NIST NCSTAR 1 full text:Chapter 6

❌

RECONSTRUCTION OF THE COLLAPSES 6.1 APPROACH The following presents an overview of the methods used to reach the accounts in Part I. The details may be found in the companion reports to this document, which are indexed in Appendix C. A substantial effort was directed at establishing the baseline performance of the WTC towers, i.e., estimating the expected performance of the towers under normal design loads and conditions. This enabled meeting the third objective of the Investigation, as listed in the Preface to this report. The baseline performance analysis also helped to estimate the ability of the towers to withstand the unexpected events of September 11, 2001. Establishing the baseline performance of the towers began with the compilation and analysis of the procedures and practices used in the design, construction, operation, and maintenance of the structural, fire protection, and egress systems of the WTC towers. The additional components of the performance analysis were: • The standard fire resistance of the WTC truss-framed floor system, • The quality and properties of the structural steels used in the towers, and • The response of the WTC towers to design gravity and wind loads. The second substantial effort was the simulation of the behavior of each tower on September 11, 2001, providing the basis for meeting the first and second objectives of the Investigation. This entailed four modeling steps: 1. The aircraft impact into the tower, the resulting distribution of jet fuel, and the damage to the structure, partitions, insulation materials, and building contents. 2. The spread of the multi-floor fires. 3. The heating of the structural elements by the fires. 4. The response of the damaged and heated building structure, and the progression of structural component failures leading to the initiation of the collapse of the towers. For such complex structures and complex thermal and structural processes, each of these steps stretched the state of the technology and tested the limits of software tools and computer hardware. For example, the investigators advanced the state-of-the-art in the measurement of construction material properties and in structural finite element modeling. New modeling capability was developed for the mapping of firegenerated environmental temperatures onto the building structural components. For the final analyses, four cases were used, each involving all four of the modeling steps. Case A and Case B were for WTC 1, with Case B generally involving more severe impact and fire conditions than Chapter 6 82 NIST NCSTAR 1, WTC Investigation Case A. For WTC 2, Case D involved more severe impact and fire conditions than Case C. The results of the two cases for each tower provided some understanding of the uncertainties in the predictions. There were substantial uncertainties in the as-built condition of the towers, the interior layout and furnishings, the aircraft impact, the internal damage to the towers (especially the insulation), the redistribution of the combustibles, and the response of the building structural components to the heat from the fires. To increase confidence in the simulation results, NIST used information from an extensive collection of photographs and videos of the disaster, eyewitness accounts from inside and outside the buildings, and laboratory tests involving large fires and the heating of structural components. Further, NIST applied formal statistical methods to identify those parameters that had the greatest effect on the model output. These key inputs were then varied to determine whether the results were reasonably robust. The combined knowledge from all the gathered data and analyses led to the development of a probable collapse sequence for each tower,13 the identification of factors that contributed to the collapses, and a list of factors that could have improved building performance or otherwise mitigated the loss of life. 6.2 DEVELOPMENT OF THE DISASTER TIMELINE Time was the unifying factor in combining photographic and video information, survivor accounts, emergency calls from within the towers, and communications among emergency responders. The visual evidence was the most abundant and the most detailed. The destruction of the WTC towers was the most heavily photographed disaster in history. The terrorist attacks occurred in an area that is the national home base of several news organizations and has several major newspapers. New York City is also a major tourist destination, and visitors often carry cameras to record their visits. Further, the very height that made the towers accessible to the approaching aircraft also made them visible to photographers. As a result there were hundreds of both professional and amateur photographers and videographers present, many equipped with excellent equipment and the knowledge to use it. These people were in the immediate area, as well as at other locations in New York and New Jersey. There was a surprisingly large amount of photographic material shot early, when only WTC 1 was damaged. By the time WTC 2 was struck, the number of cameras and the diversity of locations had increased. Following the collapse of WTC 2, the amount of visual material decreased markedly as people rushed to escape the area and the huge dust clouds generated by the collapse obscured the site. There is a substantial, but less complete, amount of material covering the period from the tower collapses to the collapse of WTC 7 late the same afternoon. 13 The focus of the Investigation was on the sequence of events from the instant of aircraft impact to the initiation of collapse for each tower. For brevity in this report, this sequence is referred to as the “probable collapse sequence,” although it does not actually include the structural behavior of the tower after the conditions for collapse initiation were reached and collapse became inevitable. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 83 There were multiple sources of visual material: • Recordings of newscasts from September 11 and afterward, documentaries, and other coverage provided information and also pointed toward other potential sources of material. • Web sites of the major photographic clearinghouses. • Local print media. • NYPD and FDNY. • Collections of visual material assembled for charitable or historical purposes. • Individuals’ photographs and videos that began appearing on the World Wide Web as early as September 11, 2001. • Responses to public appeals for visual material by the Investigation Team. Investigation staff contacted each of the sources, requested the material, made arrangements for its transfer, and addressed copyright and privacy issues. Emphasis was placed on obtaining material in a form as close as possible to the original in order to maintain as much spatial and timing information as possible: direct digital copies of digital photographs and videos, high resolution digitized copies of film or slide photographs, and direct copies from the original source of analog video. The assembled collection included: • 6,977 segments of video footage, totaling in excess of 300 hours. The media videos included both broadcast material and outtakes. Additionally, NIST received videotapes recorded by more than 20 individuals. • 6,899 photographs from at least 200 photographers. As with the videos, many of the photographs were unpublished. This vast amount of visual material was organized into a searchable database in which each frame was characterized by a set of attributes: photographer (name and location), time of shot/video, copyright status, content (including building, face(s), key events (plane strike, fireballs, collapse), the presence of FDNY or NYPD people or apparatus, and other details, such as falling debris, people, and building damage). The development of a timeline for fire growth and structural changes in the WTC buildings required the assignment of times of known accuracy to each video frame and photograph. Images were timed to a single well-defined event. Due to the large number of different views available, the chosen event was the moment the second plane struck WTC 2, established from the time stamps in the September 11 telecasts. Based on four such video recordings, the time of the second plane impact was established as 9:02:59 a.m. The TV network clocks were quite close to the actual time since they were regularly updated from highly accurate geopositioning satellites or the precise atomic-clockbased timing signals provided by NIST as a public service. Chapter 6 84 NIST NCSTAR 1, WTC Investigation Absolute times were then assigned to all frames of all videos that showed the second plane strike. By matching photographs and other videos to specific events in these initially assigned videos, the time assignments were extended to visual materials that did not include the primary event. Times were also cross-matched using additional characteristics, such as the appearance and locations of smoke and fire plumes, distinct shadows cast on the buildings by these plumes, the occurrence of well-defined events such as a falling object, and even a clock being recorded in an image. By such a process, it was possible to place photographs and videos extending over the entire day on a single timeline. As the time was assigned to a particular photograph or video, the uncertainty in the assignment was also logged into the database. In all, 3,032 of the catalogued photographs and 2,673 of the video clips in the databases were timed with accuracies of ± 3 s or better. This process enabled establishing the times of four major events of September 11, listed in Table 6–1. The building collapse times were defined to be the point in time when the entire building was first observed to start to collapse. Table 6–1. Times for major events on September 11, 2001. Event Time First Aircraft Strike 8:46:30 a.m. Second Aircraft Strike 9:02:59 a.m. Collapse of WTC 2 9:58:59 a.m. Collapse of WTC 1 10:28:22 a.m. There were additional sources of timed information. Phone calls from people within the building to relatives, friends, and 9-1-1 operators conveyed observations of the structural damage and developing hazards. Communications among the emergency responders and from the building fire command centers contributed further information about the areas where the external photographers had no access. 6.3 LEARNING FROM THE VISUAL IMAGES The photographic and video images were rich sources of information on the condition of the buildings following the aircraft impact, the evolution of the fires, and the deterioration of the structure. To enable analysis of this information, a shorthand notation (based on the building design drawings) was used to label the exterior columns and windows of the buildings: • First, the faces of the towers were numbered in a manner identical to those used in the original plans: WTC 1: north: 1 east: 2 south: 3 west: 4 WTC 2: west: 1 north: 2 east: 3 south: 4 • The 59 columns across each tower face were assigned three-digit numbers. Following the floor number, the first digit was that of the face, and the remaining two digits were assigned consecutively from right to left as viewed from outside the building. Thus, the fourth column from the right on the east face of the 81st floor of WTC 1 was labeled 81-204. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 85 • Each of the 58 windows on each floor and tower face was assigned the number of the column to its right as viewed from the outside of the building and was also identified by its floor. Thus the rightmost window on the east face of the 94th floor of WTC 1 was labeled 94-201. As an example of information that was extracted, Figure 6–1 shows an enhanced image of the east face of WTC 2. Figure 6–2 expands a section of interest. The amount of detail available is evident. For instance, large piles of debris are present on the north side of the tower on the 80th and 81st floors, and locations where fires are visible or where missing windows are easily identified. Many details of each frame were important in tracking the evolution of the fires and the damage to the buildings. Note: Enhancements by NIST. Figure 6–1. 9:26:20 a.m. showing the east face of WTC 2. In each photograph and each video frame, each window was also coded to indicate whether the window was still in place or not and the extent to which flames and smoke were visible. Color-coded graphics of the four façades of the two towers were then constructed. Examples of these graphics were shown in Chapters 2 and 3. The results of the visual analysis included: • The locations of the broken windows, providing information on the source of air to feed the fires within. • Observations of the spread of fires. • Documentation of the location of exterior damage from the aircraft impact and subsequent structural changes in the buildings. Chapter 6 86 NIST NCSTAR 1, WTC Investigation Note: Enhancements by NIST. Figure 6–2. Close-up of section of Figure 6–1. • Identification of the presence or absence of significant floor deterioration at the building perimeter. • Observations of certain actions by building occupants, such as breaking windows. The near-continuous observations of the externally visible fires provided input to the computer simulations of fire growth and spread. The discrete observations of changes in the displacement of columns and, to a far lesser degree, floors became validation data for the modeling of the approach to structural collapse of the towers. Table 6–2 lists the most important observations. 6.4 LEARNING FROM THE RECOVERED STEEL 6.4.1 Collection of Recovered Steel NIST had two reasons for obtaining specimens of structural steel from the collapsed towers. The primary objective was characterizing the quality of the steel and determining its properties for use in the structural modeling and analysis of the collapse sequences. The second reason was obtaining information regarding the behavior of the steel in the aircraft impact zone and in areas which had major fires. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 87 Table 6–2. Indications of major structural changes up to collapse initiation. Tower Time (a.m.) Observation 10:18 Smoke suddenly expelled on the north face (floors 92, 94, 95 to 98) and west face (92, 94 to 98). 10:23 Inward bowing of perimeter columns on the east side of the south face from floors 94 to 100; maximum extent: 55 in. ± 6 in. at floor 97. WTC 1 10:28:22 First exterior sign of collapse (downward movement of building exterior). Tilting of the building section above the impact and fire area to due south as the structural collapse initiated. First exterior sign of downward movement of building at floor 98. 