Heat transfer problem

[manarfadul]

Hi,
I am trying to perform a heat transfer analysis on a column consists of three layers. only the middle region is heated and the top and bottom region are the same as at the air temperature. the three layers were tied together and everything should be set right. However, when I perform the analysis I noticed that the temperature values at some regions, "particularly the boundaries when I have rapid change in temperature"are going below the air temperature even go down to negative then go up again. I used one layer only to detect where the error is, and it worked fine. then I used the all three layers with all column heated and it worked fin too. It sounds that I have this problem only with more than one part tied together and with the mid part heated only. Any idea what could cause this problem?

Manar

 

[corus]

You have an unstable thermal transient solution most probably caused by too low a time step, or too coarse a mesh, presuming you're using finite elements in your analysis. Try refining the mesh towards the region where the transient is most severe and reduce the initial time step.

 

[manarfadul]

Thanks for your suggestions. And yes I already refined the mesh, the results got much better but still though drop down around 17 to 19 below the initial temperature(20) in the first 400 seconds in my analysis and then go up as it should be. regarding the step time I used linear elements and as far as I know this oscillation could occur only in higher order elements, besides I used Abaqus automatic increments which should be unconditionally stable. tried to reduce initial time step as you said, did not work at all. tried to refine the mesh more and more but my laptop was about to blow up!

 

[corus]

When you refine the mesh, bias it towards the position/surface where the transient is most severe. In that way you avoid having millions of elements in your model. Abaqus isn't unconditionally stable but will 'wobble' under severe transients. You could add options under *heat transfer line where the change in temperature is limited to a set value per iteration, otherwise it will cut back the time step.

 

[manarfadul]

Yes, your are absolutely right. but having hard time now matching the meshes of the three parts together with bias mesh. regarding temperature change option, my first setting was 10 and then tried to reduce it to 5 which did not enhance the results either. I think reducing temperature change per increment to be less than 5 will be so expensive.