WTC7: Is AE911's (and NIST's) Focus on A2001 Justified if it Was Not "Key" in NIST's Global Model?

So if beams are collapsing... I suppose it's not global until a whole load of beams and columns collapse... before that it's local.... This of course is as I wrote kinda silly. The fire was undoing the frame and at some point enough of the frame was "undone" such that the rest rapidly was went Kapooey.

The details are academic really... all motions are known results of fire/heat. So it seems it like this bone's connect to that bone and away we go...
 
That's the global, asynchronous editing then - perhaps LS-DYNA work was already too advanced (and had eaten too much of the budget) by the time they decided to foucs on A2001?
Quite possibly. The LS-DYNA simulations took eight weeks to run. They only show two of them: this one with the ANSYS damage plus impact damage. And the one without impact damage. Ideally you'd be running scores of them to see what effects various variations have.
 
It's sagging in the middle. It's attached at the ends. It seems to fail because the beams pull it down. It has not failed in the vertical direction at the start of that simulation.

Why is it unthinkable? Regardless, it very clearly happened. SIX other girders collapsed in the global model. A2001 only fell later.

In the frame above A2001 on Floor 13 has not yet visible moved.

I've attached a 5fps full size version of the girder tracking with is easier to frame advance.

It is unthinkable that NIST would not include the trigger that started the collapse as you propose. Furthermore, the LS-DYNA model confirmed that was the cause. That is what Sunder was talking about at the press conference.

Figure 12-42 shows floor 14 falling first. That’s consistent with the LS-DYNA animation video. The girder between columns 79 and 44 fails shortly after that as I pointed out on post 25. The other failures also happened before the failure of the key girder. NIST knows that. But they are clear that the collapse of the girder between columns 79 and 44 on the 13th floor was the one that set off the cascade of floors that left column 79 unbraced over eight floors and cause it to buckle.

Your opinion is based on your observations of a two dimensional animation of a three dimensional collapse and claiming that you know more about what the LS-DYNA model shows than NIST.

Sunder clearly tells us what the LS-DYNA model showed in the press conference that I posted in post #33. He knows what the LS-DYNA model shows and you don’t.
 
Last edited:
Figure 12-42 shows floor 14 falling first. That’s consistent with the LS-DYNA animation video. The girder between columns 79 and 44 fails shortly after that as I pointed out on post 25. The other failures also happened before the failure of the key girder. NIST knows that. But they are clear that the collapse of the girder between columns 79 and 44 on the 13th floor was the one that set off the cascade of floors that left column 79 unbraced over right floors and cause it to buckle.

But it didn't. It fell to floor 12 and stopped there. It did this AFTER six other girder and entire sections of floor had already fallen from floors 13 and 14.
See all those collapses going on? They are nothing at all to do with A2001.Metabunk 2018-01-17 16-39-38.jpg
 
They have the actual data, not just the animation, and you don't.
Furthermore, as I pointed out, we can see the key girder failing.
The animation shows what happened in the LS-DYNA simulation, does it not? Please explain how A2001(13) dropping from floor 13 to floor 12 leads to the buckling of C79. Point out the floor collapses that were caused by the fall of A2001(13)
 
The animation shows what happened in the LS-DYNA simulation, does it not? Please explain how A2001(13) dropping from floor 13 to floor 12 leads to the buckling of C79. Point out the floor collapses that were caused by the fall of A2001(13)
After further review I realized that my last post, that you justifiably removed, was incorrect.

Thank you for pointing out that Figure 12-44 shows that the collapse of floor 13 in the NE corner of WTC 7 stopped at the 12th floor.

NIST knew this but still said that the girder failure on floor 13 was the primary cause of the cascading floor failures on the final report and at the Aug. 21, 2008 press conference. They have the data and we don’t so there is no legitimate reason to doubt that is the case.


Source: https://vimeo.com/11955064

At 12:55 “Eventually, a girder on floor 13, lost its connection to a particular interior column that provided support for the long girder and other local – The displace girder and other local fire induced damage caused floor 13 to collapse. So this was the girder that failed, connecting column 79 to column 44. (points to graphic) This began a cascading chain of failures of eight additional floors.”
Content from External Source
While reviewing the LS-DYNA video animation again, I noticed that the falling 13th floor girder appears to take out the girder between column 76 and 79 on floor 12 which would start a cascading floor failure on the west side of column 79. NIST must have known that. That would explain why NIST said that the displaced girder between columns 79 and 44 on the 13th floor was the trigger that set off a cascade of floor failures that left column 79 unbraced over 9 floors.



[T]he girder between Columns 44 and 79 buckled and walked off the bearing seat between 3.7 h and 4.0 h. NCSTAR 1-9 Vol 2 p. 527 [PDF p. 189/593]

We don’t know when the following occurred but NIST does and that is probably the reason they cited the girder on floor 13 as the trigger that caused the cascading floor failures.

Building Response at 4.0 h
Floors 10, 11, and 12. On Floors 10, 11, and 12 (Figure 11–32, Figure 11–33, and Figure 11–34), the girder between Columns 76 and 79 failed due to a tensile weld failure in the knife connection on the west side of Column 79. Temperatures in this region were less than 100 °C on these floors. The tensile force in the connection was due to an eastward lateral displacement of Column 79, which was primarily caused by thermal expansion of the girder between Column 76 and Column 79 at Floor 13.
Content from External Source
NCSTAR 1-9 p. 504 [PDF p. 166/570]
 
Last edited by a moderator:
The problem is this: NIST's "analysis" of the north eastern section's response to fire was used to inform both the ANSYS 16 floor model, and the LS-DYNA 47 floor model.

