The 9/11 Forum

OneWhiteEye wrote:You're quite handy with OpenGL; very attractive!

femr2 wrote:a) The 'bottom' of the rendered upper block does not change, as the 'compacted' floors are actually above that level.

I learned something here, should've been more careful before I spoke. It doesn't all get added to the bottom, for sure. But neither does it all get added to the top. Had to stop and think about this a while.

Thanks. Feedback is always good.

I may be wrong, but there may be a slight issue with interpretation, which follows from moving from zero floor thickness to floor thickness (z).

For my model, none gets added to the 'bottom'...

For zero thickness, by implication, the impact is between the bottom of the descending mass and the 'bottom' of the impacted zero thickness floor.
For thickness (z), the impact is between the bottom of the descending mass and the top of the impacted floor.
For the methods you are using, this has the implication of a 'jump' in the lowest point of the descending mass after each impact, of distance (z).

I note in your code that you appear to be adding the additional distance, but in your conclusion interpretation seems a bit grey. Have you included this factor in your conclusions ? Have I interpreted the code correctly ? (I'm not entirely sure if the code takes the jump into account either, which would imply some error in the graphed lower positions.)

This only becomes an issue when graphing the displacements btw...

My spreadsheet model has been updated to include the floor thickness, so all of the existing mechanisms are also available, allowing inclusion of mass loss, energy to fail the structure, crush energy, lateral ejection energy, specific floor by floor masses, specific floor heights...all that sort of stuff. I'll generate a few additional scenarios with the inclusion of floor thickness. As the thickness parameter is variable, I'll also include a couple of different thicknesses.

I think a very useful addition to the spreadsheet and 3D app will be graphing the relative vertical displacements of the roofline and lower edge of the 'crushing front', as, as you've indicated, the inclusion of floor thickness implies that the rate of descent of the roofline is actually slower than the rate of descent of the lower edge of the 'crushing front'.

Quite a useful addition I think, as it cleanly highlights the potential 'error' in rate of descent calculations which use the roofline as the point of focus. (For those which imply that crush-down is valid or near reality at least. If your view is, like mine, that the visual evidence indicates early destruction of the 'cap', then this point is slightly moot.)

(A few edits later )

Heiwa wrote:

femr2 wrote:The model has been updated ... the following video illustrates the new behaviour:
http://www.youtube.com/watch?v=xJSjbxFoJ8M

3 seconds to crush 13 floors and 10.52 seconds to crush 98! All nicely piled up - pancakes? - while upper part is 100% intact all the time! It doesn't look real. Are you working for Hollywood?

No, I'm not working for Hollywood. I'm progressing a 'crush-down' model as far and as openly as possible within the limitations of a step-wise framework. I make it very clear that I do not think the cap of each tower remained in any way intact, and also that in reality a 'crush front' is a bit of a mis-nomer too. May I recommend you have a trawl through some of my YT video's so you can get a gist of my position on the wider '9/11' thing.

The scenario being discussed is 'free-fall'. There is zero structural resistance. No mass loss. No energy conversions through crushing materials. Nothing. Simply magically suspended floors which coalesce with the falling mass upon impact, and the application of conservation of momentum (for the purpose of velocity change only in this context).

So... one factor the scenario is highlighting is that 'free-fall' for the tower is ~10.5s.
Lower quoted free-fall figures are simply wrong, as they do not take the fact that impacted bodies of mass must be accelerated, even when facing ZERO resistance.
Such quoted figures are also relative to the roofline. The most recent change to the model also highlights the additional error which this incurs, as the 'collapse time' really should be the time the 'crush front' reaches ground level IMO, and that the rate of descent of that crush front is faster than the rate of descent of the 'roofline'.

Consider the upper and lower parts as assemblies of material points linked together in structures with 95% air volume wise. The upper part is floating, the lower one is fixed on ground.
The material points can be your floors and the links can be the columns. Then remove the links between floors 99/98 and allow floor 98 to hit floor 98 due to a gravity drop. What happens?
Forces develop and are transmitted via the links to all material points and ground at high speed, say 5000 m/s. This means all material points/links are affected at first contact. The material points move!

Such transmission is rather difficult to express in a step-wise spreadsheet model, but if you can think of a way of including, say, the energy absorption into the underlying structure, or the induced additional velocity change on impact due to extrapolated conservation of momentum, then I'll be happy to look at how to include such. What you are talking about is a much more FEA based simulation, which is beyond the scope of what I'm doing here. I'd thoroughly like to get hold of the NIST WTC LS-Dyna models and run the simulation past 'initiation', but the data is not forthcoming. Do what you can....you know.

