WTC7 structural damage analysis (long contribution)
Pls note introduction of problem in Comment No. 1 below. Below is a draft of submittal of comments to NIST and JREF members are welcome to comment prior my sending it to NIST.
Submittal of Comments
Name: Anders Björkman, 6 rue Victor Hugo, F 06240 France
Affiliation: President, Heiwa Co, European Agency for Safety at Sea (address as above)
Contact: +336 61725424,
anders.bjorkman@wanadoo.fr
Report Number: NIST NCSTAR 1-9
Page Number: 455-536
Paragraph/Sentence: Chapter 11 - 11.2 ANSYS Model, 11.3 Analysis results
Comment No. 1: In structural damage analysis - as opposite to structural design analysis - it is not a load path of the intact structure that is of interest, but the path of failures from the first small local failure due to a known cause (e.g. fire) to the end of destruction including all structural failures in between as a consequence of the first, small failure. Such a damage analysis shall identify the critical failure in the path that caused the end result, i.e. could the critical failure be avoided, then the destruction would have been arrested there.
The NIST WTC7 draft report fails to do this proper structural damage analysis:
It is not clear in what order the various local structural failures take place in the ANSYS model, what elements/nodes are affected, details of failure, cause of failure and consequence of failure (serious or can be ignored?) and how the boundary conditions (loads on columns at floor 16) are affected.
Reason for Comment: The ANSYS model consists of primary (vertical columns connected to ground), secondary (horizontal and sloping beams connected to primary parts) and tertiary parts (e.g. floor elements connected to secondary parts; beams) and associated connections, and it is of vital importance to know the order of failures. We know that the structure at ambient temperature is very low stressed and thus looks very safe.
Heat/thermal expansion may affect a tertiary member that heats up quicker than adjacent secondary members and the local connections may fail and the tertiary part is out of action but it hardly affects the effectiveness of the secondary part, which of course is verified when the FEA analysis is re-done after each failure. The secondary parts and their connections to primary parts are much stronger than tertiary parts and will deflect with the primary parts and it is highly unlikely that thermal expansion will produce forces that break the much stronger connections between secondary and primary parts.
Suggestions for Revision: Chapter 11.2 to be expanded with a list of local failures in order of occurrence with details and seriousness as outlined above in Comment No. 1. After each failure the condition of the model is evidently re-analysed by FEA and the results of each element (primary, secondary and tertiary) summarized in Chapter 11.2. Doing that we will know when the situation becomes really serious, e.g. when/if primary parts start to get affected. Evidently a global collapse is only possible if primary parts are affected and we are interested in all local failures leading to that. The draft report is incomplete in this respect.
Also the boundary conditions at floor 16 may be affected by local, initial failures, e.g. if a column fails for any reason you cannot assume that the (boundary) load on it applied at floor 16 remains constant. It would be very easy to extend the ANSYS model to 47 floors to solve that uncertainty. Then you can see how each failure below floor 16 affects the load distribution above floor 16. Thus it is suggested that the ANSYS model is extended to floor 47.
Report Number: NIST NCSTAR 1-9
Page Number: 537-600
Paragraph/Sentence: Chapter 12
Comment No. 2: The LS-DYNA 47 floors model is very big with >3 million elements and >3.5 million nodes and the data of a partly damaged ANSYS model incl. boundary conditions is copy/pasted into it to represent the starting condition. As shown in Comment No. 1 the details of damaged ANSYS model are not clear and it is not certain if it represents a realistic start condition. Another question is if you can copy/paste data of a damaged structure into an undamaged one? What about the boundary conditions at floor 16? Another question is the reliability of the software. Has it been tested properly?
Reason for Comment: The LS-DYNA, like the ANSYS model, consists of primary (vertical columns connected to ground), secondary (horizontal and sloping beams connected to primary parts) and tertiary parts (e.g. floor elements connected to secondary parts; beams) and associated connections, and it is again of vital importance to know the order of failures. We know that the structure at ambient temperature is very low stressed and thus looks very safe.
Heat/thermal expansion may at this time have affected tertiary parts below floor 16, the local connections of which may have failed and the tertiary part is out of action and it has apparently affected the effectiveness of the secondary parts, but the situation is not clear. The secondary parts and their connections to primary parts are much stronger than tertiary parts and will deflect with the primary parts and it is highly unlikely that thermal expansion will produce forces that break the connections between secondary and primary parts. However, it is assumed that some secondary and primary parts below floor 16 have actually failed and shifted out of initial locations affecting the boundary conditions at floor 16.
It is of course of interest to know how these local failures below floor 16 immediately affect the virtually undamaged structure above, when (A) the analysis starts and (B) every further failure that follows and finally, (C) at the end when all parts are rubble.
It is evidently possible that when a primary column fails below floor 16, the load on it is transmitted to adjacent columns via intact structure above floor 16, i.e. the boundary conditions must be modified in the ANSYS model analysis. Was it done?
It is not clear how the software LS-DYNA can keep track of parts that are completely disconnected from the structure due to multiple failures.
Suggestion for Revision: The method to copy/paste details of the ANSYS (Chapter 11) model at end of assumed failures below floor 16 produces uncertainties. It would be better to start afresh with the LS-DYNA model and input all the local - serious (?) failures - one by one - as identified in the ANSYS model below floor 16 and see what happens everywhere at every initial, local failure and then proceed with the further failures, one by one, away from the first failures..
Chapter 12 thus to be expanded with a list of all further local failures above and below floor 16 in order of occurrence with details and seriousness as outlined above in Comment No. 1. After each failure the condition of the model is evidently re-analysed by FEA and the results of each element (primary, secondary and tertiary) summarized in Chapter 12.4. Doing that we will know when the situation becomes really serious, e.g. when/if other primary parts start to get affected and why and what the real failure of total collapse is. Had that failure been avoided, the destruction would have stopped then and there. But apparently the failures continue until total destruction. Then details of elements/parts getting completely detached from the structure can be given and how these loose parts are assumed to load the structure then.
Evidently a global collapse is only possible if all primary parts of the structure are affected and we are interested in all local failures leading to that. The draft report is incomplete in this respect. I am personally quite surprised that the local failures down below around a few columns are not arrested, when running out energy to produce further failures up top. Just because one or two column fails due to local failures, should not cause other, complete intact columns to fail far away.
Actually, if the LS-DYNA software can produce what is suggested, it should be able to simulate all the structural conditions from (A) the completely intact, prior fire, cold condition, (B) all the part damaged conditions with still intact structure left and not least, (C) the end condition, when all structural parts or sub-assemblies are disconnected in the rubble at equilibrium on the ground
There are (C) huge blocks of structure in the rubble with broken primary parts (columns). The LS-DYNA software apparently can simulate how these big blocks bounded by failed elements came about, e.g. how the primary columns were sheared off away from bolted and welded connections, and ended up as seen on many photos. It is also a good test to verify the reliability of the LS-DYNA software, details of which are completely unknown to me. (Only reference to LS-DYNA is a user's manual of little value).
If the LS-DYNA software is as good as suggested, it can be used in analysing structural damages in ship collisions and thus improve safety at sea (my principal interest).
