hi all.

I am using a 3D SIMPLE code to simulate a laminar natural convection. I found for some time steps, it took above 20,000 iterations to achieve a converge. Is it a little strange ? What is the normal iteration numbers to achieve a convergence using SIMPLE?

My case is:

Natural convection combined with both heat and mass transfer. 3D Mesh: 10*10*30. Tolerance: Max((fi(iter1)-fi(iter2))/fi(iter1))<10^-6

Thank you.

Dear Tom:

The problem of natural convection is somewhat difficult for one in achieving convergence, although the problem may look extremely simple!! The convergence, I would say, it depends on a lot of factors. The most notable one is the Rayleigh number; if Rayleigh number is high, it is difficult to achieve a fully-converged steady solution. In that case, we have to resort to the transient approach.

It seems to me that your error tolerance has been dropped to a considerably low value. What is your iterative method for the matrix-solver? It will play a significant impact on the rate of convergence. Check your Reynolds number as well. Most probably the flow is already turbulent, and hence the laminar flow model is not appropriate. You will encounter convergence problem if you force yourself using the laminar flow model in computing turbulent flow.

FYI, in a case of laminar natural convection Rayleigh number=1e6, I need only 300-400 iterations to achieve convergence. SIMPLE is used. The mesh is 30x30x10.

-khai ching-