1. Mick West

    Mick West Administrator Staff Member


    Last year I created a physical model to illustrate some aspects of the WTC Tower collapses. The basic structure was columns connected by floors. The floors supply lateral stability to the columns, so when the floors are removed the columns are very unstable. This worked out reasonably well, and I got this nice demonstration from it:


    Source: https://www.youtube.com/watch?v=flo62pdaIMI

    I'm currently re-creating my physical model as a virtual model using Blender physics to allow for a larger variety of experiments, and also to illustrate the difference that scale makes.

    The first version I tried quite closely resembles the first physical model


    The intent behind these models is to help people understand more about the actual mechanisms of the collapse speed of the Twin Towers. One of the reasons the 9/11 "Truth" movement has been relatively popular is the non-intuitive nature of the collapse where a smaller top block appeared to crush a much larger lower part at quite high speeds. This has led to some quite inaccurate "models" of the what happened, like this one from the organization "Architects and Engineers for 9/11 Truth"

    The problem is that these things ignore the scale of the building, the internal structure of the building, and how it actually collapsed.

    The collapse of the towers (once started) was a "zipper" progressive collapse. A wave of very fast falling debris stripped the horizontal members away from the columns, which eventually collapsed - rapidly in the case of the outer walls, and less rapidly in the case of the inner core. Any model of the collapse needs to reflect that reality, and that's what I'm trying to illustrate here.
    Last edited: Mar 31, 2017
    • Like Like x 3
    • Informative Informative x 2
    • Winner Winner x 1
  2. Mick West

    Mick West Administrator Staff Member

    An early version with some arbitrary values. The beauty of a virtual model is that the reset time is seconds, and you can easily modify the model and repeat the experiment
    Last edited: Mar 25, 2017
  3. Mick West

    Mick West Administrator Staff Member

    Like with my the physical model I've used assemblies of three floors. Here I just make the top assembly a bit smaller so it collapses onto the lower section.

    This is 18 assemblies, so 54 floors, half the number of floors of the towers themselves.
    Last edited: Mar 25, 2017
    • Like Like x 4
  4. NoParty

    NoParty Senior Member

    The things a bloke will do for 80 quid! We woulda chipped in, Mick, if times are hard...

    All kidding aside...how much work goes into the Blender model in post #3?

    Is doubling it to the actual floor count easy? Would one expect anything different in a 108 floor collapse?
    What are the variables you can tweek?
    • Funny Funny x 1
  5. Mick West

    Mick West Administrator Staff Member

    I've put about a day into it so far. I have practically zero experience with Blender or Python (the scripting language Blender uses).

    Here's 108 floors:

    Source: https://www.youtube.com/watch?v=MlKhIhTbzWM

    This is to scale, i.e. 1362 feet high, standard gravity. The beam on the right is in freefall and takes just under ten seconds to fall from the height of the tower.

    Several problems with this approach were revealed. The biggest seems to be Blender's physic being a bit too rigid. If you drop too much on the top of a column it tends to jolt free all the floors all the way down. In the real world there's some elasticity that prevents this.
  6. Mick West

    Mick West Administrator Staff Member

    Attached is the Blender Source.

    You'll need to run the script in script view, then run the animation (Alt-A in the view window)

    @Cube Radio, can I have my $100 now? :)

    Attached Files:

    • Funny Funny x 3
    • Like Like x 2
  7. Mick West

    Mick West Administrator Staff Member

    Some Technical Details

    Each assembly is made from two "columns" and three "girders". These are just Blender cuboids (created by scaling cubes). They are scaled so the distance between the floors is correct, but otherwise relatively arbitrary sizes.

    The weight of each element is calculated at 1/10th the volume of the its cuboid, as I'm guessing that's about the cross sectional area of actual steel in an I-beam or box column. Obviously there's a lot of variation in the real world, but here there's only two objects: columns and girders.