9:02:59 Exterior fireball from the east face of floor 82 and from the north face from floors 79 to 82. The deflagration prior to the fireballs may have caused a significant pressure pulse to act on floors above and below. 9:21 Inward bowing of exterior wall columns on most of the east face from floors 78 to 83; maximum extent: 7 in. to 9 in. at floor 80. WTC 2 9:58:59 First exterior sign of collapse (downward movement of building exterior). The northeast corner tilted counterclockwise around the base of floor 82. Column buckling was then seen progressing across the north face and nearly simultaneously on the east face. Tilting of the building section above the impact and fire area to the east and south prior to significant downward movement of the upper building section. The tilt to the south did not increase any further as the upper building section began to fall, but the tilt to the east did increase until dust clouds obscured the view. Within weeks of the destruction of the WTC, contractors of New York City had begun cutting up and removing the debris from the site. Members of the FEMA-sponsored and ASCE-led Building Performance Assessment Team, members of the Structural Engineers Association of New York, and Professor A. Astaneh-Asl of the University of California, Berkeley, CA, with support from the National Science Foundation, had begun work to identify and collect WTC structural steel from various recycling yards where the steel was taken during the clean-up effort. The Port Authority of New York and New Jersey (Port Authority) also collected structural steel elements for future exhibits and memorials. Over a period of about 18 months, 236 pieces of steel were shipped to the NIST campus, starting about six months before NIST launched its Investigation. These samples ranged in size and complexity from a nearly complete three-column, three-floor perimeter assembly to bolts and small fragments. Figures 6–3 through 6–5 show some of the recovered steel pieces. Seven of the pieces were from WTC 5. The remaining 229 samples represented roughly 0.25 percent to 0.5 percent of the 200,000 tons of structural steel used in the construction of the two towers. The collection at NIST included samples of all the steel strength levels specified for the construction of the towers. The locations of all structural steel pieces in WTC 1 and WTC 2 were uniquely identified by stampings (recessed letters and numbers) and/or painted stencils. NIST was successful in finding and deciphering these identification markings on many of the perimeter panel sections and core columns, in many cases using metallurgical characterization to complete missing identifiers. In all, 42 exterior panels were positively identified: 26 from WTC 1 and 16 from WTC 2. Twelve core columns were positively identified: eight from WTC 1 and four from WTC 2. Twenty-three pieces were identified as being parts of trusses, although it was not possible to identify their locations within the buildings. Chapter 6 88 NIST NCSTAR 1, WTC Investigation Source: NIST. Figure 6–3. Examples of a WTC 1 core column (left) and truss material (right). Source: NIST. Figure 6–4. WTC 1 exterior panel hit by the fuselage of the aircraft. Overlaying the locations of the specimens with photographs of the building exteriors following the aircraft impact (for perimeter columns and spandrels) and the extent-of-damage estimates (Section 6.8) (for core columns) enabled the identification of steel pieces near the impact zones. These included five specimens of exterior panels from WTC 1 and two specimens of core columns from each of the towers. 6.4.2 Mechanical and Physical Properties NIST determined the properties of many of the recovered pieces for comparison with the original purchase requirements, comparison with the quality of steel from the WTC construction era, and input to the structural models used in the Investigation. Structural steel literature and producers’ documents were used to establish a statistical basis for the variability expected in steel properties. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 89 The properties of the steel samples tested were consistent with the specifications called for in the steel contracts. In particular, the yield strengths of all samples of the floor trusses were higher than called for in the original specifications. This was in part because the truss steels were supplied as a higher grade than specified. Overall, approximately 87 percent of all perimeter and core column steel tested exceeded the required minimum yield strengths specified in design documents. Test data for the remaining samples were below specifications, but were within the expected variability and did not affect the safety of the towers on September 11, 2001. Furthermore, lower strength values measured by NIST could be expected due to (a) differences in test procedures from those used in the qualifying mill tests and (b) the damaged state of the samples. The values of other steel properties were similar to typical construction steels of the WTC construction era. The limited tests on bolts indicated that their strengths were greater than the specified minimum, and they were stronger than contemporaneous literature would suggest as typical. The tested welds performed as expected. NIST measured the stress-strain behavior at room temperature (for modeling baseline performance), high temperature strength (for modeling structural response to fire), and at high strain rates (for modeling the aircraft impact). Based on data from published sources, NIST estimated the thermal properties of the steels (specific heat, thermal conductivity, and coefficient of thermal expansion) and creep behavior for use in the structural modeling of the towers’ response to fire. 6.4.3 Damage Analysis NIST performed extensive analyses of the recovered steel specimens to determine their damage characteristics, failure modes, and (for those near the fire zones) fire-related degradation. In some cases, assessment of enhanced photographic and video images of the towers enabled distinguishing between damage that occurred prior to the collapse and damage that occurred as a result of the collapse. Because the only visual evidence was from the outside of the buildings, this differentiation was only possible for the perimeter panel sections. The observations of fracture and failure behavior, confirmed by an Investigation contractor, were also used to guide the modeling of the towers’ performance during impact and subsequent fires and to evaluate the model output. Source: NIST. Figure 6–5. WTC 1 exterior panel hit by the nose of the aircraft. Chapter 6 90 NIST NCSTAR 1, WTC Investigation For two of the five exterior panels from the impact zone of WTC 1, the general shape and appearance of the recovered pieces matched photographs taken just before the building collapse. Thus, NIST was able to attribute the observed damage to the aircraft impact. NIST also made determinations regarding the connections between structural steel elements: • There was no evidence to indicate that the joining method, weld materials, or welding procedures were inadequate. Fractures of the columns in areas away from a welded joint were the result of stretching and thinning. Perimeter columns hit by the plane tended to fracture along heat-affected zones adjacent to welds. • The failure mode of spandrel connections varied. At or above the impact zone, bolt hole tearout was more common. Below the impact zone, it was more common for the spandrels to be ripped from the panels. There was no evidence that fire exposure changed these failure modes. • The exterior column splices at the mechanical floors, which were welded in addition to being bolted, generally did not fail. The column splices at the other floors generally failed by bolt fracture. • The perimeter truss connectors (or seats) below the impact zone in WTC 1 were predominantly bent down or torn off completely. Above the impact zone, the seats were as likely to be bent upward as downward. Core seats could not be categorized since their asbuilt locations could not be determined. • Failure of core columns was a result of both splice connection failures and fracture of the columns themselves. Examination of photographs showed that 16 of the exterior panels recovered from WTC 1 were exposed to fire prior to the building collapse. None of the nine recovered panels from within the fire floors of WTC 2 were directly exposed to fire. NIST used two methods to estimate the maximum temperatures that the steel members had reached: • Observations of paint cracking due to thermal expansion. Of the more than 170 areas examined on 16 perimeter column panels, only three columns had evidence that the steel reached temperatures above 250 °C: east face, floor 98, inner web; east face, floor 92, inner web; and north face, floor 98, floor truss connector. Only two core column specimens had sufficient paint remaining to make such an analysis, and their temperatures did not reach 250 °C. NIST did not generalize these results, since the examined columns represented only 3 percent of the perimeter columns and 1 percent of the core columns from the fire floors. • Observations of the microstructure of the steel. High temperature excursions, such as due to a fire, can alter the basic structure of the steel and its mechanical properties. Using metallographic analysis, NIST determined that there was no evidence that any of the samples had reached temperatures above 600 ºC. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 91 These results were for a very small fraction of the steel in the impact and fire zones. Nonetheless, these analyses indicated some zones within WTC 1 where the computer simulations should not, and did not, predict highly elevated steel temperatures. 6.5 INFORMATION GAINED FROM OTHER WTC FIRES There had been numerous fires in the towers prior to September 11, 2001. From these, NIST learned what size fire WTC 1 and WTC 2 had withstood and how the tower occupants and the responders functioned in emergencies. While The Port Authority’s records of prior fires were lost in the collapses, FDNY provided reports on 342 fires that had occurred between 1970 and 2001. Most of these fires were small, and occupants extinguished many of them before FDNY arrival. Fortyseven of these fires activated one to three sprinklers and/or required a standpipe hose for suppression. Only two of the fires required the evacuation of hundreds of people. There were no injuries or loss of life in any of these fires, and the interruptions to operations within the towers were local. A major fire occurred in WTC 1 on February 13, 1975, before the installation of the sprinkler system. A furniture fire started in an executive office in the north end of an 11th floor office suite in the southeast corner of the building. The fire spread south and west along corridors and entered a file room. The fire flashed over, broke seven windows, and spread to adjacent offices north and south. The air conditioning system turned on, pulling air into the return air ducts. Telephone cables in the vertical shafts were ignited, destroying the fire-retarded wood paneling on the closet doors. The fire emerged on the 12th and 13th floors, but there was little nearby that was combustible. The fire also extended vertically from the 9th to the 19th floors within the telephone closet. Eventually the fire was confined to 9,000 ft2 of one floor, about one-fourth of the total floor area. The trusses and columns in this area had been sprayed with BLAZE-SHIELD D insulation to a specified ½ in. thickness. Four trusses were slightly distorted, but the structure was not threatened. Only one major fire incident resulted in a whole-building evacuation. At 12:18 p.m. on February 26, 1993, terrorists exploded a bomb in the second basement underground parking garage in the WTC complex. The blast immediately killed six people and caused an estimated $300 million in damage. An intense fire followed and, although the flames were confined to the subterranean levels, the smoke spread into four of the seven buildings in the WTC complex. Most of the estimated 150,000 occupants evacuated the buildings, including approximately 40,000 from the affected towers. In all, 1,042 people were injured in the incident, including 15 who received blast-related injuries. The evacuation of the towers took over 4 hours. The incident response involved more than 700 firefighters (approximately 45 percent of FDNY’s on-duty personnel at the time). In addition, there was a fire on the 104th floor of WTC 1 on September 11, 2001, that apparently did not contribute to the eventual collapse, yet was quite severe. At 10:01 a.m., flames were first observed on the west face, and by 10:07 a.m., intense flames were emanating from several windows in the southern third of that face. The fire raged until the building collapsed at 10:28 a.m. Thus, the tower structure was able to withstand a sizable fire for about 20 min, presumably with the ceiling tile system heavily damaged and the truss system exposed to the flames. The 104th floor was well above the aircraft impact zone, so there should have been little damage to the sprayed fire-resistive material, which was the same (Table 5–3) as Chapter 6 92 NIST NCSTAR 1, WTC Investigation on the floors where the fires led to the onset of the collapse. The photographic evidence showed no signs of column bowing or a floor collapse. 6.6 THE BUILDING STRUCTURAL MODELS 6.6.1 Computer Simulation Software Structural modeling of each tower was required in order to: • Establish the capability of the building, as designed, to support the gravity loads and to resist wind forces; • Simulate the effects of the aircraft impacts; and • Reconstruct the mechanics of the aircraft impact damage, fire-induced heating, and the progression of local failures that led to the building collapse. The varied demands made different models necessary, and different software packages were used for each of these three functions. The reason for the choice in each case is presented in the next three sections of the report. 