Why is my random-vibration (PSD) analysis failing on second PFACT?


You may be hitting defect 60605. MPC contact is not being treated correctly for the second and subsequent PFACT commands. The error has been fixed for 12.0.


Defect 60605 Description:

This problem was crashing in 11.0 because the first PFACT command builds the MPC contact and element matrices, assembles them, and then solves them using the Sparse solver. The second and third PFACT commands solve the same system of equations but with different boundary conditions. These different boundary conditions do not change the structure of the matrix solved with the 1st PFACT command, thus the second and later solves only need to perform a backsubstitution with the new load vector when using the Sparse solver.

However, the bug existed because the 2nd and 3rd solves were rebuilding the MPC contact. The MPC contact was changing between solves and this caused a crash inside the Sparse solver. The Sparse solver in 11.0 uses the FULL file header saved in memory for the first PFACT command. When the matrix structure does not change and we are doing a backsubstitution pass only, then we reuse the FULL file header from the first solution. In other words, we do not re-read the FULL file. So when the FULL file contents changed, because the MPC contact had changed which changed the structure of the matrix, then we hit the crash. In 12.0, we now read the FULL file every time to avoid this sort of incosistentcy since the Sparse solver should only run off the FULL file contents and NOT the ANSYS key values which can easily get all muddied up.

Also, at 12.0 the MPC contact will not be rebuilt.





Show Form
No comments yet. Be the first to add a comment!