    The connections are the key. Each girder is attached to the columns on either side, and each column is attached to the one below it. These just use the Rigid Body Constraint in an empty object that's positioned at the center of one of the connecting objects. I'm not sure how important this position is. Here's an example connection, of the bottom left floor connected to a center column:

    The whole thing is created via script, like the creating and connecting of girders
    	for n in range(0,N):
    		girder1 = physicsBox(Vector((colSize/2+girLength/2,0,girZ))+offV,(girLength, girSize*depth*5, girSize))
    		girder1.name = "girder"
    		girZ -= spacing
    with connection creating helper functions
    def connect(a,b,strength=200):
    # try actually creating the connecton
    #	bpy.ops.object.empty_add(type='PLAIN_AXES', radius=1, view_align=False, location=(0,0,0))
    	bpy.ops.object.empty_add(type='PLAIN_AXES', radius=1, view_align=False, location=b.location)
    	connection = bpy.context.object;
    	bpy.context.object.rigid_body_constraint.object1 = a
    	bpy.context.object.rigid_body_constraint.object2 = b
    	bpy.context.object.rigid_body_constraint.use_breaking = True
    	bpy.context.object.rigid_body_constraint.breaking_threshold = strength
    	bpy.context.object.name = "MickConnection"
    	connection.hide = True
    def connect3(a,b,c,strength=200*2):
    The "strength" here is the value used by Renderware, but I don't know how (or if) this relates to the physical reality.

    I had to really bump up the steps per second with the taller towers, otherwise the large forces involved made things explode. So I have:
    bpy.context.scene.rigidbody_world.steps_per_second = 1000
    Where the default is 60.

    Things to do:
    • Investigate different constraints - breakable hinge spring for the columns?
    • What is Kai doing in his model? Can we use non-rigid bodies?
    • Would a different framework be better? Unity?
    • Render some nicer animations with lighting and textures
    • Is 3D doable with this approach?
    Last edited: Mar 22, 2017
  8. Cube Radio

    Cube Radio Member

    By all means. However my original remarks in leading to your $100 bet, which I then turned into a challenge to you, were these:

    "the fact remains that absolutely nothing has been done in terms of 3D modelling to resolve this question [of how the towers could have collapsed in the incredibly destructive way they collapsed]"

    ...and I was therefore seeking

    "a high-rise tower using a virtual model even approximately representing the twin towers"

    ...on the understanding that, in your case

    "the conditions of the build were agreed beforehand with @psikeyhackr and @aka"

    (Since @psikeyhackr has long been a stranger to these parts, perhaps @Tony Szamboti could be encouraged to "referee" along with @aka instead, and we can retroactively waive my requirement that the conditions of the build be agreed beforehand.)

    If your tower here can be made to fulfil my requirements to a reasonable degree, I will pay you $100 of my hard-earned cash -- preferably to a mutually agreeable charitable cause we can decide on later.

    I would at least require a third dimension and the input of the two "referees" I mention, though, so I will see if I can contact them now -- provided you deign to publish this post, as all my contributions are now moderated before they are allowed to appear.

    As an aside, I felt many of my posts in the previous thread were very unreasonably moderated into oblivion. :) So, although I will continue to pop in on occasion to see how you're all getting on -- and stand by my promise to pay you $100 if my requirements are reasonably fulfilled -- do not expect to see me often unless there are developments that take my interest here.

    I'll be out in the real world, earning you your $100 :)
    Last edited: Mar 23, 2017
    • Dislike Dislike x 1
    • Disagree Disagree x 1
    • Funny Funny x 1
  9. Mick West

    Mick West Administrator Staff Member

    So what's the simplest topology that you'd consider to be a 3D representation of the towers? You must have something in mind. What does it look like? Can you draw a diagram?

    My "2D" model consists of two walls and one floor per level. If we were to take this to the next level then we'd need at least something like five columns (four corner and one center). Then ideally we'd have girders that support floor panels.

    But draw me a diagram. What would convince you?
  10. Cube Radio

    Cube Radio Member

    This isn't a direct answer to your question, but as you've decided to pre-moderate my posts I'm not inclined to take things too seriously at the moment.

    Big Brother will get back to you: it may take a little while for me to contact my referees. I wouldn't want you to waste your effort. Certainly we'd need more columns, as you say.

    In the meantime, have you seen this? It's free, although made with lightwave and strictly a visualisation I think. I guess it represents something like what I'd hope for, although to make something similar in blender strikes me as a ridiculous amount of work, and I certainly wouldn't expect it of you -- especially as Kostack is already doing something like it. https://www.turbosquid.com/3d-models/free-world-trade-center-3d-model/517182

    However, this is another model that shows what can be simulated in a mere game.

