The purpose of this thread is to cast light on what I'll call NISTian black boxes, as well as (what I believe) is the blindness of the NIST fan base regarding these black boxes. A NISTian black box is an insufficiently documented procedure, wherein a 'plan B' was pursued instead of an intial 'plan A', which affects the outcome of NIST's analysis in ways that can't be determined by the reader of the NIST docs. Indeed, I don't even think NIST knows - it's not just a question of documentation. Be that as it may, certainly the NIST apologists should be able to readily explain to us what's inside these black boxes, if their faith is of the scientific rather than religious type.
I should admit, from the beginning, that I haven't done my homework, in that the person to put the questions regarding the NISTian black boxes (NBB hereafter) should be somebody who understands the modeling technology. However, as nobody else has stepped forward to do this (including the now numerous engineering and architect professional members of ae911truth), I guess that I will plunge in where angels fear to tread! I have taken a graduate course in numerical methods, many years ago, but that in no way makes me suitable for the task at hand - even if I actually remembered most of what I had learned, which I don't.
The algorithm for producing the NBB's is pretty simple - load the NIST pdf file, search for the string "converge" (ignoring it in executive summaries), and record in this thread those instances where the plan B workaround for making the hitherto un-converging model converge. This will be followed by two question per incident, viz.,
Q1) what is the physical (as in "inherent physics captured in the software") difference in the two approaches, in general? The preferred form of this answer is in terms of list of various properties/behavior over a domain large enough to encompass the WTC scenario to which it is applied. I will guess that graphs are probably the best way to convey the information, though they'd ideally be accompanied by the equations from which they were derived.
Q2) what was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied? In particular, what was the quantitative effect of using the convergent method on the overall modeling behavior?
Some variation in this routine is expected, depending on the text, which I just happen to come across while using the "converge" string process. I don't recall NIST using the term "workaround", but it should be clear that I am talking about the computational method that was actually used, since the first method(s) contemplated either did not converge, or converged too slowly.
Now, 2) seems like a trick question, even to me. For if the first method, presumably involving the best guess of software widgets/algorithms/approximations that could be made by the NIST engineers, would not converge, how is answering 2) even possible? I don't know, but the answer may well be that it is never possible. OTOH, it may be possible to compute formulas for converging deltas between the two methods, some or all of the time.
It may turn out (I haven't looked at the NIST documents in a long time, I'm about to now) that even 1) is a trick question, of sorts. And the reason is that it may often be the case that even the convergent methods are not adequately documented. If the reader of NIST's 10,000 page report still cannot determine which elements of the toolbox were used at a given step in the modeling process (as well as whatever ancillary info was used), they sure as heck can't tell us what the deltas are with the (presumably) less documented non-converging methods.
I plan on only adding to this thread on weekends. This is not my life's work, even if this has proven to be a huge life diversion....
Also, I will not be going in order through the NCSTAR documents. I start with 5 examples in NCSTAR 1-6C.
OK, NIST apologists, this is your chance to really shine! Otherwise, some "twoofers" may come to the conclusion that proclaiming the NIST documents as an explanation was mostly an exercise of faith in NIST, rather than a fair (and of necessity, knowledgeable) evaluation of their work.
NBB #1: p. 12 / acrobat p.60
=======================
"To improve convergence in analysis, the negative slope in the stress-strain relationship after cracking or crushing in compression was removed, and the concrete was assumed to be plastic after cracking or crushing"
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #2: p. 73 / acrobat p.121
========================
"To improve convergence in analysis, the negative slope in the stress-strain relationship after cracking or crushing in compression was removed, and the concrete was assumed to be plastic after cracking or crushing"
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #3: p 77 / acrobat p. 125
===================
Gravity Plus Thermal Loading: The analysis of the truss model subjected to temperature time history was carried out statically; however, when the solution process did not converge, to overcome the convergence problem, the problem was solved dynamically with a 5 percent Rayleigh damping. The static analysis was then resumed when the acceleration and velocity became small.
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #4: p. 84 / Acrobat p. 132
=========================
Creep in shell was included in the simplified truss model; however, it was not included when the simplified truss model was incorporated in the full floor model, because of convergence problems inherent in BEAM188 elements.
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #5: p. 84 / Acrobat p. 132
========================
The concrete slab was modeled by SHELL181 elements with a temperature-dependent bilinear material model that had the same yield stength in both tension and compression. The yield strength was set to the compressive strength.
