CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   convergence in tetramesh (http://www.cfd-online.com/Forums/main/81472-convergence-tetramesh.html)

 khaiching October 27, 2010 21:30

convergence in tetramesh

Hi colleagues:

I would like to gather comments from you all regarding the convergence performance using tetrahedral mesh.

I build up my 3D tetrahedral mesh using Delaunay method and find that the convergence is kinda irregular. For a simple incompressible flow condition in a flow distributor at a certain tetrahedral mesh setup, I could not achieve convergence at all. The pressure correction residual error is getting unstable ( residual error goes down to a certain level and it spikes up again).

Subsequently, I extract the cells (problematic cell) that contain the largest residual error and try to remesh the region of those problematic cells. In this case, the convergence can be attained.

Dear colleagues, I would like to know whether you have come across this phenomenon before? Something come across my mind: for the matrix solver of pressure correction equation, I utilise the Bicgstab method with preconditioner. Could this problem due to the characteristic of the Bicgstab method itself, whereby the error cannot be pushed out from the domain as efficiently as the AMG method, particularly in the unstructured tetrahedral environment? I have not tried the AMG yet.

Or, could it be due to any other reasons, which is unknown to me?

Thanks.

 eplanting October 27, 2010 23:01

Variable dt or fixed dt? Also, are you using upwinding or midpoint, midpoint is a bit more accurate.

 khaiching October 28, 2010 01:57

Hi:

I am using steady-state algorithm to solve the equations. SIMPLE algorithm in fact.

I am using 1st order upwind scheme for all the convective terms. Therefore, I think the instability occured is not due to my convection scheme.

 eplanting October 28, 2010 05:10

ick, while upwinding is powerful, it tends to give more pretty pictures than accurate ones. Anyways, I still think it's probably something to do with a value of dt occuring that's too high for the solution, that's a pretty common problem around the graduate level at least, one gradient is pretty steep, and it looks like the rest of the solution is converging, but one area is developing an error, anyway, I'd recheck ur code for the calculation of dt, or you could post it.

 All times are GMT -4. The time now is 11:02.