Hardfire: Physics of 9/11

Hi femr2, welcome.

I understand the limitations of your simulator, but I was curious if you've tweaked it to get conditions where the collapse is arrested? If so, what were they?

Also, can the simulator be tweaked to include crush up? Is the upper block considered rigid in the simulator or is there a way to crush one floor of the upper section for every 2 of the lower?

The free standing core would be hard to incorporate into the simulator, but I assume there would be a way to adjust the strength of the core by progressively removing columns from the core beginning at the 60th floor. Has this been attempted?
 
If you google my name, it's the second link - WTC Collapse Simulator.

All the calculations are housed in a relatively simple, open, xl spreadsheet, which is focussed upon the generation of a collapse time for a gravity only driven collapse. The spreadsheet is available for download.

The visualisation module simply imports data directly from the spreadsheet, so until I 'finalise' the model, it's not currently available for download, though there are a few YT video scenarios available.

The forum is linked on the home page, and includes fairly detailed answers to Ryans' initial queries.

Thanks for the tip. I'll take a look (probably tomorrow).
 
Welcome, Femr2.

Thank you for your efforts to model the tower collapses. Since I cannot analyze your models, can you briefly state how they were created and what you think they show?

I'm confused by your choices in the forum you created. You are replying to Ryan Mackey's questions to someone else in a different forum, addressing Ryan as if you were having a discussion with him:
I thank you for your questions, and acceptance of the model in your challenge, and hope we can progress and clarify it with mutually satisfactory results.

Why did you not contact Ryan directly?
 
Last edited:
Hi femr2, welcome.

I understand the limitations of your simulator, but I was curious if you've tweaked it to get conditions where the collapse is arrested? If so, what were they?

Also, can the simulator be tweaked to include crush up? Is the upper block considered rigid in the simulator or is there a way to crush one floor of the upper section for every 2 of the lower?

The free standing core would be hard to incorporate into the simulator, but I assume there would be a way to adjust the strength of the core by progressively removing columns from the core beginning at the 60th floor. Has this been attempted?

I assume you have read my responses to Ryan before commenting ? If so, I suggest that rather than simply stating you understand the limitations, that you expand and state exactly what you actually mean.

Collapse arrest conditions are many, and include setting small concrete crush scale (or high specific constant energy sink), cap mass loss and reduced initial 'drop' height'.

Future extension will include cap crush-up for the cap.

Currently the cap is considered rigid, though cap mass loss can be included for each floor impact.
My personal version of the model (to be released shortly) includes more advanced mass loss mechanics, including the ability to specify the behaviour mentioned (though not limited to such, as it is more likely that the weaker floors within the cap would 'suffer' more than the stronger impacted floors) and also floor-by-floor specific impacted floor mass loss.

The new version does include the ability to specify specific core portion remaining on a floor by floor percentage basis.
 
Welcome, Femr2.

Thank you for your efforts to model the tower collapses. Since I cannot analyze your models, can you briefly state how they were created and what you think they show?

I'm confused by your choices in the forum. You are replying to Ryan Mackey's questions to someone else in a different forum, addressing Ryan as if you were having a discussion with him:


Why did you not contact Ryan directly?

I have not modeled the tower 'collapses'. I have modeled a gravity only driven collapse. We have very different views on the cause I am sure, but as I have made clear within my forum, I do not wish to engage in any form of biased discussion about the wider subject, and instead prefer to focus on the specific details of the model itself.

Why can you not analyse the model ?

It is impossible to briefly state how the model was created.

It is pointless to begin a conversation about the validity of a model by jumping to the very end of the conversation and stating opinion on what it 'shows', which is entirely dependant upon point of view, final mechanics and final data used.

I have made my views regarding email conversation very clear, and have only, reluctantly, begun conversation here. It is still my intention to conduct this discussion directly with Ryan on my forum, to enable one-on-one direct dialogue.
 
I assume you have read my responses to Ryan before commenting ? If so, I suggest that rather than simply stating you understand the limitations, that you expand and state exactly what you actually mean.

