TO: Penny Beebe
FROM: Casey Stevenson
SUBJECT: Study of the collapse of the World Trade Center towers
DATE: November 5, 2004
The focus of my research project is the collapse of the World Trade Center (WTC) towers fol-lowing the terrorist attacks of September 11, 2001. Specifically, I hope to establish the sequence of events that took place inside the towers that eventually led to their total collapse. This focus has not changed since my last progress report, dated October 15. In this progress report, I will provide all of the information I have obtained since then, although some references to the previ-ous report will be necessary. I have spent an average of 4 hours per week in preparation of this report.
I have acquired a progress report issued by the National Institute of Standards and Technology (NIST) in June 2004 that contains all of their findings to that date. NIST is the government agency performing a complete investigation of the WTC collapse. The Federal Emergency Management Agency (FEMA), whose findings were included in my first progress report, per-formed only the initial investigation; NIST is in the process of a complete investigation.
In addition to a progress report, investigation leaders from NIST made updated presentations of their findings on October 19 to the National Construction Safety Team. The most recent findings are included in the presentations, which have the most current information available. I have also
acquired Forensic Engineering: Proceedings of the Third Congress, a collection of technical pa-pers, two of which deal directly with the structural response of the WTC towers to the terrorist attacks.
III. Information Acquired
A. Tower Design
As mentioned in my proposal, the WTC towers were designed to withstand the accidental impact of a large aircraft. Specifically, the designers planned for a Boeing 707 traveling at 600 mph crashing into the 80th floor. The analysis performed by the designers indicated that such an im-pact would cause local damage to the impact floors but would not cause an entire tower to col-lapse. However, “[t]he effect of the fires due to jet fuel dispersion and ignition of building con-tents was not considered.” (NIST, 2004, p. 4) Skeptics have said that designers planning for an aircraft impact could not have ignored the ensuing fires, but documentation of the design process indicates just that.
B. Computer Modeling
The initial investigation of the collapses revealed almost everything that could be deduced from simple observation of the towers on September 11. One of the most effective ways to learn more about the structural events that took place inside the towers is to use computer programs to model the towers, the aircraft, and the impact of the two bodies. A discussion of the computer modeling of collapse follows.
Abboud et al. used two different computer models in their analysis. First, they modeled the air-craft and towers in a computer program known as FLEX. They went to great lengths to ensure that the towers and aircraft were modeled accurately and that the analysis procedure was consis-tent with the events of an impact. For example, many analysis programs allow structural ele-ments to deform only slightly and do not allow them to crush or break apart. The program used in this analysis was manipulated to allow just such behavior to occur; the aircraft and building elements could be crushed or torn into fragments. Overall, the analysis was very realistic and a good representation of what actually might have happened during the impact. While FLEX was the appropriate program to analyze the impact, a program known as SAP2000 was used to ana-lyze the structure after impact. The results of the FLEX analysis were used as the input for the SAP2000 model. SAP2000 allows for temperature variations (in this case fires) to be included in an analysis and is thus better suited to examine the behavior of the damaged structures under the load of the fires. (2003, pp. 362-364)
NIST used a program known as LS-DYNA to analyze the impact of the aircraft into the towers. They then used SAP2000 and a program called ANSYS to analyze the structure in its damaged state by using the results of the LS-DYNA analysis as input. A typical ANSYS model of a WTC tower core and hat truss is shown in Figure 1. The green areas represent steel members and the white letters are member labels. A fire dynamics program known as FDS was used to simulate the ignition and spread of fires after impact. (Sunder, 2004, p. 18)
C. Aircraft Impact
In my first progress report, I stated that the damage to the cores of the WTC towers could not be quantified and would likely never be known for sure. I have found sources that attempt to de-scribe the damage hidden in the cores by using computer modeling and simulation of the impacts of the aircraft. A more complete analysis of the structure of the WTC towers can be performed by making use of the information in these sources.
