐 + ( 𝐴) 𝑇2 + ( 𝐴) 𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝛿𝑥
If we compare the above equation with equation (10), the fixed temperature boundary
condition enters the calculation as a source term (𝑆𝑢 + 𝑆𝐷 𝑇𝐷 ). Therefore,
2𝑘
𝑆𝑢 = ( 𝐴) 𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴
𝛿𝑥
2𝑘
𝑆𝐷 = − ( 𝐴)
𝛿𝑥
Coefficient for 𝑎𝑐 equal to zero because theoretically it’s located beyond the boundary.
Therefore, discretised equation for point 1 can be written as;
𝑎𝐷 𝜑𝐷 = 𝑎𝐶 𝜑𝐶 + 𝑎𝐸 𝜑𝐸 + 𝑆𝑢
𝑎1 𝑇1 = 𝑎𝐶 𝑇𝐶 + 𝑎2 𝑇2 + 𝑆𝑢
𝑎𝐶 𝑎𝐸 = 𝑎2 𝑎𝐷 = 𝑎1 𝑆𝐷 𝑆𝑢
𝑘 2𝑘 2𝑘
0 𝐴 𝑎𝐶 + 𝑎𝐸 − 𝑆𝐷 − 𝐴 𝐴𝑇
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐴

For point 5;

𝑑 𝑑𝑇 𝑑𝑇 𝑑𝑇
∫ (𝑘 ) 𝑑𝑉 = (𝑘𝐴 ) − (𝑘𝐴 ) = 0
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐵 𝑑𝑥 𝐺′
∆𝑉
 FVM – Diffusion problems | Dr Mohd Faizal Mohamad

𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐵 − 𝑇5 𝑇5 − 𝑇4
𝑘𝐴 ( ) − 𝑘𝐴 ( )=0
𝛿𝑥⁄2 𝛿𝑥
𝑘 𝑘 𝑘 𝑘
𝐴𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐵 − 𝐴𝑇5 − 𝐴𝑇5 + 𝐴𝑇4 = 0
(𝛿𝑥⁄2) (𝛿𝑥⁄2) 𝛿𝑥 𝛿𝑥
𝑘 2𝑘 𝑘 2𝑘
( 𝐴+ 𝐴) 𝑇5 = ( ) 𝑇4 + (0)𝑇𝐸 + ( 𝐴) 𝑇𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐵
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝛿𝑥
Discretised equation for point 5;
𝑎𝐷 𝜑𝐷 = 𝑎𝐶 𝜑𝐶 + 𝑎𝐸 𝜑𝐸 + 𝑆𝑢
𝑎5 𝑇5 = 𝑎4 𝑇4 + 𝑎𝐸 𝑇𝐸 + 𝑆𝑢
𝑎𝐶 = 𝑎4 𝑎𝐸 𝑎𝐷 = 𝑎 5 𝑆𝐷 𝑆𝑢
𝑘 2𝑘 2𝑘
𝐴 0 𝑎𝐶 + 𝑎𝐸 − 𝑆𝐷 − 𝐴 𝐴𝑇
𝛿𝑥 𝛿𝑥 𝛿𝑥 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝐵

Substitute the numerical values for k, A and δx into each equation yields;
300𝑇1 = 100𝑇2 + 200𝑇𝐴
200𝑇2 = 100𝑇1 + 100𝑇3
200𝑇3 = 100𝑇2 + 100𝑇4
200𝑇4 = 100𝑇3 + 100𝑇5
300𝑇5 = 100𝑇4 + 200𝑇𝐵
In matrix form

 300  100 0 0 0  T1  200TA 

 100 200  100
 0 0  T2   0 
 0  100 200  100 0  T3    0 
    
 0 0  100 200  100 T
  
4 0 
 0 0 0    
 100 300  T5  200TB 

T1  140 
T  220
 2  
T3   300 
   
T
  
4 380 
T5  460