Sorry, that was in regards to tilt only. I'm not fully aware of all the limitations in the simulator.

I assume this actually means the simulator is also limited to perfect collisions of intact structural members, and cannot take into account damage that may have occurred before the upper mass makes contact with the lower floors. It can't project this. In order to do this, you would have to estimate the amount of damage to each floor, then plug it in to the spreadsheet.

I assume this actually means the simulator can't take into account the stagger in the 3 story lengths of steel, in either the exterior or interior columns.
 
Sorry, that was in regards to tilt only. I'm not fully aware of all the limitations in the simulator.

I assume this actually means the simulator is also limited to perfect collisions of intact structural members, and cannot take into account damage that may have occurred before the upper mass makes contact with the lower floors. It can't project this. In order to do this, you would have to estimate the amount of damage to each floor, then plug it in to the spreadsheet.

I assume this actually means the simulator can't take into account the stagger in the 3 story lengths of steel, in either the exterior or interior columns.

The basic model is an extended version of the 'normal' crush-down models I'm sure we are all aware of, but with the addition of further parameters such as mass loss, much more specific masses, accurate floor heights, etc...

Collisions are currently inelastic, yes. A future version may include a form of elastic collision, but the intention is to provide users with a tool to understand how various factors affect collapse time rather than approach a finite element analysis.

Inclusion of damage above and below the 'failure' floor could be included by variation of the support energy sink (the amount of energy required to collapse the structural supports) on a per floor basis on my current version, but in the version you have, it is not included.

Staggered support lengths are not taken into account. A simple per-floor energy sink is implemented, which ramps towards the base to account for column thickening.
 
Last edited:
Ryan, you have now been fully aware of the model I have created since the 12th April. You have posted links to it and obviously read the content, which states quite clearly that the model is put forth to your 'challenge'.

In your own words it 'is clearly the best entry in response to my "Hardfire Modeling Challenge."', but you are still claiming that no-one has taken you up on said challenge.

Nobody has. You have fallen short of several criteria, as I already told 'your friend.'

As you are also aware, I have placed your prior questions (received via a YT user who contacted you about my site) along with appropriate answers in a simple forum. I have also clearly stated that the forum would be locked to accept posts from only yourself and I until full conclusion of any ensuing discussion, negating your concerns about derailment.

I have no reason to trust a forum that you run, and no interest in a forum that nobody else reads.

Well, here is the contact you have 'requested'. I put the model forth to your challenge. I do not wish to engage in 'behind closed doors' private email conversation, and would prefer to discuss through my forum. If you will only discuss here, then that will have to do, but I am sure you will agree that this is far from a suitable location.

No. I told you a couple of weeks ago that this forum is not a suitable location thanks to incessant derails that have totally overrun the moderators here, and I already told you and 'your friend' how to proceed. Any time you want to stop playing games, it's your move.
 
I have not modeled the tower 'collapses'. I have modeled a gravity only driven collapse. We have very different views on the cause I am sure, but as I have made clear within my forum, I do not wish to engage in any form of biased discussion about the wider subject, and instead prefer to focus on the specific details of the model itself.
Ok, you've modeled a gravity-only-driven collapse. What predictions about the tower collapses do you believe can be made based on this model?

Why can you not analyse the model?
Perhaps if I devoted a lot of time to your spreadsheet, I could. Since you've already done that, can you please explain in layman's terms what you've attempted to accomplish and what you believe the results to be?

It is impossible to briefly state how the model was created.
It is? That's surprising to me.

It is pointless to begin a conversation about the validity of a model by jumping to the very end of the conversation and stating opinion on what it 'shows', which is entirely dependant upon point of view, final mechanics and final data used.
That's OK. What, in your point of view, does your model show?