Figure 1. ANSYS Computer Model
(Adapted from Gross, 2004, p. 7)
In addition to repeating the findings of the FEMA team investigating damage to the perimeter columns, NIST reports damage to the core columns. In WTC 1, the aircraft severed 3 core columns and severely damaged 10 more in the center of the north face of the core (Sunder, 2004, p. 22). According to analysis by Abboud et al., half of the core columns of WTC 1 were severed or damaged so severely that they lost all their load-carrying capacity (2003, p. 364). In WTC 2, NIST reports that 5 core columns were severed and 5 damaged at the east end of the south face and at the southeast corner of the core (Sunder, 2004, p. 22).
The aircraft impacts damaged other parts of the towers besides the core and perimeter columns. As stated in my first progress report, local floor damage and collapse was visible along the im-
pact faces of the towers. Abboud et al. report that this visible collapse was the extent of floor damage and that floors across the majority of the towers remained intact after the impacts (2003, p. 364). Significant damage to the ceilings occurred, allowing “unabated” heat transfer from the fires to trusses and other structural members normally concealed above the ceiling (NIST, 2004, p. 6). The FLEX model of the impact also defined a debris field, or an area of the tower where flying debris could have removed fireproofing from structural members (Abboud, 2003, p. 364). Later analyses of the fires and the performance of the tower structures reveal that a significant amount of fireproofing in the debris field must have been removed by the aircraft impacts.
D. Fire Development
In my first progress report, I described various methods that investigators used to estimate the size of the fires in the towers. These analyses resulted in estimates of heat output as well as a maximum temperature of the fires. I have acquired new sources that looked more closely at the available evidence and more accurately reconstruct the fires in the towers.
Examination of video evidence and fire modeling in FDS shows that the fires started near the impact regions and progressed across the towers. In both towers, little flaming was observed in the impact areas, but flames were seen moving along the faces of the towers away from the im-pact zone. In WTC 1 significant flames were observed on the south and west faces. In WTC 2, which was impacted on the south face, the most consistent flames were observed along the east side and in the northeast corner of the tower. (Beyler, 2003, p. 373) The fires would ignite in a given location and burn for about 20 minutes, then move down the face of the tower, igniting a new area (NIST, 2004, p.11). This fire progression pattern means that areas of the towers were subject to intense fires for only a short time period, not the entire time between impact and col-lapse.
NIST’s fire simulations indicate that temperatures approached 1000° C in fire areas, but such high temperatures were not sustained over the entire tower area at a given time (Sunder, 2004, pp. 26-27). Beyler et al. agree, but they are very careful to point out that such temperatures were only present where fires were actively burning. In an attempt to dispel media claims that mas-sive, raging fires brought the towers down, the authors show that, on the whole, the fires were 5
less intense than a standard fire. Large amounts of incombustible debris from the towers’ me-chanical systems and from the aircraft would slow the development of the fires. Incombustible gypsum and concrete dust created by the impacts would further slow fire development. A maximum compartment temperature (temperature of the open space on a WTC floor) of 400 – 700° C is estimated, which is structurally significant but not detrimental. Widespread tempera-tures above 1000° C did not develop and fires and extreme temperatures were localized. (2003, pp. 372-380)
E. Effect on Structure
In my first progress report, I discussed how the perimeter columns redistributed loads to adjacent perimeter columns and core columns after the aircraft impact severed several perimeter columns. Damage to core columns was not considered in my first progress report and will thus be de-scribed here. It is the damage to the core columns, combined with that to the perimeter columns, that ultimately caused the towers to collapse.
As mentioned above, the aircraft impacts severed or damaged a number of columns in the core of both towers. The structure responded by redistributing loads previously carried by the damaged core columns to other parts of the structure. First, load was transferred to adjacent core columns via the core framing. Second, load was transferred from the core to the perimeter by the floor system, which directly connected the two. Last, loads were redistributed to intact core and pe-rimeter columns via the hat truss. (NIST, 2004, p. 5) Severed or damaged core columns hung from the hat truss, acting as tension members that suspended the floors above the impact region. The towers remained stable following impact and would collapse only after the fires had had a significant effect on the remaining intact structure.
