Incorrect results of modal analysis with PaStiX


on the PrePoMax forum, a user reported issues with modal analysis. It turned out that they occur only when the default solver (PaStiX) is used. With Spooles the results are correct while with PaStiX proper torsional modes are missing and there are some weird unsymmetric modes instead.

Here’s the input file: Dropbox - modal.inp - Simplify your life

Just add SOLVER=SPOOLES to *FREQUENCY to see how it works with Spooles. I checked it with the current version of CalculiX (2.20).

I wonder if we should notify Guido or someone else responsible for the PaStiX solver.

1 Like

maybe this is special cases for unrestrained BC models, did not happen when one of end faces are fixed.

fortunately, on Windows OS and frequency analysis the solver automatically switch to Pardiso when the dynamic libraries are provide.

I’ve had similar issues with free-free modal extractions where the results are not correct. In some cases, even the fixed modal extraction gets non-physical mode shapes. However, even switching to spooles does not solve the problem. I believe it must be something else.

this may need to concern for all user, i’ll further study with another software at these specific domains (OpenSees) by converting the mesh initially.

can you share the model to downloads?

Some time ago I had issues with free-free modal analysis but of a single element and Guido said that he fixed it (we’ll see in the next release): Free-free modal analysis on CPS4 element not working

Apparently, there are more problems with this kind of analysis.

… only a problem for small models (e.g one element)

as notify i did testing for CPS4/CPS3 element models with moderate mesh, shown no error message. also the result is similar when switching between Spooles and PaStiX solver.