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 The information provided may be out of date.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.

Strange behaviour of interpolation in heat transfer

Please login with a confirmed email address before reporting spam

Dear people,
I have a strange behaviour with my problem of heat transfer. I try to explain it:

- the object is a block of wood of 30x30x5 cm
- there is a convective cooling with a coefficient of 5 W/(m^2K)
- the heating source comes from a file of 30x30x5 values of power density arranged as a grid structure; the file is used as interpolating function (this file is the result of a FDTD simulation of electromagnetic fields)

The target of COMSOL simulation is the study of temperature distribution in function of mean power and time, so I multiply my file with a coefficient (say K).

THE STRANGE BEHAVIOUR:
the simulation normally work well and there is a proportionality between K and Delta T (mean temperature change), but some time happen that a little change of coefficient K (1%) bring a large change of Delta T (30%): seems a step change in the process heating.
I have tried to change the mesh size from normal to finer but the problem remains.

I can post all my project if necessary.

Someone can help me?
Bye Ale.

4 Replies Last Post 19 oct. 2011, 14:05 UTC−4

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 19 oct. 2011, 12:38 UTC−4
Hi people,
I've some news about my problem.
The "time dependent" simulation is split in two times:
the first with heating and the second without (the heat source heating has the expression:

DenPot(x,y,z)*(t<=tstop1-deltat1)
where DenPot is my interpolation function
tstop and deltat1 are fixed parameters

in the time dependent point the "Times" field is:

range(tiniz1,deltat1,tstop1) range(tiniz2,deltat2,tstop2)
where tiniz1, deltat1 tstop1, tiniz2,deltat2,tstop2 are fixed parameters

in my despair I have deleted the second range of the time and magical!:
now mysimulation works fine (but without the cooling phase).

Some comments?
Perhaps a bug of software?
bye Ale
Hi people, I've some news about my problem. The "time dependent" simulation is split in two times: the first with heating and the second without (the heat source heating has the expression: DenPot(x,y,z)*(t

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 19 oct. 2011, 13:11 UTC−4
Hi

before blaiming a "bug" have yu checed if your transition is "smooth" and derivable ? because asking a solver to take a "dirac step" is NOT a goo programming appraoch. the solver needs to evaluate the Jacobian of yur system (i.e. take at least frst order derivatives of the variables to get the trend and to better know where to through next solver point, but if you enter an abrupt step you have a very different left/right side derivative and no continuity Hence mostly solver issues.

Try to use the V4 (smoothed) "step" function and place few "hard points" on the time of the transition (see the solver doc, and i.e. use an "intermediate stepping" instead ofthe default "automatic").

--
Good luck
Ivar
Hi before blaiming a "bug" have yu checed if your transition is "smooth" and derivable ? because asking a solver to take a "dirac step" is NOT a goo programming appraoch. the solver needs to evaluate the Jacobian of yur system (i.e. take at least frst order derivatives of the variables to get the trend and to better know where to through next solver point, but if you enter an abrupt step you have a very different left/right side derivative and no continuity Hence mostly solver issues. Try to use the V4 (smoothed) "step" function and place few "hard points" on the time of the transition (see the solver doc, and i.e. use an "intermediate stepping" instead ofthe default "automatic"). -- Good luck Ivar

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 19 oct. 2011, 13:27 UTC−4
The results at the first “range” should obviously not be affected by the presence of the second. Look at the time dependent solver log with and without the second range removed. COMSOL may be taking bigger steps with the longer time range that affect the solution accuracy, especially since your heat source is discontinuous at time tstop1-deltat1. If true, reduce the maximum time step size, and smooth the step function as Ivar mentioned.

Nagi Elabbasi
Veryst Engineering
The results at the first “range” should obviously not be affected by the presence of the second. Look at the time dependent solver log with and without the second range removed. COMSOL may be taking bigger steps with the longer time range that affect the solution accuracy, especially since your heat source is discontinuous at time tstop1-deltat1. If true, reduce the maximum time step size, and smooth the step function as Ivar mentioned. Nagi Elabbasi Veryst Engineering

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 19 oct. 2011, 14:05 UTC−4
Thank you a lot to you both for interest and for the answers!
I'm sorry if I blame the software, I did not want to discredit anyone or anything.
I had not thought about the abrupt cessation of excitation.
How can I put at zero the excitation (heat source)?
Can work this expression where t1,t2 is the transition interval?:

DenPot(x,y,z)*((t<=t1)+(t>t1 && t<=t2)*(t-t1)/(t2-t1))

Soon I will do the tests that you have suggested, then I will tell you.

Thanks again for your interest and sorry again for doubting!

cordially Ale

Thank you a lot to you both for interest and for the answers! I'm sorry if I blame the software, I did not want to discredit anyone or anything. I had not thought about the abrupt cessation of excitation. How can I put at zero the excitation (heat source)? Can work this expression where t1,t2 is the transition interval?: DenPot(x,y,z)*((tt1 && t

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.