6.6.2 The Reference Models Under contract to NIST, Leslie E. Robertson Associates (LERA) constructed a global reference model of each tower using the SAP2000, version 8, software. SAP2000 is a software package for performing finite element calculations for the analysis and design of building structures. These global, three-dimensional models encompassed the 110 stories above grade and the six subterranean levels. The models included primary structural components in the towers, resulting in tens of thousands of computational elements. The data for these elements came from the original structural drawing books for the towers. These had been updated through the completion of the buildings and also included most of the subsequent, significant alterations by both tenants and The Port Authority. LERA also developed reference models of a truss-framed floor, typical of those in the tenant spaces of the impact and fire regions of the buildings, and of a beam-framed floor, typical of the mechanical floors. LERA’s work was reviewed by independent experts in light of the firm’s earlier involvement in the WTC design. It was that earlier work, in fact, that made LERA the only source that had the detailed knowledge of the design, construction, and intended behavior of the towers over their entire 38-year life span. The accuracy of the four models was checked in two ways: • The two global models were checked by Skidmore, Owings & Merrill (SOM), also under contract to NIST, and by NIST staff. This entailed ensuring consistency of the models with the design documents, and testing the models, for example, to ensure that the response of the models to gravity and wind loads was as intended and that the calculated stresses and deformations under these loads were reasonable. • The global model of WTC 1 was used to calculate the natural vibration periods of the tower. These values were then compared to measurements from the tower on eight dates of winds Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 93 ranging from 11.5 mph to 41 mph blowing from at least four different directions. As shown in Table 6–3, the N-S and E-W values agreed within 5 percent and the torsion values agreed within 6 percent, both within the combined uncertainty in the measurements and calculations. • SOM and NIST staff also checked the two floor models for accuracy. These reviews involved comparison with simple hand calculations of estimated deflections and member stresses for a simply supported composite truss and beam under gravity loading. For the composite truss sections, the steel stress results were within 4 percent of those calculated by SAP2000 for the long-span truss and within 3 percent for the short-span truss. Deflections for the beams and trusses matched hand calculations to within 5 percent to 15 percent. These differences were within the combined uncertainty of the methods. Table 6–3. Measured and calculated natural vibration periods (s) for WTC 1. Direction of Motion N-S E-W Torsion Average of Measured Data 11.4 10.6 4.9 Original Predicted Values 11.9 10.4 – Reference Global Model Predictions 11.4 10.7 5.2 The few discrepancies between the developed models and the original design documents, as well as the areas identified by NIST and SOM as needing modification, were corrected by LERA and approved by NIST. The models then served as references for more detailed models for aircraft impact damage analysis and for thermal-structural response and collapse initiation analysis. NIST also used these global reference models to establish the baseline performance of the towers under gravity and wind loads. The two key performance measures calculated were the demand-to-capacity ratio (DCR) and the drift. • Demand is defined as the combined effects of the dead, live, and wind loads imposed on a structural component, e.g., a column. Capacity is the permissible strength for that component. Normal design aims at ensuring that DCR values for all components be 1.0 or lower. A value of DCR greater than 1.0 does not imply failure since designs inherently include a margin of safety. • Drift is the extent of sway of the building under a lateral wind. Excessive deflection can cause cracking of partitions and cladding, and, in severe cases, building instability that could affect safety. Using SAP2000, NIST found that, under original WTC design loads, a small fraction of the structural components had DCR values greater than 1.0. (Most DCR values of that small fraction were less than 1.4, with a few as high as 1.6.) For the perimeter columns, DCR values greater than 1.0 were mainly near the corners, on floors near the hat truss, and below the 9th floor. For the core columns, these members were on the 600 line between floors 80 and 106 and at core perimeter columns 901 and 908 for much of their height. (See Figure 1–5 for the column numbers.) One possible explanation to the cause of DCRs in excess of 1.0 may lie in the computer-based structural analysis and software techniques employed for this Chapter 6 94 NIST NCSTAR 1, WTC Investigation baseline performance study in comparison with the relatively rudimentary computational tools used in the original design nearly 40 years ago. As part of its wind analysis, NIST calculated the drift at the top of the towers to be about 5 ft in a nearly 100 mph wind—the wind load used in the original design. Common practice was, and is, to design for substantially smaller deflections; but drift was not, and still is not, a design factor prescribed in building codes. The estimation of wind-induced loads on the towers emerged as a problem. Two sets of wind tunnel tests and analyses were conducted in 2002 by independent laboratories as part of insurance litigation unrelated to the NIST Investigation. The estimated loads differed by as much as 40 percent. NIST analysis found that the two studies used different approaches in their estimations. This difference highlighted limitations in the current state of practice in wind engineering for tall buildings and the need for standards in the field of wind tunnel testing and wind effects estimation. 6.6.3 Building Structural Models for Aircraft Impact Analysis Ideally, the Investigation would have used the reference global models of the towers as the “targets” for the aircraft. However, this was not possible. The impact simulations required inclusion of both a far higher level of detail of the building components and also the highly nonlinear behavior of the tower and aircraft materials, and the larger model size could not be accommodated by the SAP2000 program. There were also physical phenomena for which algorithms were not available in this software. Another finite element package, LS-DYNA, satisfied these requirements and was used for the impact simulations. Early in the effort, it became clear to both NIST and to ARA, Inc., the NIST contractor that performed the aircraft impact simulations, that the model had to “fit” on a state-of-the-art computer cluster and to run within weeks rather than months. To minimize the model size while keeping sufficient fidelity in the impact zone to capture the building deformations and damage distributions, various tower components were depicted with different meshes (different levels of refinement). For example, tower components in the path of the impact and debris field were represented with a fine mesh (higher resolution) to capture the local impact damage and failure, while components outside the impact zone were depicted more coarsely, simply to capture their structural stiffness and inertial properties. The model of WTC 1 included floors 92 through 100; the model of WTC 2 extended from floor 77 through floor 85. The combined tower and aircraft model of more than two million elements, at time steps of just under a microsecond, took approximately two weeks of computer time on a 12-noded computer cluster to capture the needed details of the fraction of a second it took for the aircraft and its fragments to come to rest inside the building. The structural models, partially shown in Figures 6–6 through 6–9, included: • Core columns and spliced column connections; • Floor slabs and beams within the core; • Exterior columns and spandrels, including the bolted connections between the exterior panels in the refined mesh areas; and • Tenant space floors, composed of the combined floor slab, metal decking, and steel trusses. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 95 They also included representations of the interior partitions and workstations. The live load mass was distributed between the partitions and cubicle workstations. Figure 6–6. Structural model of the 96th floor of WTC 1. Figure 6–7. Model of the 96th floor of WTC 1, including interior contents and partitions. Chapter 6 96 NIST NCSTAR 1, WTC Investigation Figure 6–8. Multi-floor global model of WTC 1, viewed from the north. Figure 6–9. Multi-floor global model of WTC 2, viewed from the south. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 97 Within these models, it was critical that the structural and furnishing materials behaved correctly when impacted by the aircraft or debris. For each grade of steel, the stress-strain behavior and the yield strength were represented using data from tests conducted at NIST. The weakening and failure of the concrete floor slabs were simulated using material models embedded in LS-DYNA. The primary influence of the nonstructural components on the impact behavior was their inertial contribution. Values for the resistance to rupture of gypsum panels and the fracture of the wood products in the workstations were obtained from published studies. In order to complete the global models of the two towers, models of sections of the buildings were developed. As shown in Section 6.8.1, these submodels enabled efficient identification of the principal features of the interaction of the buildings with specific aircraft components. 6.6.4 Building Structural Models for Structural Response to Impact Damage and Fire and Collapse Initiation Analysis The structural response and collapse analysis of the towers was conducted in three phases by NIST and Simpson Gumpertz & Heger, Inc. (SGH), under contract to NIST. The first phase included component and detailed subsystem models of the floor and exterior wall panels. The objectives of Phase 1 were to gain understanding into the response of the structure under stress and elevated temperatures, identify dominant modes of failure, and develop reductions in modeling complexity that could be applied in Phase 2. The second phase analyzed major subsystem models (the core framing, a single exterior wall, and full tenant floors) to provide insight into their behavior within the WTC global system. The third phase was the analysis of global models of WTC 1 and WTC 2 that took advantage of the knowledge gained from the more detailed and subsystem models. A separate global analysis of each tower helped determine the relative roles of impact damage and fires with respect to structural stability and sequential failures of components and subsystems and was used to determine the probable collapse initiation sequence. Phase 1: Component and Detailed Subsystem Analyses Floor Subsystem Analysis The floors played an important role in the structural response of the WTC towers to the aircraft impact and ensuing fires. Prior to the development of a floor subsystem model, three component analyses were conducted, as follows: • Truss seats. Figure 6–10 shows how an exterior seat connection was represented in the finite element structural model. The component analysis determined that failure could occur at the bolted connection between the bearing angle and the seat angle, and the truss could slip off the seat. Truss seat connection failure from vertical loads was found to be unlikely, since the needed increase in vertical load was unreasonable for temperatures near 600 °C to 700 °C. • Knuckles. The “knuckle” was formed by the extension of the truss diagonals into the concrete slab and provided for composite action of the steel truss and concrete slab. A model was developed to predict the knuckle performance when the truss and concrete slab acted compositely. Chapter 6 98 NIST NCSTAR 1, WTC Investigation Figure 6–10. Finite element model of an exterior truss seat. • Single composite truss and concrete slab section. A floor section was modeled to investigate failure modes and sequences of failures under combined gravity and thermal loads. The floor section was heated to 700 °C (with a linear thermal gradient through the slab thickness from 700 °C to 300 °C at the top surface of the slab) over a period of 30 min. Initially the thermal expansion of the floor pushed the columns outward, but with increased temperatures, the floor sagged and the columns were pulled inward. Knuckle failure was found to occur mainly at the ends of the trusses and had little effect on the deflection of the floor system. Figure 6–11 shows that the diagonals at the core (right) end of the truss buckled and caused an increase in the floor system deflection, ultimately reaching approximately 42 in. Two possible failure modes were identified for the floor-truss section: sagging of the floor and loss of truss seat support. Figure 6–11. Vertical displacement at 700 oC. Stand-off Plates Seat angle 5/8 in. Diameter bolt Truss top chord Gusset plate Strut Bearing angle MN MX -42.11 -37.357 -32.603 -27.849 -23.095 -18.342 -13.588 -8.834 -4.081 .673211 Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 99 A finite element model of the full 96th floor of WTC 1 was translated from the SAP2000 reference models into ANSYS 8.1 for detailed structural evaluation (Figure 6–12)14. The two models generated comparable predictions of the behavior under dead or gravity loads. Figure 6–12. ANSYS model of 96th floor of WTC 1. The model was used to evaluate structural response under dead and live loads and elevated temperatures, identify failure modes and associated temperatures and times to failure, and identify reductions in modeling complexity for global models and analyses. The structural response included thermal expansion of steel and concrete members, temperature-dependent properties of steel and concrete that affected material stiffness and strength, and bowing or buckling of structural members. The deformation and failure modes identified were floor sagging between truss supports, floor sagging resulting from failure of a seat at either end of the truss, and failure of the floor subsystem truss supports. Exterior Wall Subsystem The exterior walls played an important role in each tower’s reaction to the aircraft impact and the ensuing fires. Photographic and video evidence showed inward bowing of large sections of the exterior walls of both towers just prior to the time of collapse. A finite element model of a wall section was developed in ANSYS for evaluation of structural response under dead and live loads and elevated structural temperatures, determination of loads that would have caused buckling, and identification of reductions of modeling complexity for global models and analyses. The modeled unit consisted of seven full column/spandrel panels (described in Section 1.2.2) and portions of four other panels. The model was validated against the reference model developed by LERA (Section 6.6.2) by comparing the stiffness for a variety of loading conditions. 