Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

FRF using Modal Summation with parametric sweep, can Eigensolutions be stored and reused to avoid repetition of Eigenanalysis?

Please login with a confirmed email address before reporting spam

Hi,

I am rather new to Comsol and have a question. Currently I am studying stiffened plates with the shell physics of the structural mechanics module and perform a modal analysis and calculation of frequency response functions, using the "Frequency Domain, Modal" study in Comsol 5.4. I have added a point grid on my plate to sample the frequency response functions at discrete points with respect to the structural wavelength, which I also need for some more post-processing, later on.

I am performing a parametric sweep to change the forcing position on the plate (which is linked to a coordinate on the plate). I am wondering whether it is possible to do the Eigenanalysis (which in my case takes quite a while since I include all modes up to 10 kHz) only once and store the solutions (eigenvalues and eigenvectors) to then perform the modal summation for all forcing points via sweep. Currently for each defined forcing point, when using the parametric sweep, my model is doing the Eigenanalysis again. My geometry and mesh do basically not change, but the geometry needs to be recalculated to correctly assign the new forcing position. I have seen there is a option in the "Frequency Domain" study node called "reuse solution from previous step" but when I activate it, the calculation time does not go down and according to the progress log I am following, it still repeats the Eigenanalysis. Maybe my way is not the most straight-forward way and the problem lies within the way I designed my model. But, if there is any solution to calculate the Eigenanalysis once and reuse it and someone has an idea how to, I would be very happy to receive some help. Many thanks in advance.

Kind regards, Christopher

PS: In case you need more detailled information, please let me know and I will add them in a reply.


3 Replies Last Post Jul 4, 2020, 10:12 a.m. EDT
Henrik Sönnerlind COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 4 years ago Jun 16, 2020, 4:03 p.m. EDT

As long as you are using the same geometry and mesh, it is easy to reuse the eigenmodes:

Rather than using two study steps within the same study, you should use two different studies, one for the eigenfrequency analysis, and one where you do the parametric sweep for the mode superposition.

The only thing to look out for is that when you use two different studies, you must make the connection from the mode superposition to the eigenfrequency solution manually. That is, you must point to the relevant eigenmodes in the Eigenpair section in the Modal Solver node.

If, however, you are moving the forcing point around, you will get a new geometry and a new mesh, so that you cannot reuse the eigenmodes.

There are different ways to deal with this. The easiest is to create all loading points once so that the geometry does not change. Then, you just load different points in the parametric sweep. Basically, there is one Point Load with all points as selection, but the value is controlled by the parameter.

As an alternative, you can use a load which is not bound to the mesh. This requires some more advanced modeling. This has been discussed in for example

https://www.comsol.com/forum/thread/179182/adding-a-point-load-without-a-geometrical-point
https://www.comsol.com/forum/thread/260403/how-to-simulate-a-point-wise-concentrate-load-traveling-along-a-structure

-------------------
Henrik Sönnerlind
COMSOL
As long as you are using the same geometry and mesh, it is easy to reuse the eigenmodes: Rather than using two study steps within the same study, you should use two different studies, one for the eigenfrequency analysis, and one where you do the parametric sweep for the mode superposition. The only thing to look out for is that when you use two different studies, you must make the connection from the mode superposition to the eigenfrequency solution manually. That is, you must point to the relevant eigenmodes in the **Eigenpair** section in the **Modal Solver** node. If, however, you are moving the forcing point around, you will get a new geometry and a new mesh, so that you cannot reuse the eigenmodes. There are different ways to deal with this. The easiest is to create all loading points once so that the geometry does not change. Then, you just load different points in the parametric sweep. Basically, there is one **Point Load** with all points as selection, but the value is controlled by the parameter. As an alternative, you can use a load which is not bound to the mesh. This requires some more advanced modeling. This has been discussed in for example

Please login with a confirmed email address before reporting spam

Posted: 4 years ago Jul 4, 2020, 9:29 a.m. EDT

Dear Henrik, Im also new to Comsol and have similar interests as Christoffer Knuthe. As I work with automotive and rail vehicle applications I have large structures say a truck floor or a battery bpack for an EV, for which I want to calculate transmission loss TL. Excitation can be a diffuse field, a point force or a fluctuating pressure field (TBL or CFD calculated). Frequency range 50- 3 kHz with, say, a few hundred frequency lines are typically needed. Typically, in my applications, one wants to know the improvement in TL from addition of foams and structural damping layers for such structures.

I have tried to run the benchmarks examples you provide regarding standardised (more or less) TL set ups for a doubleglazed window and a concrete wall. But the solve times are very long. My main question is if there are some guidelines for calculation approaches within Comsol for such problems to manage memory and CPU time in a clever way: when to use modal approach, when direct approach, when how many angles of incidence are needed to represent a diffuse field etc. Also, Im curious if there are benefits/limitations using the EFEM appoach for such problems. A few years back I analysed a train floor with an Energy Flow Method with reasonable results.

I should probably mention that my background is 15 years with industrial acoustic predictions using dedicated vibro-acoustic tools from ESI, LMS, MSC (FEM/BEM/SEA) as well as TMM methods from e.g. AlphaCell. I started looking into Comsol for an electic motor application but got curious of the capabilities alsofor transmission problems.

I know this may be a challenging and rather complex questions but I was just hoping that you have some documents you can recommend. with kind regards Ulf Orrenius

Dear Henrik, Im also new to Comsol and have similar interests as Christoffer Knuthe. As I work with automotive and rail vehicle applications I have large structures say a truck floor or a battery bpack for an EV, for which I want to calculate transmission loss TL. Excitation can be a diffuse field, a point force or a fluctuating pressure field (TBL or CFD calculated). Frequency range 50- 3 kHz with, say, a few hundred frequency lines are typically needed. Typically, in my applications, one wants to know the improvement in TL from addition of foams and structural damping layers for such structures. I have tried to run the benchmarks examples you provide regarding standardised (more or less) TL set ups for a doubleglazed window and a concrete wall. But the solve times are very long. My main question is if there are some guidelines for calculation approaches within Comsol for such problems to manage memory and CPU time in a clever way: when to use modal approach, when direct approach, when how many angles of incidence are needed to represent a diffuse field etc. Also, Im curious if there are benefits/limitations using the EFEM appoach for such problems. A few years back I analysed a train floor with an Energy Flow Method with reasonable results. I should probably mention that my background is 15 years with industrial acoustic predictions using dedicated vibro-acoustic tools from ESI, LMS, MSC (FEM/BEM/SEA) as well as TMM methods from e.g. AlphaCell. I started looking into Comsol for an electic motor application but got curious of the capabilities alsofor transmission problems. I know this may be a challenging and rather complex questions but I was just hoping that you have some documents you can recommend. with kind regards Ulf Orrenius

Please login with a confirmed email address before reporting spam

Posted: 4 years ago Jul 4, 2020, 10:12 a.m. EDT

Hello Henrik,

thank you very much for your response and your helping comments. It sounds very reasonable. I was a few weeks off from working with COMSOL, so apologies for the rather late response from my side. I will soon have another go with it.

Kind regards, Christopher

Hello Henrik, thank you very much for your response and your helping comments. It sounds very reasonable. I was a few weeks off from working with COMSOL, so apologies for the rather late response from my side. I will soon have another go with it. Kind regards, Christopher

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.