A finite element analysis of the northeast corner floor system was conducted to evaluate its response to elevated temperatures and to confirm which failure modes needed to be accounted for in the 16-story ANSYS model.
...
This analysis demonstrated possible failure mechanisms that were used to develop the leading collapse hypothesis further. The failure modes in this model were incorporated into the 16 story ANSYS and 47 story LS-DYNA analyses.
Content from External Source
NCSTAR 1-9, ch 8.8, p349 & 353 http://ws680.nist.gov/publication/get_pdf.cfm?pub_id=861611

Any errors in the A2001/column 79/column 44 area were therefore incorporated into analysis of other areas of the building, including the other girders which fail simultaneously as the model is run. If thermal expansion does not cause failure at column 79, the study's methodology, integrity and validity are undermined.

NC1.png

NC2.png
 
Any errors in the A2001/column 79/column 44 area were therefore incorporated into analysis of other areas of the building, including the other girders which fail simultaneously as the model is run. If thermal expansion does not cause failure at column 79, the study's methodology, integrity and validity are undermined.

That's a bit vague. We don't know what was carried over from the earlier investigations. There only one they discussed in detail was the rocking off to the East. We know THAT wasn't the initiating event in the LS-DYNA model either.

The modeling that led to the global collapse simulations we see was the 16 story ANSYS model which then transferred to the global LS-DYNA model. It it did not include the A2001 walk-off
 
We know that their analysis of the northeastern corner was wrong, and NIST states that they used it to inform the ANSYS and LS-DYNA models. Specifically, that analysis was used to support the 'failure modes' incorporated into the bigger models. If the only failure mode they show does not work, we have a problem. Do we need anything else to establish the relevance of A2001?
 
We know that their analysis of the northeastern corner was wrong, and NIST states that they used it to inform the ANSYS and LS-DYNA models. Specifically, that analysis was used to support the 'failure modes' incorporated into the bigger models. If the only failure mode they show does not work, we have a problem. Do we need anything else to establish the relevance of A2001?

You need to be more specific. You seem to be making some very loose connections.
 
We know that their analysis of the northeastern corner was wrong, and NIST states that they used it to inform the ANSYS and LS-DYNA models. Specifically, that analysis was used to support the 'failure modes' incorporated into the bigger models. If the only failure mode they show does not work, we have a problem. Do we need anything else to establish the relevance of A2001?
Mick has no justification for his arguments about other failures in the 47 story model being the cause of the collapse, because NIST used the 16 story ANSYS model as its input and has not justified the root causes of these other alleged failures in the 16 story model, just like they didn't with the alleged A2001 girder failure. The point being, if you don't know if the input was justified how can you know the output even has a basis?

Given the clear evidence of manipulation in the NIST alleged A2001 girder failure it is not overreacting to suspect the same was done elsewhere. Since these other alleged failures shown in the 47 story model had to originate in the 16 story model, and have not been justified, there is no basis to lend them credibility in the 47 story model.
 
Last edited:
Given the clear evidence of manipulation in the NIST alleged A2001 girder failure it is not overreacting to suspect the same was done elsewhere. Since these other alleged failures shown in the 47 story model had to originate in the 16 story model, and have not been justified, there is no basis to lend them credibility in the 47 story model.

As far as I'm concerned, that's case closed. We need a new investigation.
 
That's a bit vague. We don't know what was carried over from the earlier investigations.

Here is what the assumed for the ANSYS model, fom NIST NCSTAR 1-9, WTC Investigation, p. 482 (p. 548 in pdf):
The bolts could fail in shear under a tensile, compressive or lateral force, but the beam would only walk off the seat under a tensile or lateral force. The failure of the bolts and weld was a prerequisite for the beam end walking off the seat. The travel distance for walk off was 6.25 in. along the axis of the beam and 5.5 in. lateral to the beam. A walk off failure occurred when the bolts sheared at the seat connection, the bolts sheared or the weld failed at the top clip angle, and the beam walked off the seat. A control element was used to model beam walk-off in the axial direction. Beam walk off in the lateral direction was monitored during the analysis.
Content from External Source
I was wondering if the tinkering with the image came from some comments the NIST got after the presentation of their draft version, however, the wording in the draft version is similar and the parameters the same:

NIST SCSTAR 1-9 Volume 2, Draft version, p. 482 (p.144 in pdf)
When the connection was subjected to a large horizontal tension, a walk off failure occurred when the bolts sheared at the seat connection, the bolts sheared or the weld failed at the top clip angle, and the beam walked off the seat. The failure of the bolts and weld was a prerequisite for the beam end walking off the seat. The travel distance for walk off was 6.25 in. along the axis of the beam and 5.5 in. lateral to the beam. A control element was used to model axial walk-off. Lateral walk-off was monitored during the analysis.
Content from External Source
So the question would have been if that was something that came up during the draft presentation meeting…or even later when they corrected their supposedly typographic errors, see Erratum of June 2012:

NIST has made the following changes to the report on the collapse of World Trade Center Building 7:
  1. In Chapter 11, page 482, Analytical Model for Seated Connection at Columns 79 and 81

    The fourth sentence in the 3rd paragraph should be modified as follows:

    The travel distance for walk off was 6.25 5.5 in. along the axis of the beam and 5.5 6.25 in. lateral to the beam.

    The 5.5 in. dimension was the length of the girder bearing on the seat connection that had to slide off the seat axially to the girder. The 6.25 in. dimension accounted for the length from the flange tip to the far side of the web, so that the web was no longer supported on the bearing plate. This change corrects a typographical error which showed a lateral displacement of 5.5 in. instead of the correct value of 6.25 in., which was used in the analyses.
  2. In Chapter 11, page 527, Thermal Effects on Connections for Floor Beams and Girders

    The third and fourth sentences in the 3 rd paragraph should be modified as follows:

    The bearing seat at Column 79 was 11 12 in. wide. Thus, when the girder end at Column 79 had been pushed laterally at least 5.5 6.25 in., it was no longer supported by the bearing seat.

    The 16-story model of WTC 7 used a 12 in. bearing plate on the north side of Column 79, consistent with Frankel drawing 1091. The 5.5 in. dimension was incorrectly cited, as the 6.25 in. dimension accounted for the lateral walk-off distance. These changes correct typographical errors. The dimensions and lateral displacements used in the analyses were correct.
Content from External Source
This now begs the question if the manual edit to the figures has been done after the official release of the report in 2008. If that would be the case this would be opening a whole can of worms. Does anyone have a copy of the reports that is older than June 2012 in order to verify that issue? Also, does somebody know what the background for the release of that erratum is?
 