Assuming then, as you do, that links between floor 98/97 fail and floor 98 drops on floor 97, the same thing is repeated again. New forces are developing and are interfering with the first set of forces.
I would suggest that all these forces affect the upper part to some extent.

Again, it is a crush-down model which I'm extending to include as many valid factors as possible. It's not limited to floor 98. Any floor can be set to 'fail. The physical implication is that all structure between the failed floor and the one below instantly 'vanishes' to allow the first impact to occur.
Of course all energy transfers are, in reality, going to be transmitted in highly complex manners, but again, modelling such is FEA. My model does not even include elastic collisions...but I don't need to make excuses or apologies for that. It is what it is. It's not intended to 'prove' anything, but to aid in understanding some of the underlying energetics involved, and to see how certain factors such as mass loss can affect collapse time...

I'm aware of your position on these matters, but I suggest it would be more productive for you to build an appropriate model, real or virtual, and use it to highlight your position, or...to understand why I have built this model and help to improve and extend it as far as possible.

femr2 wrote:
Again, it is a crush-down model which I'm extending to include as many valid factors as possible. It's not limited to floor 98. Any floor can be set to 'fail. The physical implication is that all structure between the failed floor and the one below instantly 'vanishes' to allow the first impact to occur.
Of course all energy transfers are, in reality, going to be transmitted in highly complex manners, but again, modelling such is FEA. My model does not even include elastic collisions...but I don't need to make excuses or apologies for that. It is what it is. It's not intended to 'prove' anything, but to aid in understanding some of the underlying energetics involved, and to see how certain factors such as mass loss can affect collapse time...

I'm aware of your position on these matters, but I suggest it would be more productive for you to build an appropriate model, real or virtual, and use it to highlight your position, or...to understand why I have built this model and help to improve and extend it as far as possible.

But no structure crushes-down due to a part C of it dropping on the rest, part A. It is not possible! Reason is simply that the rest, part A, is stronger than part C. Part A can absorb much more strain energy than part C. Part A always stops or destroys part C in the complex collision process. This I have learnt from analyzing 200+ ship collisions and tests with many structures.

Heiwa wrote:But no structure crushes-down due to a part C of it dropping on the rest, part A. It is not possible! Reason is simply that the rest, part A, is stronger than part C. Part A can absorb much more strain energy than part C. Part A always stops or destroys part C in the complex collision process. This I have learnt from analyzing 200+ ship collisions and tests with many structures.

Like I said, I'm aware of your position. It's not fundamentally different to my own, but that's far from the point...

One factor you do seem to avoid, with the repetition of the 'part c, part a, part x...' thing, is that if you get a big enough mass and drop it onto one floor of the WTC, you're going to get progressive floor failures. The majority of the floor connections are, after all, the same all the way down (no increase in strength). The mechanical floor regions are going to mess with that, as are certain floors with solid column supports, and retention of mass inside the perimeter is also pretty much required for such behaviour.
The behaviour of the core and external walls is entirely separate, as in that scenario (progressive floor connection failure) the external columns and core would remain standing...but, the point I'm trying to make with you is that simplifying the 'model' to part a, b & c is a much more limiting over-simplification of 'reality' than even a well thought through 'crush-down' model IMO.

Progressive 'collapse' of the outside core floor sections IS possible, IF enough mass drops to overcome the floor support strength.

That wasn't actually the case (initially there was not enough descending mass for this process to occur), but again, that's not the point.

In reality... The core did not survive. The external walls did not survive. Significant mass was lost. Significant quantities of energy were 'consumed' by additional processes such as crushing, etc...so there is still a BIG problem with each descent.

I think we will both agree that the scenario I've outlined (progressive floor connection failure resulting in intact core and external columns) doesn't match the actual behaviour of the descent of either tower, but it does show that the primitive rigid part a/b/c position is also an extreme and inapplicable over-simplification. Yes ?

femr2 wrote:Thanks. Feedback is always good.

You're welcome. This is very interesting work, and all the more interesting with the pretty packaging.

I may be wrong, but there may be a slight issue with interpretation...

I believe you're right about that part - at least!

For my model, none gets added to the 'bottom'...

I may have been misleading in some of my wording. I guess it depends on your definitions of 'added' and 'bottom' but please bear in mind that this is a problem in accretion, and geometrically, it occurs at the bottom of the collection of moving upper block/slabs I'll just call the block. The block grows in extent with each collision, and does so discretely, at the bottom. That's my definition of 'added' and 'bottom' and why I gave that impression. However, I didn't intend to give the impression that some displacement is added to the bottom of something that isn't there, rather the displacement is defined by where the bottom of something is. That something is the block.

