Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Damped Eigenfrequency Analysis

Please login with a confirmed email address before reporting spam

I was trying a model from the COMSOL gallery which used Damped Eigenfrequency Analysis for their model solution. But I am using 4.2a where it has only Eigenfrequency Analysis. This could be the reason why my solution is not showing desired values (no damping part in eigenfrequency values, although using thermoelastic effect). How to do a Damped Eigenfrequency Analysis in 4.2a? Can someone please explain. :(

13 Replies Last Post 26.06.2012, 13:28 GMT-4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.06.2012, 09:03 GMT-4
Hi

have you added some damping right click the linear elastic material node and add damping, then select the right type, you can start with isotropic loss factor (eta = 1/2/Q) but its worth to study the others more in detail too, check the doc

--
Good luck
Ivar
Hi have you added some damping right click the linear elastic material node and add damping, then select the right type, you can start with isotropic loss factor (eta = 1/2/Q) but its worth to study the others more in detail too, check the doc -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.06.2012, 09:50 GMT-4
Hi

Thanks for the reply. I haven't added any damping. I am particularly looking to evaluate Q value for a particular mode. My model is based on the attached example, where damped eigenfequency analysis was used.

Should I use analytical or experimental Q value For isotropic loss factor?
Hi Thanks for the reply. I haven't added any damping. I am particularly looking to evaluate Q value for a particular mode. My model is based on the attached example, where damped eigenfequency analysis was used. Should I use analytical or experimental Q value For isotropic loss factor?


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.06.2012, 10:00 GMT-4
Hi

if you do not add any material damping, and have no other damping mechanism then I do not see how you can do a damped eigenfrequency analysis ;)

If you have a measurement of eta or Q_factor and then derive an eta, you could use that to start with, but often the Rayleigh (alpha, beta) are more suited, but it requires more material property data measurements as these are frequency dependent.

And damping is very dependent on the means of assembly: bolt, screw, welding, and contact surface finish + pressure loads, w.r.t micro-slipping, and how these parameters vary with time after many temperature cycles or whatever your parts might survive

A very interesting and challenging domain, unfortunately too little documented and it's difficult to use a model validation for another different case

--
Good luck
Ivar
Hi if you do not add any material damping, and have no other damping mechanism then I do not see how you can do a damped eigenfrequency analysis ;) If you have a measurement of eta or Q_factor and then derive an eta, you could use that to start with, but often the Rayleigh (alpha, beta) are more suited, but it requires more material property data measurements as these are frequency dependent. And damping is very dependent on the means of assembly: bolt, screw, welding, and contact surface finish + pressure loads, w.r.t micro-slipping, and how these parameters vary with time after many temperature cycles or whatever your parts might survive A very interesting and challenging domain, unfortunately too little documented and it's difficult to use a model validation for another different case -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.06.2012, 18:57 GMT-4
Hi Ivar

I have added a loss factor based on Q calculated using an analytical solution. The eigenfrequency analysis gives a Q value closer to the on I had used for loss factor. As I am interested in thermoelastic damping, I am not getting a reasonable temperature profile of my model. I don't understand why the mechanics and heat transfer modes are not coupling. Do you have any advice for me?
Hi Ivar I have added a loss factor based on Q calculated using an analytical solution. The eigenfrequency analysis gives a Q value closer to the on I had used for loss factor. As I am interested in thermoelastic damping, I am not getting a reasonable temperature profile of my model. I don't understand why the mechanics and heat transfer modes are not coupling. Do you have any advice for me?

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.06.2012, 05:48 GMT-4
Hi

Indeed if you enter a Q value as the eta function for isotropic damping you will get a response very close to the theory, what might make some differences, is that in COMSOL you take into account the full 3D shape of your model, and in many "theory" approaches one uses "ideal modes", this can make some differences.

I believe there is a thermoelastic damping axample somewhere in the model library, there was at least, or was that from one of the courses ?
You need to couple the stress or deformation to the heat dissipating mode, and then couple back the temperatre field to the stress etc, I do not believe this is done "as is " in calssical "solid" physics, you need to look after some of the coupled physics.
Now I also see in V4.3 there are many new energy variables, to probably help us validate the physics coupling. But I still need my full holiday time to read the 60MB of pdf we received with the new release :)

--
Good luck
Ivar
Hi Indeed if you enter a Q value as the eta function for isotropic damping you will get a response very close to the theory, what might make some differences, is that in COMSOL you take into account the full 3D shape of your model, and in many "theory" approaches one uses "ideal modes", this can make some differences. I believe there is a thermoelastic damping axample somewhere in the model library, there was at least, or was that from one of the courses ? You need to couple the stress or deformation to the heat dissipating mode, and then couple back the temperatre field to the stress etc, I do not believe this is done "as is " in calssical "solid" physics, you need to look after some of the coupled physics. Now I also see in V4.3 there are many new energy variables, to probably help us validate the physics coupling. But I still need my full holiday time to read the 60MB of pdf we received with the new release :) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.06.2012, 13:48 GMT-4
Thanks for the reply. I have tried the example of thermoelastic damping in model gallery but I didn't get the same result. The documentation is not clear as it is for 3.5.