14 ANSYS allowed including the temperature-varying properties of the structural materials, a necessary feature not available in SAP 2000. Chapter 6 100 NIST NCSTAR 1, WTC Investigation The model was subjected to several gravity loads and heating conditions, several combinations of disconnected floors, and pull-in from sagging floors until the point of instability. In one case, the simulation assumed three disconnected floors, and the top of the wall subsystem was subjected to “pushdown” analysis, i.e., an increasing force to provide a measure of remaining capacity in the wall section. The model captured possible failure modes due large lateral deformations, column buckling from loss of support at floor truss seats and diagonal straps, failure of column splice bolts, and failure of spandrel splice bolts or tearing of spandrel or splice plates at bolt holes. The model also showed: • Large deformations and buckling of the spandrels could be expected at high temperatures, but they did not significantly affect the stability of the exterior columns and generally did not need to be precisely modeled in the tower models. • Partial separations of the spandrel splices could be expected at elevated temperatures, but they also did not significantly affect the stability of the exterior columns. • Exterior column splices could be expected to fail at elevated temperatures and needed to be accurately modeled. • Plastic buckling of columns, with an ensuing rapid reduction of load, was to be expected at extremely high loads and at low temperatures. • The sagging of trusses resulted in approximately 14 kip of inward pull per truss seat on the attached perimeter column. Phase 2: Major Subsystem Analyses Building on these results, ANSYS models were constructed of each of the three major structural subsystems (core framing, a single exterior wall, and full composite floors) for each of the towers. The models were subjected to the impact damage and elevated temperatures from the fire dynamics and thermal analyses to be described later in this chapter. Core Framing The two tower models included the core columns, the floor beams, and the concrete slabs from the impact and fire zones to the highest floor below the hat truss structure: from the 89th floor to the 106th floor for WTC 1 and from the 73rd floor to the 106th floor for WTC 2. Within these floors, aircraft-damaged structural components were removed. Below the lowest floors, springs were used to represent the stiffness of the columns. In the models, the properties of the steel varied with temperature, as described in Section 5.5.2. This allowed for realistic structural changes to occur, such as thermal expansion, buckling, and creep. The forces applied to the models included gravity loads applied at each floor, post-impact column forces applied at the top of the model at the 106th floor, and temperature histories applied at 10 min intervals with linear ramping between time intervals. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 101 Under these conditions, the investigators first determined the stability of the core under impact conditions and then its response under thermal loads: • In WTC 1, the core was stable under Case A (base case) impact damage, but the model could not reach a stable solution under Case B (more severe) impact damage. • The WTC 1 core became unstable under Case A impact damage and Case B thermal loads as it leaned to the northwest (due to insulation dislodged from the northwest corner column); the core model was restrained in horizontal directions at floors above the impact zone half way through the thermal loads. • The WTC 2 core was stabilized for Case C (base case) by providing horizontal restraint at all floors representing the restraint provided by the perimeter wall to resist leaning to the southeast. A converged, stable solution was not found for Case D (more severe) impact damage. • The WTC 2 stabilized core model for Case C impact damage was subjected to Case D thermal loads. Following each simulation, a pushdown analysis was performed to determine the core’s reserve capacity. The analysis results showed that: • The WTC 1 isolated core structure was most weakened from thermal effects at the center of the south side of the core. (Smaller displacements occurred in the global model due to the constraints of the hat truss and floors.) • The WTC 2 isolated core was most weakened from thermal effects at the southeast corner and along the east side of the core. (Larger displacements occurred in the global model as the isolated core model had lateral restraints imposed that were somewhat stiffer than in the global model.) Composite Floor The composite floor model was used to determine the response of a full floor to Case A and B thermal loads for WTC 1 floors and Case C and D thermal loads for WTC 2 floors. It included: • A reduced complexity truss model, validated against the single truss model results. • Primary and bridging trusses, deck support angles, spandrels, core floor beams, and a concrete floor slab. • Fire-generated local temperature histories applied at 10 min intervals with linear ramping between time intervals. • Temperature-dependent concrete and steel properties, except for creep behavior. Chapter 6 102 NIST NCSTAR 1, WTC Investigation • Restraint provided by exterior and core columns, which extended one floor above and below the modeled floor. The potential for large deflections and buckling of individual structural members and the floor system were included. The results showed that: • At lower elevated temperatures (approximately 100 °C to 400 °C), the floors thermally expanded and displaced the exterior columns outward by a few inches; horizontal displacement of the core columns was insignificant. None of the floors buckled as they thermally expanded, even with the exterior columns restrained so that no horizontal movement was allowed at the floors above and below the heated floor, which maximized column resistance to floor expansion. Even with this level of column restraint, the exterior columns did not develop a sufficient reaction force (push inward to resist the expansion outward) to buckle any of the floors. • At higher elevated temperatures (above 400 °C), the floors began to sag as the floors’ stiffness and strength were reduced with increasing temperature, and the difference in thermal expansion between the trusses and the concrete slab became larger. As the floor sagging increased, the outward displacement of the exterior columns was overcome, and the floors exerted an inward pull force on the exterior columns. • Floor sagging was caused primarily by either buckling of truss web diagonals or disconnection of truss seats at the exterior wall or the core perimeter. Except for the truss seat failures near the southeast corner of the core in WTC 2 following the aircraft impact, web buckling or truss seat failure was caused primarily by elevated temperatures of the structural components. • Analysis results from the detailed truss model found that the floors began to exert inward pull forces when floor sagging exceeded approximately 25 in. for the 60 ft floor span. • Sagging at the floor edge was due to loss of vertical support at the truss seats. The loss of vertical support was caused in most cases by the reduction in vertical shear capacity of the truss seats due to elevated steel temperatures. • Case B impact damage and thermal loads for WTC 1 floors resulted in floor sagging on the south side of the tower over floors that reasonably matched the location of inward bowing observed on the south face. Case A impact damage and thermal loads did not result in sagging on the south side of the floors. • Cases C and D impact damage and thermal loads for WTC 2 both resulted in floor sagging on the east side of the tower over floors that reasonably matched the location of inward bowing observed on the east face. However, Case D provided a better match. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 103 Exterior Wall Exterior wall models were developed for the south face of WTC 1 (floors 89 to 106) and the east face of WTC 2 (floors 73 to 90). These sections were selected based on photographic evidence of column bowing. Many of the simulation conditions were similar to those for the isolated core modeling: removal of aircraft-damaged structural components, representation of lower floors by springs, temperature-varying steel properties, gravity loads applied at each floor, post-impact column forces applied at the 106th floor, and temperature histories applied at 10 min intervals with linear ramping between time intervals. The analysis results showed that: • Inward pull forces were required to produce inward bowing consistent with the displacements measured from photographs. The inward pull was caused by sagging of the floors. Heating of the inside faces of the exterior columns also contributed to inward bowing. • Exterior wall sections bowed outward in a pushdown analysis when several consecutive floors were disconnected, the interior face of the columns was heated, and column gravity loads increased (e.g., due to load redistribution from the core and hat truss). At lower temperatures, thermal expansion of the inside face was insufficient to result in inward bowing of the entire exterior column. At higher temperatures, outward bowing resulted from the combined effects of reduced steel strength on the heated inside face, which shortened first under column gravity loads, and the lack of lateral restraint from the floors. • The observed inward bowing of the exterior wall indicated that most of the floor connections must have been intact to cause the observed bowing. • The extent of floor sagging observed at each floor was greater than that predicted by the full floor models. The estimates of the extent of sagging at each floor was governed by the combined effects of insulation damage and fire; insulation damage estimates were limited to areas subject to direct debris impact. Other sources of floor and insulation damage from the aircraft impact and fires (e.g., insulation damage due to shock and subsequent vibrations as a result of aircraft impact or concrete slab cracking and spalling as a result of thermal effects) were not included in the floor models. • Case B impact damage and thermal loads for the WTC 1 south wall, combined with pull-in forces from floor sagging, resulted in an inward bowing of the south face that reasonably matched the observed bowing. The lack of floor sagging for the Case A impact damage and thermal loads resulted in no inward bowing for the south face. • Cases D impact damage and thermal loads for the WTC 2 east wall, combined with pull-in forces from floor sagging, resulted in an inward bowing of the east face that reasonably matched the observed bowing. Chapter 6 104 NIST NCSTAR 1, WTC Investigation Phase 3: Global Modeling The global models were used for the two final simulations and provided complete analysis of results and insight into the subsystem interactions leading to the probable collapse sequence. Based upon the results of the major subsystem analyses, impact damage and thermal loads for Cases B and D were used for WTC 1 and WTC 2, respectively. The models extended from floor 91 for WTC 1 and floor 77 for WTC 2 to the roof level in both towers. Although the renditions of the structural components had been reduced in complexity while maintaining essential nonlinear behaviors, based on the findings from the component and subsystem modeling, the global models included many of the features of the subsystem models: • Removal of aircraft-damaged structural components. • Application of gravity loads following removal of aircraft damaged components and prior to thermal loading. • Temperature-dependent concrete and steel properties. • Creep strains for column components. • Representation of lower floors by springs. • Local temperature histories applied at 10 min intervals with linear ramping between time intervals. There were several adjustments to the models based on the findings from the subsystem modeling: • Removal of thermal expansion from the spandrels and equivalent slabs in the tenant area to avoid local buckling that affected convergence but had little influence on global collapse initiation. • Representing the WTC 2 structure above the 86th floor as a single “super-element” to reduce model complexity. The floors above the impact zone had only exhibited linear behavior in the previous analyses. This modification assumed linear behavior of the hat truss, which was checked as part of the review of analysis results. • Representation of the lower part of the tower (starting several floors below the impact damage) as a super-element. This prevented the use of construction sequence in applying gravity loads to the model (where loads are applied in stages to simulate the construction of the building). The lack of construction sequence increased the forces on the exterior columns slightly, and decreased those on the core columns slightly. The inclusions of creep for column components was necessary for the accuracy of the models, but its addition also greatly increased the computation time. As a result, the simulations of WTC 1 took 22 days and those of WTC 2 took 14 days on a high-end computer workstation. The results of these simulations are presented in Section 6.14. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 105 6.7 THE AIRCRAFT STRUCTURAL MODEL Due to their similarity, the two Boeing 767-200ER aircraft were represented by a single, finite element model, two views of which are shown in Figure 6–13. The model consisted of about 800,000 elements. The typical element dimensions were between 1 in. and 2 in. for small components, such as spar or rib flanges, and 3 in. to 4 in. for large parts such as the wing or fuselage skin. Structural data on which to base the model were collected from the open literature, electronic surface models and CAD drawings, an inspection of a 767-300ER, Pratt and Whitney Engine Reference Manuals, American Airlines and United Airlines, and the Boeing Company website. Figure 6–13. Finite element model of the Boeing 767-200ER. Chapter 6 106 NIST NCSTAR 1, WTC Investigation More detailed models of subsections of the aircraft were constructed for the component level analyses described below. Special emphasis was placed on modeling the aircraft engines, due to their potential to produce significant damage to the tower components. The element dimensions were generally between 1 in. and 2 in., although even smaller dimensions were required to capture some details of the engine geometry. The various components of the resulting engine model are shown in Figure 6–14. Fuel was distributed in the wing as shown in Figure 6–15 based on a detailed analysis of the fuel distribution at the time of impact. Figure 6–14. Pratt & Whitney PW4000 turbofan engine model. Figure 6–15. Boeing 767-200ER showing the jet fuel distribution at time of impact. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 107 6.8 AIRCRAFT IMPACT MODELING 6.8.1 Component Level Analyses Prior to conducting the full simulations of the aircraft impacting the towers, a series of smaller scale simulations was performed to develop understanding of how the aircraft and tower components fragmented and to develop the simulation techniques required for the final computations. These simulations began with finely meshed models of key components of the tower and aircraft structures and progressed to relatively coarsely meshed representations that could be used in the global models. Examples of these component-level analyses included impact of a segment of an aircraft wing with an exterior column, impact of an aircraft engine with exterior wall panels, and impact of a fuel-filled wing segment with exterior wall panels. Figure 6–16 shows two frames from the last of these analyses, with the wing segment entering from the left, being fragmented as it penetrates the exterior columns, and spraying jet fuel downstream. t = 0.0 s t = 0.04 s Figure 6–16. Calculated impact on an exterior wall by a fuel-laden wing section. The Investigation Team gained valuable knowledge from these component impact analyses, for example: • Moving at 500 mph, an engine broke any exterior column it hit. If the engine missed the floor slab, the majority of the engine core remained intact and had enough residual momentum to sever a core column upon direct impact. • The impact of the inner half of an empty wing significantly damaged exterior columns but did not result in their complete failure. Impact of the same wing section, but filled with fuel, did result in failure of the exterior columns. Chapter 6 108 NIST NCSTAR 1, WTC Investigation 6.8.2 Subassembly Impact Analyses Next, a series of simulations were performed for intermediate-sized sections of a tower. These subassembly analyses investigated different modeling techniques and associated model sizes, run times, numerical stability, and impact response. Six simulations were performed of an aircraft engine impacting a subassembly that included structural components from the impact zone on the north face of WTC 1, exterior panels, truss floor structures, core framing, and interior contents (workstations). One response of the structure to the engine impact is shown in Figure 6–17. Figure 6–17. Response of a tower subassembly model to engine impact. Typical knowledge gained from these simulations were: • The mass of the concrete floor slab and nonstructural contents had a greater effect on the engine deceleration and subsequent damage than did the concrete strength. • Variation of the failure criteria of the welds in the exterior columns did not result in any noticeable difference in the damage pattern or the energy absorbed by the exterior panels. 6.8.3 Aircraft Impact Conditions From the NIST photographic and video collection, the speed and orientation of the aircraft (Table 6–4) were estimated at the time of impact. The geometry of the wings, different in flight from that at rest, was estimated from the impact pattern in the photographs and the damage documented on the exterior panels by NIST. United Airlines and American Airlines provided information on the contents of the aircraft, the mass of jet fuel, and the location of the fuel within the wing tanks. Table 6–4. Summary of aircraft impact conditions. Condition AA 11 (WTC 1) UAL 175 (WTC 2) Impact Speed (mph) 443 ± 30 542 ± 24 Vertical Approach Angle 10.6° ± 3° below horizontal (heading downward) 6° ± 2° below horizontal (heading downward) Lateral Approach Angle 180.3° ± 4° clockwise from Plan Northa 13° ± 2° clockwise from Plan Northa Roll Angle (left wing downward) 25° ± 2° 38° ± 2° a. Plan North is approximately 29 degrees clockwise from True North. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 109 6.8.4 Global Impact Analysis From the component and subassembly simulations, it became apparent that each computation of the full tower and aircraft would take weeks. Furthermore, the magnitude and location of damage to the tower structure were sensitive to a large number of initial conditions, to assumptions in the representation of the collision physics, and to any approximations in the numerical methods used to solve the physics equations. Thus, it was necessary to choose a manageable list of the factors that most influenced the outcome of a simulation. Careful screening was conducted at the component and subassembly levels, leading to identification of the following prime factors: • Impact speed, • Vertical approach angle of the aircraft, • Lateral approach angle of the aircraft, • Total aircraft weight, • Aircraft materials failure strain, • Tower materials failure strain, and • Building contents weight and strength. Guided by these results and several preliminary global simulations, two global simulations were selected for inclusion in the four-step simulation of the response of each tower, as described in Section 6.1. The conditions for these four runs are shown in Table 6–5. The computers simulate the aircraft flying into the tower, calculated the fragments that were formed from both the aircraft and the building itself, and then followed the fragments. The jet fuel, atomized upon impact into about 60,000 “blobs” averaging one pound, dispersed within and outside the building. Each simulation continued until the debris motion had reduced to a level that was not expected to produce any significant further impact damage. Table 6–5. Input parameters for global impact analyses. WTC 1 WTC 2 Analysis Parameters Case A Case B Case C Case D Impact Speed 443 mph 472 mph 542 mph 570 mph Vertical Approach Flight Parameters Angle 10.6° 7.6° 6.0° 5.0° Lateral Approach Angle 180.0° 180.0° 13.0° 13.0° Weight 100 % 105 % 100 % 105 % Aircraft Parameters Failure Strain 100 % 125 % 100 % 115 % Failure Strain 100 % 80 % 100 % 90 % Live Load Weighta Tower Parameters 25 % 20 % 25 % 20 % Contents Strength 100 % 100 % 100 % 80 % a. Live load weight expressed as a percentage of the design live load. Chapter 6 110 NIST NCSTAR 1, WTC Investigation These simulations each took about 2 weeks on a 12-node computer cluster. Figure 6–18 shows six frames from the animation of one such simulation. (a) Time=0.00 s (b) Time=0.10 s (c) Time=0.20 s Figure 6–18. Side view of simulated aircraft impact into WTC 1, Case B. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 111 (d) Time=0.30 s (e) Time=0.40 s (f) Time=0.50 s Figure 6–18. Side view of simulated aircraft impact into WTC 1, Case B (Cont.) Chapter 6 112 NIST NCSTAR 1, WTC Investigation 6.9 AIRCRAFT IMPACT DAMAGE ESTIMATES 6.9.1 Structural and Contents Damage Each of the four global simulations generated information about the state of the structural components following the impact of the aircraft. The four degrees of column damage are defined as follows and shown graphically in Figure 6–19. The unstrained areas are blue and the highly strained areas are red. • Lightly damaged column: column impacted, but without significant structural deformation; • Moderately damaged column: visible local distortion, but no deformation of the column centerline; • Heavily damaged column: Permanent deflection of the column centerline; and • Failed column: Column severed. (a) Light (b) Moderate (c) Heavy (d) Severed Figure 6–19. Column damage levels. Figure 6–20 shows the calculated damage to a floor slab. Figure 6–21 shows the response of the furnishings and the jet fuel to the impact. Figures 6–22 through 6–25 show the combined damage for all floors for the four global simulations. The latter proved useful in visualizing the extent of aircraft impact in one graphic image. Figure 6–20. Case B damage to the slab of floor 96 of WTC 1. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 113 (a) Pre-impact configuration (b) Calculated impact response (c) Calculated impact response (fuel removed) Figure 6–21. Case B simulation of response of contents of 96th floor of WTC 1. Chapter 6 114 NIST NCSTAR 1, WTC Investigation Figure 6–22. Combined structural damage to the floors and columns of WTC 1, Case A. Figure 6–23. Combined structural damage to the floors and columns of WTC 1, Case B. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 115 Figure 6–24. Combined structural damage to the floors and columns of WTC 2, Case C. Figure 6–25. Combined structural damage to the floors and columns of WTC 2, Case D. Chapter 6 116 NIST NCSTAR 1, WTC Investigation 6.9.2 Validity of Impact Simulations Assessment of the aircraft impact simulations of exterior damage to the towers involved comparing the predicted perimeter wall damage near the impact zone with post-impact photographs of the walls. Figure 6–26 shows a photograph of the north face of WTC 1 after impact and the results of the Case A simulation. The calculated silhouettes capture both the position and shape of the actual damage. Figures 6–27 and 6–28 depict more detailed comparisons between the observed and calculated damage. The aircraft hole is shown in white. The colored dots characterize the mode in which the steel or connection failed (e.g., severed bolt, ripped weld) and the magnitude of the deformation of the steel: • Green: proper match of failure mode and magnitude • Yellow: proper match in the failure mode, but not the magnitude • Red: neither the failure mode nor the magnitude matched • Black: the observed damage was obscured by smoke, fire, or other factors The predominance of green dots and the scarcity of red dots indicate that the overall agreement with the observed damage was very good. The agreement for Cases B and D was slightly lower. Assessment of the accuracy of the predictions of damage inside the buildings was more difficult, as NIST could not locate any interior photographs near the impact zones. Three comparisons were made: • The Case A simulation for WTC 1 predicted that the walls of all three stairwells would have been collapsed. This agreed with the observations of the building occupants. The Case A simulation for WTC 2 showed that the walls of stairwell B would have been damaged, but that Stairwell A would have been unaffected. Stairwell C was not included in the WTC 2 model, but is adjacent to where damage occurred. The building occupants reported that Stairwells B and C were impassable; Stairwell A was damaged but passable. • The two simulations of WTC 2 showed accumulations of furnishings and debris in the northeast corner of the 80th and 81st floors. These piles were observed in photographs and videos. • Two pieces of landing gear penetrated WTC 1 and landed to the south of the tower. The Case B prediction showed landing gear penetrating the building core, but stopping before reaching the south exterior wall. For WTC 2, a landing gear fragment and the starboard engine penetrated the building and landed to the south. The Case D prediction correctly showed the main landing gear emerging from the northeast corner of WTC 2. However, Case D showed that engine not quite penetrating the building. Minor modifications to the model (all within the uncertainty of the input data) would have resulted in the engine passing through the north exterior wall of the tower. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 117 (a) Observed Damage = (b) Calculated damage Figure 6–26. Observed and Case A calculated damage to the north face of WTC 1. Chapter 6 118 NIST NCSTAR 1, WTC Investigation Figure 6–27. Schematic of observed damage (top) and Case A calculated damage (lower) to the north face of WTC 1. Figure 6–28. Schematic of observed damage (above) and Case C calculated damage (right) to the south face of WTC 2. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 119 Not all of the observables were closely matched by the simulations due to the uncertainties in exact impact conditions, the imperfect knowledge of the interior tower contents, the chaotic behavior of the aircraft breakup and subsequent debris motion, and the limitations of the models. In general, however, the results of the simulations matched these observables sufficiently well that the Investigation Team could rely on the predicted trends. Simulations of the damage to the core columns had been performed previously by staff of Weidlinger Associates, Inc. (WAI) and the Massachusetts Institute of Technology (MIT). Each developed a range of numbers of failed and damaged columns, as did NIST. The range of the MIT results straddled the NIST results. WAI’s analysis resulted in more failed and damaged columns, with WTC 2 being unstable immediately following impact. 6.9.3 Damage to Thermal Insulation The dislodgement of thermal insulation from structural members could have occurred as a result of (a) direct impact by debris and (b) inertial forces due to vibration of structural members as a result of the aircraft impact. The debris from the aircraft impact included the fragments that were formed from both the aircraft (including the contents and fuel) and the building (structural members, walls, and furnishings). In interpreting the output of the aircraft impact simulations, NIST assumed that the debris impact dislodged insulation if the debris force was strong enough to break a gypsum board partition immediately in front of the structural component. Experiments at NIST confirmed that an array of 0.3 in. diameter pellets traveling at approximately 350 mph stripped the insulation from steel bars like those used in the WTC trusses. Determining the adherence of SFRM outside the debris zones was more difficult. There was photographic evidence that some fraction of the SFRM was dislodged from perimeter columns not directly impacted by debris. NIST developed a simple model to estimate the range of accelerations that might dislodge the SFRM from the structural steel components. As the SFRM in the towers was being upgraded with BLAZESHIELD II in the 1990s, The Port Authority had measured the insulation bond strength (force required to pull the insulation from the steel). The model used these data as input to some basic physics equations. The resulting ranges of accelerations depended on the geometry of the coated steel component and the SFRM thickness, density and bond strength. For a flat surface (as on the surface of a column), the range was from 20g to 530g, where g is the gravitational acceleration. For an encased bar (such as used in the WTC trusses), the range was from 40g to 730g. NIST estimated accelerations from the aircraft impacts of approximately 100g. In determining the extent of insulation damage in each tower, NIST only assumed damage where dislodgement criteria could be established and supported through observations or analysis. Thus, NIST made the conservative assumption that insulation was removed only where direct debris impact occurred and did not include the possibility of insulation damage or dislodgement from structural vibration. This assumption produced a lower bound on the bared steel surface area, thereby making it more difficult to heat the steel to the point of failure. Chapter 6 120 NIST NCSTAR 1, WTC Investigation An intact ceiling tile system could have provided the floor trusses with approximately 10 min to 15 min of thermal protection from ceiling air temperatures near 1,000 °C. These temperatures would quickly heat steel without thermal insulation to temperatures for reduction of the strength of structural steels. 