The 9/11 Forum

I guess the mpg is a direct stream cut from the .vob, it is good enough. The vob is (as far as I know) compressed mpeg2

Additional Comments by F.R. Greening on NIST’s Global Collapse
Analysis in Chapter 12 of NCSTAR 1-9:

In Chapter 12 of NCSTAR 1-9 NIST describes the main findings of its computer modeling of the collapse of WTC 7. Some of the key results of this modeling are also presented in Figures 12-48 to 12-69 to be found on pages 571 to 595 of Chapter 12.

In Section 5.0 of the Comments I submitted to NIST September 11th, 2008, I addressed some of my concerns with the material presented in Chapter 12 of the Draft Report. Here I wish to consider NIST’s core collapse sequence in more detail especially the results illustrated by Figures 12-48 et seq. and Figures 12-66 et seq. in Chapter 12 of NCSTAR 1-9. These Figures show the vertical progression of core column failures on the east side of WTC 7 that, according to NIST, led to the collapse of the East Penthouse. NIST states that the WTC 7 core collapse started with the buckling of column 79 about 14 seconds into the computer simulation, and was followed by the descent of the East Penthouse below the roofline of WTC 7 at about 16.5 seconds into the simulation.

According to NIST, the global collapse of WTC 7 began 6.9 seconds after the East Penthouse collapse or about 23 seconds into the simulation. Now consider NIST’s Figures 12-66, 12-67 and 12-69 and in particular the images showing the alleged state of the core 17.5, 19.5, 20.7, 21.8, 24.1, 26.8 and 28.8 seconds into the collapse simulation. These images represent NIST’s view of what the core looked like at ~ 1 - 2 second intervals following the collapse of the East Penthouse. What is most significant about these images is that around the time of global collapse initiation NIST’s simulation shows that the eastern half of the core had completely collapsed while the western half of the core remained standing and relatively undamaged. This is quite remarkable since videos of the collapse of WTC 7 show that up to and well beyond the moment that the roofline of WTC 7 exhibited its first downward movement, the exterior of the building revealed absolutely no signs of NIST’s proposed partial collapse of the core even though the core was connected to the exterior walls of Building 7 by dozens of horizontal beams on every floor.

NIST’s proposed collapse of the eastern half of the core would have completely removed the lateral restrains normally acting on the eastern exterior columns of WTC 7. Indeed, NIST assert that in the moments before global collapse initiation, “the exterior façade on the east quarter of the building was just a hollow shell.” This would have caused the eastern façade to buckle well before global collapse ensued. This buckling would have been visible as a bowing of the northeast corner of the building. Needless to say, such pre-collapse buckling or bowing of WTC 7 was not observed. However, the problems with NIST’s simulations only get worse after global collapse initiation. Thus Figure 12-68 of NCSTAR 1-9 shows that the western end of the WTC 7 core only started to collapse about 25 seconds into NIST’s simulation, a full 2 seconds after NIST claims that global collapse was underway.

NIST states in NCSTAR 1-9 that 24.5 seconds into its simulation, the roof of WTC 7 was falling with a velocity of approximately 10 to 15 m/s, while Figure E-4 of NCSTAR 1-9A shows that the roofline of WTC 7 had descended about 3 stories at 24.6 seconds of the simulation, or 1.6 seconds into global collapse. These Figures are in complete contradiction to Figure 12-68 of NCSTAR 1-9 which shows a computer-generated image of WTC 7 taken 24.1 seconds into NIST’s collapse simulation with the eastern end of the core completely collapsed, but the western end still standing, essentially undamaged, to within 0.5 meters of its full 186 meter height.

The images 26.8 and 28.8 seconds into NIST’s simulation, or 3.8 and 5.8 seconds after the initiation of global collapse, deliver a final blow to the credibility of NIST’s collapse simulation. Thus the 26.8 and 28.8 second images included in Figure 12-68 reveal a collapsing core with its eastern side a full eight stories, or about 32 meters, below its western side. This would indicate a roofline collapse that started at the eastern end of Building 7 and progressed over a period of about 4 seconds to the western end.

