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.

Riddle about the density

Please login with a confirmed email address before reporting spam

Hi guys,

I would like you to help me work out something I have been discussing a couple of times with a local COMSOL expert, who couldn't help me owing he is not in an expert in my specialization field.

I am modeling a thin beam quasi clamped (3 nodes prescribed displacement), which is clamped so at on one side that it can expand thermally free. At first I do a thermal expansion from 0 to dT and then a linearized eigenfrequency analysis. All material parameters are constant. When regarding the lowest order TE mode at f = v/(4*L)=sqrt(E/roh)/(4*L), I experience a temperature coefficient of frequency TCF=1*alpha, where alpha is the linear thermal expansion coefficient. Until here I agree with the local COMSOL expert and he gets the same result.

However, from the analytical analysis it is known that TCF=0.5*TCE-0.5*TCroh-alpha. All material parameters being constant TCE=0, and with rho=rho0/(1-3*alpha*dT), the temperature coefficient of density becomes TCroh=-3alpha. The overall TCF therefore should be TCF=TCrho-alpha=1.5*alpha-alpha=0.5*alpha.

Btw, the same result I get for non-clamping for the f = v/(2*L) mode. Overall my only conclusion would be that COMSOL violates the mass law and calculates for free expansion rho=rho0/(1-4*alpha*dT).

Another annotation: For the same problem Comsol 3.5 with ALE(and manual density adjustment) obtains the right TCF=0.5*alpha.

Of course I am strongly willing to believe that COMSOL calculates right, so where are the riddlers amongst you able to point out the mistake?

Best regards,

Florian


4 Replies Last Post 18 févr. 2011, 09:41 UTC−5
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 17 févr. 2011, 16:05 UTC−5
Hi

in v4 the deformed frame is "on" (spatial frame, in contrary to v3.5) could this be the reason ?
that you are using the wrong frame ? and observing a "double" effect ?

--
Good luck
Ivar
Hi in v4 the deformed frame is "on" (spatial frame, in contrary to v3.5) could this be the reason ? that you are using the wrong frame ? and observing a "double" effect ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17 févr. 2011, 16:42 UTC−5
Hi Ivar,

thanks for your valuable comment, which already could be seen as demystification of one issue:
Calculating the same problem (thermal expansion+linearized eigenfrequency analysis of the beam) gave different results in Comsol 3.5, namely the temperature coefficient of frequency was TCF=-1*alpha, which would mean that the thermal effect on the density was not taken into account by Comsol 3.5 or as you say the "spatial frame" is off and the only effect on the resonance frequency would be the elongation of the rod.

Calculating that example with Comsol 4, at first I was pretty happy that suddenly the TCF is positive, however, after a moment of cheering I had to notice that the TCF=1*alpha instead of 0.5*alpha.

Could you explain to me, how to chose the frame in the linearized eigenfrequency analysis? I did not see an option...

However, for having a double effect, I would assume the TCF to be even more off, e.g. TCF=-alpha+2*TCroh=2*alpha.

Florian





Hi Ivar, thanks for your valuable comment, which already could be seen as demystification of one issue: Calculating the same problem (thermal expansion+linearized eigenfrequency analysis of the beam) gave different results in Comsol 3.5, namely the temperature coefficient of frequency was TCF=-1*alpha, which would mean that the thermal effect on the density was not taken into account by Comsol 3.5 or as you say the "spatial frame" is off and the only effect on the resonance frequency would be the elongation of the rod. Calculating that example with Comsol 4, at first I was pretty happy that suddenly the TCF is positive, however, after a moment of cheering I had to notice that the TCF=1*alpha instead of 0.5*alpha. Could you explain to me, how to chose the frame in the linearized eigenfrequency analysis? I did not see an option... However, for having a double effect, I would assume the TCF to be even more off, e.g. TCF=-alpha+2*TCroh=2*alpha. Florian