Last edited:
As the discussions of this thread got all a bit confused and the focus of discussion moved between different questions. I will try to summarize what I understood so far from it.

The question of the thread "Is AE911's (and NIST's) Focus on A2001 Justified if it Was Not "Key" in NIST's Global Model?" is actually two separate questions

1. Is A2001 a key element in NIST's global model?
2. If it were not, is the focus on it justified?

In fact, the thread explicitly asks only the second question under the premise that A2001 is not a key element. This can be clearly denied. However, inevitably the first question comes up implicitly then.

The answer to this question, however, is not straightforward. There is, as @Mick West pointed out correctly, apparently a wide misconception about how NIST conducted their study. It would therefore be necessary to first remove these misunderstandings before moving on. IMO it would be probably best to summarize the NIST approach either in a separate new thread that is pinned or to add it to the thread "
Have You Actually READ the NIST Report on Building 7?".

Then, in the middle of the thread, @Mick West and @Oystein pointed out some inconsistencies within the NIST report about how important the failure of the girder-column connection of girder A2001 to column 79 actually is.
That lead then to the following question:

3. Where does all this confusion about girder A2001 in the NIST report come from?
and after finding a manual edit in a figure in the final report version, the related question

4. Why and when did NIST alter the image manually showing the resulting damage from the ANSYS simulation?
I guess these second two question could be considered off-topic relating to the thread, however, I am not sure if it makes sense to open a new thread and move posts there, as the discussions are intertwined.

In the following, I try to tackle the questions separately thus, so sorry for the multiple posting but I hope it makes things easier to follow.

 
Here, I try to answer the first question,
1. Is A2001 a key element in NIST's global model?

For this I'll quote the relevant section of NCSTAR1-9, Ch. 12, p. 563 (p. 629 pdf):
12.3.2 Model Initialization and Loading Sequence
The global model was initialized as follows to minimize any dynamic effects associated with loading sequence.
  • First, gravity was applied slowly to the 47 floor structure over 4.5 s of elapsed simulation time to damp residual vibrations and eliminate dynamic response. The loads were applied smoothly with a sinusoidal load curve.
  • Then, the debris impact damage from the collapse of WTC 1 was applied to the structure instantaneously by removing damaged elements from the model that were no longer capable of bearing their loads. The structure was then allowed to damp residual vibrations for 2 s.
  • Next, the structural temperatures were applied smoothly with a sinusoidal curve and allowed to damp residual vibrations for 2 s.
  • Last, the fire-induced damage obtained from the 16 story ANSYS analysis, including damage to floor beams, girders, and connections, was applied instantaneously. The heated, damaged structure was then free to react.
Content from External Source
So the last point suggests that all the damage obtained by the ANSYS simulation is applied simultaneously and instantaneously! That confirms the point made by @Mick West in post #9 and @Oystein in post #10 (however, none of them quoted that section which explicitly describes the NIST approach; that's why I answer this again). So there is no single initial failure event in the global simulation.

I also agree that this is a point of valid criticism. It could very likely make a difference in the simulation if these damages are applied sequentially at the time as they occurred respectively in the ANSYS simulation, or if they are applied all at once.

However, NIST also made the global simulation with column 79 removed for the height of two floors only without any other damage (and also without heating as far as I can see), showing that this scenario also leads to a total collapse of the building in Section 12.4.7. This clearly shows how vulnerable WTC7 was around column 79, which could maybe explain the focus NIST put there.

So the difference if they apply all the damage sequentially or all at once will probably be quantitative, not qualitative, i.e. the way the building collapses would change, but not the fact that it collapses eventually. And with high confidence the same reasoning goes with a simulation that would keep the connection between girder A2001 and column 79 completely intact. There would be probably a quantitative change of how the building collapses en detail, e.g. the girder A2001 will fail due to buckling instead of walk-off off the seat, but the fact that it collapses eventually can certainly be concluded from the implications of the simulation results of Section 12.4.7 assuming the same instantaneous load sequence as before.
 
Last edited:
So the last point suggests that all the damage obtained by the ANSYS simulation is applied simultaneously and instantaneously! That confirms the point made by @Mick West in post #9 and @Oystein in post #10 (however, none of them quoted that section which explicitly describes the NIST approach; that's why I answer this again). So there is no single initial failure event in the global simulation.

I also agree that this is a point of valid criticism. It could very likely make a difference in the simulation if these damages are applied sequentially at the time as they occurred respectively in the ANSYS simulation, or if they are applied all at once.

I think you're doing good work in slogging through the report and clearly outlining the potential issues. On this particular point, however, note that the application of the damage state from the ANSYS model at that point was not arbitrary: that was the point in time at which the ANSYS model saw buckling in column 79. While it is a simplification to then model as if all of the damage up until that point had happened at once, once you have the column 79 buckling I think the reality is that it wouldn't likely make much of a difference in the local area around 79 whether you had ramped up the damage over time or applied it all at once. (I think this is the main driver of the divergence between the ANSYS description and the output of the LS-DYNA model--the LS DYNA failure essentially started with column 79 because its damage state from the ANSYS model was that it was buckled.) The point of the global model is to look at what then happens as a result of the column buckling in the rest of the building given the total damage from the fires.
 
Last edited:
Mick: They did, they used that damage pattern. But that damage pattern didn't have that particular ... column connection failure, A2001 to column 79, That wasn't the first thing that actually failed. There was actually a number of simultaneous collapses.



Mick: Which you can see here, there's collapses over here and over here [indicates the two circled areas above]. This say the "ANSYS-based damage application is resulting in floor structure failures around Column 79 to 81", which as you know is all the interior East side of the building, so they didn't actually..."

