As for discussing the details of how, in the NIST simulations, the WTC7 perimeter is distorted: Be cautious.
We cannot hope to ever simulate the real collapse to an arbitrarily fine level of detail without knowing far better than we possibly can the starting conditions.
And this is the best reason why NIST should not even have tried to replicate the behaviour of the shell, as indeed I assume they DID not really try.
The question NIST sought to answer was NOT "can you replicate in great accuracy the collapse of the walls?"
The real question was: "Can the initial, local failures predicted by the ANSYS model result in a progressive collapse of the whole thing, as observed?" where the "as observed" part includes the major steps: 1. East core drops as a unit, 2. West core drops as a unit, 3. Perimeter drops as a unit.
(No, I do not cite NIST's objectives here; I do not even paraphrase them. This is to save me and you a bit of time. I assume we are already on the same page here, that, essentially, you all agree that this is what NIST did and did not go after. Challenge my assumption, if you must, and I can dig for the cites)
Remember that NIST published animations of two main global collapse runs: One WITH initial damage from debris impact, one WITHOUT such initial damage.
I hope you all recall that the latter resulted in a collapse with very gross deformations of the entire building's shell, while the former resulted in a "tamer" collapse with less distortion of the shell and thus much more "like" the collapse filmed in reality.
The thing here is that no one knows the exact extent of prior damage! NIST only plugged in the damage that is documented in images, in addition to damage in the lower 14 floors of the East core as results from the ANSYS simulation. But there were large parts of the building's shell that no camera ever captured after the fall of the North Tower, and there surely was also accrued structural damage from fires in the West core, and above the 16th floor, that NIST did not model.
And so even the simulation WITH impact damage is a WRONG model because it starts out with TOO LITTLE damage.
But seeing how wildly different the "with" and "without" impact damage animations are, it stands to reason that correcting the model "with" damage, albeit too little and perhaps some wrong damage, with a model run that has all the right (alas unknowable) damage would look quite different still, and would, or would perhaps not, resemble the real collapse even better.
In any case, the point remains: Because it makes, obviously, a significant difference for the LS-DYNA model whether and what initial damage you put in, and because we cannot and do not know the actual amount and nature of initial (and accrued) damage, we MUST expect the LS-DYNA simulaton to be visibly different from the real collapse - and so there being a difference is not a legitimate ground to dismiss the NIST animation wholesale.
NIST got the main features right.
(And since the Hulsey study has been brought up as a reference, a quick reminder of the main difference between NIST's and Hulsey's animations:
- In the NIST report, the individual failures, their timing and the eventual collapse animation are the RESULT of an unbroken chain of causes and effects, starting with the structure as built minus impact damage plus initial fire location, through a realistic fire simulation with realistic fuel loads, through a full simulation of rising and falling temperatures in the structural materials as fires come and wane, through a detailed tallying of structral damage that accrues as a result of that heating, and how connection strength is impacted, how beams expand and contract, and sag, and through a full dynamic 47-floor model that takes all the results of the prior modelling - AND NOTHING MORE. This model in fact explains the final collapse sequence as a result of physics running its course from true starting conditions.
- In the Hulsey model, there is the structure as built, but he models no initial damage, no fires, no heating, no accrual of connection damage - and then makes disappear, as if by magic, entire columns, many of them, at the times he needs to remove them to mimick the outer appearance of the collapse. This is an appeal to magic (or to "demolition devices" of which there is neither evidence nor even a theory) and entirely non-explanatory: The removal (failure) of columns here is not a result of the modelling, it's input to the model.