    Source: https://youtu.be/UFIeI45k8H0
    • Funny Funny x 1
  11. Mick West

    Mick West Administrator Staff Member

    Looks like a regular grid internally, not at all like the WTC structure.

    The idea here is to get a model that will demonstrate the mechanism for the collapse of the WTC - i.e. the stripping of floors leading to the columns being unsupported. A regular grid like that is not really going to demonstrate that.
  12. Mick West

    Mick West Administrator Staff Member

    Source: https://www.youtube.com/watch?v=_S9VQ_7e2JI

    In this version
    • I've placed two sets of assemblages side by side, more like my physical garage version.
    • I've adjusted the strength of the connections so that the bottom is 10x as strong as the top.
    • The width is adjusted to the actual width of the tower
    • Collapse initiation is done on a lower level, allowing some tilt of the upper block
    Obviously the speed of the collapse is going to vary based on how strong you make the connections, but I really think this is a reasonable approximation of what happened.

    And again this is full scale model. Note the separate girder taking just under 10 seconds to fall the full distance.

    Blender file attached. The key vars beings:

    N=3  # Number of floors per assemblage
    F=3.77 # Spaceing of floors in m
    Assemblages=36 # number of assemblages, 36 for full tower with three floors per assembly
    # Strenght of the connection on the top floor
    #how much this increases as we get to the bottom (mutiplier, so 1.0 = no changes, 2.0=2x)
    girderStrengthScale = 10.0
    columnStrengthScale = 10.0
    colSize = 1  # column width in meters
    girderLength = 30 # towers were 61 m wide, so if we have two assemblies wide 30 is about right
    towers = 2

    Attached Files:

    • Like Like x 1
    • Agree Agree x 1
  13. Mick West

    Mick West Administrator Staff Member

    Regarding scaling. In the above I scale the strength of the connections, but not the weight of the beams and columns. This seems incorrect, as the horizontal members (i.e. floors, girders) carry similar loads on each floor, but the columns are what change.

    So also need to change the weight of the columns as they get closer to the ground - Presumably they would also have stronger connections. But not so much with the floors.
  14. NoParty

    NoParty Senior Member

    I'm guessing that--between innumerable conditions and goalpost moving--Mick's gonna have
    to put in more time to get that $100 than a 7 year-old Bangladeshi garment seamstress...
    • Funny Funny x 3
    • Informative Informative x 1
  15. Mick West

    Mick West Administrator Staff Member

    Attached Files:

  16. Cube Radio

    Cube Radio Member

    What kind of "goalpost moving" do you anticipate? Do you think it's unreasonable to attempt to build a 3D model?

    [off topic complaints removed. Deirdre]
    Last edited by a moderator: Mar 24, 2017
  17. deirdre

    deirdre Moderator Staff Member

    @Cube Radio please stop complaining that the software threw you into moderation mode for continuously violating posting guidelines. "Big brother" isn't out to get you. It's YOU who knowingly keeps breaking the rules. If you complain one more time I am going to ban you for a week until the software clears itself because frankly all your violations do is cause more work for mods.
    • Like Like x 1
  18. NoParty

    NoParty Senior Member

    I'm just speculating here, Cube, based on years o' internet experience...
    already a wee bit off topic, though, so I won't go further down that road.
    Prove my quip wrong with a prompt payment, and I'll be the first to admit my speculation was in error. :)
    • Like Like x 1
  19. Mick West

    Mick West Administrator Staff Member

    Moving towards a more three dimensional structure:

    With the addition of floor panels over the office areasI think this is the minimum that would satisfy the overall topology being similar enough to the actual towers

    Unfortunately it takes the number of elements in a three floor assembly from 10 (in my double wide model) to 112 (16 columns, 3x8 floors, 3x24 beams). The way Blender works this means a 100x time in model generation, and at least 10x time in physic simulation. So generating and running simulations on a full-height model will be time consuming. I'll get it working on a smaller model first.

