kotawsu,
Following on from Pams idea, simply add another step to your analysis. Fix the displacements from your first thermal analysis step using '*boundary,fixed' and fix the degrees of freedom as pam suggests. This should hopefully give you the loads you're after as reactions.
Matt
Ewoks,
Pam is right, you need to perform an eigenvalue buckling analysis on the tube first to find the buckling modes.
You can then use these modes together with the *IMPERFECTION command in Abaqus explicit using the *DYNAMIC, EXPLICIT command to allow the tube to collapse.
Matt
