Hai all,

I am doing a completely Solid CHT problem. My geometry consists of different domains with a size of 800,000 elements(hexa and tetra). The hexa and tetra interfaces are linked using GGI method. After a transient simulation of 300 seconds with 0.1 second timestep, I get the results with the following Domain imbalances for all the five domains.

Domain 1 4.8%

Domain 2 -0.1%

Domain 3 0%

Domain 4 1.7%

Domain 5 3.3%

The residul convergence was set to Max 1e05.

Does this amount of Domain imbalances affect my results?

sounds like your case isn't converged, and this will certainly reduce the accuracy of your results. your imblances should be less than 1%, ideally less than 0.01%. Try using a larger timestep for the solid domains if you are running steady state. if you are running transient, keep running the case.

Hi,

I agree with Johnny, it looks like you have not fully converged yet. I recommend turning on convergence testing by imbalances for CHT simulations as the residual converges much faster than the imbalance for this case.

Glenn Horrocks

Johnny and Glenn, Thank for your replies.

I am presently running the case with conservation target of 0.01 .

Johnny,

You said that "if you are running transient, keep running the case".

Could you please explain the concept behind it? My interest of this simulation is to find the Temperature gradient of the various domains in my problem only until 300 seconds. After 300 seconds, the domains have imbalance. How does it change the solution if I continue running after 300 seconds?

Thank you.

I have the results of the solution after using conversation target of 0.01 . I haven't changed the timesteps. The results are better than the previous one.

Domain1 -0.8%

Domain2 -0.02%

Domain3 0%

Domain4 -0.5%

Domain5 2.05%

What is the value that the Domain imbalance can be, for a transient solution?

Imbalances are still high for Domain 5 in my opinion. If you are running transient, you may need to run more coefficient loops if you need a time accurate solution. Running it for more time will simply get an accurate steady state solution at the end of the run (which may take more than 300 [s] if you don't increase the number of coefficient loops). I guess my initial point was that you cannot run with different timesteps for the solid and fluid domains if you are running a transient solution.