Convergence in each time step means you are able to reduce the residuals of the equations up to get theoretically vanishing. If you fails the convergence in a single time-step you are providing a different and numerically error-affected initial conditions for the next time step. If your problem admits a steady state, the final convergence towards the steady state should not be affected. Conversely, if you are solving a physical unsteady problem that generates an error that eventually propagates in a numerical but not physical transient.

I would suggest to use under relaxation factor to slow down convergence in particular for unsteady case and try to use medium mesh.

Thank you.

I endorse answer of sir Ahmed Abdelhameed

Hello Soumitra,

While I agree with Ahmed Abdelhameed, I suggest use a finer mesh near to the surfaces and critical edges.

Good luck

I just found out (on simplifying the case, i.e. skipping condensation) that, the incompressible phase which is liquid water when used alone (single phase), it is not converging and on top of that causing backflow at the outlet.

on the other hand, compressible (-vapor) phase, when used alone, it converges easily without backflow.

Hence , maybe it was obvious for my complex condensation (multiphase-phase change) simulation to have convergence and backflow problems due to presence of the incompressible (-liquid) phase at the outlet.

So, in order to tackle the problem, maybe I should first find out that why the simplified case with incompressible (liquid) phase alone is not converging.

Do you think now it is because of the mesh?

P.S. The geometry is a simple C-shape tube with bend radius at corners. And the geometry is meshed uniformly with very good quality hex elements. There is const. mass flow rate inlet at specified pressure and temperature. Outlet pressure and tube fluid temperature (along length) is known from the experiment.