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.
Eigen Frequency simulation results does not compare well with frequency response smulation
Posted 19 mag 2016, 22:31 GMT-4 Version 5.0 5 Replies
Please login with a confirmed email address before reporting spam
Hi All,
I am facing a problem with my simulation and was wondering whether I am doing something wrong or this is normal. I came up with my design and carried out Eigen frequency response for the first 6 modes. The modes in the lateral directions (in-plane) are the 5th (7 kHz) and 6th (7.2 kHz) modes. The first 4 modes are out-of-plane modes.
I there after carried a frequency response for the lateral mode by applying a small sinus wave and the dominant mode was about 6.8 kHz. My guess was that the frequency response was going to be 7 kHz too as predicted by the Eigen frequency analysis but unfortunately this is not the case. Could I be doing something wrong? Or how can we explain this phenomena?
I use Comsol 5.0
Any suggestions or advise will be highly appreciated.
Kind Regards,
Ben
I am facing a problem with my simulation and was wondering whether I am doing something wrong or this is normal. I came up with my design and carried out Eigen frequency response for the first 6 modes. The modes in the lateral directions (in-plane) are the 5th (7 kHz) and 6th (7.2 kHz) modes. The first 4 modes are out-of-plane modes.
I there after carried a frequency response for the lateral mode by applying a small sinus wave and the dominant mode was about 6.8 kHz. My guess was that the frequency response was going to be 7 kHz too as predicted by the Eigen frequency analysis but unfortunately this is not the case. Could I be doing something wrong? Or how can we explain this phenomena?
I use Comsol 5.0
Any suggestions or advise will be highly appreciated.
Kind Regards,
Ben
5 Replies Last Post 10 giu 2016, 01:25 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi
it's difficult to give a clear reply, without a better view of your model and the settings you are using.
If you are concerned about the difference 6.8 versus 7 kHz, I find that a rather good match, 10% of difference is not necessarily that much for a coarse FEM model. It's all depending on the model settings, but you can certainly also get a better match.
A few things that might differ: an eigenfrequency analysis does not take into account any external forces loads, while the frequency response does, and a good frequency sweep implies to add some damping, and damping do change the frequency (it actually lowers it, how much depends on the damping value)
--
Good luck
Ivar
it's difficult to give a clear reply, without a better view of your model and the settings you are using.
If you are concerned about the difference 6.8 versus 7 kHz, I find that a rather good match, 10% of difference is not necessarily that much for a coarse FEM model. It's all depending on the model settings, but you can certainly also get a better match.
A few things that might differ: an eigenfrequency analysis does not take into account any external forces loads, while the frequency response does, and a good frequency sweep implies to add some damping, and damping do change the frequency (it actually lowers it, how much depends on the damping value)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Thank you Ivar for your prompt response and insight, It was of great help. But just a clarification from you. In the frequency response analysis, you mentioned about the damping. How exactly does one change damping (add damping) in comsol frequency response analysis? I have failed to do this I the past.
Kind Regards,
Ben
Kind Regards,
Ben
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi
Adding damping is an endless discussion, indeed. And far from obvious (well not to add damping, in COMSOL you simply add a "damping" sub node to the linear material or other relevant physics node, or even you add a complex spring constant, or you couple in any dissipative energy source ...)
The issue is which value to use, and for which parts/entities, is this an assembled structure, or a welded one, or a modern Additive Manufactured single piece item ... ?
You can also use the Q factor, the quality of a measured ringing effect to estimate the damping to use, but I know no true rule or source that gives you "damping" factors as a material property as we have for density, Young modulus etc. Anyhow, as usual this means extra VV&C (verifications, validations and Calibrations) something one should do extensively for EACH model we make ;)
I believe this lack of "damping" values is mainly coming from historical reasons, adding damping made the models too complicated to solve some 50-20 years ago, one stayed also with "linear FEM" but today, with our capability to design and manufacture very slender and small structures, this knowledge of material damping, assembly damping etc becomes increasingly important, just as non-linear geometrical studies becomes the norm and must for most MEMS design.
Another issue is that you can have difficulties to simulate "through" a little damped resonance peak, the solver might fail on its way to the peak response. My way there is to start from both side, even if I miss the peak, I can get enough points to extrapolate, particularly if I make some algorithms to capture more and more points as we close in on the resonance, something some FEM software's have built in, COMSOL not, it relies on our ingenuity to find the "right" way.
--
Good luck
Ivar
Adding damping is an endless discussion, indeed. And far from obvious (well not to add damping, in COMSOL you simply add a "damping" sub node to the linear material or other relevant physics node, or even you add a complex spring constant, or you couple in any dissipative energy source ...)
The issue is which value to use, and for which parts/entities, is this an assembled structure, or a welded one, or a modern Additive Manufactured single piece item ... ?
You can also use the Q factor, the quality of a measured ringing effect to estimate the damping to use, but I know no true rule or source that gives you "damping" factors as a material property as we have for density, Young modulus etc. Anyhow, as usual this means extra VV&C (verifications, validations and Calibrations) something one should do extensively for EACH model we make ;)
I believe this lack of "damping" values is mainly coming from historical reasons, adding damping made the models too complicated to solve some 50-20 years ago, one stayed also with "linear FEM" but today, with our capability to design and manufacture very slender and small structures, this knowledge of material damping, assembly damping etc becomes increasingly important, just as non-linear geometrical studies becomes the norm and must for most MEMS design.
Another issue is that you can have difficulties to simulate "through" a little damped resonance peak, the solver might fail on its way to the peak response. My way there is to start from both side, even if I miss the peak, I can get enough points to extrapolate, particularly if I make some algorithms to capture more and more points as we close in on the resonance, something some FEM software's have built in, COMSOL not, it relies on our ingenuity to find the "right" way.
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi Invar,
Thank you so much for you in-depth explanation. With this therefore I can be able to do more relevant studies on the topic.
Best Regards,
Ben
Thank you so much for you in-depth explanation. With this therefore I can be able to do more relevant studies on the topic.
Best Regards,
Ben
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi Invar,
Good morning. I am confused about a result I found for the phase information (solid.uPhZ - Displacement phase, Z component) which I suppose should vary between lets say 0 degrees to maybe a value above 500 degrees for the frequency response analysis. The values I get are so large in the tune of 20,000 which I think are abnormal. The units I selected was degrees. Am I doing something wrong? Your suggestions and clues will be highly appreciated. Thanks for your help till now.
Best Regards,
Ben
Good morning. I am confused about a result I found for the phase information (solid.uPhZ - Displacement phase, Z component) which I suppose should vary between lets say 0 degrees to maybe a value above 500 degrees for the frequency response analysis. The values I get are so large in the tune of 20,000 which I think are abnormal. The units I selected was degrees. Am I doing something wrong? Your suggestions and clues will be highly appreciated. Thanks for your help till now.
Best Regards,
Ben