I think everyone is familiar with the soda pop can analogy for the towers' demise. You balance on top of an empty soda can, tap the side of the can with your finger, and it immediately collapses. It seems that this collapse would be fairly easily modeled numerically -- the can has regular geometry, known strength parameters, etc., and the weight on the top and the sudden deformation provided by the finger tap could be adjusted until "global collapse ensued." A back of the envelope computation should be capable of demonstrating both why the can doesn't collapse immediately, and what size the deformation would have to be to initiate collapse.
Would it be possible to start with a model of this type, and work up the complexity to create something similar to the WTC 7 geometry? Then, one could remove interior members (or simulate other types of deformations) and/or add weight to the top until failure was induced.
I am a modeler, but not of the structural type. The process of working from simplicity to complexity is something I use quite often.