Q1) What is the physical difference between these two approaches, in general?
Comment Even a non-engineer like me knows that concrete does much better in compression than tension!
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
I should admit, from the beginning, that I haven't done my homework, in that the person to put the questions regarding the NISTian black boxes (NBB hereafter) should be somebody who understands the modeling technology. However, as nobody else has stepped forward to do this (including the now numerous engineering and architect professional members of ae911truth), I guess that I will plunge in where angels fear to tread! I have taken a graduate course in numerical methods, many years ago, but that in no way makes me suitable for the task at hand - even if I actually remembered most of what I had learned, which I don't.
The algorithm for producing the NBB's is pretty simple - load the NIST pdf file, search for the string "converge" (ignoring it in executive summaries), and record in this thread those instances where the plan B workaround for making the hitherto un-converging model converge. This will be followed by two question per incident, viz.,
Q1) what is the physical (as in "inherent physics captured in the software") difference in the two approaches, in general? The preferred form of this answer is in terms of list of various properties/behavior over a domain large enough to encompass the WTC scenario to which it is applied. I will guess that graphs are probably the best way to convey the information, though they'd ideally be accompanied by the equations from which they were derived.
Q2) what was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied? In particular, what was the quantitative effect of using the convergent method on the overall modeling behavior?
Some variation in this routine is expected, depending on the text, which I just happen to come across while using the "converge" string process. I don't recall NIST using the term "workaround", but it should be clear that I am talking about the computational method that was actually used, since the first method(s) contemplated either did not converge, or converged too slowly.
Now, 2) seems like a trick question, even to me. For if the first method, presumably involving the best guess of software widgets/algorithms/approximations that could be made by the NIST engineers, would not converge, how is answering 2) even possible? I don't know, but the answer may well be that it is never possible. OTOH, it may be possible to compute formulas for converging deltas between the two methods, some or all of the time.
It may turn out (I haven't looked at the NIST documents in a long time, I'm about to now) that even 1) is a trick question, of sorts. And the reason is that it may often be the case that even the convergent methods are not adequately documented. If the reader of NIST's 10,000 page report still cannot determine which elements of the toolbox were used at a given step in the modeling process (as well as whatever ancillary info was used), they sure as heck can't tell us what the deltas are with the (presumably) less documented non-converging methods.
I plan on only adding to this thread on weekends. This is not my life's work, even if this has proven to be a huge life diversion....
Also, I will not be going in order through the NCSTAR documents. I start with 5 examples in NCSTAR 1-6C.
OK, NIST apologists, this is your chance to really shine! Otherwise, some "twoofers" may come to the conclusion that proclaiming the NIST documents as an explanation was mostly an exercise of faith in NIST, rather than a fair (and of necessity, knowledgeable) evaluation of their work.
NBB #1: p. 12 / acrobat p.60
=======================
"To improve convergence in analysis, the negative slope in the stress-strain relationship after cracking or crushing in compression was removed, and the concrete was assumed to be plastic after cracking or crushing"
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #2: p. 73 / acrobat p.121
========================
"To improve convergence in analysis, the negative slope in the stress-strain relationship after cracking or crushing in compression was removed, and the concrete was assumed to be plastic after cracking or crushing"
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #3: p 77 / acrobat p. 125
===================
Gravity Plus Thermal Loading: The analysis of the truss model subjected to temperature time history was carried out statically; however, when the solution process did not converge, to overcome the convergence problem, the problem was solved dynamically with a 5 percent Rayleigh damping. The static analysis was then resumed when the acceleration and velocity became small.
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #4: p. 84 / Acrobat p. 132
=========================
Creep in shell was included in the simplified truss model; however, it was not included when the simplified truss model was incorporated in the full floor model, because of convergence problems inherent in BEAM188 elements.
Q1) What is the physical difference between these two approaches, in general?
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
NBB #5: p. 84 / Acrobat p. 132
========================
The concrete slab was modeled by SHELL181 elements with a temperature-dependent bilinear material model that had the same yield stength in both tension and compression. The yield strength was set to the compressive strength.
Q1) What is the physical difference between these two approaches, in general?
Comment Even a non-engineer like me knows that concrete does much better in compression than tension!
Q2) What was the physical, quantitative difference which accrued by using the workaround in the specific WTC scenarios to which this workaround was applied?
). However, I also have little doubt that there is not even enough documentation presented by NIST to allow an individual with sufficient expertise to do this.
