[chriss85]

I'm currently experiencing a problem with sonicFoam in regards to relaxation.

I'm getting different results with sonicFoam when changing the relaxation factors of U/e. From my understanding the relaxation factors should only affect the stability and convergence speed of the iterative procedure during one time step.
I make sure that the result is fully converged by looking at the evolution of the residuals and maximum field values and then running some more iterations.

This result is mostly noticeable when using very high temperatures (my case: dense plasmas, T~50000 K). It is easily visible in the tutorial cases like forward step when the temperature is increased to 50000 K and different relaxation factors are tested. Convergence speed and stability behaves as expected but results are different by a small amount. I suspect this is a numerical problem as this error seems to get larger with increasing temperatures.

Can anyone shed some light on this? If I don't get unique solutions, which are most valid?

I have attached a test case based on sonicFoam. If you monitor the maximum field values for a few time steps you will see a difference when relaxation factors are modified (say 0.1 vs. 0.3, but also with higher values). Furthermore I attached some screens that show the numerical difference (on another mesh though). I'm currently testing if this error is adding up by calculating larger time spans, as the significance of this error is not known for the final result of a calculation.

[ziemowitzima]

Dear Chriss

I encountered different, but similar problem to yours. Could you look here:
http://www.cfd-online.com/Forums/ope...tml#post572034

best
ZMM