For zero thickness, by implication, the impact is between the bottom of the descending mass and the 'bottom' of the impacted zero thickness floor.

Or, just the floor/story? After all, there's nothing but empty space in between. How's that different from impacting the 'top' of the next floor down, since the two are synonymous by definition in this context?

For thickness (z), the impact is between the bottom of the descending mass and the top of the impacted floor.

Yes, at it was in the prior case above, though trivially so. The position of the crush front is the bottom of the block, what else would it be? At the very moment before collision, the bottom of the block is where? About to touch the top of the impacted slab, as you point out. A moment later, when impact occurs, the bottom of the impacted floor becomes the bottom of the block. This is the new, post-collision location of the crush front because the impacted floor has joined the block. It took no time to get there because nothing moved, it was already there.

For the methods you are using, this has the implication of a 'jump' in the lowest point of the descending mass after each impact, of distance (z).

Exactly. What would you propose as an alternative? I'm at a loss for choices.

I note in your code that you appear to be adding the additional distance, but in your conclusion interpretation seems a bit grey.

The code does indeed add two distance numbers together, but at issue is the notion of 'additional'. Sorry if you get this already and I'm beating a dead horse... If not, let me introduce some contrast.

Story height is h, slab thickness is z and distance between slabs is h-z. The bottom of the upper block must fall though an initial height h-z to collide with the topmost slab below. Imagine two structures side by side, one z0=0 and the other z1=h/2. For convenience of visualization, the point-like floor locations of the z0 structure are aligned at the same elevations as the bottom of the floors in the z1 structure. The initial offset doesn't matter in the dynamics.

Anyway, both descend at freefall and reach a displacement of h/2 at the same time. z0 keeps freefalling, z1 has an impact. Now the bottom of the impacted slab, which has not moved in an instantaneous collision, is the crush front. This is a discontinuous advance in the downward direction, exclusively, involving no motion. Omitting it would be an error, it's not adding something in from nowhere. Nothing bounces up afterward, either. Yes, the upper block is slowed in the impact but it's immediately in freefall again, with double the effective displacement of the other block.

It should be easy to convince yourself that if the initial upper block mass is made MUCH greater than that of the individual impacted floors, the velocity reduction can be minimized away; i.e., the impacted floors do not slow the accreting upper block at all and each floor immediately assumes the current velocity of the block. In this case, the z1 collapse is complete in exactly the time it takes an object to freefall half the distance to the ground. The z0 structure, on the other hand, has a crush front halfway to ground. The rooflines, a fixed height above (height of upper block), are at the same elevation for both structures. Except z1 is done!

Both upper blocks experienced freefall over the same distance, sweeping up the floors like flies on a windshield. Except z1 gets quite a crusty buildup on the way, thick enough to stop it after it has gone only half the distance to the ground.

Quite a useful addition I think, as it cleanly highlights the potential 'error' in rate of descent calculations which use the roofline as the point of focus. (For those which imply that crush-down is valid or near reality at least. If your view is, like mine, that the visual evidence indicates early destruction of the 'cap', then this point is slightly moot.)

I favor early destruction of much of it, most of it before too much time has passed, and very top remnant staying somewhat together through most of the fall, but eventually falling off.

This is a very important point. Roofline measurements, even without tilt, do not tell the whole tale. I've taken a different approach to modeling this stuff but it should be somewhat complementary. One thing I run into a lot is simultaneous crush-up. See here and tell me whether you think the difference in stretch is as big a deal as simultaneous crush-up!

...which would imply some error in the graphed lower positions.

Wouldn't be surprised, I'm not sweating the small details. Which graph - first, last or both?

femr2 wrote:One factor you do seem to avoid, with the repetition of the 'part c, part a, part x...' thing, is that if you get a big enough mass and drop it onto one floor of the WTC, you're going to get progressive floor failures. The majority of the floor connections are, after all, the same all the way down (no increase in strength). The mechanical floor regions are going to mess with that, as are certain floors with solid column supports, and retention of mass inside the perimeter is also pretty much required for such behaviour.

....

I think we will both agree that the scenario I've outlined (progressive floor connection failure resulting in intact core and external columns) doesn't match the actual behaviour of the descent of either tower, but it does show that the primitive rigid part a/b/c position is also an extreme and inapplicable over-simplification. Yes ?

Progressive floor failures? The 'floor' consists of a thin concrete shell held by a thin plate in turn held by a wire truss, lower flange of which (another wire) carries false, inner ceiling panels.

I would say that that 0.75 m thick 'floor" is again >90% air and to compress it into 0.5 m is ???? What are you compressing? The air or the concrete?