6.9.4 Damage to Ceiling System The aircraft impact modeling did not include the ceiling tile systems. To estimate whether the tiles would survive the aircraft impact, the University at Buffalo, under contract to NIST, conducted tests of WTC-like ceiling tile systems using their shake table (Figure 6–29) and impulses related to those induced by the aircraft impact on the towers. The data indicated that accelerations of approximately 5g would most likely result in substantial displacement of ceiling tiles. Given the estimated impact accelerations of approximately 100g, NIST assumed that the ceiling tiles in the impact and fire zones were fully dislodged. This was consistent with the multiple reports of severely damaged ceilings (Chapter 7). Source: NIST Figure 6–29. Ceiling tile system mounted on the shaking table. 6.9.5 Damage to Interior Walls and Furnishings As shown in Figure 6–18, the aircraft impact simulations explicitly included the fracture of walls in the debris path and the “bulldozing” of furnishings. Damage to the impacted furnishings was not modeled. Walls and furnishings outside the debris paths were undamaged in the simulations. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 121 6.10 THERMAL ENVIRONMENT MODELING 6.10.1 Need for Simulation Following the impact of the aircraft, the jet-fuel-ignited fires created the sustained and elevated temperatures that heated the remaining building structure to the point of collapse initiation. The photographic evidence provided some information regarding the locations and spreading of the fires. However, the cameras could only see the periphery of the building interior. The steep viewing angles of nearly all of the photographs and videos further limited the depth of the building interior for which fire information could be obtained. NIST could not locate any photographic evidence regarding the fire exposure of the building core or the floor assemblies. The simulations of the fires were the second computational step in the identification of the probable sequences leading to the collapse of the towers. The required output of these simulations was a set of three-dimensional, time varying renditions of the thermal and radiative environment to which the structural members in the towers were subjected from the time of aircraft impact until the onset of building collapse. The rigor of the Investigation placed certain requirements on the computational tool (model) used to generate these renditions: • Resolution of the varying thermal environment across key dimensions, e.g., the truss space; • Representation of the complex combustibles; • Computation of flame spread across the large expanses of the WTC floors; and • Confidence in the accuracy of the predictions. 6.10.2 Modeling Approach The time frame of the Investigation and the above requirements led to the use of the Fire Dynamics Simulator (FDS). Under development at NIST since 1978, FDS was first publicly released in February 2000 and had been used worldwide on a wide variety of applications, ranging from sprinkler activation to residential and industrial fire reconstructions. However, it had never before been applied to spreading fires in a building with such large floor areas. Figure 6–30 shows how FDS represented the eight modeled floors (92 through 99) of the undamaged WTC 1. A similar rendition was prepared for floors 78 through 83 of WTC 2. The layout of each floor was developed from architectural drawings and from the information described in Section 5.8. There was a wide range of confidence in the accuracy of these floor plans, varying from high (for the floors occupied by Marsh & McLennan in WTC 1, for which recent and detailed plans were obtained) to low (for most of the space in WTC 2 occupied by Fuji Bank, for which floor plans were not available). The effects of the aircraft impact were derived from the simulations described in Section 6.8. The portions of walls and floors that were “broken” in those simulations were simply removed from the FDS models of the towers. The furnishings outside the aircraft-damaged regions were assumed to be unmoved and undamaged. The treatment of furnishings within the impact zone is discussed later in this section. Chapter 6 122 NIST NCSTAR 1, WTC Investigation FDS represented the spaces in which the fires and their effluent were to be modeled as a grid of rectangular cells. These grids included the walls, floors, ceilings, and any other obstructions to the movement of air and fire. In the final simulations, the grid size was 0.5 m x 0.5 m x 0.4 m high (1.6 ft x 1.6 ft x 1.3 ft.). Each floor contained about 125,000 grid cells, and the nature of each cell was updated every 10 ms (100 times every second). The computations were performed using parallel processing, in which the fires on each floor were simulated on a different computer. At the end of each 10 ms update, the processors exchanged information and proceeded to the computations for the next time interval. Each simulation of 105 min of fires for WTC 1 took about a week on eight Xeon computers with a combined 16 GB of memory. The simulations for WTC 2, with fewer floors and 60 min of real time fires, took about half the time. Figure 6–30. Eight floor model of WTC 1 prior to aircraft impact. The fires were started by ignition of the jet fuel, whose distribution was provided by the aircraft impact simulations. The radiant energy from these short-lived fires heated the nearby combustibles, creating flammable vapors. When these mixed with air in the right proportion within a grid cell, FDS burned the mixture. This generated more energy, which heated the combustibles further, and continued the burning. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 123 The floors of the tower on which the dominant burning occurred were characterized by large clusters of office workstations (Figure 1–11). NIST determined their combustion behavior from a series of single-workstation fire tests (Figure 6–31). In these highly instrumented tests, the effects of workstation type, the presence of jet fuel, and the presence of fallen inert material (such as pieces of ceiling tiles or gypsum board walls) on the burning surfaces were all assessed. While FDS properly captured the gross behavior of these fires, the state of modeling the combustion of real furnishings was still primitive. Thus, the results of this test series were used to refine the combustion module in FDS. The accuracy of FDS predictions was then assessed using two different types of fire tests. In each case, the model predictions were generated prior to conducting the test. The first series provided a measure of the ability of FDS to predict the thermal environment generated by a steady state fire. A spray burner generating 1.9 MW or 3.4 MW of power was ignited in a 23 ft by 11.8 ft by 12.5 ft high compartment. The temperatures near the ceiling approached 900 °C. FDS predicted: • Room temperature increases near the ceiling to within 4 percent. • Gas velocities at the air inlet to the compartment (and thus the air drawn into the compartment by the fire) within the uncertainty in the experimental measurements. • The leaning of the fire plume due to the asymmetry of the objects within the compartment. The extent of the leaning was underestimated. • Radiant heat flux near the ceiling to within 10 percent, within the uncertainty of the experimental measurements. The second series was a preamble to the modeling of the actual WTC fires. Arrays of three WTC workstations were burned in a 35.5 ft by 23 ft by 11 ft high compartment (Figure 6–32). The tests examined the effects of the type of workstation, the presence of jet fuel, and the presence of fallen inert material on the burning surfaces. In one of the tests, the workstations were rubblized (Figure 6–33). Figure 6–34 depicts the intensity of the test fires. Figure 6–35 shows the measured and predicted heat release rate data from one of the tests in which there was no jet fuel nor inert material present. Source: NIST. Figure 6–31. Fire test of a single workstation. The large fires discussed in this report are characterized by heat release rate, or burning intensity, (in MW), by total energy released (in GJ), and by the heat flux, or radiant intensity (in kW/m2). Chapter 6 124 NIST NCSTAR 1, WTC Investigation Figure 6–32. Interior view of a three-workstation fire test. Source: NIST. Figure 6–33. Rubblized workstations. Source: NIST. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 125 Source: NIST Figure 6–34. Three-workstation fire test, 2 min after the start. Figure 6–35. Measured and predicted heat release rate from the burning of three office workstations. The differences in the fire behavior under the different experimental conditions were profound in these roughly hour-long tests. The jet fuel greatly accelerated the fire growth. Only about 60 percent of the Chapter 6 126 NIST NCSTAR 1, WTC Investigation combustible mass of the rubblized workstations was consumed. The near-ceiling temperatures varied between 800 °C and 1,100 °C. Nonetheless, FDS successfully replicated: • The general shape and magnitude of the time-dependent heat release rate. • The time at which one half of the combustion energy was released to within 3 min. • The value of the heat release rate at this time to within 9 percent. • The duration of the fires to within 6 min. • The peak near-ceiling temperature rise to within 10 percent. All these predictions were within the combined uncertainty in the model input data and the experimental measurements. Combined, these results led to the assessment that the uncertainty in the thermal environment predictions of the WTC fires would be dominated less by the FDS errors and more by the unknowns in such factors as the distribution of the combustibles, ventilation, and building damage. 6.10.3 The Four Cases Four fire scenarios (Case A and Case B for WTC 1 and Case C and Case D for WTC 2) were superimposed on the four cases of aircraft-driven damage of the same names (Section 6.9). A number of preliminary simulations had been performed to gain insight into the factors having the most influence on the severity of the fires. The most influential was the mass of combustibles per unit of floor area (fuel load); second was the extent of core wall damage, which affected the air supply for the fires. The aforementioned workstation fire tests had also indicated that the damage condition of the furnishings also played a key role. The scenario variables and their values are shown in Table 6–6. Table 6–6. Values of WTC fire simulation variables. WTC 1 WTC 2 Variable Case A Case B Case C Case D Tenant combustible fuel loada 4 lb/ft2 5 lb/ft2 4 lb/ft2 5 lb/ft2 Distribution of disturbed combustibles Even Weighted toward the core Heavily concentrated in the northeast corner Moderately concentrated in the northeast corner Condition of combustibles Undamaged except in impact zone Displaced furniture rubblized All rubblized Undamaged except in impact zone Representation of impacted core wallsb Fully removed Soffit remained Fully removed Soffit remained a. In addition, approximately 27,000 lb of solid combustibles from the aircraft were distributed along the debris path. b. In Cases A and C, the walls impacted by the debris field were fully removed. This enabled rapid venting of the upper layer into the core shafts and reduced the burning rate of combustibles in the tenant spaces. In Cases B and D, a more severe representation of the damage was to leave a 4 ft gypsum wallboard soffit that would maintain a hot upper layer on each fire floor. This produced a fire of longer duration near the core columns and the attached floor membranes. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 127 FDS contained no algorithm for breaking windows from the heat of the fires. Thus, during each simulation, windows were removed at times when photographs indicated they were first missing. Damage to the ventilation shafts was derived from the aircraft impact simulations. For undamaged floors, all the openings to the core area were assumed to total about 50 ft2 in area. 6.10.4 Characterization of the Fires For each of the four scenarios, FDS was used to generate a time-dependent gas temperature and radiation environment on each of the floors. The results of the FDS simulations of the perimeter fire were compared with the fire duration and spread rate as seen in the photographs and videos. For ease of visualization, contour plots of the room gas temperature 1.3 ft below the ceiling slab (in the “upper layer” of the compartment) were superimposed on profiles of the photographed fire activity. An example is shown in Figure 6–36. The stripes surrounding the image represent a summary of the visual observations of the windows, with the black stripes denoting broken windows, the orange stripes denoting external flaming, and the yellow stripes denoting fires that were seen inside the building. Fires deeper than a few meters inside the building could not be seen because of the smoke obscuration and the steep viewing angle of nearly all the photographs. Figure 6–36. Upper layer temperatures on the 94th floor of WTC 1, 15 min after impact. Chapter 6 128 NIST NCSTAR 1, WTC Investigation Given the uncertainties in some of the floor plans, the damage to the internal walls, and movement of the office furnishings, the intent of the simulations was to capture the magnitudes of the fires and the broad features of their locations and movement; and they did so. The following sections summarize the simulated behavior of the fires (which was used in the following stages of the disaster reconstruction) and their correlation with the analysis of the photographic evidence. WTC 1 Much of the fire activity was initially in the vicinity of the impact area in the north part of the building. As a result of the orientation of the impacting aircraft and its fuel tanks, the early fires on the 92nd through 94th floors tended toward the east side of the north face, while the early fires on the 97th through 99th floors tended toward the west side of the north face. The fires on all the floors spread along the east and west sides and were concentrated in the south part of the building at the time of collapse, as depicted in Figure 6–37. Figure 6–37. Direction of simulated fire movement on floors 94 and 97 of WTC 1. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 129 The fire simulation results for Case A and Case B were similar, indicating only a modest sensitivity to the fuel load and the degree of aircraft-generated damage. This was because, in general, the size and movement of the fires in WTC 1 were limited by the supply of air from the exterior windows. Since the window breakage pattern was not changed in Case B, the additional and re-distributed combustibles within the building did not contribute to a larger fire. The added fuel did slow the spread slightly because the fires were sustained longer in any given location. Although there was generally reasonable agreement between the simulated and observed fire spread rates, there were instances where the fires burned too quickly and too near the windows. This resulted from an artifact of the model: the combustible vapors burned immediately upon mixing with the incoming oxygen. Simulations performed with doubled fuel loads slowed the fire spread well below the observed rates. Combined with the above results, this suggested that the estimated overall combustible load of 4 lb/ft2 was reasonable. The simulations showed high temperatures in some of the elevator shafts. The late fire observed on the west face of the 104th floor may have started from fuel gases in the core shafts that had accumulated over the course of the first hour of fires below. The presence of fire in the shafts on the 99th floor provided some support for this hypothesis, but no simulations were performed for floors higher than the 99th. The predictions of maximum temperatures (e.g., red zones in Figure 6–37) were consistent with those in the three-workstation fire tests. The use of an “average” gas temperature was not a satisfactory means of assessing the thermal environment on floors this large and would also have led to large errors in the subsequent thermal and structural analyses. The heat transferred to the structural components was largely by means of thermal radiation, whose intensity is proportional to the fourth power of the gas temperature. At any given location, the duration of temperatures near 1,000 °C was about 15 min to 20 min. The rest of the time, the calculated temperatures were near 500 °C or below. To put this in perspective, the radiative intensity onto a truss surrounded by smoke-laden gases at 1,000 °C was approximately 7 times the value for gases at 500 °C. WTC 2 Simulating the fires in WTC 2 posed challenges in addition to those encountered in simulating the fires in WTC 1. The aircraft, hitting the tower to the east of center, splintered much of the furnishings on the east side of the building and plowed them toward the northeast corner. Neither the impact study nor the validation experiments performed at NIST could be completely relied upon to predict the final distribution, condition, and burning behavior of the demolished furnishings. In addition, only the layouts of the 78th and 80th floors were available to the Investigation; the other floors were only roughly described by former occupants. As a result of these unknowns, the uncertainty in these calculations was distinctly greater than in those for WTC 1. To help mitigate gross differences between the simulations and the observables, NIST made floor-specific adjustments, based on the results of preliminary computations. In particular, the fuel load and volatility on the 80th floor were reduced, and the fuel load on the 81st and 82nd floors was increased. Chapter 6 130 NIST NCSTAR 1, WTC Investigation In contrast with WTC 1, in WTC 2 there was less movement of the fires. The major burning occurred along the east side, with some spread to the north. There was no significant burning on the west side of the tower. Also unlike WTC 1, changing the combustible load in WTC 2 had a noticeable effect on the outcome of the simulations. Because so many windows on the impact floors in WTC 2 were broken out by the aircraft debris and the ensuing fireballs, there was an adequate supply of air for the fires. Thus, the burning rate of the fires was determined by the fuel supply. In the Case D simulation, the office furnishings and aircraft debris were spread out over a wider area, and the furnishings away from the impact area were undamaged. Both of these factors enabled a higher burning rate for the combustibles. In general, the Case D simulations more closely approximated the observations in the photographs and videos, although there was still some prediction of burning too close to the perimeter, especially on the east side of the 78th, 79th, 81st and 83rd floors. The burning in the northeast corner of the 81st and 82nd floors was more intense in Case D than in Case C. The fire in the east side of the 79th floor burned more intensely and reached the south face sooner. Nothing in the simulations explained the absence of fires in the “cold spot,” the 10-window expanse toward the east of the north face of the 80th, 81st, and 82nd floors. 6.10.5 Global Heat Release Rates Much of the information needed to simulate the fires came from laboratory-scale tests. While some of these involved enclosures several meters in dimension and fires that reached heat release rates of 10 MW and 12 GJ in total heat output, they were still far smaller than the fires that burned in the WTC towers. Figure 6–38 shows the heat release rates from the FDS simulations of the WTC fires. The peak plateau heat release rates were about 2 GW for WTC 1 and 1 GW for WTC 2. Integrating the areas under these curves produced total heat outputs from the simulated fires of about 8,000 GJ from WTC 1 and 3,000 GJ from WTC 2. Time (min) 0 20 40 60 80 100 Heat Release Rate (GW) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 WTC 1, Case A WTC 2, Case C WTC 1, Case B WTC 2, Case D Figure 6–38. Predicted heat release rates for fires in WTC 1 and WTC 2. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 131 6.11 DATA TRANSFER The following data from FDS were compiled for use as boundary conditions for the finite-element calculation of the structural temperatures: • The upper and lower layer gas temperatures, time-averaged over 100 s and spatially averaged over 3 ft. The upper layer gas temperatures were taken 1.3 ft (one grid cell) below the ceiling. The lower layer temperatures were taken 1.3 ft above the floor. • The depth of the smoke layer. • The absorption coefficient of the smoke layer 1.3 ft below the ceiling. 6.12 THERMAL MAPPING 6.12.1 Approach Simulating the effect of a fire on the structural integrity of a building required a means for transferring the heat generated by the fire to the surfaces of the insulation on structural members and then conducting the heat through those members. In the Investigation, this meant mapping the time- and space-varying gas temperatures and radiation field generated by FDS onto and throughout the (insulated) columns, trusses and other elements that made up the tower structure. This process was made difficult for these large, geometrically complex buildings by the wide disparity in length and time scales that had to be accounted for in the simulations. FDS generated thermal maps with dimensional resolution of the order of a meter and temperatures fluctuating on a time scale of milliseconds. The finite element models for thermal analysis resolved length of the order of ½ in. on a time scale of seconds. Devising a computation scheme to accommodate the finest of these scales, while simulating the largest of these scales, presented a software challenge in order to avoid unacceptably long computation times. 6.12.2 The Fire-Structure Interface NIST developed a computational scheme to overcome this difficulty, the Fire Structure Interface (FSI). These computations began with the structural models of each WTC tower as described in Section 6.6.4, damaged by the aircraft as described in Section 6.8.4 and exposed to fire-generated heat, as described in Section 6.10.4. For a particular tower and damage scenario, FSI “bathed” each small section of each structural member in an air environment that had been generated by FDS. For efficiency of computation, two simplifications were made: • The fluctuating environment was averaged over 30 s intervals, and The transfer of radiant energy from a hot mass to a cool mass is proportional to the absolute temperature (Kelvin) to the fourth power. Thus, the contribution of the hot upper layer dominates the overall radiative heat transfer. Convective heat transfer is linearly proportional to the difference in temperature between the hot gas and the cool solid. Chapter 6 132 NIST NCSTAR 1, WTC Investigation • The local environment was represented by a hot, soot-laden upper layer and a cooler, relatively clear lower layer. FSI then calculated the radiative and convective heat transfer to each of these small sections using conventional physics. Finally, the temperature data were read into the ANSYS 8.0 finite element program, which applied the temperature distribution to the structural elements. 6.12.3 Thermal Insulation Properties Equivalent Uniform Thickness of SFRM Preliminary simulations with FSI explored the extent to which bare steel structural elements would heat more rapidly than the same elements would if they were well insulated. In one such calculation experiment, one of the largest columns in the tower structure was immersed in a furnace at 1,100 °C. Uninsulated, it took just 13 min for the steel surface temperatures to reach 600 °C, in the range where substantial loss of strength occurs. When insulated with 1 1/8 in. of SFRM, the same column had not reached that temperature in 10 hours. This established that the fires in WTC 1 and WTC 2 would not be able to significantly weaken the insulated core or perimeter columns within the 102 min and 56 min, respectively, after impact and prior to collapse. Thus, it was important to know whether the insulation was present or removed and much less important to know the exact thickness of the SFRM. It was likely that the thinner steel bars and angles in the floor trusses would be more sensitive to the condition of the insulation. If the insulation were present, but too thin or imperfectly applied, these components might have been heated to failure in times on the order of an hour. NIST performed additional simulations to probe the effect of gaps in the truss insulation and of variations in the thickness, similar to those observed in real SFRM application (Figure 5–6). It was evident that incorporation of these small-scale variations into the description of the structural members would have lengthened the FSI computations to an extreme. Furthermore, there was insufficient information to determine how the thickness varied over the length of the structural members. NIST combined the measured variations in the SFRM thickness (as described in Section 5.6.2) with simulations of the heat transfer through the uneven material. This led to the identification of a uniform thickness that provided the same insulation value as did the measured coatings. These values, shown in Table 5–3, were used in the thermal calculations. They were found to be greater than the specified thicknesses but slightly smaller than the average measured thicknesses, as they should be. SFRM Thermophysical Properties When the Investigation began, there were few published data on the insulating properties of SFRMs, especially at elevated temperatures. It was expected, and soon confirmed, that the fires could generate temperatures up to 1,100 °C. Therefore, NIST contracted for measurement of the key SFRM thermophysical properties that, along with coating thickness, determine the insulating effect of the coatings. These properties included thermal conductivity, specific heat capacity, and density. These were measured for each SFRM at temperatures up to 1,200 °C. Since there were no ASTM test methods developed specifically for characterizing the thermophysical properties of SFRMs as a function of temperature, ASTM test methods developed for other materials were used. Samples were prepared by the Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 133 manufacturers of the fire-resistive material, which included BLAZE-SHIELD DC/F and BLAZE-SHIELD II. • The thermal conductivity measurements were performed according to ASTM C 1113, Standard Test Method for Thermal Conductivity of Refractories by Hot Wire (Platinum Resistance Thermometer Technique). The room temperature values were in general agreement with the manufacturer’s published values for both materials. The thermal conductivities increased with temperature. • Specific heat capacity was measured in accordance with ASTM E 1269, Standard Test Method for Determining Specific Heat Capacity by Differential Scanning Calorimetry (DSC). By including DSC measurement of a NIST Reference Material (sapphire), the measured SFRM quantities were directly traceable to NIST standards. • The densities of the SFRMs were calculated from measurements of changes in the mass and dimensions of samples as their temperatures were increased. The length-change measurements were performed according to ASTM E 228, Standard Test Method for Linear Thermal Expansion of Solid Materials. The mass loss measurements were performed according to ASTM E 1131, Standard Test Method for Compositional Analysis by Thermogravimetry. It was not known which type(s) of gypsum wallboard were used to enclose the core columns. Therefore, the thermophysical properties of four types of gypsum panels were examined. • Thermal conductivity was measured using the heated probe technique described in ASTM D 5334, Standard Test Method for Determination of Thermal Conductivity of Soil and Soft Rock by Thermal Needle Probe Procedure. In general, the thermal conductivity initially decreased as the temperature increased to 200 °C and then increased with increasing temperature above 300 °C. • Specific heat capacities of the cores of the four gypsum panel samples were measured using a differential scanning calorimeter at NIST according to ASTM E 1269, Standard Test Method for Determining Specific Heat Capacity by Differential Scanning Calorimetry. The four panels had nearly identical specific heat capacities as a function of temperature. • The variation of density with temperature was determined from the change in volume of the gypsum material and the mass loss. The linear expansion was determined using a dilatometer and the mass loss from thermogravimetric analysis. All four materials showed the same trend as a function of temperature. 