I have made my views regarding email conversation very clear, and have only, reluctantly, begun conversation here. It is still my intention to conduct this discussion directly with Ryan on my forum, to enable one-on-one direct dialogue.
This strikes me as odd. You say you're responding to his challenge, but you didn't contact him, yet you addressed him on a forum you created, as if you two were having a discussion, and you say that you intend to conduct this discussion on your forum. I ask you again: what is the purpose of such an oblique approach? If I respond to someone's challenge, I respond directly to that person. Isn't that a good idea?
 
Nobody has. You have fallen short of several criteria, as I already told 'your friend.'

I have no reason to trust a forum that you run, and no interest in a forum that nobody else reads.

No. I told you a couple of weeks ago that this forum is not a suitable location thanks to incessant derails that have totally overrun the moderators here, and I already told you and 'your friend' how to proceed. Any time you want to stop playing games, it's your move.

I have made it perfectly clear that the YT user contacted you of his own volition.
I have made it perfectly clear that I will not be entering into a private email conversation.
This is our first direct communication, and you appear to be 'derailing' your own thread.
I am playing no games, and your insinuation is far from acceptable in my opinion.
If you wish to continue discussion here, then I will continue responding.
I would hope that other users may decide to minimise derailment if the conversation begins...
 
Staggered support lengths are not taken into account. A simple per-floor energy sink is implemented, which ramps towards the base to account for column thickening.

This may be a significant factor in favour of collapse retardation. Not so much over the entire height of the building, but during the first 6-9 floors of the collapse.
 
This may be a significant factor in favour of collapse retardation. Not so much over the entire height of the building, but during the first 6-9 floors of the collapse.

It would appear that Ryan does not wish to conduct discussion of this model.

The only reason for creating an ID here was to facilitate that discussion.

I agree that extension of the model to include such a factor would be a positive step.

Without any 'purpose' to this discussion, which is after-all the Hardfire Model Challenge, I'm not sure what the point of continuing is.

I'm more than happy to continue discussion on my forum, which regardless of Ryans opinion, I will be conducting with utmost integrity.

Thanks 'all' for the responses. I'm sure it could result in improving the model...
 
No. I told you a couple of weeks ago that this forum is not a suitable location thanks to incessant derails that have totally overrun the moderators here, and I already told you and 'your friend' how to proceed. Any time you want to stop playing games, it's your move.

It would appear that Ryan does not wish to conduct discussion of this model.
What were his conditions?
 
What were his conditions?

That's rather redundant Gravy. We all have our own conditions, and I have relaxed my position to conduct discussion here.

I created the forum to facilitate a direct one-on-one discussion free from derail as I do not enter into hidden from view email conversation.

If that's the end of the story and ('the best entry in response to my "Hardfire Modeling Challenge."') goes ignored due to such a simple 'conflict', then so be it.

I think the best route now is to post the questions Ryan posed, along with my answers, and see what happens...
 
1. Do you consider energy loss due to conservation of momentum and due to destruction of materials independently? It appears that you do. If so, this is wrong. The energy that "disappears" in the inelastic collisions goes primarily into deformation of materials, some ultimately ending up as heat but damaging steel and concrete on the way. Note that this energy is about 20-30% of the total potential energy -- it's huge. If this energy isn't enough to destroy materials, then you should of course take out an additional amount to reach that threshold, but if you consider momentum loss and damage as independent energy sinks, you are double-counting the energy loss due to collisions. That has the effect of artificially slowing your collapse an extra 0.2 g or so all the way down the stack.

Energy loss due to conservation of momentum is indeed treated separately to other energy sinks such as the 'destruction of materials'.

This is not incorrect.

Consider the virtual model of the tower as virtual slabs of titanium held aloft with no support (floating). A titanium mass is dropped and upon each impact an inelastic collision occurs and the impacted mass 'welded' as per the normal construct for the crush-down models. Conservation of momentum applied to this model will result in kinetic energy 'consumption', even though there is no resistance, no supporting structure to deform, and no destruction of the titanium slabs themselves.

The kinetic energy sink is into heat, sound and (primarily) acceleration of the additional mass, not destruction of materials.

There is no destruction of materials in the model just described.
 
