jdm aks eigenmode solver

I see in its manual, we should require more modes than needed to ensure accuracy, because solver may give poor accurary for higher order modes.

But one time, to confirm the convergency of results given by the solver, I increase the density of mesh step by step and inspect changes.
To some extent, I think the mesh is not very dense, just 900000+ cells, but even some lower modes have poor accuracy (marked with * by CST).

Anyone encountered such a situation? What's the reason?

Thank you for ur attention!

Hi,

which CST Eigenmode solver do you use? AKS or JD-Solver? If you are only interested in the first few modes ,I belive it is better to use the JD Solver.

F.

Yes, I used AKS method.

Then that's the shortcoming of AKS method?

I'll try JDM.

Thank u!

Hi

guess this no really a shortcoming. The JD Solver is just more robust as far as I know. The Playing around with the AKS Solver Parameters might also give you good results.

The JD Solver acaluates Mode after Mode. The calcualtion time goes up with the numbers of modes you need. THe AKS Solver calcualtes several modes is one shot. For a larger numbers of modes the AKS Solver is therefore much faster...

F.

What do you mean by "playing around with the AKS Solver Parameters"? Which parameters can be adjusted to obtain better results?

On the other hand, the mode number I want is not large, just 5 or 6.
But now CST consumes 1.5-2 G Ram, and at last abnormally terminated.
I don't know why, maybe terminated by OS?

I heared Windows can handle at most 2G Ram, then what OS should be used for more complex simulations and bigger Ram?

Thank u all!

Hi wandering heart,

in the solver special settings you find some parameters Like Estimation of the highgest eigenfrequncy e.g.. The default value is 0 if you do not not the higherst freq. Howver, if you alreday have a pretty good guess (maybe from some simulations before) you can enter this number there.

How large is your calculation in terms of mesh cells?)

Maybe you should send your structure to th CST hotline. DId you ever contact them?

F.

Yes, I also see the parameter Estimation of the highgest eigenfrequncy.
I just gave a better estimation, solver is running now.

My model has more than 900M mesh cells.

You mean I can get help from CST homepage?

Hi!
Either from the support section (you need a login for this) or with a e-mail to
info@cst.com.

F.

ok

thanks!