CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   Natural Convection - Problem in achieving converged solution (https://www.cfd-online.com/Forums/fluent/122551-natural-convection-problem-achieving-converged-solution.html)

 howhs August 21, 2013 23:05

Natural Convection - Problem in achieving converged solution

Hi People,

I have some enquries regarding natural convection simulation for my final year project. Basically, I am now doing a validation study by making comparison with the experimental study using my computational study. This experimental study will have a simple enclosed cylindrical duct with a temperature difference at both sides of the duct. A high temperature and a low temperature will be fixed at both ends of the ducts. The circumferential face will be assigned to be adiabatic (assuming face is well insulated). This will induce a natural convection within the duct.

I am using ANSYS Workbench to generate my mesh only for the fluid domain. I am using 'inflation' method to create the boundary layer mesh. Thus, I am using the y+ value in the order of 1 to calculate the y value as I will be using the first layer thickness in the ANSYS Mesh with 20 layers and the growth rate of 1.2. For the rest of the mesh, I am using the tetra mesh with the element size of 10mm.

As for the boundary condition, I am setting a fixed temperature at both ends (82 and 14 degree celsius) of the duct and the circumferential face as adiabatic. Fluid used will be distilled water with the operating temperature of 25 degree celsius. Since the temperature difference at both ends is not large, I am applying Bossinesq theory to calculate the buoyancy effect, thus the operating density need not be stated.

SIMPLE Scheme and First order Scheme are used initially achieve convergence first.

Since the boundary condition will lead to turbulent natural convection, it is adviced from the following link to reduce the order of Ra number by 2 first.
http://aerojet.engr.ucdavis.edu/flue...ug/node572.htm
I did not apply turbulence model for the first run since the Ra number is in the order of magnitude of 10e7. I started with this but I could not get the a converge solution especially for continuity. I have tried to toggle around with the relaxation factors by reducing the body force and momentum and when it gets converged, i slowly increase the relaxation value again. I still have difficulty getting the solution converged.

Do you have any advice for this?