2. Do you require that all concrete is destroyed uniformly, and during the collapse? If so, this is also wrong. Even Dr. Steven Jones has remarked on this for years, noting that at best only a small fraction of the concrete was actually turned to dust, and even of the dust much of it is quite coarse. You could make a more rigorous mass estimate by looking at the dust study by Lioy et. al. and looking at the mass fraction relative to, say, dust from drywall, assume that most or all of that was turned to dust, and treat the rest of the concrete as uncrushed. Also, what concrete is ground up does not have to be ground up during the collapse. If it hits the ground with some leftover kinetic energy, this energy also contributes to grinding of materials. Furthermore, concrete exposed to fire (i.e. the fire floors, and practically everything in the Pile) will spall, and this will allow grinding with hardly any kinetic energy expenditure. In summary, the amount of concrete turned to dust during the collapse is small.

No. There is no requirement for uniform destruction of all concrete.

The concrete crush energy sink can be switched off entirely, or either of two distinct selectable methods can be applied:

1) A user specifiable constant energy usage (in joules) for the crushing of concrete. Applied per-floor.
or
2) A user specifiable particle scale range (top to bottom) (in microns) applied on a floor-by-floor linear ramped scale, with a current limit of up to 500 micron (although of course this can be easily extended by the user. It's a linear energy usage scale)

The first method does not relate to concrete volume at all, only energy consumed per floor, and so is a fully specifiable constant allowing the user to use a value equivalent to the crushing of zero concrete thru to all concrete, and using their own external methods for determining the amount of energy required to do so.

Mechanically the second method does apply to the full concrete volume of each impacted floor, but again, the scale of the crush can be changed as described above. Many sources state an 'average' concrete crush scale well below the current maximum of 500 micron, which is why the limit was set as such. It could easily be extended to a maximum of say 1m if the user chose to.

If it hits the ground with some leftover kinetic energy, this energy also contributes to grinding of materials.
In this situation the KE should be removed from the system via mass loss. Applying it's energy to post-descent material destruction is erronious.

Furthermore, concrete exposed to fire (i.e. the fire floors, and practically everything in the Pile) will spall, and this will allow grinding with hardly any kinetic energy expenditure.

The model does not include energy sink for any concrete in the cap, only for the impacted floors below. Your point is redundant within the model.

I assume that by 'pile' you mean the descending mass I refer to as the 'cap' ? (As in pile-driving mass)

In summary, the amount of concrete turned to dust during the collapse is small.

This is a much discussed point of contention within 9/11 debates, and does not belong here.

Further extension of the model will include the ability to specify mass loss on a per-floor basis for each possible material.
This extension will allow per-floor specification of the concrete volume crushed and per-floor max/min crush scale.
Additional energy sinks will also be included for the crushing of other materials, which are currently not included.
At this point, only the crush energy for concrete is included (if non zero). Crush energy required for dry wall, furnishings, fittings and a wealth of other required crush 'bodies' will also be included.
 
3. In your ejected mass fraction, are you considering the ejected mass of the new floor hit, or of the entire debris pile? If the latter, then you are going to overpredict mass ejection in the later stages of collapse. This is because the mass scales superlinearly with floors hit (as floors get heavier as you move down), and the rate at which floors are hit scales roughly quadratically with time. This means your mass ejection fraction scales somewhere around the three-and-a-half to fourth power of time. This isn't right. Admittedly, estimating this is quite difficult. I would suggest that you put in loss as a given fraction per unit time rather than unit distance as a rough estimate. The much more sophisticated way is to assume mass ejection scales with the resistance of lower floors (c.f. equal and opposite), since what is really happening is mass gets ejected because of momentum flow from the lower floors through the debris pile to loose elements near the edges. I haven't thought about it in much depth and this could use some more work.

Mass loss can be omitted entirely, or included as described below.

The ejected mass fraction is specified by two separate values:

1) A percentage of the 'Cap' mass, pre-impact.
and/or
2) A percentage of the impacted floor mass. (Storey)

Both of these values are clearly labelled.

