You might say that it's about the difference between mathematics and physics. Numerical methods are iterative, and in order to claim you've solved some equations on a computer, you have to show that an error term can be made small enough to suit you, after having performed however many iterations it has taken to achieve this.
What if you run your computer for a week, through zillions of iterations, and your error term is still not suitably small? In other words, what if your numerical method won't converge?
Well, instead of buying a faster computer, waiting another week, trying to find a more efficient algorithm, etc., one thing you could do is to solve a similar problem, using equations that are similar, but still not the same.
In so doing, you are actually making different physical (as in physics) assumptions - i.e., you have changed the physics.
I am essentially putting NIST, and NIST'ians, on the spot by challenging them to quantify the differences between what the oiginal physics approach taken by NIST - which I believe is generally more realistic (contrary to what rwguin seems to believe) - and the one settled upon ('plan B', so to speak). Those that cannot be so quantified, I will continue to refer to as NBB's. These are analogous to religious "mysteries of the faith".
It seems apparent that such quantitative determinations don't exist - I don't recall them being convincingly demonstrated anywhere, though there may be some handwaving assertions that all is well. As I noted above, there's no footnotes in NCSTAR 1-6C, so the NBB's of 1-6C surely are not documented there.
I expect that, before all is said and done, there will be scores of instances of NBB.
Even "realistic" physics equations can (and generally will) have assumptions 'hidden' therein, that physicists are likely to forget about. Case in point - the basic wave equation for ideal stretched string! If your equation or equations are realistic 'enough', then they will have predictive power when experimentally tested. In the case of a WTC model, since so little of the forensic evidence was kept, just how are we supposed to know that there's any significant correlation between reality and NIST's models? (For floor elements, there wasn't much to keep.
But for columns, I will guess that creep was measurable or estimable. I'd appreciate it if knowledgeable engineers would inform us as to whether or not this is correct.)Another point - which I don't particularly want to get into in this thread - is how do you know that even NIST'ian first choices which converged were adequate for the job? As sophisticated as it may have been, the computer modeling software still could not be used to predict the wall getting pulled in - it had to be added as an ad hoc assumption. That raises serious questions in my mind as to what other deficiencies were inherent in the entire NISTian enterprise.