    Attached Files:

  20. Spectrar Ghost

    Spectrar Ghost Senior Member

    Could you replace the center four columns with a 'core' consisting of four thickish plates? That would reduce your object count and better approximate the structure.
  21. Mick West

    Mick West Administrator Staff Member

    I had planned to just make the center columns stronger. But I might reconsider

    This only has a third of the beams, and no floors and still took about an hour to generate. The full model is probably looking at 5+ hours. The physics runs reasonably fast though.

    I might have to hack Blender a bit to speed it up. No reason for it to take that long. It's because it does a full scene update each time an object is added.
  22. Mick West

    Mick West Administrator Staff Member


    Added floor slabs
  23. Mick West

    Mick West Administrator Staff Member

    So in theory something like this:
    That's a mockup without any connections, if I run the physics on it it does this:

    Still it runs at 0.25 fps, not unmanageable. The problem is in generating the model with the right connections.
    • Like Like x 2
  24. Mick West

    Mick West Administrator Staff Member

    Source: https://www.youtube.com/watch?v=iGbYEHC-9I8

    Collapse of a 36 floor (12 assembly) model. Pretty much ad-hoc values for everything as it takes a long time between runs. Generating the model took over an hour, the physics took maybe ten minutes.


    Like in the other model the column all span three floors with the join being between floors. In the run above the columns are just held together with friction.
    Last edited: Mar 25, 2017
    • Like Like x 2
  25. H William Taylor

    H William Taylor New Member

    I have to admit, i am liking your dedication to this. Surprised this hasn't already been done. While i suspect it will take several weeks to get the structure right in Blender, I like what you have done so far
    • Agree Agree x 1
  26. Cube Radio

    Cube Radio Member

    Great work, Mick. Are you continuing with this? Either way, could you provide the .blend file for this latest model as you have with the others, please?
  27. Mick West

    Mick West Administrator Staff Member

    Got distracted with other things for a week, then I was fiddling with methods of speeding up the model creation, then looked into the much more accurate Bullet Constraint Builder which Kai Kostock has been updating on GitHub.

    I doubt that's going to speed things up. But might be useful for smaller models, or a really long run.


    The precise run in the video above is lost to history (other than the code in the image) as I never saved it (I now know it autosaves to /tmp/ but that's been flushed :( all very suspicious)
  28. Cube Radio

    Cube Radio Member

    What do you mean by "suspicious"?

    So there is absolutely no .blend file available for the most recent model? Do you expect me to pay you $100 without being able to examine and download that file?
  29. Mick West

    Mick West Administrator Staff Member

    It's a joke about how my losing the files might be seen as suspicious to overly conspiratorial people. Sorry.

    No, of course not. I expected you to pay $100 for the previous file, but you made clear you wanted something more 3D. I'd wait for a taller model.
    • Like Like x 1
  30. Cube Radio

    Cube Radio Member

    Haha. At least you didn't lose the files "for reasons of public safety"...

    Yes, I do, please. As you have noted, I've been banging on about this experiment (or my perception of the need for something like it) for years. My requirements are pretty much as they've always been: 3D, tall (50+ stories, ideally) and capable of holding itself up before collapse is initiated -- these would be the minimum although, as you say, factors like wind load would also be good to take into account.

    I would also greatly appreciate and welcome input from other forum members along the lines of Spectrar Ghost's remarks above, so that we can move collaboratively toward a structure that has some broadly representative features of the actual towers themselves, and support you in the spirit of co-operation (even if you're actually doing the heavy lifting on your own).

    I do ask for all .blend files to be made available for download, of course -- for each iteration of the model you produce, if you would be so kind.

    Thank you, truly, for continuing to develop the tower model. At this point I would be willing to pay you the $100 merely in return for your commitment to this project and to encourage you to continue with it.

    However, that would remove the element of challenge, which I hope is also some small motivating factor (I do not imagine for a second that it is even close to your primary reason for doing this).

    Looking forward to the next developments: thank you again.
    Last edited: Mar 31, 2017
    • Like Like x 1
  31. Mick West

    Mick West Administrator Staff Member

    I've recreated the code as used in the previous video - amusingly the changes I'd lost were mostly visible in the image I'd posted. You'll need to set "numAssemblies to an appropriate value. The video uses 12. I'd try smaller first, like 3 or 4, to get a sense of things.

