Hi,

I have a problem with timestep control in user-fortran Subroutines! My task is to write a subroutine for the fluid-structure-interaction (FSI) without using the given MPCCI code.

Now I planned to write a routine, which is called at the end of each timestep and check the convergence of the calculated solution. If the convergence is OK, the simulation keeps on going and the next timestep is calculated.

But if the convergence is bad, the solver should jump to the end of the timestep before and repeat calculating the actual timestep.

Therefore I will have to manipulate the solver-stack and I need more informations about the solver data-structure!

So here are my questions to specialists in CFX:

1. Where can I find detailed informations about the solver data-structure?

2. Which data has to be manipulated, to set the solver to the end of a before calculated timestep? (f.e. copy /FLOW/SOLUTION/LATEST to /FLOW/SOLUTION/TSTEPn/CLOOPm, or setting ATSTEP to a special value)

Thanks for any reply,

Gabor

The only MMS information available is in the CFX Solver Modeling manual under User Fortran.

I wouldn't hold my breath about backing up a time step without a LOT of help from ANSYS. There is an adaptive time stepping example in the above mentioned manual, but it only cuts the time step if the current step uses more than a recommended number of coefficient loops, then proceeds on to the next step. This has worked pretty well for me without needing to go back and re-run a timestep. I'd try this first.

I think this is also implemented in CFX-10 without using the additional Fortran.

Jeff