CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   help for the Fastran (http://www.cfd-online.com/Forums/main/81212-help-fastran.html)

june_zhazha October 20, 2010 01:51

help for the Fastran
 
I use CFD-Fastran to compute the lift force and drag force of the AGARD-B model at Ma=1.5, total temperature T0=288k, dynamic pressure Q=62Kpa conditions, at first it converges very well. But after perhaps hundreds or thousands steps(according to the CFL number), the residual errors suddenly increase in few steps, and then, in some steps ,the process was terminated for the divergence regardless of magnitude of the CFL number. I think it was irrelevant with the grid, because i used the same grid to compute by Fluent, it could converge.

Could someone give me some advices?

CapSizer October 20, 2010 02:14

It's been a while since I've worked with CFD-Fastran, but my experience is that it works well with grids that are very different to the ones that Fluent likes. Is this a structured or unstructured grid? Things may well have changed since I last used it, but the structured Fastran solver worked much better for me than the unstructured one.

june_zhazha October 20, 2010 06:38

Quote:

Originally Posted by CapSizer (Post 279915)
It's been a while since I've worked with CFD-Fastran, but my experience is that it works well with grids that are very different to the ones that Fluent likes. Is this a structured or unstructured grid? Things may well have changed since I last used it, but the structured Fastran solver worked much better for me than the unstructured one.


I use structured grid. Y+~40

CapSizer October 20, 2010 11:42

Try running it to full convergence with a first order differencing scheme, then go to a higher order scheme (Van Leer?).

june_zhazha October 21, 2010 05:04

Quote:

Originally Posted by CapSizer (Post 279999)
Try running it to full convergence with a first order differencing scheme, then go to a higher order scheme (Van Leer?).

I usually use fisrst order differencing scheme (Roe or Vanleer) to initialize the flow field, and then go to a higher order scheme. But finally the result is divergence. Does anyone meet this problem? And how is it resolved?

simulationman October 30, 2010 11:41

Same troubles
 
I am using ACE+ which is essentially from the same group. And I do face such problems.

Now ACE+ does provide an output file. Seeing that I found that the problem actually started a few steps before the actual increase of the residuals and consequent divergence. The problems were like limits of variables were reached or the maximum number of sweeps for a particular variables were reached.

So when that happens, I usually just increase the relaxation and it works out fine. Maybe the same solution applies for Fastran.


All times are GMT -4. The time now is 02:26.