@Mick West, you argue here that there are two different collapse zones, however, please note that one of them is coming from the debris impact from the collapse of WTC1, as one can see in the image [click to zoom in]. that damage did not cause a collapse sequence. NIST did also a simulation without that debris impact which also showed total collapse, as you certainly are aware of, but your line of argument is then not valid for that particular case, of course.

This is just for your info. You did IMO a great job in that podcast, so don't worry me nitpicking on these details ;)
 
On this particular point, however, note that the application of the damage state from the ANSYS model at that point was not arbitrary: that was the point in time at which the ANSYS model saw buckling in column 79.

The ANSYS model did not see buckling in C79.
NCSTAR 1-9 page 476, pdf 542

Column splices were also not modeled for interior columns, as the purpose of the ANSYS model was to accumulate local
failures up to the point of buckling in a column. When column buckling appeared to be imminent, the
analyses were continued in the LS-DYNA 47 story model.
Content from External Source
and

The purpose of the ANSYS model was to simulate the accumulation of local damages and failures up to the initiation of overall global collapse due to fire.
Content from External Source

11.3.2
The building response is examined at 3.5 h and 4.0 h of heating. At 3.5 h, the floor systems had fireinduced
damage and failures of some connections, beams, and girders. After 4.0 h of heating, there was
substantially more damage and failures in the WTC 7 structural floor system, particularly in the northeast
region surrounding Column 79. The structural condition at these two times illustrates how the structure
developed sufficient fire-induced damage to reach the collapse initiation event
.
Content from External Source
Summary of the 4.0h damage:

Summary. After 4.0 h of heating, Columns 79, 80, and 81 had lost lateral support in the north-south
direction at Floor 13, due to failure of the girders between the columns. The girders between Columns 80
and 81 had buckled and the girders between Columns 79 and 44 and Columns 26 and 81 had walked off
the bearing seat at Column 79 and 81, respectively. In addition, all of the bolts had sheared at Column 79
on Floor 14, and two to three bolts had sheared on Floor 12. Approximately one-half to three-quarters of
the east floor beams had a connection damage index of 0.75-0.99 on Floors 11, 12, and 14 and all of the
east floor beams had failed on Floor 13. After 4.0 h of heating, there was substantially more damage and
failures in the WTC 7 structural system than after 3.5 h of heating. Columns 79, 80, and 81 had lost
lateral support at one or more floors.
Content from External Source
How they decided what damage to transfer:


As noted previously, the ANSYS analysis used a non-linear static procedure with an implicit solution
algorithm that solved for equilibrium at each time step, but did not account for the dynamic effects of
falling debris from framing failures in the floor systems. Based on engineering judgment and preliminary
analyses in LS-DYNA
, which included the effects of dynamics caused by debris impact from failed floor
sections, the level of failures, damage, and thermal weakening in ANSYS at 4.0 h was identified as likely
to result in an initiating event in an LS-DYNA analysis.
Content from External Source
"did not account for the dynamic effects" means that nothing moved. When a connection failed it was just recorded, when a girder or beam failed it was recorded and removed, nothing fell onto lower floors, and the column never got close to buckling.

The transfer:

11.5 TRANSFER OF ANSYS RESULTS TO THE LS-DYNA ANALYSIS
The fire-induced floor damage was input to the LS-DYNA model to determine if the predicted floor
failures would lead to an initiating collapse event, followed by a failure progression to global collapse.
When the ANSYS analysis had reached a point where buckling instability of columns appeared imminent,
the accumulated damage due to thermally-induced failures of floor framing was summarized as input for
the LS-DYNA model. The transferred data also included the temperature of structural components at that
time and buckled beams and girders.
Content from External Source
So basically they used their "engineering judgement" to pick the 4.0 h damage set to run the full LS-DYNA as the one to transfer over (as opposed to the 3.5 and 4.5 hour sets - 3.5 was not enough, 4.5 was basically the same as 4.0). They did this in part by running some simpler LD-DYNA runs, probably with all the outputs of the ANSYS runs, and seeing which ones started to collapse. They took that one, and did the full run.
 
@Mick West, you argue here that there are two different collapse zones, however, please note that one of them is coming from the debris impact from the collapse of WTC1, as one can see in the image [click to zoom in]. that damage did not cause a collapse sequence. NIST did also a simulation without that debris impact which also showed total collapse, as you certainly are aware of, but your line of argument is then not valid for that particular case, of course.

This is just for your info. You did IMO a great job in that podcast, so don't worry me nitpicking on these details ;)

I wanted to convey that the initiating event was not a single girder walk off. There's several simultaneous initiating events around C79, none of which were girder walk-offs. But it was hard to show that in the podcast, so I pointed to the other area to show that more than one thing was going on. I should probably had focussed more on the multiple beam and girder failures that actually led to the collapse (in that simulation)
 
However, NIST also made the global simulation with column 79 removed for the height of two floors only without any other damage (and also without heating as far as I can see), showing that this scenario also leads to a total collapse of the building in Section 12.4.7. This clearly shows how vulnerable WTC7 was around column 79, which could maybe explain the focus NIST put there.

NIST put their focus there because of the way the penthouse collapsed, in particular the "kink", and the lack of failure of the exterior.
Metabunk 2018-02-07 09-32-26.jpg
 
I stand corrected re column 79 buckling in the ANSYS model. It is correct that the damage state was taken when such buckling was "imminent" , not after it happened. I misremembered.
 
So basically they used their "engineering judgement" to pick the 4.0 h damage set to run the full LS-DYNA as the one to transfer over (as opposed to the 3.5 and 4.5 hour sets - 3.5 was not enough, 4.5 was basically the same as 4.0).

They could have recorded the times when which element failed in the ANSYS theoretically. However, I guess as the 25 seconds of the LS-DYNA simulations took up to 8 weeks on the cluster each, they could not apply for that practical reason each single damage separately and wait until it reaches equilibrium - even if they recorded the times the elements failed in the ANSYS simulation.