Actually when a floor hits another floor it is ceiling panels hitting a concrete shell, etc, etc. What breaks?

Sometimes I wonder if something was put on top of the false ceiling panels to assist the destruction.

OneWhiteEye wrote:

For zero thickness, by implication, the impact is between the bottom of the descending mass and the 'bottom' of the impacted zero thickness floor.

Or, just the floor/story? After all, there's nothing but empty space in between. How's that different from impacting the 'top' of the next floor down, since the two are synonymous by definition in this context?

No different. Just a verbal way of highlighting the point, which you get.

For the methods you are using, this has the implication of a 'jump' in the lowest point of the descending mass after each impact, of distance (z).

Exactly. What would you propose as an alternative? I'm at a loss for choices.

Nothing. Just checking that you were taking it into account.

The code does indeed add two distance numbers together

Cool. Job done.

I favor early destruction of much of it, most of it before too much time has passed, and very top remnant staying somewhat together through most of the fall, but eventually falling off.

Prepared to estimate the number of floors of that top remnant, and at what point it fell off ? Your estimations for WTC 1 & 2 will be significantly different I assume ?

One thing I run into a lot is simultaneous crush-up. See here and tell me whether you think the difference in stretch is as big a deal as simultaneous crush-up!

As soon as I can get around the implications of asynchronous collisions (or find a synchronous method that is acceptable) I'll most definitely be including 'crush-up' in the model. Uber significant factor.

...which would imply some error in the graphed lower positions.

Wouldn't be surprised, I'm not sweating the small details. Which graph - first, last or both?

Would have been the 'bottom' of block displacements, but as you are including the 'jump', there's no error.

Heiwa wrote:Progressive floor failures?

Yes. (Within the context of the discussion)

The 'floor' consists of a thin concrete shell held by a thin plate in turn held by a wire truss, lower flange of which (another wire) carries false, inner ceiling panels.

I would say that that 0.75 m thick 'floor" is again >90% air and to compress it into 0.5 m is ???? What are you compressing? The air or the concrete?

I included a list of some elements in the post...the floor truss, office furniture such as tables/chairs/filing cabinets/partition walls, HVAC...etc...escalators for mechanical floors.... The value you are referencing includes the core in it's context, which includes additional core located elements such as lift shafts, stairwells, restrooms, 'safes', all that kind of thing. The thickness is a variable parameter, so I'll be looking at various values anyway. 0.55m was just an example.

Actually when a floor hits another floor it is ceiling panels hitting a concrete shell, etc, etc. What breaks?

If the floor itself is taken in isolation, then there would be the compaction and deformation of the truss frameworks, probably inter truss initial breakage of the concrete slab etc. The subsequent detachment must be the connections between floor assemblies and the external and outer core columns, or 'breakage' of the floor away from those connections. I'm not including the behaviour of the external and outer core columns themselves in the description. Please remember that there is a maximum load for each floor, which even in a static mode will result in floor 'failure' if significantly exceeded. (Again, in context of current discussion)

Sometimes I wonder if something was put on top of the false ceiling panels to assist the destruction.

I doubt it. The strength of the connections between suspended ceiling and floor truss would not support much additional mass, and any uneven load would be very noticable. (I'm not precluding the installation of 'devices' in the inter-floor spaces btw, but I think there are simpler and more effective alternatives in such regions...gravity for one. Special floors (mechanical etc) are a different case.)

OneWhiteEye wrote:

femr2 wrote:Anyone fancy explaining how the debris ejecta on the right hand side of the Tower is so far ahead of the simulated free-fall descent ?

Zone B thickness?


Edit: not trying to be flippant, the real collapse is a heterogeneous 3D affair, I don't expect strong conformance to 1D simplifications. Of course, the differences are worth exploring, with each step being more complicated. It is a good question, but one that should be examined carefully and in detail without rush to judgement. All I'm saying is, while you're on a 1D model and looking at direct comparisons, a more accurate comparison might include greater thickness in the crush layer.

Edit2: at 5 seconds, crush down only, the thickness is only about a third of an upper block, so my earlier comments have limited applicability.

Back to this I think, now that zone B thickness has been added and clarified.


http://www.youtube.com/watch?v=rikfw4cdOBo

http://femr2.ucoz.com/photo/1-0-59

I've also looked at the effect of, say, half mass free-fall descent (to get an idea of how a south-side separated collapse could have behaved) and the timing is only marginally different..0.25s...so I'm still asking the question:

Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?

That's very nice, femr2. This is extremely helpful. I've never seen any better side-by-side. It has advanced me considerably, no joke.

Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?

It's still a good question. Looking at the 5 second mark, it seems to be about an 11-12 story discrepancy, about the size of (rigid) Zone C. Is that about what you'd estimate?

