Grid independence
 

Hi,

I've been trying to study how a heat source changes through a duct. My problem is that the solution seems to change as I refine the mesh (i.e. the value of the temperature rise changes with mesh).

It's an axisymmetric geometry. I first changed the grid sizing on the radial direction, which changed the solution. But then someone told me that since I'm using Spallart-Allamaras model without wall functions I have to ensure that my first grid node from the wall lies within y+<5. And this actually fixed the problem for the radial direction, i.e. when I increase the radial nodes it gives me the same temperature rise.

However, axially the solution still changes with the number of grids. I have tried a large number of nodes and it still seems to change with grid spacing. I know that my grid is very fine already. Is there anything else I'm missing?

Re: Grid independence
 

From what I have seen when dealing with heat transfer, typical values of y+ at the walls need to be much less than 1. It may be that you just need a finer grid than you currently have.

Re: Grid independence
 

James

If you're using a turbulence model without wall functions then you will nned about 20 cells across the boundary layer i.e. a very fine mesh indeed. I never had much sucess using the spallart-allamaras turbulence model with wall functions it always gave low prediction of heat transfer. I had more success with k-w sst model.

allan