Either or both can be set to a user specified percentage, including zero, or switched off entirely (Equating to zero for both, but allowing any percentages entered to be retained in the sheet).

These factors negate the verbose implications you describe.

The model will be extended in the future to include per-floor specifiable mass loss, per material, enabling much more precise control. I'm intending upon identifying specific ejecta from video footage as part of the exercise. Separation of external column structure mass will also be included within the model to account for the probable quantities of external columns which played no part in further descent, and did not form part of the 'pile driving' mass.

since what is really happening is mass gets ejected because of momentum flow from the lower floors through the debris pile to loose elements near the edges

This natural behaviour inference is not directly conducive to this discussion, as I am sure you can understand given our differences of opinion on the cause of each descent. I intend on extending the model to include the energy required to laterally accelerate the ejected mass.

I agree that a mechanism relating mass loss to collapse progression is a good idea. I shall have a think about the best way to do so, but first thought would be to relate to impact KE.
 
4. In computing energy to fail the steel, what mechanism are you assuming? Compressive failure, buckling failure, or fracture? It appears you are using compressive strain since you talk about plastic deformation, but I am not certain. In the actual collapses, plastic deformation is slight, as evinced by the relatively straight condition of nearly all structural elements after collapse. In all previous models -- Greening's and Bazant's -- they overpredict resistance, deliberately, rather than consider lower energy buckling or fracture modes.

The 'energy to fail the steel', termed 'Support Collapse Energy' in the model, can be omitted entirely. or included by two methods.

1) A user specifiable per-floor constant (in joules)
or
2) A per-floor calculated value based on several factors.

The user specifiable value is completely open, allowing any value supplied by the user to be used, for each floor.

The calculated value uses a base figure for the collapse of one floor (the failure floor), which is split between the external columns and the core columns, and then ramped on a scaled multiplier towards the tower base.

The base figure, and each split value can be set by the user.

The ramping formula can be found in the following spreadsheet column: TowerMass:AJ

The ramping formula accounts for the thickening of the steel towards the base, but admittedly it is an estimate (it underestimates the strength increase significantly). If I can get hold of more accurate steel mass and scale source data the model will be updated to use specific steel mass-to-collapse energy conversions.

In the actual collapses, plastic deformation is slight, as evinced by the relatively straight condition of nearly all structural elements after collapse.

Again, this natural behaviour inference is out of place in this discussion. Modelling should be performed on expected behaviour, not forced implementation of mechanics which are limited by observed behaviours.

Future update of the model will include an implementation of elastic collision, including much more refined steel deformation mechanics, and will include the appropriate time and energy requirements.
 
5. What observed collapse time are you trying to match? The correct figure seems to be about 15-18 seconds, though you appear to be measuring all the way to the bottom of the basement, whereas we can't see into the basement and usually measure to grade. This will probably only add a few percent to your estimate, but you'll still overpredict collapse time.

I am not trying to match any collapse time. The intention is to model the effect of the various parameters upon collapse time.

This hopefully provides the user with an understanding of how such factors affect the time of collapse, and assists users in making their own conclusions about probable causes.

The measurement is to ground level ONLY (above grade).

Figures are included for mass below grade simply because they are available, and for completeness.

In addition, the post crush-down collapse of the cap is modelled without any energy sinks or resistance, which produces a slightly short full collapse time. This will be addressed in a future update and may also include below grade mechanics.

Official collapse times vary, though it is clear that Dr Shyam Sunder, Lead Investigator for NIST stated the times as 9 and 11 seconds. This clearly cannot be a timing for the descent of all debris though, as both are faster than pure free-fall. We shall of course disagree on the mode of descent, which complicates what is actually meant by the descent time, and I suggest that focus is placed on refining the model rather than specifying what value it is supposed to be 'generating'.

With all parameters set to mutually acceptable values, and the possible inclusion of additional factors and more complex mechanics, the resulting time will, I am sure, be a topic of further, separate, discussion in itself.
 

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom