[alexksei]

Hello everyone

It seems such type of question has been asked many times here, but I still can't find proper answer.

I have: fluid domain, solid domain inside fluid domain, the solid is hot, the fluid is cool. Natural convetion occurs. I run transient simulation, because steady-state solution for natural convection does not give proper residuals.
So, in transient simulations I have proper residuals: 10^-5 and 10^-6 for RMS-Mom and Heat Transfer. Imbalance for H-Energy is good (0%), but IMBALANCE IS LARGE up to 20% for T-Energy (heat conduction is solid domain).

For steady state they recommend increasing Solid Time Scale factor, but what should I do for transient?

Total time is 30s, time step is 0.1s

[ghorrocks]

You cannot use acceleration factors like solid time scale factor in transient runs. Your large imbalance means you are not converging well each time step, so I would recommend adding imbalances as a convergence criteria. This means you will do more iterations per time step.

[alexksei]

Thank you, Glenn, for your reply.

If you mean setting convergence criteria in Solver Control, then I am already using Conservation Target of 1%, but, as I wrote, imbalance for T-Energy is up to 20%.

After I'd posted question here I've found TimeScale Factor Control in Domain Control for solid domain. Making it 1000 gave me 4% imbalance for T-Energy (instead of 20%). Should I consider increasing this parameter or you think for transient it is not appropriate?

I actually don't understand why such TimeScale control can be appled for transient, as timestep is specified explicitly in Analisys Type.

[ghorrocks]

If you set a conservation target of 1% but it is only getting to 20% then you must have some other factor stopping it. Have you set a maximum number of iterations? You will need to increase it - but a preferable path forwards is likely to be to reduce the time step so it converges easier. Either way, you need a good imbalance convergence, and if you are not getting it you need to fix the simulation so you achieve it.

I am not sure what the time scale factor does in this case, so read the documentation and do some testing to confirm it does not affect results before using it for important simulations.

[alexksei]

Originally Posted by ghorrocks

If you set a conservation target of 1% but it is only getting to 20% then you must have some other factor stopping it. Have you set a maximum number of iterations? You will need to increase it - but a preferable path forwards is likely to be to reduce the time step so it converges easier. Either way, you need a good imbalance convergence, and if you are not getting it you need to fix the simulation so you achieve it.

I am not sure what the time scale factor does in this case, so read the documentation and do some testing to confirm it does not affect results before using it for important simulations.

I don't set maximum number of iterations. For transient analysis what I specify is the total simulation time and time step delta t.

Let's clarify one thing: for transient simulation the imbalance must be close to zero at every time step (as RMS should have converged values at each time step), is it correct? Or only the value of imbalance at the very last timestep shoud be considered?

As for the domain timescale factor for solid domain: I have found it in official CFX tutorial named "Time Transformation Method for a Transient Rotor-stator Case with Conjugate Heat Transfer", in which they consider rotor blades with conjugate heat transfer. This is what they wrote for blades:
"Increasing the timescale factor to 1.0e6 from the default 1 increases the physical time scale of this transient run inside the solid domain by a factor of 1.0e6"

I did not find anything else related to this in the documentation. Only for steady-state there is smth.

[Opaque]

If I am not mistaken, Glenn was referring to the maximum number of coefficient loops within a given timestep.

If you set a MaxCoefLoop to say 5, and your imbalance is set to 1%, the software will finish the timestep at 5 inner iterations regardless of the imbalance value at that iteration/loop. If the imbalance is critical for you, you must set a higher number for the maximum number of coefficient loops.

Just try some settings, and you can see how it behaves for a few timesteps. Once you understand how it works, and what you need of the solution, go ahead and do the full calculation.