As the change in tempearature in thermoelastic damping of a material is due to rapid change in stresses, the heatsource equation for Isotropic materials consists of derivatives of strains in spatial direction. I am unable to implement the heat source equation in the model. For which the temperature profile is not obtained.

Thanks for the reply. I have tried the example of thermoelastic damping in model gallery but I didn't get the same result. The documentation is not clear as it is for 3.5. As the change in tempearature in thermoelastic damping of a material is due to rapid change in stresses, the heatsource equation for Isotropic materials consists of derivatives of strains in spatial direction. I am unable to implement the heat source equation in the model. For which the temperature profile is not obtained.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.06.2012, 02:34 GMT-4
Hi

unfortunately I do not have any examples to propose, but it is a subject I'll have to catch, hopefully its better implemented in v4.3 in the new non-linear material mode, as I have an important modelling to do with wheat exchange in NiTi material under vibration loads, and the realted thermal effects.

So for the moment I'm in the reading phase, and desparately looking for good COMSOL examples even "just" on classical shape memory models

--
Good luck
Ivar
Hi unfortunately I do not have any examples to propose, but it is a subject I'll have to catch, hopefully its better implemented in v4.3 in the new non-linear material mode, as I have an important modelling to do with wheat exchange in NiTi material under vibration loads, and the realted thermal effects. So for the moment I'm in the reading phase, and desparately looking for good COMSOL examples even "just" on classical shape memory models -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.06.2012, 07:01 GMT-4
Hi

Thanks for the reply. I contacted the support and they have sent me the TED model in 4.2a. Pretty happy with that.

As I am new to COMSOL, I am having some problem implementing some equation. Could you please help how would I implement the follwoing equation (this is a part of my heat source equation)? Is it possible to write in this form?

δ/δt(εxx+εyy+εzz)
Hi Thanks for the reply. I contacted the support and they have sent me the TED model in 4.2a. Pretty happy with that. As I am new to COMSOL, I am having some problem implementing some equation. Could you please help how would I implement the follwoing equation (this is a part of my heat source equation)? Is it possible to write in this form? δ/δt(εxx+εyy+εzz)

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.06.2012, 07:48 GMT-4
Hi

you need to get the full COMSOL names with the related prefixed (depends on the physics used. I have noticed that in the new 4.3 doc there are detailed description of all (most ?) internal variables you can find in the Results pull down list

--
Good luck
Ivar
Hi you need to get the full COMSOL names with the related prefixed (depends on the physics used. I have noticed that in the new 4.3 doc there are detailed description of all (most ?) internal variables you can find in the Results pull down list -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.06.2012, 11:39 GMT-4
Hi

I have tried to do the exactly the same modelling that was sent to me by COMSOL. But still I can't get the results. I have double checked everything between my model and COMSOL model. Still no result. I am pasting the log of the rtwo simulations. According to my log something is not right. But don't know how to correct it. Can you please help what is wrong... :(

Why the number of degrees of freedom is different from mine?

COMSOL MODEL:

Eigenvalue Solver 1 in Solver 1 started at 25-Jun-2012 16:37:14.
Eigenvalue solver
Number of degrees of freedom solved for: 2808.
Nonsymmetric matrix found.
Scales for dependent variables:
mod1.u: 0.00094
mod1.T: 21
mod1.wZ: 2.9e+002
Iter ErrEst Nconv
1 0.018 1
2 0.019 1
3 0.024 1
4 0.015 1
5 0.033 1
6 0.015 1
7 0.012 1
8 0.012 1
9 0.0022 1
10 0.0012 1
11 0.018 1
12 0.0072 1
13 0.031 1
14 0.00019 1
15 0.00027 1
16 0.017 1
17 0.00024 1
18 0.00017 1
19 0.0098 1
20 0.0063 1
21 0.0032 1
22 0.0025 1
23 0.0014 1
24 1.7e-005 1
25 7e-006 1
26 9.5e-006 1
27 1.5e-006 1
28 0.00032 1
29 3e-005 1
30 0.015 1
31 6.9e-008 2
530 linear system solutions.
530 matrix multiplications.
529 re-orthogonalizations.
Eigenvalue Solver 1 in Solver 1: Solution time: 3 s.


