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 The information provided may be out of date.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.

What is an "internal" degree of freedom compared with a regular degree of freedom?

Josh Thomas Certified Consultant

Please login with a confirmed email address before reporting spam

Could anyone explain the difference between an "internal" degree of freedom and a regular degree of freedom?


11 Replies Last Post Jun 27, 2012, 10:29 a.m. EDT
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 Jun 26, 2012, 1:52 p.m. EDT
Hi

it depends a bit in which context you use it,

But for example to fix a temperature on a boundary you might create an internal dependent variable that drives a heat flux Q such that the Twall-Tset = 0 value is achieved, this would be for me an internal DoF

--
Good luck
Ivar
Hi it depends a bit in which context you use it, But for example to fix a temperature on a boundary you might create an internal dependent variable that drives a heat flux Q such that the Twall-Tset = 0 value is achieved, this would be for me an internal DoF -- Good luck Ivar

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 26, 2012, 3:36 p.m. EDT
I cannot find the term “internal degree of freedom” in the manuals? What do you mean?
I cannot find the term “internal degree of freedom” in the manuals? What do you mean?

Josh Thomas Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 26, 2012, 3:46 p.m. EDT
Haha! Neither could I!

I never saw the term before but now in v4.3 when I solve some models, under Messages when COMSOL reports the number of degrees of freedom solved for sometimes it will have in parentheses the text "(internal XXX)".

So that it looks something like this:

Number of degrees of freedom solved for: 2300 (internal 46000)

This came up for me when I added plasticity to my linear elastic material model. The internal DOF were about 20X the regular DOF.
Haha! Neither could I! I never saw the term before but now in v4.3 when I solve some models, under Messages when COMSOL reports the number of degrees of freedom solved for sometimes it will have in parentheses the text "(internal XXX)". So that it looks something like this: Number of degrees of freedom solved for: 2300 (internal 46000) This came up for me when I added plasticity to my linear elastic material model. The internal DOF were about 20X the regular DOF.

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 Jun 26, 2012, 4:17 p.m. EDT
Hi
but I believe COMSOl is adding now more internal variables to resolve the commplex equations of non linear material formulas, there were already some for Solid rigid body too, but I agree not really mentioned so far

--
Good luck
Ivar
Hi but I believe COMSOl is adding now more internal variables to resolve the commplex equations of non linear material formulas, there were already some for Solid rigid body too, but I agree not really mentioned so far -- Good luck Ivar

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 26, 2012, 5:18 p.m. EDT
They are probably components of the plastic strain tensor or something similar. That would explain the large number since it is a tensor quantity and it is present at every integration point. However, in my experience, you should not see any significant increase in solution time or memory requirements due to these internal variables. I am curious if your example shows otherwise.

Nagi Elabbasi
Veryst Engineering
They are probably components of the plastic strain tensor or something similar. That would explain the large number since it is a tensor quantity and it is present at every integration point. However, in my experience, you should not see any significant increase in solution time or memory requirements due to these internal variables. I am curious if your example shows otherwise. Nagi Elabbasi Veryst Engineering

Josh Thomas Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 9:02 a.m. EDT
Nagi and Ivar -

Thanks for your replies.

Nagi, I ran the plasticity model in v4.2a and v4.3. Here are the solution time results:

v4.2a - Sol'n time = 2.05 hrs
v4.3 - Sol'n time = 2.61 hrs

That's about a 25% increase. I don't think this is due to the added internal variables that Ivar mentioned. I think the internal variables were there in v4.2a just not posted in the Messages tab as being internal. Here's the output in the Messages tab:

v4.2a - Number of degrees of freedom solved for: 46.25 MDoF
v4.3 - Number of degrees of freedom solved for: 2.35 MDoF (Internal 46.25 MDoF)

Now, I am wondering why the slow down. Has anyone seen similar slow-down when moving to v4.3?

-Josh

P.S. I ran the v4.2a solution overnight so I wasn't able to monitor the memory usage.
Nagi and Ivar - Thanks for your replies. Nagi, I ran the plasticity model in v4.2a and v4.3. Here are the solution time results: v4.2a - Sol'n time = 2.05 hrs v4.3 - Sol'n time = 2.61 hrs That's about a 25% increase. I don't think this is due to the added internal variables that Ivar mentioned. I think the internal variables were there in v4.2a just not posted in the Messages tab as being internal. Here's the output in the Messages tab: v4.2a - Number of degrees of freedom solved for: 46.25 MDoF v4.3 - Number of degrees of freedom solved for: 2.35 MDoF (Internal 46.25 MDoF) Now, I am wondering why the slow down. Has anyone seen similar slow-down when moving to v4.3? -Josh P.S. I ran the v4.2a solution overnight so I wasn't able to monitor the memory usage.

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 9:57 a.m. EDT
Hi Josh,

That is actually a small increase. If you refine the mesh using the elastic material to reach 46 million DOFs instead of 2.4 you will see several orders of magnitude increase in solution time! I really like that distinction COMSOL made in V4.3 between internal and regular DOFs for that reason.

That extra 25% time is probably going to evaluation of plasticity equations and extra solver iterations since the model is more nonlinear with plasticity. You can check the number of iterations from the log output.

Nagi Elabbasi
Veryst Engineering
Hi Josh, That is actually a small increase. If you refine the mesh using the elastic material to reach 46 million DOFs instead of 2.4 you will see several orders of magnitude increase in solution time! I really like that distinction COMSOL made in V4.3 between internal and regular DOFs for that reason. That extra 25% time is probably going to evaluation of plasticity equations and extra solver iterations since the model is more nonlinear with plasticity. You can check the number of iterations from the log output. Nagi Elabbasi Veryst Engineering

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 10:01 a.m. EDT
Oh, I did not realize that the solution times you provided where for plasticity in V4.2A and V4.3. I thought they were between models with and without plasticity.
Oh, I did not realize that the solution times you provided where for plasticity in V4.2A and V4.3. I thought they were between models with and without plasticity.

Josh Thomas Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 10:02 a.m. EDT
Yeah. That was what surprised me. I didn't expect such an increase in sol'n time for the same model file.
Yeah. That was what surprised me. I didn't expect such an increase in sol'n time for the same model file.

Ed Gonzalez COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 10:21 a.m. EDT
Hi,
The new Double Dogleg solver addresses a larger class of highly nonlinear simulations and complements an earlier Newton solver, and it is the default solver for plasticity in version 4.3. That might be the reason for the increase in computational time in your model.
The Double Dogleg solver might be slightly slower (for a given set of loads, mesh, material parameters, etc.), but it is much more robust. You can change the settings and use the nonlinear Newton solver if you prefer.

Also, Nagi is correct, the "internal" degree of freedom are due to the plasticity variables.
Hi, The new Double Dogleg solver addresses a larger class of highly nonlinear simulations and complements an earlier Newton solver, and it is the default solver for plasticity in version 4.3. That might be the reason for the increase in computational time in your model. The Double Dogleg solver might be slightly slower (for a given set of loads, mesh, material parameters, etc.), but it is much more robust. You can change the settings and use the nonlinear Newton solver if you prefer. Also, Nagi is correct, the "internal" degree of freedom are due to the plasticity variables.

Josh Thomas Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 27, 2012, 10:29 a.m. EDT
Thanks, Ed. I appreciate the response.
Thanks, Ed. I appreciate the response.

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.