[amir.a.aliabadi]

Hello Dear Colleagues,

I have successfully set up a case for simulating airflow over a heated cavity (urban canyon) using buoyantPimpleFoam. I have cyclic boundary conditions in the span-wise and stream-wise direction and I initiate flow by applying a pressure jump using fixedJump BC at inlet and outlet.

The simulation runs fine on single core but when I parallelize it using decomposeParDict and MPI after some iterations my Courant number increases and finally the solution diverges. I am using the "simple" method with 4 to 27 subdomains. I have tried various combinations such as (4 1 1), (1 4 1), (1 1 4), (2 2 1) etc. for 4 subdomains, ... and (3 3 3) for 27 subdomains. No matter how fine I make the time step, I do get the solution diverging after some iterations.

Is there any recommendation or references I can read to help me parallelize this solver effectively?

Thank You!
Amir