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 18 févr. 2011, 02:30 UTC−5
Hi

take a look and try at the example given here

www.comsol.eu/community/forums/general/thread/14096/#p38830

These are parts of the remarks I made at last COMSOl Conference in Paris, unfortunately I find that many of these trivial, but ESSENTIAL things with COMSOL are not well explained to the user, and most users want to resolve their complex cases without doing the basics ;) we are all impatient, me too, but still, it's so easy to do errors in such complex but really nice software

--
Have fun COMSOLING
Ivar
Hi take a look and try at the example given here http://www.comsol.eu/community/forums/general/thread/14096/#p38830 These are parts of the remarks I made at last COMSOl Conference in Paris, unfortunately I find that many of these trivial, but ESSENTIAL things with COMSOL are not well explained to the user, and most users want to resolve their complex cases without doing the basics ;) we are all impatient, me too, but still, it's so easy to do errors in such complex but really nice software -- Have fun COMSOLING Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18 févr. 2011, 09:41 UTC−5
Ivar,

words of wisdom from your side: I totally agree that one of the biggest deficits of COMSOL in comparison to other FEM Software like ANSYS is the obvious lack of documentation.
I already addressed that once to COMSOL directly, when I have been struggling with a model including two PML (a PML material model is only applicable for one space direction; if you have more than one one PML in different directions you need for each space direction an additional PML material model).

I also have to agree that FEM is to tempting to overrun those simple verification problems and just rush to the complex problem itself, while violating model abstraction, boundary conditions or simply misuse the software. That is not only a problem of the industry, where you need some fast results, but also of Academia. FEM is really dangerous (even more with COMSOL, owing to its ease of use), you always get a result, but you never get a measure for the result quality.

Back to the problem:
I have been writing back and forth with the local section of COMSOL now for around 20 mails. It took me ten mails to accept them the problem, but they still refuse to accept that the problem and analytical derivation (which is even in any basics acoustic book) is valid. So I ask again, where is the error in my thinking and/or may it not be that in such a complex software package as COMSOL there are not some remaining bugs in the fundamental code. As we agree, only few people do really compare simple analytically dealable models with their FEM results.

Some questions:
1.) If it would be possible to use the material frame as in COMSOL 3.5 I would prefer that and manually adjust the density over (1-solid.eX)*(1-solid.eY)*(1-solid.eZ). This worked in the past.

2.) How would I use ALE in v4 to calculate that problem. I have no idea how to assign ALE to a specific physics and/or to a specific solver.

3.) In the attachment is the example file. The resonance frequency is at 4.15 MHz and like I stated before it should have a TCF=0.5*alpha. Guys have a try yourself.

Florian
Ivar, words of wisdom from your side: I totally agree that one of the biggest deficits of COMSOL in comparison to other FEM Software like ANSYS is the obvious lack of documentation. I already addressed that once to COMSOL directly, when I have been struggling with a model including two PML (a PML material model is only applicable for one space direction; if you have more than one one PML in different directions you need for each space direction an additional PML material model). I also have to agree that FEM is to tempting to overrun those simple verification problems and just rush to the complex problem itself, while violating model abstraction, boundary conditions or simply misuse the software. That is not only a problem of the industry, where you need some fast results, but also of Academia. FEM is really dangerous (even more with COMSOL, owing to its ease of use), you always get a result, but you never get a measure for the result quality. Back to the problem: I have been writing back and forth with the local section of COMSOL now for around 20 mails. It took me ten mails to accept them the problem, but they still refuse to accept that the problem and analytical derivation (which is even in any basics acoustic book) is valid. So I ask again, where is the error in my thinking and/or may it not be that in such a complex software package as COMSOL there are not some remaining bugs in the fundamental code. As we agree, only few people do really compare simple analytically dealable models with their FEM results. Some questions: 1.) If it would be possible to use the material frame as in COMSOL 3.5 I would prefer that and manually adjust the density over (1-solid.eX)*(1-solid.eY)*(1-solid.eZ). This worked in the past. 2.) How would I use ALE in v4 to calculate that problem. I have no idea how to assign ALE to a specific physics and/or to a specific solver. 3.) In the attachment is the example file. The resonance frequency is at 4.15 MHz and like I stated before it should have a TCF=0.5*alpha. Guys have a try yourself. Florian

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.