Free Courses Fluids Structures 3D Design Electronics Materials Photonics Student Teams STEM Python Missions Indo Kursus My Courses

[FAQ]

This behavior is most usually due to a very ill-conditioned system. Try changing the solver from Block Lanczos (direct) to subspace. Subspace better handles indefinite matrices than the Block Lanczos (BL) method. This can be done in Analysis Settings -> Solver Type -> Subspace. Usually what is happening is that there is a disagreement between the number of modes to be found between what the BL found and what the Sturm sequence check finds. So a new BL cycle is done with a new shift, but then the same disagreement is found but in the opposite direction. the subspace does not perform a Sturm check by default, so it can bypass the issue altogether and ‘fix’ what is causing the ill conditioned model. But at the expensive of a possible less accurate model. There is also an option to turn off the Sturm checking for the BL method via the new LANBOPT command.