Has anyone ever had this problem? I've been running a transient simulation for about 2 days using 4 processors on an AMD Opteron machine. I've have stopped the simulation a few times and restarted from the previous .res files without any problems. The simulation is managing to converge during each time step with only 1 inner iteration - which is great. However, if I restart the simulation (with interpolation turned off) using only three processors and hence three partitions the simulation now takes more than 20 iterations per time step to converge!! I have a feeling that you cannot re-start in CFX-11 using different numbers of partitions â€" possibly because some information is being lost in the re-start initialisation. Can anyone offer any advice with might solve this issue?

James

Dear James,

Have you stopped the 3 partitions simulation, and restart? Does it still takes ~20 inner iterations per time step?

You should be able to restart from N partitions to M partitions w/o that kind of trouble.. A minor jump in the residuals, or a minor increase in the number of iterations is possible. The linear solver iterative behavior is not identical between different number of partitions..

Without more information on your setup, it is nearly impossible to diagnose what could be the problem.. Are you doing transient-rotor stator?

Opaque

Opaque

My problem is a simple flow through a piping arrangement. The convergence with four partitions seems to be fine - (1 iteration for each time step). I have yet to run the simulation from the start again using three partitions, due to time constraints. It's pretty frustrating to have a four partition run converge quickly and then find that a restart using three partitions fails to converge as quickly! The behaviour seems to be identical to a restart using interpolation.

James