Is it possible to change the problem size during solution process?

[Ziyu]

Hi all,

I saw in the user's manual (P192 Section14. that "It is not permitted to change the size of the problem by adding new nodes or elements". But I am still wondering if anyone, by any chance, has some expericence about changing the problem size during the solution process. We are currently engaged in adaptive mesh refinement for large deformation problems using FEAP. It will be desirable if the mesh can be refined at each time step based on an error estimator (which we are already able to do by using our own estimator code and the mesh generator Tetgen), and then FEAP can keep loading to the next time step with the newly refined mesh (usually a mesh of larger size). It can be expected that if the mesh changes, numerous arrays in FEAP have to change correspondingly, which may require a lot of work...

Best,

Ziyu

[Prof. S. Govindjee]

Have a look at program/pnewprob.f .  This is the routine that is called when one is solving multiple problems in a single input file.  You could potentially duplicate this routine for in-core adjustment of your problem.  pnewprob  gets things going but you will also have to mimic the rest of the mesh input in core.  Or you could write a new input file to disk and use pnewprob as part of the effort to get it loaded etc.
-sg