Troubleshooting
- It stops with a floating exception error
- Time step is dropping rapidly
- Divergence
- Hourglass excitation
- Incompatible boundary condition
Looking at time history curves like kinetic energy, turbulent energy, rotational energy, hourglass energy, material variables, etc., try to identify the time when an unexpected behavior can be observed.
If Starter does not go through with a clean message, open the output file and look for the word "ERROR", which will define a more detailed explanation of the problem.
- Rerun a restart between your last restart file and "just before the divergence"
- Run a new restart until the divergence with the following options in the
Engine
file:
/PRINT/-1 /ANIM/DT Ti 1.E-30 (where, Ti is the initial time of this run + epsilon) /TFILE/ 1.E-30
You will then have a printout in the listing, an animation file and a time history sample for every cycle. Use them to investigate your problem. Very often looking at the velocity field in the animation shows some irregularity; which can easily be connected with a hole in the boundary conditions.
Divergence
Divergence means some variable is getting too large. This might occur if your problem is poorly defined (mathematically), or if the time step is bigger than the theoretical critical time step (numerical).
Try to locate the problem in time and space by running the time history and animation programs. Look for anomalous velocities, turbulent energy and viscosity.
- Kinematic constraints
- Elementary boundaries (material TYPE11)
Verify classical assumptions of fluid mechanics.
Look for leaks in your mesh (non-coincident nodes, non-plane symmetries, omitted boundary conditions). If the turbulence model is being used, verify you have at least one turbulent wall.
Look for hourglass velocity patterns. This might occur if you have concentrated fluxes, elements with bad aspect ratios, or if the hourglass coefficient is set to a very small value (see material parameters).
Check to make sure you do not have a shell thicker than the element size: flexural stability is a priority assumed in the code (not calculated) and this assumption is conservative only as long as the thickness does not exceed the length of the shell.