1. WTC 1
As the fires spread across the floors of WTC 1 they heated the core columns, causing them to lose stiffness and buckle. Loads were again redistributed, as they were after impact, with the core framing, floor system, and hat truss redirecting loads to remaining core columns and pe-rimeter columns. As more core columns lost stiffness and strength, three things occurred.
First, the ability of core framing to transfer loads to adjacent core columns diminished. The core framing was being heated just like the columns, and elements were losing their stiffness and strength. Additionally, if all adjacent columns were themselves buckling and failing, the core framing could not redistribute load among them. (NIST, 2004, p. 5)
Second, connections between core columns and the hat truss failed. They were not designed to resist the enormous tension load required to support the floors hanging below. As each hat truss connection failed, the structure was forced to redistribute loads in a different manner. The only element still capable of transferring loads away from the core was the floor system. (NIST, 2004, p. 5)
Third, the hat truss itself began to fail. The diagonal members of the hat truss (a normal truss, defined in my first progress report) buckled in compression. As the diagonals yielded to the ever-increasing loads, the ability of the hat truss to transfer loads among core columns and be-tween the core and the perimeter diminished. The floor system again was the only remaining structural element that could redistribute loads, and it too was being weakened by the fires. (NIST, 2004, p. 5)
One may ask how the floor system (see Figure 2) could transfer loads among columns. Besides supporting the floors of the WTC towers, the floor system acted as a structural element known as a diaphragm, which is a wide, flat element (think of a piece of plywood as a diaphragm). It is very strong in the plane of the diaphragm, but relatively weak out of the plane. If one pushes down in the center of a piece of plywood, it will deflect downwards. However if one pushes or pulls on the edge of a piece of plywood, the wood will not bend or deflect. The purpose of the floor system in the undamaged structure was to connect the perimeter and core and transfer hori-zontal loads between them (horizontal loads would push and pull on the edge of the floor system, in its strong direction). The floor system was supported at its edges by the perimeter columns and at the center by the core.
As the core columns lost their stiffness and as the hat truss became progressively weaker, the core columns began to displace downward, pulling one end of the floor trusses down with them. The floor system diaphragm now looked like a shallow bowl; the perimeter columns supported its edges but the core columns were pulling its center down. The floor system was now pulling upward on the core from the perimeter columns, acting much like a cable in tension.
Figure 2. Perimeter, Floor System and Core
(Sunder, 2004, p. 33)
As the floor diaphragm began to act as a cable, the localized fires and elevated compartment temperatures were heating it. It began to lose stiffness and strength and started to sag. My first progress report describes floor system sag in response to heating. As the core moved downward and as the floor system lost its stiffness, it acted more and more like a cable in tension; part of the core load was hanging from the floor system that was supported by the perimeter. This cable action, along with the action of the hat truss and core framing, redistributed loads from the core to the perimeter as the core col-umns buckled.
The floor system pulled up on the core columns, also pulling inward on the perimeter columns, which were designed to carry forces along their length, not perpendicular to their length. They bowed in, as seen in Figure 3, near the impact floors. (Sunder, 2004, p. 11) At the same time, the perimeter columns were being heated by the fires, which did not have the same effect on the pe-rimeter columns as they did on the core columns. Three sides of the perimeter columns were ex-posed to the atmosphere and ambient temperatures, while only the interior side was heated. As described in my first progress report, when a member is heated it tends to expand. As the inside face of the perimeter columns was heated, it expanded. The outside face remained at its initial length. Inward bowing of the perimeter columns resulted from this thermal load, augmenting the inward bowing caused by the pull of the floor system (NIST, 2004, p.7).
From the discussion of buckling in my first progress report, we see how columns that are bowed are more likely to buckle. The perimeter columns were still in relatively good condition com-pared to the core columns, but were forced to carry more and more of the load as the core col-umns lost strength. As they became more heavily loaded, they were also pulled inward. Col-lapse initiated when the perimeter columns buckled.
