[Hadi Yavari]

I am performing CHT simulation on a gas turbine blade. The mesh includes about 40M polyhedral cells. The flow is fully turbulent (2 equations RANS) and compressible .

1- when I initialize the solution and run it immediately, the max RAM usage reaches 12GB / 1M cells.

2- If I initialize the solution, save and close the case and open it again, the max RAM usage reaches 5GB / 1M cells in each iteration.

is something wrong?

[LuckyTran]

There's a ton of temporary variables that get created when you run if that's what you're wondering. All the gradients need to be calculated, and then limited gradients, and then limited fluxes calculated from limited gradients.

[Hadi Yavari]

In both approaches the solution is initialized and it evolves in a same way through each iteration.But the RAM requirement of the 1st approach is more than twice. This can be problematic in large cases (500G vs 200G for my case)

Is there anyway no to close/reopen the case and still avoid excessive RAM requirement?

[LuckyTran]

Well, what kind of unconventional initialization are you doing for the RAM requirements to be that high? And does it involve mapping data from a different mesh using a stencil?

[Hadi Yavari]

I use a table of velocity components, static pressure, static temperature and viscosity ratio for initialization. I extracted the table from a converged solution on a quite similar geometry.