MY MODEL:

Eigenvalue Solver 1 in Solver 1 started at 25-Jun-2012 16:31:35.
Eigenvalue solver
Number of degrees of freedom solved for: 2247.
Symmetric matrices found.
Scales for dependent variables:
mod1.u: 3.5e-006
mod1.wZ: 1
Iter ErrEst Nconv
1 0 1
20 linear system solutions.
20 matrix multiplications.
19 re-orthogonalizations.
Eigenvalue Solver 1 in Solver 1: Solution time: 0 s.
Hi I have tried to do the exactly the same modelling that was sent to me by COMSOL. But still I can't get the results. I have double checked everything between my model and COMSOL model. Still no result. I am pasting the log of the rtwo simulations. According to my log something is not right. But don't know how to correct it. Can you please help what is wrong... :( Why the number of degrees of freedom is different from mine? COMSOL MODEL: Eigenvalue Solver 1 in Solver 1 started at 25-Jun-2012 16:37:14. Eigenvalue solver Number of degrees of freedom solved for: 2808. Nonsymmetric matrix found. Scales for dependent variables: mod1.u: 0.00094 mod1.T: 21 mod1.wZ: 2.9e+002 Iter ErrEst Nconv 1 0.018 1 2 0.019 1 3 0.024 1 4 0.015 1 5 0.033 1 6 0.015 1 7 0.012 1 8 0.012 1 9 0.0022 1 10 0.0012 1 11 0.018 1 12 0.0072 1 13 0.031 1 14 0.00019 1 15 0.00027 1 16 0.017 1 17 0.00024 1 18 0.00017 1 19 0.0098 1 20 0.0063 1 21 0.0032 1 22 0.0025 1 23 0.0014 1 24 1.7e-005 1 25 7e-006 1 26 9.5e-006 1 27 1.5e-006 1 28 0.00032 1 29 3e-005 1 30 0.015 1 31 6.9e-008 2 530 linear system solutions. 530 matrix multiplications. 529 re-orthogonalizations. Eigenvalue Solver 1 in Solver 1: Solution time: 3 s. MY MODEL: Eigenvalue Solver 1 in Solver 1 started at 25-Jun-2012 16:31:35. Eigenvalue solver Number of degrees of freedom solved for: 2247. Symmetric matrices found. Scales for dependent variables: mod1.u: 3.5e-006 mod1.wZ: 1 Iter ErrEst Nconv 1 0 1 20 linear system solutions. 20 matrix multiplications. 19 re-orthogonalizations. Eigenvalue Solver 1 in Solver 1: Solution time: 0 s.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.06.2012, 14:11 GMT-4
Hi

the "Comsol" model seem to have 3 dependent variables yours only 2, the T is missing, this could explain it , but there might be many other things too

--
Good luck
Ivar
Hi the "Comsol" model seem to have 3 dependent variables yours only 2, the T is missing, this could explain it , but there might be many other things too -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.06.2012, 05:10 GMT-4
Hi

Thanks. FIxed the problem. Could you please explain why rescaling of PDEs is necessary for equation based modelling?
Hi Thanks. FIxed the problem. Could you please explain why rescaling of PDEs is necessary for equation based modelling?

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.06.2012, 13:28 GMT-4
Hi

rescaling is motly required all the time, its linked with the limited resolution of the binary numers, you get easily numerical overflows or underflows if you do not make all your values behave close to "1" via a constant scale, that you undo once you have inverted all your matrices. Its standard numerical techniques, but often far from trivial. It means more or less oyu should have a good knowledge of your results before you solve, which is not always the case if you have not done your homeworks of first estimatinmg the values by hand. Hence you might ned to run your case a few times to better converge with the scaling, at each new run

--
Good luck
Ivar
Hi rescaling is motly required all the time, its linked with the limited resolution of the binary numers, you get easily numerical overflows or underflows if you do not make all your values behave close to "1" via a constant scale, that you undo once you have inverted all your matrices. Its standard numerical techniques, but often far from trivial. It means more or less oyu should have a good knowledge of your results before you solve, which is not always the case if you have not done your homeworks of first estimatinmg the values by hand. Hence you might ned to run your case a few times to better converge with the scaling, at each new run -- Good luck Ivar

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.