As the CAD tips mentioned in the other thread are related to some meshing issues, RAM availability, etc.; and all this stuff would be too long for a forum post, I preferred to put a link to the report (pdf @ 1.7MB).
https://docs.google.com/open?id=0B8BwIN ... GlrT19Xa00
Certainly meshing is very important but we would agree that the final goal is a reliable FEA result. So, for this new "How To" section, I'd like to propose another subject. This is "How To" consider that a FEA output is good enough. At one of the recent conversations, MNeidnicht mentioned the "Confidence Interval". It would be possible to develop this concept?
Have a good day