# Help needed with convergence

 Register Blogs Members List Search Today's Posts Mark Forums Read

 April 13, 2007, 20:26 Help needed with convergence #1 Shashank Guest   Posts: n/a Iam working on a incompressible viscous flow problem, and i have been using Mars along with simple numerical solution method. My convergence tolerance is 1e-4. My residual values come down upto 2e-3 and sit there and oscilate just a bit they neither go down nor they diverge, i have tried all the tricks playing around with convergence factors and tolerances and blending factors but my solution refuses to go below 1e-4. I have checked the mesh a 100 times and it looks quite good, iam just not able to understand where i need to look into in order to solve this problem. I would really appreciate any help in this regard

 April 14, 2007, 04:46 Re: Help needed with convergence #2 Ben Guest   Posts: n/a Are the boundary conditions well posed? Is the outlet too close to any features, do you have high gradients across it? Are there any undsteady features in the simulation?

 April 14, 2007, 18:29 Re: Help needed with convergence #3 Shashank Guest   Posts: n/a I Have used prescribed flow split BC's , I created extrusions at the inlets and outlets and they don't seem to have high gradients.

 April 15, 2007, 07:22 Re: Help needed with convergence #4 Ben Guest   Posts: n/a Try using pressure boundaries instead, they tend to be a little more stable, flow split can over-constrain the system

 April 15, 2007, 13:20 Re: Help needed with convergence #5 Shashank Guest   Posts: n/a The problem is that i have no idea about the pressures at the boundary, thats why i had to use the flow split boundary conditions, but thank you for the reply. What else do you think might be wrong

 April 15, 2007, 14:50 Re: Help needed with convergence #6 Balduin Bankerotti Guest   Posts: n/a It is very likely that your flow problem is not steady state and the oscillations are normal, especially if the mesh if fine enough. Run it transient.

 April 15, 2007, 15:27 Re: Help needed with convergence #7 Ben Guest   Posts: n/a Just put the pressure as 0, try it....

 April 17, 2007, 02:31 Re: Help needed with convergence #8 sidd Guest   Posts: n/a I think, if the residual is getting stablized ( neither converging nor diverging), then results can be post-processed. And very much this can be the issue of geometry variation or mesh quality for capturing gradients or the boundary condition specified as Ben said. You may have to do some tests with these features.

 April 17, 2007, 19:30 Re: Help needed with convergence #9 Shashank Guest   Posts: n/a Thank you very much guys for all these suggestions i will work on them and let you know how it goes.

 April 22, 2007, 17:37 Re: Help needed with convergence #10 star user Guest   Posts: n/a try UD initially and then change to MARS. it will help u.

 May 2, 2007, 05:14 Re: Help needed with convergence #11 Scibor Guest   Posts: n/a Ttry to restart the run with SMPISO solution procedure (With under-relaxation for press of 0.5) and choose AMG alghoritm, remember that this solution is really heavy for your processor, but I think it's quite sure you're getting results!