Applications of Finite Element Method To Scalar Transport Problems
Applications of Finite Element Method To Scalar Transport Problems
Applications of Finite Element Method To Scalar Transport Problems
Lecture 23
APPLICATIONS OF FINITE ELEMENT METHOD TO
SCALAR TRANSPORT PROBLEMS
d d
S 0 (23.1)
dx dx
where is the diffusion coefficient and S is the source term. Boundary values of at end-
points 0 and L are specified as
(0) A , ( L) B , (23.2)
Let us divide the domain in two node linear finite elements. In any element, the unknown
function can be approximated as
e ( x) N1 ( x)1e N 2 ( x)2e (23.3)
where N1 and N2 are linear shape functions, and , are values of at local nodes 1 and 2
1
e e
2
Performing integration by parts, and requiring that the weight function vanishes at the end-
points, we get the following weak form:
dw d
L
0 dxi dx wi S dx 0 (23.5)
Ku = f (23.6)
where K is the stiffness matrix, u is the vector of nodal unknowns ϕi and f is called the load
vector. Elements of the matrices in the preceding equation are given by
K mn K mn
e
, and f m f me (23.7)
e e
where
N and f me N
e
K mn m,k N n ,k dx, m S dx. (23.8)
e e
System of linear equations is solved using suitable linear solver (e.g. TDMA would be the
best solver for the present example as K would be tri-diagonal for linear finite elements).
Next example illustrates the preceding process explicitly for a sample heat conduction
problem.
Example 23.1
Consider the steady state heat conduction in a slab of width l = 0.5 m with heat generation.
The left end of the slab (x = 0) is maintained at T = 373 K. The right end of the slab (x = 0.5
m) is being heated by a heater for which the heat flux is 1 kW/m2. The heat generation in the
slab is temperature dependent and is given by Q = (1273 – T) W/m3. Thermal conductivity is
constant at k = 1 W/(m-K).Write down the governing equation and boundary conditions for
the problem. Use the finite difference method (central difference scheme) to obtain an
approximate numerical solution of the problem. For the first order derivative, use forward or
backward difference approximation of first order. Choose element size h = 0.1, and use the
TDMA. (We have chosen the same as the one solved previously in Example 15.1 using FDM
to illustrate comparison between FEM, FVM and FDM.)
Solution
Let us recall that governing equation for the steady state heat conduction with constant heat
generation is the slab is
d 2T
k 2 Q 0 (i)
dx
Given: Q = 1273 – T. Thus, Eq. (i) becomes
d 2T
k 2 ( f T) 0 (ii)
dx
where f = 1273. Left end of the slab is maintained at constant temperature; hence boundary
condition at this end is given by
T (0) 373 (iii)
At the right end, heat influx is specified. Thus, boundary condition at this end is
dT
k ( L) 1000 (iv)
dx
For discretization, let us use a finite element mesh of linear elements. Global nodes are 1
(x=0), 2 (x=0.1), 3(x=0.2) , 4(x=0.3), 5(x=0.4) and 6 (x=0.5). Each finite element is of width
h = 0.1, whereas finite volumes around boundary nodes 1 and 6 are of width 0.05.
1 2 3 4 h 5 6
Galerkin weighted residual statement (23.4) for this problem takes the form
L
d 2T
0 i k dx 2 ( f T ) dx 0
w (v)
x2 x x x1 dN 1 dN 2 1
N1 ( x) , N 2 ( x) 1 , (viii)
h h dx h dx h
Inserting the approximation (vii), weighted residual formulation for an element can be written
as
L
dwi dN1 e dN 2 e e dT
dx dx 1 dx 2 i 1 1 2 2 dx i 0 e fwi dx (ix)
e
k T T w ( N ( x )T N ( x )T ) dx k w
e
For Galerkin formulation, weight function wi will be taken as one of the shape
functions, Ni. Hence, the preceding equation can be written as in matrix form as
K11e K12e T1e b1e
e e e
e (x)
K e1 K 22 T2 b2
where
dN m dN n
N m N n dx and bme N m f dx,
e
K mn (xi)
e
dx dx e
with an additional contribution to b for the first and last element coming from specified flux.
Thus,
dN dN i 1 h
Kiie i Ni Ni dx , i 1, 2
e
dx dx h 3
dN i dN j 1 h
Kije
e
dx dx
Ni N j dx ,
h 6
i, j 1, 2
h
b1e b2e f (for interior elements) (xii)
2
h dT h
b11 f k , b21 f ,
2 dx x 0 2
h h dT
b1N
f , b2N f k
2 2 dx x L
Assembling all elemental equations, we get
KT = b , where K mn K mn
e
, and f m f me (xiii)
e e
1 h 1 h
h3
h 6
0 0 0 0
1 h 1 h 1 h h dT
h 6 2 0 0 0 2 f k dx
h 3 h 6 T
1 x 0
1 h 1 h 1 h T2 hf
0 2 0 0
h 6 h 3 h 6 T hf
3
1 h 1 h T hf
0 0 0 2 0 4 (xiv)
h 3 h 6 T5 hf
1 h 1 h
1 h T h dT
0 0 0 2 6 f k
h 6 h 3 h 6 2 dx xL
1 h 1 h
0 0 0 0
h 6 h 3
Let us note that temperature is specified at x = 0 (flux kdT/dx is unknown). Hence, we modify
the first equation in Eq. (xiv) by incorporating known temperature value. Further, substitute h
= 0.1, f = 1273, value of flux at right-end of the domain, and divide last five equations by 10
(to make all the diagonal elements to be of same order of magnitude). The resulting discrete
system is
1 0 0 0 0 0 T1 373
a d a 0
0 0 T2 12.73
0 a d a 0 0 T3 12.73
0 0 0 d a 0 T4 12.73 (xv)
0 0 0 a d a T5 12.73
0 0 0 0 a d / 2 T6 106.365
where a = 0.9983333, d = 2.0066667.
The preceding system is very similar to the discrete system (x) obtained using finite
difference discretization in Example 15.1 (only coefficients in last row differ). Numerical
calculations using TDMA are given in the following table:
Comparison of results obtained using FEM, FVM and FDM using the same grid size is given
below:
Exact FEM FVM FDM %Error FEM %Error FVM %Error FDM
373.000 373.000 373.000 373.000 0.000 0.000 0.000
498.950 499.006 498.931 497.289 0.011 0.004 0.333
617.180 617.259 617.121 613.821 0.013 0.009 0.544
728.850 728.944 728.753 723.762 0.013 0.013 0.698
835.080 835.179 834.942 828.209 0.012 0.017 0.823
936.920 937.029 936.751 928.209 0.012 0.018 0.930
We can clearly observe that the finite element and finite volume results using identical grid
spacing are more accurate than those obtained using FDM. The primary reason is use of first
order backward difference method used in finite difference solution for incorporation of flux
boundary condition.
REFERENCES/FURTHER READING
Chung, T. J. (2010). Computational Fluid Dynamics. 2nd Ed., Cambridge University Press,
Cambridge, UK.
Muralidhar, K. and Sundararajan, T. (2003). Computational Fluid Dynamics and Heat
Transfer, Narosa Publishing House.
Reddy, J. N. (2005). An Introduction to the Finite Element Method. 3rd Ed., McGraw Hill,
New York.
Zienkiewicz, O. C., Taylor, R. L., Zhu, J. Z. (2005). The Finite Element Method: Its Basis
and Fundamentals, 6th Ed., Butterworth-Heinemann (Elsevier).