Question on adjusting relaxation factor
 

hi everyone, I have two questions about adjusting the relaxation factor (on a steady state problem). 1) Is it true that for a given mesh, let' say for the first few hundreds iterations or so, if using the code default relaxation, I am seeing oscilation on the residuals and they no longer reduces further, then reducing the relaxation factor will get rid of the oscilation and bring the residual lower?

2) Let say that from the very beginning (1st iteration), I have over reduced the relaxation factor, and now it takes "forever" to converge. Can I, for example after few hundreds iterations, increase the relaxation factor to speed up the convergence? But by doing so, would I also get into the above situation and have to reduce the relaxation once again another few hundreds iterations later on?

Thanks in advance!!

Re: Question on adjusting relaxation factor
 

No straightforward answer except to say "try it and see" because "it depends."

No, I'm not trying to be a smart-***, that's the straight answer.

Good Luck!

Re: Question on adjusting relaxation factor
 

Hi

Jim's right its not that straight forward. In my experience if you have to play around with the relaxation factors too much or make them very small, your convergence problems lie elsewhere i.e. poor grid quality, badly possed boundary conditions, differencing scheme issues to mention but a few.

Have a play and see what happens.

James

Re: Question on adjusting relaxation factor
 

Thanks for the feedback!