I have some ideas about closing the gap. Maybe not all of it... probably not within the limitations of the kind of model we're talking about here.

The effort and results are superb, thank you. One request? Spin that puppy around a little, pull the camera back a ways and drop it down. Just eyeball the perspective roughly, all that needs to show is the red and green bands for these purposes.

...half mass free-fall descent (to get an idea of how a south-side separated collapse could have behaved)

Can you describe this in more detail?

OneWhiteEye wrote:Looking at the 5 second mark, it seems to be about an 11-12 story discrepancy, about the size of (rigid) Zone C. Is that about what you'd estimate?

Yes. Nearer to 11 than 12.

I have some ideas about closing the gap. Maybe not all of it... probably not within the limitations of the kind of model we're talking about here.

Fire away...

One request? Spin that puppy around a little, pull the camera back a ways and drop it down. Just eyeball the perspective roughly, all that needs to show is the red and green bands for these purposes.

Is this what you mean ? (Quick fit. Placement might not be as accurate as is possible with a bit more time. A higher resolution version of the source video would be very handy)

...half mass free-fall descent (to get an idea of how a south-side separated collapse could have behaved)

Can you describe this in more detail?

Just testing to see how much quicker a reduced floor mass descent would traverse, to try and explain some of the 'gap'. The highlighted ejecta appears to be emanating from the southern 'half' of the tower. I'll gather some specific data together to show the effect of gradually reduced floor mass on the descent time.

I should remind (myself) that we're still in zero resistance free-fall territory, and should begin including some of the additional available factors soon, but I'd like to try and get some form of crush-up in place before I do...

femr2 wrote:Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?

Am I mis-interpreting what you have presented? Is not momentum being conserved in your simulation? The debris ejecta is the only thing in free fall here, it appears to me.

Un-hyphenated wrote:

femr2 wrote:Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?

Am I mis-interpreting what you have presented? Is not momentum being conserved in your simulation? The debris ejecta is the only thing in free fall here, it appears to me.

That depends upon your perspective.

Conservation of momentum is indeed applied to the inelastic collisions between the descending mass and each impacted floor, but zero resistance is included.

I'm aware of the paradox in using rigid bodies of mass, perfectly inelastic collisions and zero resistance in the 'real' world, but the bottom line in using the term free-fall is this:

Conservation of momentum results in a velocity change when two bodies of mass collide within the bounds of an inelastic collision.

A fundamental reason for this is the fact that the impacted mass must be accelerated from rest upon impact, so as there is no resistance, the fall is still validly called 'free' (IMO). The effect of the collisions cannot be ignored, even when the progression is facing zero resistance.

It is not valid to compare a free-fall timing of an object with no collisions during it's descent to the time an object which experiences a series of collisions traverses the same distance. That is to say that the "free"fall time each tower could collapse in is NOT the same as the free-fall time of a singular object from the same height. That would only become valid if every floor of the tower began to descend at the same time, and no mass collided with any other mass during that time.

If you object to this route, another would be to not apply conservation of momentum and instead sum 82 individual from rest ~3.7m free-fall descents...82*0.86s=70.5s

I could call it a 'zero resistance' descent instead, but that's just semantics IMO.

Add: As the model makes the energy 'loss' through the COM calc available for subsequent deformation of materials, including a simple constant 'resistance' sink for failing structural support only results in an additional 2s to the descent time, though of course it's not that simple and a constant value is not valid either. The model allows various methods and numerous other factors to be taken into account. Whenever additional factors are included, I don't term it 'free-fall'.

femr:

I was attempting to answer your question . You asked, "Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?"

Then, in response to my reply, I believe you answered your own question:

"Conservation of momentum results in a velocity change when two bodies of mass collide within the bounds of an inelastic collision. A fundamental reason for this is the fact that the impacted mass must be accelerated from rest upon impact.."

Bottom line: the debris ejecta is not colliding with anything on the way down. So it falls faster.

Un-hyphenated wrote:femr:

I was attempting to answer your question . You asked, "Why is the debris ejecta on the right hand side of the Tower so far ahead of the simulated free-fall descent ?"

Then, in response to my reply, I believe you answered your own question:

"Conservation of momentum results in a velocity change when two bodies of mass collide within the bounds of an inelastic collision. A fundamental reason for this is the fact that the impacted mass must be accelerated from rest upon impact.."

Bottom line: the debris ejecta is not colliding with anything on the way down. So it falls faster.

Well the very simple answer is, yes, you are misinterpreting the question.

It does not refer to the free-falling debris, it refers to the debris ejecta on the right hand side of the tower: