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.
Transient heat transfer analysis
Posted 12.05.2010, 12:51 GMT-4 8 Replies
Please login with a confirmed email address before reporting spam
Attached is a file showing the model and solution.
I am trying to do a simple analysis initially to understand the flaw. My actual problem is a coupled diffusion-thermal analysis.
I tried creating a simple block using one of the library materials (copper). did a mapped mesh. solved it using a transient heat transfer analysis.
the initial condition is 100 K all over the block.
the boundary condition is 0K on all the outer boundaries.
I modeled a 1/8th geometry of the block. so only 3 faces will be at 0K.
when I check the temperature distribuitoni after 5 sec, the boundary plot shows a max T value around 101.628K
when I do an extrusion plot, the temperature is 112 K.
there is no heat transfer occuring, no external or internal heat sources. the expected solution is temperature distribution to be within 0 and 100 K, but never go above 100 K.
I have tried using different mesh options, time stepping options and different materials. I have also tried using finer mesh towards the boundatries and coarse smesh towards the center of the block to better capture the sudden change in the temperature at the boundary.
In all cases, the max T value in the solution is above 100.
Comments please.
Thank you for your time
I am trying to do a simple analysis initially to understand the flaw. My actual problem is a coupled diffusion-thermal analysis.
I tried creating a simple block using one of the library materials (copper). did a mapped mesh. solved it using a transient heat transfer analysis.
the initial condition is 100 K all over the block.
the boundary condition is 0K on all the outer boundaries.
I modeled a 1/8th geometry of the block. so only 3 faces will be at 0K.
when I check the temperature distribuitoni after 5 sec, the boundary plot shows a max T value around 101.628K
when I do an extrusion plot, the temperature is 112 K.
there is no heat transfer occuring, no external or internal heat sources. the expected solution is temperature distribution to be within 0 and 100 K, but never go above 100 K.
I have tried using different mesh options, time stepping options and different materials. I have also tried using finer mesh towards the boundatries and coarse smesh towards the center of the block to better capture the sudden change in the temperature at the boundary.
In all cases, the max T value in the solution is above 100.
Comments please.
Thank you for your time
Attachments:
8 Replies Last Post 14.05.2010, 01:58 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I agree that thermal effects should not make "waves" and go above the initial temperature, but looking at your images I have the feeling you are just a few instants after having started to run your transient with 100K inside and 0K on the external borders. This means that you started from a square temperature profile, with so to say a singularity. then I'm not astonished if the smoothing fonctions of the polynomial developments on the results manages to make some "unphysical" or rather "numerical" ripples. Try to set the graphs to linear and look directly at the mesh profile. If you still go above 100K I agree it's suspect, if not my explanation could be valid.
Around sharp edges: always check the meshing and the results, I often go back to non-interpolated direct mesh observations, then I plot it in Comsol standard foramt
Have fun Comsoling
Ivar
I agree that thermal effects should not make "waves" and go above the initial temperature, but looking at your images I have the feeling you are just a few instants after having started to run your transient with 100K inside and 0K on the external borders. This means that you started from a square temperature profile, with so to say a singularity. then I'm not astonished if the smoothing fonctions of the polynomial developments on the results manages to make some "unphysical" or rather "numerical" ripples. Try to set the graphs to linear and look directly at the mesh profile. If you still go above 100K I agree it's suspect, if not my explanation could be valid.
Around sharp edges: always check the meshing and the results, I often go back to non-interpolated direct mesh observations, then I plot it in Comsol standard foramt
Have fun Comsoling
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
also does turning "recover" ON on the surface plot have any effect?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks for your response .
I agree the initial and boundary conditions are not realistic. a sudden jump is definitely expected. I have tried running the model for longer times (500, 1400 sec etc.) to see if the initial jump eventually smooths out. But it did not.
I did the same exact analysis in ANSYS and got the result as expected (temperature distribution within 0 to 100) I have attached the result.
I have to use COMSOL for my actual problem since I need the multiphysics option, so I cannot continue with ANSYS.
I am checking concentrations to the order of 4 decimal, hence any little jump as observed in the solution is not acceptable.
I have another question here, ANSYS usually plots out a nodal solution. I dont see the same option in COMSOL. All I can get is a booundary plot or isosurface. Is there a nodal plot available in COMSOL?
I am using COMSOL 3.5 a, and I dont find the 'recover' option . Please elaborate on 'Try to set the graphs to linear'. I did not understand that. Thanks. (I am a new user)/
Thanks a lot for your time and patience.
Your help is appreciated.
I agree the initial and boundary conditions are not realistic. a sudden jump is definitely expected. I have tried running the model for longer times (500, 1400 sec etc.) to see if the initial jump eventually smooths out. But it did not.
I did the same exact analysis in ANSYS and got the result as expected (temperature distribution within 0 to 100) I have attached the result.
I have to use COMSOL for my actual problem since I need the multiphysics option, so I cannot continue with ANSYS.
I am checking concentrations to the order of 4 decimal, hence any little jump as observed in the solution is not acceptable.
I have another question here, ANSYS usually plots out a nodal solution. I dont see the same option in COMSOL. All I can get is a booundary plot or isosurface. Is there a nodal plot available in COMSOL?
I am using COMSOL 3.5 a, and I dont find the 'recover' option . Please elaborate on 'Try to set the graphs to linear'. I did not understand that. Thanks. (I am a new user)/
Thanks a lot for your time and patience.
Your help is appreciated.
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Here is the recover: cl.ly/65817cc36d73100ab2c0 right under smooth
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
have you tried to use linear elements ?
on the other side it reminds me about some of the comments on the Stabilisation Techniques, maybe some of the settings of the numerical diffusion are not otimum for your case. Take a read through the doc (CM refguide), even better ask support (but normally it's holidays these days ;)
Have fun Comsoling
Ivar
have you tried to use linear elements ?
on the other side it reminds me about some of the comments on the Stabilisation Techniques, maybe some of the settings of the numerical diffusion are not otimum for your case. Take a read through the doc (CM refguide), even better ask support (but normally it's holidays these days ;)
Have fun Comsoling
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
have you tried to use linear elements ?
What is going to be the effect of using linear elements? are they more stable with less overshoot/undershoots?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
I have used linear elements. The peak effect is reduced at the cost of the solution accuracy.
Thanks for the comments .
I will contact support.
Thanks for the comments .
I will contact support.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
then I suspect some of the numerical staqilisation thechnics are playing you a trick, reread the corresponding doc, it can help you to understand
Good luck
Ivar
then I suspect some of the numerical staqilisation thechnics are playing you a trick, reread the corresponding doc, it can help you to understand
Good luck
Ivar