Now, if NIST’s collapse simulations are supposed to accurately reflect what happened to Building 7 on 9/11, one is compelled to ask:

Why did WTC 7 undergo a strictly vertical collapse, with the roofline remaining essentially horizontal throughout the first 5 seconds of its downward motion, when NIST’s simulations show the eastern side of the building starting to collapse 4 seconds before the western side?

I'd not much time (theme park and stuff...) It is strange, the old video was larger than this one, it has been enlarged unless the source was
not from a dvd (the camera must be PAL or SECAM), but it looked crappy anyway.

Now we have (from a frame from the video) 226 pix is 15 stories, i.e.



The total drop we have in the video (at the position the drop is measured) is 235 pixels

I use a video with 50fps using the bob doubler filter, DEMO:

http://rapidshare.com/files/145225184/b ... O.avi.html (only 64k, xvid codec), or http://www.youtube.com/watch?v=bh7xLNqJ-XM

Here everything is slowed down, the left one is genuine interlaced video, the right part is
after using virtualdub's bob doubler filter. The reason that all frames are unique is because a single
frame contains information from 2 shots, the 1,3,5,... lines and 2,4,6,.. lines differ 1/50 second.
We see we can double the framerate, but this means that the error vertically wil become 2 times bigger.
Error analysis will not be done now in this post.

And here is the 50 frames per second movie, that is used to create the smeorograms:

http://www.megaupload.com/nl/?d=FGQPONV3

Smearograms



reduced to black and white (with nearest colour, which is a RUDE method)



cropped, rotated, mirrored:



A better analysis of the smearogram should be done, because the noise destroys the real motion.

I used an old simple script to get the data in real time/distance (using an other primitive method that converts
bitmaps into ascii...)

