asdeasoft.net

DEAR STKO,

I wanted to investigate how the results of the modal analysis change when the analysis is carried out on the structure in the deformed state (after the gravitational analysis) and before (i.e. only the modal analysis itself).
Performing modal analysis alone is not a problem.

The problem I have is when the modal analysis is done after the gravity analysis.
I've read various such problems on the forum.
I have tried various methods.
I have used and not used the wipe command to check the results.
I now that you recomended witohut the wipe command.

I get different results depending on the choice:
A) constraints
b) system
c) numInc in the analyze

The differences are not too big but I don't know which are correct. What I don't understand most is why number of analysis steps to perform effects so much on the eigen results?

Using material and geometric nonlinearity, should I use indicated constraints and system settings or a specific combination of both??

Attached is my model.

A) constraints

Maybe, if you use Penalty, it imposes the constraints approximately, using very stiff elements.

b) system

Not directly, since the eigen command internally aways assemble the tangent stiffness (what the standard newton does).
But since you are doing a nonlinear analysis, for which you accept an approximate equilibrium (using the tolerance in the test command), different system may lead to different paths to the equilibrium, leading o different small errors.

c) numInc in the analyze

Same as before, the larger the time step, the more difficult is for the nonlinear integrator to reach convergence.
Moreover, algorithms such as the IMPL-EX in the ASDConcrete3D, perform some explicit extrapolation to improve convergence, so they require small time steps.

Finally, eigenvalue analysis with a tangent stiffness is always tricky, since the tangent matrix loses its physical meaning of stiffness (it contains just derivatives).
Especially when you have large nonlinearities, some eigenvalue can become negative (buckling or strain-softening).
Also, Damage Models like ASDConcrete3D or kinematics like Corotational lead to non-symmetric stiffness matrix (not good for eigenvalue analyses).