    Attached Files:

    • Like Like x 1
  32. Jeffrey Orling

    Jeffrey Orling Active Member

    The simulations need to have "interlocked" staggered structural skin and long span floors. The models look more like jenga collapses. The interlocking panels caused them to peel off in huge sheets in most cases.
  33. Jeffrey Orling

    Jeffrey Orling Active Member

    The virtual and real models of the collapse need to treat the contents and the slabs as different class of materials from the steel. Slabs and contents DID add to the destructive mass when the fell... but they also were destroyed and through thousands of mechanical collisions broken up into small elements as fine as dust and sand. That sort of destruction is very different from the jenga type of solid elements coming undone and falling.. The building was something like 95% air... and the solid steel structural elements by volume were only a few percentage points. The entire mass close packed would only stand about 40' high or some 200'x200'. The slabs undestructed by closed packed would stand about 35' tall.

    You can do a little experiment.... get a bag of cement or plaster... spread it around on the floor. and then take a sheet of plywood standing on edge in the middle of the spread out dust. Let the panel fall and observe what happens to the dust... it is "exploded" outward from under the plywood carried on the "escaping" air.
    Last edited: Apr 4, 2017
  34. Mick West

    Mick West Administrator Staff Member

    The point of the models is to illustrate the principle of progressive collapse via floor stripping. No matter how accurate you get there's always going to be lots you can do to make it more like the actual towers.

    Unfortunately increasing the model complexity is difficult using Blender - simply because we have 110 floors, and Blender has O(n*(n-1))) problems adding geometry via script.
  35. Jeffrey Orling

    Jeffrey Orling Active Member

    Correct Mick.... And these models do show some features of the collapse. I suppose you might be able to show several characteristics of the twin tower collapse... but obviously this will add complexity. Maybe we should identify the key distinctive features,,,,, such as the "dustification, facade peeling in large panels.... floor stripping as you've done.. and so on.
  36. Cube Radio

    Cube Radio Member

    Last edited by a moderator: Apr 10, 2017
    • Funny Funny x 1
  37. Jeffrey Orling

    Jeffrey Orling Active Member

    I know this is off topic. Mick I popped over to the link above and can tell you this a "den of truthers"... If you feel like wading into that mess go for it. I wouldn't waste the time. You models are all good at showing a limited aspect/feature of the collapse. A complete faithful simulation is probably not possible. On the other hand I doubt that the insider jobbers who claim placed devices can do a model (virtual) which produces ALL or MOST of the observed building movements.

    Burden of proof, as they say, is on those who make extraordinary claims.
    Last edited: Apr 10, 2017
  38. deirdre

    deirdre Moderator Staff Member

    No he didn't. All he did in that post is whine about being moderated here.
    And imply it took Mick 2 years to build the above model(s) in Blender, when in reality it took like 2 days.

    I don't consider that a comment on Mick's model.

    If you want to quote a comment on Reddit then do so*, but please stop paraphrasing other members (ie aka) and just linking to reddit threads with no external quotes provided.

    *I fixed your above post to add an external quote, so I wouldn't have to delete your comment.
    • Like Like x 1
  39. Mick West

    Mick West Administrator Staff Member


    There's very little actual discussion going on over there. But I'm curious as to why they feature my simplistic illustrative models, and yet ignore Kai Kostack's much more impressive model from six years ago.

    Source: https://www.youtube.com/watch?v=4_J7ak_IZXk

    My physical wood+magnets model was not tall enough, but was also dismissed because they could detect "jolts", even though a lack of jolts was not a requirement of the challenge.

    I don't really anticipate much discussion of my virtual model. I'd hope that some people recognize what it is demonstrating, and how that applies to the towers. However it's quite rare for people to have such a realization and then tell people about it. They normal just fall silent to mull it over.

    I've no interest in splitting the discussion over two places. @Cube Radio, you are, I think, a reasonable proxy for the "collapses like that are physically impossible" crowd. You are also, I think, not the type to fall silent when you finally realize that the towers might actually have collapsed by the floors being stripped from the columns which then failed in part due to slender column buckling at the joints.
  40. Mick West

    Mick West Administrator Staff Member

    3 days part time, then I got frustrated by the slow speed of scripted object creation, and lost interest. I'll be back!

    I certainly don't consider it finished, but it does illustrate the general mechanism.