However, they could in principle have used the end state of the 3.5h LS-DYNA simulation as the input for the 4h LS-DYNA simulation, just as they used the structural damage from the WTC1 debris as input in one of their simulations.

I guess they did the 4.0h simulation first (it appears also as first in the report) before they discovered that the 3.5h simulation leads not to a collapse, so they probably decided it is not worth the effort to redo the 4h simulation.

They did this in part by running some simpler LD-DYNA runs, probably with all the outputs of the ANSYS runs, and seeing which ones started to collapse. They took that one, and did the full run.

Do you have any reference for this or is this just guessing? I am asking this because it is not clear to me how to make the LS-DYNA model simpler…
 
They could have recorded the times when which element failed in the ANSYS theoretically. However, I guess as the 25 seconds of the LS-DYNA simulations took up to 8 weeks on the cluster each, they could not apply for that practical reason each single damage separately and wait until it reaches equilibrium - even if they recorded the times the elements failed in the ANSYS simulation.

However, they could in principle have used the end state of the 3.5h LS-DYNA simulation as the input for the 4h LS-DYNA simulation, just as they used the structural damage from the WTC1 debris as input in one of their simulations.

I guess they did the 4.0h simulation first (it appears also as first in the report) before they discovered that the 3.5h simulation leads not to a collapse, so they probably decided it is not worth the effort to redo the 4h simulation.

What do you mean by "3.5h LS-DYNA" and "4h LS-DYNA"? Are you suggesting they should have taken the 3.5h ANSYS Damage, applied to LS-DYNA, run it to equilibrium, then applied the 4h ANSYS damage, and run that to collapse?

So you're suggesting you could apply the difference (i.e. the additional damage) between the 3.5h ANSYS damage and the 4h ANSYS damage. But ideal you'd do it for every difference along the way.

They did this in part by running some simpler LD-DYNA runs, probably with all the outputs of the ANSYS runs, and seeing which ones started to collapse. They took that one, and did the full run.
Do you have any reference for this or is this just guessing? I am asking this because it is not clear to me how to make the LS-DYNA model simpler…

It's just my interpretation of what I quoted from NIST: "Based on engineering judgment and preliminary analyses in LS-DYNA"
 
What do you mean by "3.5h LS-DYNA" and "4h LS-DYNA"? Are you suggesting they should have taken the 3.5h ANSYS Damage, applied to LS-DYNA, run it to equilibrium, then applied the 4h ANSYS damage, and run that to collapse?

Exactly. By 3.5h LS-DYNA and 4h LS-DYNA I mean the first two simulations they run, listed as

Four simulations were performed with the global LS-DYNA model.
  • The first was based on the best estimate of both the debris impact damage from the collapse of WTC 1 and the fire-induced damage based on the ANSYS analyses (Case B at 4 h).
  • The second simulation differed only in the input of an earlier state of fire-induced damage with Case B at 3.5 h. The purpose of this LS-DYNA simulation was to determine whether a lesser degree of fire-induced damage could lead to the collapse of WTC 7.
  • The third simulation was the same as the first, except that no debris impact damage was included. The purpose of this analysis was to determine the contribution of debris impact to the WTC 7 global collapse sequence and whether WTC 7 would have collapsed solely due to the effects of the fires.
  • In the fourth simulation, the building experienced no debris or fire-induced damage. A section of Column 79 between Floors 11 and 13 was removed. The purpose of this analysis was to determine the potential for a classic progressive collapse, i.e., disproportionate structural damage from a single failure, regardless of the cause of that failure.
Content from External Source
If the simulation at 3.5h does not lead to a collapse, it should be static. Then one should be able to extract the information of the damage after the simulation, and put it as input data for the next model, just analogous as they applied the damage from debris impact, see

The global LS-DYNA model had the following input data:
  • Extent of initial damage to the building due to debris impact from the collapse of WTC 1 (Chapter 5).
  • Mechanical properties of steel (Appendix E and NIST NCSTAR 1-3D) and concrete (NCSTAR NIST 1-6A) used in the construction of WTC 7.
  • Temperature-dependent mechanical properties of steel (Appendix E and NIST NCSTAR 1-3D).
  • Temperatures of structural components and connections, at the time when the ANSYS results were transferred to the LS-DYNA analysis (Chapter 10).
  • Fire-induced damage to floor beams, girders, and their connections from the 16 story ANSYS analysis (Chapter 11).
Content from External Source
So you're suggesting you could apply the difference (i.e. the additional damage) between the 3.5h ANSYS damage and the 4h ANSYS damage. But ideal you'd do it for every difference along the way.

Well if they could input the debris impact damage, they should in principle be able to transfer the result from the 3.5h simulation.

The only reasons that i can imagine why they did not do it is that
  1. one maybe can't automatically take the data from the output and use it as input, but one is forced to do this by hand,
  2. they were time pressed to present the results and they didn't want to simulate for another 8 weeks.
You also can't apply it to every difference along the way for the simple practical reason that it would consume too much computing time.

It's just my interpretation of what I quoted from NIST: "Based on engineering judgment and preliminary analyses in LS-DYNA"

Sorry, I was looking for that quote only in Chapter 12 and missed it. Should have rather use the document search to find that phrase. It is in Chapter 11, p.531 (p. 597 pdf). The full sentence is:
Based on engineering judgment and preliminary analyses in LS-DYNA, which included the effects of dynamics caused by debris impact from failed floor sections, the level of failures, damage, and thermal weakening in ANSYS at 4.0 h was identified as likely to result in an initiating event in an LS-DYNA analysis.
Content from External Source
So if I understand that right (there is some ambiguity), the preliminary analysis included the effects from the debris impact only.

Because I can't imagine how to run a simulation with reduced complexity otherwise. You can't just make a coarser model by changing one parameter or slider, you need to actually model the building from scratch for a different graininess (can one say that?) for which you want to simulate it.