Drop with an offset (meter),t (sec)
------------------------------------
0.02,3.19431368857089
0.04,2.94859725098851
0.06,2.94859725098851
0.08,2.94859725098851
0.1,3.19431368857089
0.12,3.19431368857089
0.14,3.19431368857089
0.16,3.19431368857089
0.18,3.19431368857089
0.2,3.19431368857089
0.22,3.19431368857089
0.24,3.19431368857089
0.26,3.19431368857089
0.28,2.94859725098851
0.3,3.19431368857089
0.32,2.94859725098851
0.34,2.94859725098851
0.36,2.94859725098851
0.38,2.94859725098851
0.4,2.94859725098851
0.42,2.94859725098851
0.44,2.94859725098851
0.46,2.94859725098851
0.48,2.94859725098851
0.5,2.94859725098851
0.52,2.94859725098851
0.54,2.94859725098851
0.56,2.94859725098851
0.58,2.94859725098851
0.6,2.94859725098851
0.62,3.19431368857089
0.64,3.19431368857089
0.66,3.19431368857089
0.68,3.19431368857089
0.7,3.19431368857089
0.72,3.19431368857089
0.74,3.19431368857089
0.76,3.19431368857089
0.78,3.19431368857089
0.8,3.19431368857089
0.82,3.19431368857089
0.84,3.19431368857089
0.86,3.19431368857089
0.88,3.19431368857089
0.9,3.19431368857089
0.92,3.19431368857089
0.94,3.19431368857089
0.96,3.19431368857089
0.98,3.44003012615327
1,3.44003012615327
1.02,3.44003012615327
1.04,3.44003012615327
1.06,3.44003012615327
1.08,3.44003012615327
1.1,3.68574656373564
1.12,3.68574656373564
1.14,3.68574656373564
1.16,3.68574656373564
1.18,3.93146300131802
1.2,3.93146300131802
1.22,3.93146300131802
1.24,3.93146300131802
1.26,4.1771794389004
1.28,4.1771794389004
1.3,4.1771794389004
1.32,4.42289587648277
1.34,4.66861231406515
1.36,4.66861231406515
1.38,4.66861231406515
1.4,4.91432875164752
1.42,4.91432875164752
1.44,5.1600451892299
1.46,5.1600451892299
1.48,5.1600451892299
1.5,5.40576162681228
1.52,5.65147806439465
1.54,5.65147806439465
1.56,5.89719450197703
1.58,6.14291093955941
1.6,6.14291093955941
1.62,6.38862737714178
1.64,6.38862737714178
1.66,6.63434381472416
1.68,6.88006025230653
1.7,7.12577668988891
1.72,7.37149312747129
1.74,7.37149312747129
1.76,7.86292600263604
1.78,8.10864244021841
1.8,8.10864244021841
1.82,8.35435887780079
1.84,8.60007531538317
1.86,8.84579175296554
1.88,9.09150819054792
1.9,9.3372246281303
1.92,9.58294106571267
1.94,9.82865750329505
1.96,10.0743739408774
1.98,10.5658068160422
2,10.5658068160422
2.02,10.8115232536246
2.04,11.0572396912069
2.06,11.3029561287893
2.08,11.5486725663717
2.1,12.0401054415364
2.12,12.2858218791188
2.14,12.5315383167012
2.16,12.7772547542836
2.18,13.2686876294483
2.2,13.5144040670307
2.22,14.0058369421954
2.24,14.2515533797778
2.26,14.7429862549426
2.28,14.7429862549426
2.3,15.2344191301073
2.32,15.7258520052721
2.34,16.2172848804368
2.36,16.4630013180192
2.38,16.7087177556016
2.4,17.2001506307663
2.42,17.2001506307663
2.44,17.6915835059311
2.46,18.1830163810958
2.48,18.1830163810958
2.5,18.4287328186782
2.52,19.1658821314253
2.54,19.1658821314253
2.56,19.6573150065901
2.58,19.9030314441725
2.6,20.1487478817548
2.62,20.3944643193372
2.64,20.885897194502
2.66,21.1316136320844
2.68,21.3773300696667
2.7,21.8687629448315
2.72,22.1144793824139
2.74,22.3601958199962
2.76,23.0973451327434
2.78,23.0973451327434
2.8,23.8344944454905
2.82,24.0802108830729
2.84,24.5716437582376
2.86,24.81736019582
2.88,25.3087930709847
2.9,25.5545095085671
2.92,26.0459423837319
2.94,26.5373752588966
2.96,27.0288081340614
2.98,27.2745245716438
3,27.5202410092261
3.02,28.0116738843909
3.04,28.5031067595556
3.06,28.748823197138
3.08,29.2402560723028
3.1,29.7316889474675
3.12,29.9774053850499
3.14,30.4688382602147
3.16,30.9602711353794
3.18,31.4517040105442
3.2,31.6974204481265
3.22,32.1888533232913
3.24,32.680286198456
3.26,33.4174355112032
3.28,33.6631519487855
3.3,33.9088683863679
3.32,34.646017699115
3.34,35.1374505742798
3.36,35.6288834494446
3.38,35.8745998870269
3.4,36.6117491997741
3.42,37.1031820749388
3.44,37.5946149501036
3.46,38.0860478252683
3.48,38.8231971380154
3.5,39.0689135755978
3.52,39.5603464507626
3.54,40.2974957635097
3.56,40.7889286386744
3.58,41.2803615138392
3.6,41.771794389004
3.62,42.5089437017511
3.64,42.7546601393335
3.66,43.4918094520806
3.68,43.9832423272453
3.7,44.4746752024101
3.72,44.9661080775748
3.74,45.4575409527396
3.76,46.1946902654867
3.78,46.6861231406515
3.8,47.1775560158162
3.82,47.668988890981
3.84,48.1604217661457
3.86,48.8975710788929
3.88,49.3890039540576
3.9,49.63472039164
3.92,50.6175861419695
3.94,51.1090190171343
3.96,51.600451892299
3.98,52.3376012050461
4,52.5833176426285
4.02,53.3204669553756
4.04,53.8118998305404
4.06,54.3033327057051
4.08,55.0404820184523
4.1,55.531914893617
4.12,56.2690642063641
4.14,56.7604970815289
4.16,57.497646394276
4.18,57.9890792694408
4.2,58.4805121446055
4.22,59.2176614573527
4.24,59.7090943325174
4.26,60.2005272076822

If I use Maple's least squares (for ALL points) it gives

4.138242015-5.34017617274030254*t+4.37629473584810834*t^2

but there it tries to fit a parabola which cannot because in the beginning
there is no movement and at the end the acceleration decreases, but even with this the value is 8.75 m/s^2

To be continued.

