[turbulencious]

dear all,

I hope everybody is doing good

I came across something quite weird and I would like to share it with you in order to listen to the experience and suggestions of other people.

I simulate a laminar flow for a quite simple (though large) geometry, I use chtMultiRegionFoam of the latest version OF-2.1.1. It is a transient conjugate heat transfer problem.

Everything is good but I would like to speed up the whole simulation because I need to simulate a lot of hours and many times. I use adjustable timestep and maximum Co 5 (have tried 10 and 15) because higher values lead to divergence (weird, considering that Crank Nicholson should work for any timestep). Therefore I am limited to a very small timestep which will take me ages to get my results. So I have decided to change the convergence criteria in order to have lower computational time per time step (very desperate?). Anyway when I reduce the convergence criteria by one order of magnitude (1e-07 to 1e-06) this influences the deltaT, which drops from 4-5seconds to 0.2 seconds.

So the question is: What is the relation between the convergence criteria and the delta? Are there any ideas in order to speed up my simulations?

Thanks a lot in advance!

cordially,
Giorgos

[olivierG]

hello,

You can't get faster since chtMultiRegionFoam is a weakly coupled solver, and trying to solve with higer tolerance residual will not help you.
You may take a look at conjugateHeatFoam (1.6-ext).

regards,
olivier

[turbulencious]

thank you Olivier for your response!

what do you mean by take a look at conjugateHeatFoam. It is not available in the version that I am using. Can I still use it? And if yes, how can I use larrger time step?

thanks a lot