This preliminary run should then be equivalent to the initialization sequence for global model up to 6.5s (see Fig. 12-32). So an estimate for the runtime for that preliminary study would be that it took about two weeks to run at most. I do not know how the simulations actually scale, but I would guess it takes less computing time to simulate a time-step as long as the response of the system is not global, i.e. before a global collapse sets on. Probably it took some days.
 
Well if they could input the debris impact damage, they should in principle be able to transfer the result from the 3.5h simulation.

One problem is they would diverge after 3.5 hours. You have two states you'd need to merge:

1) 3.5h ANSYS + 30 seconds LS-DYNA
2) 3.5h ANSYS + 0.5h ANSYS . (i.e. the 4.0h ANSYS)

I think that time limitations probably were the more important factor. They could only get in so many full runs.
 
One problem is they would diverge after 3.5 hours. You have two states you'd need to merge:

1) 3.5h ANSYS + 30 seconds LS-DYNA
2) 3.5h ANSYS + 0.5h ANSYS . (i.e. the 4.0h ANSYS)

I think that time limitations probably were the more important factor. They could only get in so many full runs.

Ah yes, I got that one wrong. Of course it does not work then. Then one needs to input the state 1) into the ANSYS to let it run the next 30 min first. But for sure the data can't be transferred automatically (how should the ANSYS handle a fallen girder or multi-floor damage etc), and also the problem is that in the LS-DYNA all damage is again applied instantaneously and simultaneously will still be there. Ok, now I see why they didn't do it.
 
Last edited:
It's just my interpretation of what I quoted from NIST: "Based on engineering judgment and preliminary analyses in LS-DYNA"
Sorry, I was looking for that quote only in Chapter 12 and missed it. Should have rather use the document search to find that phrase. It is in Chapter 11, p.531 (p. 597 pdf). The full sentence is:
Based on engineering judgment and preliminary analyses in LS-DYNA, which included the effects of dynamics caused by debris impact from failed floor sections, the level of failures, damage, and thermal weakening in ANSYS at 4.0 h was identified as likely to result in an initiating event in an LS-DYNA analysis.
Content from External Source
So if I understand that right (there is some ambiguity), the preliminary analysis included the effects from the debris impact only.

Because I can't imagine how to run a simulation with reduced complexity otherwise. You can't just make a coarser model by changing one parameter or slider, you need to actually model the building from scratch for a different graininess (can one say that?) for which you want to simulate it.

This preliminary run should then be equivalent to the initialization sequence for global model up to 6.5s (see Fig. 12-32). So an estimate for the runtime for that preliminary study would be that it took about two weeks to run at most. I do not know how the simulations actually scale, but I would guess it takes less computing time to simulate a time-step as long as the response of the system is not global, i.e. before a global collapse sets on. Probably it took some days.

Again, too much speculation leads one on the wrong path. I should rather have checked all reports, and start reading at the abstract. In that case, the information about the preliminary analyses is in NCSTAR 1-9A:

Abstract:
p. iii (6 pdf)
The modeling approach began with collapse analyses of a multi-floor subassembly, leading to a global collapse analysis of WTC 7 using a model of the entire building. The floor model was used to test modeling methods and to test initiating event hypotheses.
[…]
Several load cases were considered with the multi-floor subassembly including damage effects from heating alone and with additional imposed connection failures. Most subassembly analyses were performed with a two floor model, however, a larger 14 story subassembly was used to test the effect of debris damage application and load redistribution within the structure.
Content from External Source
The effort for modeling isn't that tremendeous, as they use an automated TrueGrid Preprocessor method (p.19 / 70 pdf)
The WTC 7 model was built parametrically with TrueGrid in a semi-automated fashion. TrueGrid uses a command driven hierarchy that can be saved and organized into files. The model was organized much like a set of drawings with components feeding into subassemblies that were controlled by higher level subassemblies and finally by a top assembly. The same files could be used to generate the entire global model or a subset of the model, as in the two floor subassembly. Model generation was performed floor-by-floor in single floor segments. The user specified which floors to generate in the top level assembly file. For the global model, all floors were specified. Figure 3-1 illustrates the file hierarchy used in the TrueGrid model of WTC 7.
Content from External Source
So this is quite neat.

The resulting two floor model is this (Chapter 3.5):
twofloormodel.png Apparently, the loaded this with the full gravitational load of the remaining upper floors, as this plot suggests
gravity_twofloors.png
They did then in Chapter 4.2 different studies of the response of the structure to different scenarios, like column removal, thermal loading only, thermal loading and manually removed girder supports at columns 79 or 81. I found the following case interesting where they observed a collapse just due to temperature, however, in the middle region of the floor.
damage_loadcaseA_twofloors.png
This is also a hint that a collapse of the region around column 79 is not necessarily the only possible trigger event for a global collapse scenario. To be clear, by that I mean that had NIST modelled the connections around column 79 stronger or incapable of girder walk-off, there surely would be other regions that could trigger a collapse, albeit maybe later n time (here 5h instead of 4h collapse as in the main study).

Also of interest is the following cross-check of thermal induced damage calculated in LS-DYNAcrosscheck_damage_twofloor.png
I do not really get why they did not calculate these completely in LS-DYNA, but probably either problems with the coarseness of the model or with the computational effort to get results in reasonable time.

Finally, they did a study of the debris impact on a 14 floor model, as well as a study of the effect of column support removal for column 79.14floor.png
 
So you're suggesting you could apply the difference (i.e. the additional damage) between the 3.5h ANSYS damage and the 4h ANSYS damage. But ideal you'd do it for every difference along the way.
Well if they could input the debris impact damage, they should in principle be able to transfer the result from the 3.5h simulation.
Here is how they applied the debris damage NCSTAR1-9A p.54 (105 pdf):
After gravity initialization, debris impact damage from the collapse of WTC 1 (NCSTAR NIST 1-9, Section 5.5.3) was applied to the global model instantaneously to approximate the dynamic event. The damage applied was isolated to two zones on the southern side of the building, as shown in Figure 3-52. In the LS-DYNA calculation, the damage was applied during a simple restart where a list of elements to
delete was specified. The list of elements to delete was generated using LS-PREPOST by blanking out the damaged areas.
Content from External Source
debrisimpact.png
So they simply deleted the elements that were observed to be damaged, thereby assuming they fell out of the building completely. This you can't assume, of course, for internal damage. So the way they applied the debris damage is not feasible for the damage the got from the ANSYS, for which they used a different method subsequently.
 
