Learning Forum

[nobrien]

Hello! I'm using pseudo-transient to help dampen the solution's oscillations for my steady-state simulation, but I'm not sure if my approach is correct, mainly with how to calculate the pseudo time step.

I know Fluent's defaults seem to do something like one-third of the flow-through time (length of domain divided by representative velocity), but is there a benefit to reducing the time-step further? The flow-through time for my model is about 1 second, so I tried 0.1 to start but didn't see much reduction in the solution's oscillations.

When I went as far down as 1e-5 seconds, I satrted to see the residuals go down to a much lower value and the oscillations seem to die out, but this seems to be extreme compared to what I've read in the fluent user guide. They suggest values within an order of magnitude of the flow-through time, but the pseudo-time-step I applied is much much smaller. 

For reference, I'm performing about 13,000 iterations, which also seems pretty extreme from what I would expect. The flow is an internal, incompressible flow and is using the k-omega sst turbulence model. 

Any guidance?

[Rob]

Reducing the pseudo time scale essentially reduces a time effect in the solver maths. So, a small value will be more stable but then requires more iterations to reach the equilibrium solution. If the value is too small, you're going to use a lot of cpu. 

However.... If the flow oscillations are real, how good is a solution where you've damped them out? This is why we use point monitors in areas of interest and the time/iteration averaging to see a "stable" solution where there isn't (quite) one.