Sunteți pe pagina 1din 3

CV7803 Numerical Methods

Assignment 2
Low Chee Meng
G1403068J
March 27, 2015

Task Desription and Objective

Given a square domain of 0.3m by 0.3m, as shown in Figure 1, the objective of the assignment is to find
the temperature distribution in the domain. Also, a mesh convergence check is to be done by comparing
the calculated temperature at 9 nodes, whose positions and labels are shown in Figure 2.

Figure 1: Domain and boundary conditions for 2D heat conduction problem.

Figure 2: Nodal positions used for mesh convergence check.


The governing equation for the problem is the steady 2D heat conduction equation without sources.
(kT ) =

2T
2T
+
=0
2
x
y 2
1

(1)

Solution Description

The boundary conditions are shown in Figure 1. On the left boundary, Neumann boundary conditions
was imposed with heat flux on the left face being 500 kW/m2 , and no heat flux on the right and top
faces. On the bottom face, the Dirichlet boundary condition was used and the temperature is set to 100

C.
The governing equations and boundary conditions were implemented in a Python script. The grid
consists of uniformly sized cells and the temperature distribution results were obtained for seven different
mesh densities. The number of cells in each mesh can be represented as 9n , where n is increased from 1
to 7. Hence the coarsest and finest meshes consist of 9 and approximately 4.8 million cells respectively.
To verify convergence of the solution and mesh independence, the temperature at the original 9 nodes
of the coarsest mesh is compared with the results given by the finer meshes. The mean and maximum
temperature change, normalized against the mean of the cell temperature values corresponding to the
meshes before and after each refinement, is shown in Table 1.
Cell count increase
Mean T /Tmean
Max T /Tmean

91 to 92
1.64E-02
2.99E-02

92 to 93
1.87E-03
4.02E-03

93 to 94
2.32E-04
5.15E-04

94 to 95
2.58E-06
5.75E-06

95 to 96
2.58E-05
5.75E-05

96 to 97
2.61E-05
5.81E-05

Table 1: Normalized temperature change for each increase in mesh density.


Based on this measure of convergence, the solution converges as the mesh 95 -cell mesh is used. As the
mesh is refined further, the change in the solution is on the order of 1e-5, as shown in Figure 3 The
solution converges monotonically. For many heat transfer problems, the accuracy is acceptable.
The temperature distribution plots from the increasingly dense meshes are shown in Figure 4. As the
mesh is refined from the 9-cell to the 95 -cell grid, the temperature gradients become sharper and better
captured. Mesh refinement beyond 95 -cells appear to have negligible impact on the resolving of the
temperature gradients. The converged temperature distribution at the 9 original nodes for the coarsest
and converged solution mesh is as shown in Table 2.

Figure 3: Normalized temperature change following each mesh refinement step.

Node
1
2
3
4
5

91 cells
197.39
177.47
133.27
166.71
150.33

95 cells
199.95
180.35
137.93
169.55
153.18

Node
6
7
8
9
-

91 cells
119.13
152.57
138.87
114.38
-

95 cells
121.36
155.51
141.46
115.71
-

Table 2: Converged temperature distribution.

(a) 3x3 grid / 9 cells

(d) 81x81 grid / 6561 cells

(b) 9x9 grid / 81 cells

(e) 243x243 grid / 59049 cells

(c) 27x27 grid / 729 cells

(f) 729x729 grid / 531441 cells

(g) 2187x2187 grid / 4782969 cells

Figure 4: Temperature distributions with varying mesh densities

S-ar putea să vă placă și