CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Residual Problem in Coupled-Implicit Solver (http://www.cfd-online.com/Forums/fluent/27686-residual-problem-coupled-implicit-solver.html)

Prateep Chatterjee May 23, 2000 09:41

Residual Problem in Coupled-Implicit Solver
 
I am running a case with approximately 80000 grid points. There are both triangular as well as quad. elements. The triangular ones are confined to the region where a laminar flame is expected to stabilize. I possibly can not coarse the grid even more. While running a cold flow with Ch4 and Air I had no problem in getting a converged solution. After I patched a 3000 K temperature very near the expected flame region, the residuals have refused to come down. The solver is coupled-implicit and I have kept the Courant No. down to 0.001 at the moment. The residuals are of the order 10 to 100.

Any suggestions, please

Thanks

Prateep

wason May 23, 2000 22:47

Re: Residual Problem in Coupled-Implicit Solver
 
Dear Prateep,

I don't know exactly your case,but I think you can try to use the segregated solver.

Best regards Wason

Detlef Schulze May 26, 2000 04:15

Re: Residual Problem in Coupled-Implicit Solver
 
Dear Prateep, there is a section on Convergence Criterions within the Fluent V5 Manual. In Fluent you can use different criterions for the convergence (scaled, normalized etc.). The that is best suited for your problem depends also on you initial solution.

Good luck. Detlef

Prateep Chatterjee May 26, 2000 13:38

Re: Residual Problem in Coupled-Implicit Solver
 
Hello Detlef !

I understand what you are saying. But residuals of the order of 100 & 1000 ? instead of 0.001 etc. ?

I have changed the solver to segregated. Of course, I got a the famous Segmentation Fault (which has been resolved in the upcoming ver 5.4).

Thanks for the suggestion

Prateep

Volker Pawlik May 30, 2000 06:04

Re: Residual Problem in Coupled-Implicit Solver
 
I had and still have similar problems with the modelling of a laminar flame and yet did not succeed in getting a reasonable result. The residuals in my case are much higher than yours...

The guys from Fluent USA say it should be possible to get a solution but ONLY with the coupled solver. See the discussion on Methane/Air combustion some time ago.

P.S. Some people did a laminar flame 1-step calculation with CFX without any big problems as they said. They admited that the finite rate model (arrhenius) including the Definition of the Diffusion coeff. they used had to be programmed by UDS (or similar). I still do research work on how they managed to get the way up...

Prateep Chatterjee June 1, 2000 16:14

Re: Residual Problem in Coupled-Implicit Solver
 
Hello Volker !

Thanks for the response. Also, thanks for your earlier response, I just forgot to write back then !

I have been successful in simulating a laminar flame inside a part of a Rijke tube. But, only with the segregated solver. With the coupled solver, the residual problem remained. Did the FLUENT, USA people really said that only the coupled solver works ? I haven't been able to make the coupled solver work.

Now, I'm doing the full tube simulation to get thermoacoustical data and I did make the grid REAL coarse. Unfortunately, again the solver betrayed me. I've gone back to using the segregated solver. The under-relaxation factor for energy has to be controlled properly. I start off with 0.01 (1% update) and then generally go up to 0.5 and after I see the temperature profile developing, I crank it up to 0.9.

Write back if you have any comments.

Prateep

Volker Pawlik June 2, 2000 03:03

Re: Residual Problem in Coupled-Implicit Solver
 
Concerning the answer from Fluent USA have a look to the contribution of Graham Goldin on my question: Methane/air combustion earlier in this forum (you can find it in the archive)

Volker


All times are GMT -4. The time now is 05:39.