Figure 3. Inward Bowing of Perimeter Columns on WTC 2 East Face.
(Sunder, 2004, p.15)
The collapse initiation was not as instant as it appears to be in videos with which the public is familiar. It began with the buckling and failure of one or two perimeter columns on the south face of WTC 1. The perimeter frame then redistributed the loads carried by the buckled columns to adjacent columns by Vierendeel action (described in my first report). This time, however, adjacent columns had no reserve capacity, as they did when the aircraft first impacted the tower; they were already approaching their ultimate capacities. As the first one or two columns failed and the load they carried was transferred to adjacent columns, these adjacent columns buckled and failed right away. Buckling progressed down the south face of WTC 1 until all columns 9
were buckled; the adjacent column buckling process then turned the corner of the tower, buck-ling columns down the east and west faces. (Sunder, 2004, p. 11)
As one entire floor level of columns failed, the floors above the failure floor began to drop. The buckling of the perimeter and core columns somewhat slowed their downward motion because it requires large amounts of energy to buckle so many columns. However, the potential energy of the upper floors was simply too great and the columns below the failure floor could not stop the mass of the upper floors once they were set in motion.
2. WTC 2
WTC 2 failed in a similar manner to WTC 1, but the failure mechanisms have important differ-ences. Floor trusses in WTC 2 along the east face were subject to sagging, a mechanism de-scribed in my first progress report. The sagging of the floor trusses increased the cable action of the floor system and pulled the perimeter columns in with an even greater force than that in WTC 1. WTC 2 failed first along the east face, with the column failures progressing quickly around the corner to the south and north faces. (Sunder, 2004, p. 12)
As one may recall, damage to the core of WTC 2 was primarily in the southeast corner. When a structure redistributes loads, corner columns are essential – they are like the legs on a table. The WTC 2 core was missing an essential leg. The ability of the WTC 2 core to carry loads was thus quickly reduced, placing the loads on the perimeter columns via the floor system. WTC 2 was also hit about 14 floors lower than WTC 1, meaning that the failure floors of WTC 2 had to carry the weight of 14 more floor levels than those of WTC 1.
IV. Work to be Completed
In the next several days, I will wrap up my research on this topic. I would like to find some of the sources that my sources used so I can get a more first-hand knowledge of the concepts that are being discussed. Among others, I would also like to briefly review an MIT analysis of the collapses that was cited by the recent NIST presentations.
The draft of my final report will require a considerable amount of time to put together from the two progress reports. While all the concepts are currently explained, they need to be tied to-gether and their interdependencies need to be illustrated. The sequence in which I will describe the structural principles will also need to be carefully thought out so that the reader can easily understand what was happening inside the towers.
V. Works Cited
Abboud, N., M. Levy, D. Tennant, J. Mould, H. Levine, S. King, C. Ekwueme, A. Jain, G. Hart. (2003) Anatomy of a Disaster: A Structural Investigation of the World Trade Center Collapses. In: Proceedings of the Third Congress on Forensic Engineering. San Diego: American Society of Civil Engineers.
Beyler, C., D. White, M. Peatross, J. Trellis, S. Li, A. Luers, D. Hopkins. (2003) Analysis of the Thermal Exposure in the Impact Areas of the World Trade Center Terrorist Attacks. In: Proceedings of the Third Congress on Forensic Engineering. San Diego: American Society of Civil Engineers.
Gross, J. (2004) Project 6 – Standard Fire Tests of WTC Tower Typical Floor Construction. National Institute of Standards and Technology presentation to the National Construction Safety Team. October 19, 2004.
National Institute of Standards and Technology (NIST). (2004) June 2004 Progress Report on the Federal Building and Fire Safety Investigation of the World Trade Center Disaster. Washington, D.C.
Sunder, S. (2004) World Trade Center Investigation Status. National Institute of Standards and Technology presentation to the National Construction Safety Team. October 19, 2004.