So they simply deleted the elements that were observed to be damaged, thereby assuming they fell out of the building completely.

I think it's more that the damage happened hours before the collapse, so any loose girders etc did not contribute dynamically to the initial collapse. There were probably some damaged elements laying or hanging inside, but since they could not know where they were, then the conservative approach was to delete them.
 
Rebuttal to comments in “Some Problems with the UAF/Hulsey/AE911Truth WTC7 Draft Report”

Source: https://www.youtube.com/watch?time_continue=793&v=7OClixCTdDw&feature=emb_logo


11:40 “Next point. Why focus on girder A2001 collapse when NIST did not use that in their global collapse analysis.”
Content from External Source
Incorrect. NIST most certainly did use the collapse of the A2001 girder collapse in their global collapse analysis. (as you acknowledge at 13:08) According to NIST, the collapse of girder A2001 on the 13th floor triggered a cascade of floor failures which left column 79 without horizontal support. Without the failure of A2001 on the 13th floor column 79 would not have lost horizontal support and buckled. And there would have been no progressive collapse.

11:50 “Now, there’s a lot of focus on this one connection, girder A2001 to columns 79 and NIST suggested that the expansion of beams next to it pushed it off its seat and that was the collapse initiation event. (unintelligible) suggest this as the probable collapse initiation event.”
Content from External Source
The whole report is about the “probable” cause of the collapse so that is misleading. There were no other “probable” collapse initiation events.
In the Tech briefing, Shyam Sunder uses the phrase "probable collapse sequence" nine times.
http://911speakout.org/NIST_Tech_Briefing_Transcript.pdf

12:09 “But NIST didn’t actually use that in their dynamic analysis.”
Content from External Source
Incorrect: As noted above.

12:47 “So as I play this you will see nothing has really happened with A2001. It’s way back there. This girder has collapsed, this girder has collapsed, girder has collapsed, this girder has collapsed, girders collapsed over here and damage down here. These are all starting to go. And if we continue to play,
13:08 now, now we see that particular girder, A2001, actually collapse at this point. So girder A2001 isn’t used in the global model, in terms of the initiating event.”
Content from External Source
NIST has the failure data and you don’t.

And they say that it was the failure of A2001 on the 13th floor that triggered the cascade of floors that left column 79 horizontally unsupported in 3 directions over 9 floors. Even though there were other girder and floor failures, they did not leave column 79 horizontally unsupported in 3 directions over 9 floors.

13:22 “It just something that will fall because all the girders fall at some point. What really happened is that other girders fell around it first.”
Content from External Source
The failure of girders to the south of (not around) A2001 on floor 13 did not cause it to fail. According to NIST, A2001 was pushed off of its seat by thermally expanding beams.
 
Last edited:
Incorrect. NIST most certainly did use the collapse of the A2001 girder collapse in their global collapse analysis. (as you acknowledge at 13:08) According to NIST, the collapse of girder A2001 on the 13th floor triggered a cascade of floor failures which left column 79 without horizontal support. Without the failure of A2001 on the 13th floor column 79 would not have lost horizontal support and buckled. And there would have been no progressive collapse.

You are misunderstanding what I'm saying, and I'm not really sure why. Watch that section again and look at the green girder.

The green girder is A2001.

The green girder, in NISTS global collapse analysis, does not start to fall until AFTER the floors around it are collapsing - triggered by the failure of several other similar sized girders.
 
Assuming the area around col 79 was engulfed in flames there is no way to know what "failed" first. Why not the beams supported by the girder leading to the loss of bracing of the girder and weakening it?
 
You are misunderstanding what I'm saying, and I'm not really sure why.
I understand that you are trying to imply that A2001 on floor 13 was not the event that “triggered” the total collapse of the building. According to NIST, it was. And they have the failure data.

Watch that section again and look at the green girder.

The green girder is A2001.
I have studied the video with A2001 on floor 13 in green. (Nice work. My compliments.)

While it is true that other failures occurred before A2001 on floor 13, they did NOT cause A2001 on floor 13 to fail. That failure was supposedly caused by the thermal expansion of the floor beams to the east of that girder which supposedly pushed it off of its seat. That is the NIST explanation of the “trigger” event (actual initiating event) that led to what they call the initiating event, the buckling of column 79.

The green girder, in NISTS global collapse analysis, does not start to fall until AFTER the floors around it are collapsing - triggered by the failure of several other similar sized girders.
The floors “around” A2001 on floor 13 did not collapse before it collapsed.

The floors to the South of column 79 on floor 13 collapsed due to the failure of girders on the 14th floor which started a cascade of floor failures to the South of column 79. But the floors and girders to the west of A2001 on the 13th floor did not collapse until it failed. And the floors to the east of A2001 did not collapse on the 12th-8th floors until column 79 buckled.

NIST says that the failure of A2001 on floor 13 set of a cascade of floor failures that left column 79 unsupported in three directions. At that point it only had lateral support in the north direction.

Since the collapse of floor 14 caused the collapse of the floors to the south of column 79, and the floors to the east of column 79 did not collapse on floors 12-8 until column 79 buckled, then all that’s left is the floors and girders to the West of A2001 of floor 13 that could have collapsed due to the failure of A2001 on floor 13. This, according to NIST, is the last straw that left column 79 unsupported in three directions over nine floors – and it buckled.

NIST’s scenario depends on A2001 on the 13th floor triggering the “initiating” event. Without that failure, there would be no global collapse.
 
...
While it is true that other failures occurred before A2001 on floor 13, ...
In the context of the Hulsey draft report, which is the context, the topic currently, this is all the facts we need.