Einsteen:

Just had a quick look at your post.

Nice work!

I think it's best to go for a 3rd order polynomial to get the acceleration as a function of t.

Anyway, I need to look at your data more closely too.....

I've tried even a kind of stepwise spline thing, nice curves but the derivatives are useless.

A 3rd order fits well, which gives

4.316434217-5.83400698486097369*t+4.66407283963638352*t^2-0.448252498112575102e-1*t^3;

Higher ones sometimes fit better in the beginning but have a more shaky derivative. The derivative of the above is relatively straight, see

http://i34.tinypic.com/2yoal2a.gif

The 2nd derviative is

g := 9.328145680 - 0.2689514988 t

I think maple doesn't have the option to fit a function

c1*(t-c2)^2+c3 for t>c2
c3, for 0<t<c2

That must be the one, although how much do you have to take from the parabola, you need more points for a good fit but more points also are going to differ from the parabola. Although it is high school physics the fine points are not really trivial.

I've asked DBB to have a look at it also.

Great stuff.

Probably already known but I just want to state explicitly:

You must try to fit to a third order or higher polynomial.

A second order fit obviously assumes your answer will be

a(t)= constant

so you've predetermined your answer in the choice of the polynomial.


You are obviously assuming your answer will take the form

a(t)= ct+d

if you choose to fit to a third order polynomial.

In a perfect world you would fit to a polynomial where the coefficients of higher order terms would prove insignificant, but the world is not perfect.

For your answer to have the accuracy you desire you must at least choose a third order polynomial.

c1*(t-c2)^2+c3 for t>c2
c3, for 0<t<c2

That must be the one, although how much do you have to take from the parabola, you need more points for a good fit but more points also are going to differ from the parabola. Although it is high school physics the fine points are not really trivial.

Great stuff, Einstein. I could be wrong but I'd give up on trying to make the answer in the form of a palabola. Not enough freedom for a(t).

My guess is a=constant is wrong.

a(t)=ct+d is a pretty good try (third order x(t))

a(t)=ct^2+dt+k would be a more colorful solution (fourth order x(t)), but would a quadratic term in a(t) serve a purpose to further our knowledge?

Thanks MT,

Yeah, a parabola will always lead to a constant acceleration, but the 3rd order I fitted is a little bit too fast in the beginning I think, but that's the law of conservation of trouble... Do you have any idea how error analysis works for the current method ?

Einsteen:

I have been looking at your data a little more closely and have done some "eyeballing" of what your numbers are telling us. First it looks like the "noise level" in the vertical direction is about 0.3 meters. This is very reasonable since I previously estimated an uncertainty of about 0.5 meters in the measurement of heights for this video.

I would put t(0) at about 0.6 seconds in the table of numbers you posted, and therefore subtract 0.6 seconds from your times; this also implies a vertical offset of about 3 meters, so I would subtract 3 meters from all the height data too. Working with this you would expect to see no motion before 0.3 seconds (your point at 0.9 seconds) because it would be within the noise level. Also, interestingly, a drop of 0.3 meters at 0.3 seconds implies an acceleration of 6.7 m/s^2. If you then consider the drop at 0.5 seconds it would be 0.7 meters (the first real data point) which would give an acceleration of 5.6 m/s^2. (Note, however, that if we move t(0) just 0.05 seconds later, to 0.65 instead of 0.6, we would have 0.7 meters at 0.45 seconds in which case the acceleration would be 6.9 m/s^2.)

At 1 second into the collapse we have a drop of 3.4 meters and an acceleration of 6.8 m/s^2. Such a drop distance would be equal to almost one floor height so this could be considered as the collapse initiation period, .... however, after this, the acceleration picks up. Thus your data at 1.8 to 2.2 seconds gives an acceleration ~ 9 m/s^2.

These "eyeballed" estimates do of course need refining and proper curve-fitting techniques should be used. Nevertheless, these are my preliminary thoughts on your excellent new smearograms and I suspect a more rigorous approach will not change things very much.

