Hi Bjorn,
The problem is on the "Advanced" tab of the "Analysis > Parameters..." window. Instead of setting the "Relaxation parameter" (last input on the tab) to 0.3, it was at the default value of 1. The "Relative tolerance" (second input from the bottom) had been set to 0.3.
P.S. lf you are viewing the interface in a language other than English, perhaps the labels are translated improperly or switched. This would explain why the 0.3 was entered in the wrong box. Let me know what langauge you are using if it is not English, and I will have the translation checked.
* * * The rest of this reply can be skipped if one is not interested in a little theory behind the process
🙂 * * *
Now might be a good time for me to describe what's going on and why a "relaxation parameter" is required. In any type of nonlinear solution, one solution method is to "guess" what the result is so that the nonlinear part of the solution can be calculated and removed, and then calculate the result. If the guess and calculated result matches, the solution has converged!
In this case of radiation in heat transfer, the heat flux due to radiation is nonlinear -- it is based on the fourth power of temperature. So the heat flux due to radiation is based on the assumed temperature result. This allows the matrices to be formed and solved for the temperature. If the assumed and calculated temperatures agree, the solution has converged! Of course, the assumed temperature is rarely accurate on the first iteration, so the calculated temperature is also "not very accurate". In some cases, the calculated temperature can be worse than the assumed temperature. Instead of using the calculated temperature for the next iteration, it is better to use something in between the assumed temperature and the calculated temperature. This is what the user-specified "relaxation parameter" does. If the value is 1, then the newly calculated temperature is used for the next iteration. If the value is 0, then the previously assumed temperature is used for the next iteration. (This is not a good choice since all iterations would be identical, and the analysis would never converge nor diverge!) A relaxation parameter between 0 and 1 will use a value in between the assumed and calculated temperature. The higher the nonlinear effect, the smaller the relaxation parameter should be.
I'm sure that the above description is not correct in all of the details; after all, I just a user like you guys. I didn't write the software ;-). But the concept is close enough to help with deciding on what relaxation parameter to use, and why the analysis may be diverging.
John Holtz
Product Design Engineer, Algor Simulation
Autodesk, Inc.
John Holtz, P.E. Global Product Support
Autodesk, Inc. If not provided already, be sure to indicate the version of Inventor Nastran you are using!"The knowledge you seek is at knowledge.autodesk.com" - Confucius 😉