It is clear, and you acknowledge, that, in the NIST simulation, WTC7 had suffered multiple structural failures from the fires by the time A2001 on floor 13 started to fall.

Once we have this down, any analysis of A2001 and C79 that does not take into account the state of the surroundings (other girders framing into C79, floor slabs all around, situation on adjacent floors) is necessarily incomplete and invalid.

I think NIST's approach can be criticized, even doubted. They apply all the connection failures from ANSYS at the same moment - that's unrealistic, of course. Failures occurred over the course of quite some time. Ideally, each time a total vertical connection failure occurred, they should have stopped the ANSYS simulation, imported the entire ANSYS situation into LS-DYNA, let LS-DYNA run its course until all motion arrests, then import the resulting LS-DYNA situation back to ANSYS.
Better yet, return to the fire modelling with the LS-DYNA result, etc

Loop and repeat this circle with every new vertical failure observed in ANSYS, until either global collapse has occurred, or all fires are out and steel is cool.

Obviously, such mutual feedback looping between the various modelling devices would have been exceedingly complex, and most likely computationally infeasible (and far too expensive).

So NIST settled on a somewhat simplified strategy - let damage accrue in ANSYS until enough has accumulated to result in LS-DYNA collapse progression.

This is then NOT (exactly) what happened in reality, but perhaps a "good-enough" approximation.


In that view, as was as in reality, the collapse was not the result of one and only one connection failure, but of a multitude of accrued failures. Perhaps (in hindsight, with a view to accommodating Truthers) it was not smart by NIST to pick out one failure as THE one.


The Truth Movement for a long time has put blinders on and focused solely on a single connection - and implicitly assumed all the rest of the structure to be pristine.

Hulsey has followed that pattern. It's one of the reasons why his report FAILS completely all its purported objectives.
 
I understand that you are trying to imply that A2001 on floor 13 was not the event that “triggered” the total collapse of the building. According to NIST, it was. And they have the failure data.
I'm not "implying" anything. I try not to imply. I try to clearly state what I want to convey.

What I'm conveying is that in the LS-DYNA global simulation, the one that NIST gave us as the end result of their investigation, the push-off of A2001 is not what makes the building collapse. We don't know what happened in real life (and we probably never will, as there are unresolvable unknowns, like exact fire spread patterns), but we know what the result of this particular run of the LS-DYNA model shows us.
I have studied the video with A2001 on floor 13 in green. (Nice work. My compliments.)
The idea for this came from a video that @gerrycan posted years ago. Seeming to show how bad the NIST simulation was (which seems rather ironic now, given Hulsey's rubber-band model canned animations).

I then did this more detailed version. I encourage you to single-step through it (pause, then use the < and > keys) at around 0:02 where A2001 is first visible. (Note there are lots of repeated frames, as I uploaded it in a higher frame-rate than the original, so you need multiple presses to advance)

I have also attached the original .MOV file I made, which you might find easier to scrub trough
https://www.metabunk.org/attachments/tracking-a2001-full-height-context-mp4.39008/

If you look at A2001 when if first become visible, you will see three things.
  1. The end of the girder attached to C79 does not fall.
  2. The middle of the girder bows downwards due to applied fire damage (removal of segments)
  3. The fall of A2001, and the floor area attached to it, is arrested by Floor 12. i.e. it's not, in this simulation, causing the floor stripping that leads to C79 buckling over floors 5-14
Tracking-WTC7-Girder-C44-79-at-Floor-13-more-CROP-PAUSE.gif

So I think it is very clear that, in this simulation, the A2001 push-off is not occurring, and hence not causing the collapse.

A similar discussion (with you, about two years ago) around the focus on this connection by both NIST and AE911 can be found here:
https://www.metabunk.org/wtc7-is-ae...f-it-was-not-key-in-nists-global-model.t9427/
 

Attachments

  • Tracking A2001 Full height context.mp4
    20.7 MB · Views: 641
Last edited:
While there undoubtedly were many "failure processes" ongoing... there would / could be one failure which was the last one before a part of the building descent was observed. How many failures would it take for the EPH movement to be observed? Surely it had to include all the axial supports on the floors below the EPH... But how far down were those failures to produce the observation of the EPH descent? And how many column lines had to fail?
 
What I'm conveying is that in the LS-DYNA global simulation, the one that NIST gave us as the end result of their investigation, the push-off of A2001 is not what makes the building collapse.
According to NIST, the push-off of A2001 on the 13th floor is what triggered the cascade of floor collapses that left column 79 unsupported horizontally in three directions over 9 floors, causing it to buckle. And that led to the total collapse of the building.

You are claiming to know better than NIST what triggered the cascade of floor collapses by looking at a video simulation. But they have the failure data and you don’t so your assumption is not valid.

[...]

I then did this more detailed version. I encourage you to single-step through it

I did just that with the original version and I have a file of every frame so I can look at it frame by frame. A2001 falls between frame 14 and 15 (by my count).
Frames 14 & 15.png
It is hard to tell from looking at these frames whether or not A2001 was pushed off its seat. However, I agree that it looks like A2001 failed due to sagging and not thermal expansion of the floor beams.

If you look at A2001 when if first become visible, you will see three things.
1. The end of the girder attached to C79 does not fall.
Incorrect. It does fall. (see Frame 15 above)
2. The middle of the girder bows downwards due to applied fire damage (removal of segments)
Correct
3. The fall of A2001, and the floor area attached to it, is arrested by Floor 12. i.e. it's not, in this simulation, causing the floor stripping that leads to C79 buckling over floors 5-14
Incorrect - The floor section to the East of column 79 is arrested at floor 12, BUT the floor section to the West of column 79 sets off a cascade of floor failures down to the 5th floor, taking out the girders to the West of column 79, leaving it unsupported on three sides.

So I think it is very clear that, in this simulation, the A2001 push-off is not occurring, and hence not causing the collapse.
I agree that it appears that way but it cannot be said that it is clear. However, we agree that that A2001 was not pushed off of its seat.
 
Last edited by a moderator:
Back
Top