Einsteen:

i will upload the VOB part i cut out. cause i used another program to cut it properly, maybe that programm messed it up. I used Pinnacle Studio 10 to cut the MPEG properly, so only the WTC7 clip is uploaded.

I am sure you know alot more about video cuting and stuff than i do. so i will just use the VOB cutprogramm you gave me to cut the VOB, so you can do the detail cut urself and maybe not loose quality etc.

but already on the DVD the quality of the WTC7 clip looks diffrent than the rest of the film, the rest is much better quality. Dunno why, i thought that clip was recorded by one of the Naudet brothers.

anyway, as soon i am at home, i will upload it.

Dr. G.

Good comments. I guess NIST has still alot of work to do with their report
or alot to ignore

Excellent work (as always), einsteen!

Dr. G, do I understand correctly from your last post that you are seeing evidence in this data of a 'collapse induction' period?

OneWhiteEye:

Well I have given up "eyeballing" Einsteen's data and started doing some real curve fitting. As a result of trying a whole bunch of "best fits" I am now unsure about a discernable "induction period"! It now looks to me that a 0.3 meter resolution creates a problem in getting any sort of meaningful descent curve over the first 0.5 seconds of the collapse and, besides, this takes us right back to the old t(0) problem. However, I would say that Einsteen's data give a series of very good fits to many different 2nd or 3rd order polynomials as t(0) is varied by about 0.5 seconds. (I have even tried varying h(0), the initial drop height, by up to 0.3 meters, since this cannot be measured to better than about 0.3 meters. These small shifts all make a difference.....)

One final comment: The fact that ONE curve, whatever that curve may be, always gives a good fit does suggest that the collapse of WTC 7, once started, was driven by a relatively constant accelerating force. That's about all I can say for now ............

I didn't spent much time on fitting curves, there is a lot to say about it, as I can see...

- Perspective correction will matter (however I don't think it will decrease the initial acceleration because we are looking at smaller stories when the camera is at the base.
- the data should also be used from several sources, the other movie in higher resolution.
- the exact height is relevant, 186m vs 174m, the roof should be subtracted from the height and maybe even a base.
- the point taken to follow the top is relevant, if the penthouses fall first then these are relevant (it gives the building an extra start), although NISTs acceleration used the roofline and that's where it is about.
- error analysis etc.
- a method to follow the curve, like O.W.E. did for wtc1 would be nice
- etc etc

DC,

The cutting tool I uploaded is a direct stream cutter and doesn't recompress it, it should be the pure dvd info. A new video from the dvd is welcome but I'm afraid it will not differ much.

Dr. G wrote:[...] One final comment: The fact that ONE curve, whatever that curve may be, always gives a good fit does suggest that the collapse of WTC 7, once started, was driven by a relatively constant accelerating force. That's about all I can say for now ............

How splineless

Over at 911Blogger.com I see that Richard Gage and his A&E Group have posted their submission to NIST, (my, oh my, NIST has really got some explaining to do!); ... the submission includes a section on the collapse time, ...... Looks like everyone has a beef with that one...

Anyway, the post includes an image that is essentially a very short smearogram. It shows the t(0) problem quite well and I would even suggest that the A&E guys have taken a t(0) that is about 2 or 3 frames late, thus shortening the collapse time ... It is interesting that NIST take their t(0) too early and the A&E guys take it too late - could each side's bias be showing a little here!

In the A & E smearogram I would go for the 75 or 76th frame as a better t(0). However, since one frame is only 0.033 seconds long, this adds a mere 0.1 seconds to their time making it 3.97 seconds to drop let's say 73 meters. This gives an average acceleration of 9.26 m/s^2, which is still amazingly fast, and pretty darn close to most other (non-NISTIAN) esimates I have seen!

I will be very, (VERY), disappointed if the NIST don't step up to bat on this one and give us its best collapse descent curve ......

If not, well Max, I would call that SPLINELESS!

Which height did AE use, 174m or 186m ? We should not forgot that the temporary data we found here is based on the first one, which is in favor of a slower collapse.