Thermal simulation gives negative result, with 0C held temp.

[John_Holtz]

Hi everyone,

I was in contact with rbame through a support case, but I wanted to post some information here so that others could learn what happens.

All analysis types can be sensitive to the mesh size, but heat transfer analysis can be particularly sensitive to the mesh size. If the mesh is too coarse (element size too large, not enough element in the model), or if the element is distorted, the result is the minimum calculated temperature can be colder than the minimum temperature applied to the model, or the maximum calculated temperature can be hotter than the maximum temperature applied. Math can be strange at times, and this is one of those time.

In this case (and many others), the solution is to create a finer mesh in the part or area of the model where the abnormal temperatures occur.

Another thought that just occurred to me is to change the applied load. Instead of applying a fixed temperature, apply a convection load with a "large" convection coefficient. "Large" is 1000 times the thermal conductivity of the material. I am not sure about Fusion (which uses Nastran for the solver), but in other simulation programs a fixed temperature applies an "infinite" load which can contribute to the inaccurate results. A large convection will force the temperature to approach the desired temperature but allow some slight flexibility. And with smaller numbers in the solution matrix, the solution of the equations will be more accurate.