CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Problems with MFIX code and Parallel Processing. (http://www.cfd-online.com/Forums/main/12065-problems-mfix-code-parallel-processing.html)

Fernando Pio August 23, 2006 13:47

Problems with MFIX code and Parallel Processing.
 
Hello,

Im use MFIX code (Multiphase Flows with Interphase eXchanges) for to simulate a gas-solid flows in a circulating fluidized bed. The main objective is to investigate the numeric diffusion. Thus, the sufficiently refined mesh is very important.

Then, parallel processing with two machines (PCs) was prepared (cluster), but the problems are not converging. The convergence is guaranteed if just one machine is used.

Could somebody tell me why the problems are not converging using the MFIX code with parallel processing (of PCs)?

The operational system is a Mandrake Linux 10.1 and the Intel Fortran Compiler (Fortran 90).

Thanks.

Thanks.


agg August 23, 2006 16:10

Re: Problems with MFIX code and Parallel Processin
 
Are you specifying gas as incompressible fluid in MFIX? If so, try to use it as compressible fluid (specify average molecular wt. and gas temperature).

Fernando Pio August 24, 2006 10:44

Re: Problems with MFIX code and Parallel Processin
 
Hello,

Yes, the gas is incompressible fluid and it can not be changed in MFIX.


agg August 24, 2006 11:23

Re: Problems with MFIX code and Parallel Processin
 
If you specified RO_g, the gas is treated as incompressible. For compressible flow, you should specify Mw_avg (average molecule weight), P_g and T_g in initial and boundary conditions.

Fernando Pio August 29, 2006 14:33

Re: Problems with MFIX code and Parallel Processin
 
Hi, thanks for your message.

I need to treat the gas as incompressible flow. I think that the divergence was caused by the parallel computing. In my opinion, something is happening with the solution when the problem is divided by the process in the cluster.

The installation this correct one, therefore has convergence when i consider only one machine. According to your experience, it will be that the problem can be in the division of the problem in the cluster? What it could be occurring?

Thanks for all.



All times are GMT -4. The time now is 11:14.