6.12.4 FSI Uncertainty Assessment As was done for FDS, it was necessary to establish the quality of FSI’s predictions of temperature profiles within insulated and bare structural steel components. This was accomplished using data from a series of six tests in which assorted steel members were exposed to controlled fires of varying heat release rate and radiative intensity. The steel members, depicted in Figures 6–39 through 6–41, were either bare or coated with sprayed BLAZE-SHIELD DC/F in two thicknesses. The fibrous insulation was applied by an Chapter 6 134 NIST NCSTAR 1, WTC Investigation experienced applicator, who took considerable care to apply an even coating of the specified thickness. As such, the insulated test subjects represent a best case in terms of thickness and uniformity. Figure 6–42 shows some of the coated components. Figure 6–39. Simple bar dimensions (in.). Figure 6–40. Tubular column dimensions (in.). 132 10 14 1/4 1/4 1 118 Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 135 Figure 6–41. Truss Dimensions (in.). Source: NIST. Figure 6–42. SFRM-coated steel components prior to a test. 31 181 1 3 3 1/4 1/4 2.5 2.5 1/4 1/4 Web Bar Top Chord Angles Bottom Chord Angles Chapter 6 136 NIST NCSTAR 1, WTC Investigation Table 6–7 shows the dimensions and variability of the insulation for the two successful tests involving coated steel. The thickness measurements were taken at numerous locations along the perimeter and length of each specimen using a pin thickness gauge specifically designed for this type of insulation. Table 6–7. Summary of insulation on steel components. Applied Thickness (in.) Test Item Specified Thickness (in.) Mean Std. Deviation 5 Bar 0.75 0.91 0.22 Column 1.50 1.61 0.12 Truss A 0.75 1.06 0.28 Truss B 1.50 1.59 0.32 6 Bar 0.75 1.00 0.18 Column 0.75 0.84 0.14 Truss A 0.75 1.02 0.27 Truss B 0.75 1.01 0.27 Temperatures were recorded at multiple locations on the surfaces of the steel, the insulation, and the compartment. As an example, Figure 6–43 shows the finite element representation of the coated truss. Figure 6–43. Finite element representation of the insulated steel truss (blue), the SFRM (violet), and the ceiling (red). Figure 6–44 compares the measured and predicted temperatures on the steel surface of the top chord of a bare truss. Figure 6–45 is the analogous plot of the measured and predicted temperatures on the steel surface of the top chord of a truss insulated with 3/4 in. of BLAZE-SHIELD DC/F. Similar curves were generated for each of the steel pieces, bare and insulated. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 137 Figure 6–44. Comparison of numerical simulations with measurements for the steel surface temperature at four locations on the top chord of a bare truss. Figure 6–45. Comparison of numerical simulations with measurements for the temperature of the steel surface at four locations on the top chord of an insulated truss. Chapter 6 138 NIST NCSTAR 1, WTC Investigation Examination of the graphs for the insulated steel pieces indicated the following: • FSI captured the shape of the temperature rise at the steel surfaces and the significant decrease in the rate of temperature rise when the SFRM was present. • The times to the peak temperature (or a near-plateau) were predicted to within about a minute in all cases. • There was no consistent pattern of overprediction or underprediction of the surface temperatures. • On the average, the numerical predictions of the steel surface temperature were within 7 percent of the experimental measurements for bare steel elements and within 17 percent for the insulated steel elements. The former was within the combined uncertainty in the temperature measurements and the heat release rate in the fire model. The increase in the latter was attributed to model sensitivity to the SFRM coating thickness and thermal conductivity. In general, the FSI added little to the overall uncertainty in the simulation of the temperatures at the outer surfaces of bare steel elements and, more importantly, at the SFRM-steel interface. An additional, important outcome of the experiments was the demonstration of the insulating effect of even 3/4 in. of SFRM. Trusses, made of relatively thin steel, were far more susceptible to heating than the perimeter and core columns. As shown in Figure 2–10, in 15 min, a bare truss reached a temperature at which significant loss of strength was imminent. An identical, but insulated truss had not reached that temperature in 50 min. 6.12.5 The Four Cases FSI imposed the thermal environment from each of the four FDS fire scenarios (Cases A and B for WTC 1 and Cases C and D for WTC 2) on the four damaged structures from the aircraft simulations, which carried the same case letters. The FSI output files carried the same case letters as the input files. The FSI calculations were performed at time steps ranging from 1 ms to 50 ms. Use of the resulting data set for structural analysis would have required a prohibitive amount of computation time. Thus, for each case, the instantaneous temperature and temperature gradient for each grid volume was provided at 10 min intervals after aircraft impact. For WTC 1, there were 10 such intervals, ending at 6,000 s; for WTC 2 there were 6 intervals, ending at 3,600 s. Comparison of these coarsely timed output files with files at 1 min resolution showed any differences to be within the combined uncertainty. Each floor in the FSI simulation provided thermal information for the floor assembly above. Thus, there was not sufficient information for FSI to model the lowest floor in the FDS simulations. For WTC 1, the global thermal response generated by FSI included floors 93 through 99; for WTC 2, the included floors were 79 through 83. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 139 For ease of visualization, two graphic representations were developed. Figure 6–46 shows an example of the temperature map for the 96th floor of WTC 1. Severed columns and broken floor segments are not shown. Figure 6–47 shows a similar map for the 81st floor of WTC 2. Figure 6–46. Temperatures ( C) on the columns and trusses of the 96th floor of WTC 1 at 6,000 s after aircraft impact, Case B. Figure 6–47. Temperatures ( C) on the columns and trusses of the 81st floor of WTC 2 at 3,000 s after aircraft impact, Case D. A third visualization tool was animation of the evolving temperatures of the columns. Frames from an example, again of the 96th floor of WTC 1, Case A, are shown in Figure 6–48. The size of the square representing a column represents its yield strength. Columns may have been heated when the fire was nearby and then cooled after the local combustibles were consumed. 140 NIST NCSTAR 1, WTC Investigation Chapter 6 (a) Time = 1000 s (b) Time = 2000 s (c) Time = 3000 s (d) Time = 4000 s (e) Time = 5000 s (f) Time = 6000 s Figure 6–48. Frames from animation of the thermal response of columns on the 96th floor of WTC 1, Case A. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 141 6.12.6 Characterization of the Thermal Profiles Tables 6–8 and 6–9 summarize the regions of the floors in which the structural steel reached temperatures at which their yield strengths would have been significantly diminished. Instances of brief heating of one or two columns early in the fires were not included. Even in the vicinity of the fires, the columns and trusses for which the insulation was intact did not heat to temperatures where significant loss of strength occurred. Unlike the simulations of the aircraft impact and the fires, there was no evidence, photographic or other, for direct comparison with the FSI results. Table 6–8. Regions in WTC 1 in which temperatures of structural steel exceeded 600 C. Trusses Perimeter Floor Columns Core Columns Number Case A Case B Case A Case B Case A Case B 93 – – – – – – 94 – – – – N, S NE, S 95 N N, S – – S NW, S 96 N N, S – S S W, S 97 N, S N, S – S N W, S 98 N N, S – – – – 99 – – – – – – Key: N, north; NE, northeast; NW, northwest; S, south; W, west. Table 6–9. Regions in WTC 2 in which temperatures of structural steel exceeded 600 C. Floor Trusses Perimeter Columns Core Columns Number Case C Case D Case C Case D Case C Case D 79 – – – – – – 80 – – – – – – 81 NE NE NE NE – NE 82 E E E E E E 83 E E – E – E Key: E, east; NE, northeast. 6.13 MEASUREMENT OF THE FIRE RESISTANCE OF THE FLOOR SYSTEM As described in Section 5.4.7, the composite floor system, composed of open-web, lightweight steel trusses topped with a slab of lightweight concrete, was an innovative feature. As further noted in Section 5.6.2, the approach to achieving the specified fire resistance for these floors was the use of a SFRM. Documents indicated that the fire performance of the composite floor system of the WTC towers was an issue of concern to the building owners and designers. However, NIST found no evidence regarding the technical basis for the selection of insulation material for the floor trusses or for the insulation thickness to achieve a 2 hour rating. Further, NIST has found no evidence that fire resistance tests of the WTC floor system were conducted. Chapter 6 142 NIST NCSTAR 1, WTC Investigation Most of the possible building collapse sequences included some contribution from the floors, ranging from their ability to transfer load to their initiating the collapse by their failure. Thus, it became central to the Investigation to obtain data regarding the limits of the insulated floors in withstanding the heat from the fires. The standard test for determining the fire endurance of floor assemblies is ASTM E 119, “Standard Test Methods for Fire Tests of Building Construction and Materials.” The conduct of the test is described in Section 1.2.2 under “Fire Protection Systems.” Accordingly, NIST contracted with Underwriters Laboratories, Inc. to conduct tests to obtain information on the fire endurance of trusses like those in the WTC towers. The objective was to understand the effects of three factors: • Scale of the test. There were no established facilities capable of testing the 60 ft lengths of the long spans that were used in the towers, but there is a history of testing reduced-scale assemblies and scaling them to practical dimensions. In the Investigation’s tests, the fullscale test specimens were 35 ft long, equal to the shorter span between the core and the perimeter of the WTC towers. Their construction replicated, as closely as possible, the original short-span floors. The reduced-scale specimens were half that length and height. All assemblies were 14 ft wide. The simulation of a “maximum load condition,” as required by ASTM E 119, involved placing a combination of concrete blocks and containers filled with water on the top surface of the floor. The load on the shorter truss was double that of the longer truss to achieve the same state of stress in both trusses. Traditionally, relatively smallscale assemblies have been tested and results have been scaled to practical floor system spans. • SFRM thickness. The Port Authority originally specified BLAZE-SHIELD D as the SFRM, applied to a ½ in. covering. The average measured thicknesses were found to be approximately 0.75 in. These two thicknesses of BLAZE-SHIELD D were used in the Investigation tests. • Test restraint conditions. In 1971, well after the design of the towers was completed, the ASTM E 119 Standard began differentiating between thermally restrained and unrestrained floor assemblies. An unrestrained assembly is free to expand thermally and to rotate at its supports; a restrained assembly is not. It is customary in the United States to conduct standard fire tests of floor assemblies in the restrained condition. The current standard describes a means to establish unrestrained ratings for floor assemblies from restrained test results. In practice, a floor assembly such as that used in the WTC towers is neither restrained nor unrestrained but is likely somewhere in between. Testing under both restraint conditions, then, is thought to bound performance under the standard fire exposure. In addition, it provided a comparison of unrestrained ratings developed from both restrained and unrestrained test conditions. The test plan included four tests, which varied the three factors: Test 1: 35 ft floor, ¾ in. insulation, restrained Test 2: 35 ft floor, ¾ in. insulation, unrestrained Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 143 Test 3: 17 ft floor, ¾ in. insulation, restrained Test 4: 17 ft floor, ½ in. insulation, restrained The results of the four tests are summarized as follows: • All four test assemblies were able to withstand standard fire conditions for between ¾ hour and 2 hours without exceeding the limits prescribed by ASTM E 119. • All four test specimens sustained the maximum design load for approximately 2 hours without collapsing. • The restrained full-scale floor system obtained a fire resistance rating of 1½ hours, while the unrestrained floor system achieved a 2 hour rating. Past experience with the ASTM E 119 test method led investigators to expect the unrestrained floor assembly to receive a lower rating than the restrained assembly. • For assemblies with a ¾ in. SFRM thickness, the 17 ft assembly’s fire rating was 2 hours; the 35 ft assembly’s rating was 1½ hours. This result raised the question of whether or not a fire rating of a 17 ft floor assembly is scalable to the longer spans in the WTC towers. • The specimen in Test 4, with a fire rating of ¾ hour, would not have met the 2 hour requirement of the NYC Building Code. The Investigation Team was cautious about using these results directly in the formulation of collapse hypotheses. In addition to the scaling issues raised by the test results, the fires in the towers on September 11, and the resulting exposure of the floor systems, were substantially different from the conditions in the test furnaces. Nonetheless, the results established that this type of assembly was capable of sustaining a large gravity load, without collapsing, for a substantial period of time relative to the duration of the fires in any given location on September 11. 6.14 COLLAPSE ANALYSIS OF THE TOWERS 6.14.1 Approach to Determining the Probable Collapse Sequences At the core of NIST’s reconstruction of the events of September 11, 2001, were the archive of photographic and video evidence, the observations of people who were on the scene, the assembled documents describing the towers and the aircraft, and Investigation-generated experimental data on the properties of construction and furnishing materials and the behavior of the fires. Information from all of these sources fed the computer simulations of the towers, the aircraft impacts, the ensuing fires and their heating of the structural elements, and the structural changes that led to the collapse of the towers. To the extent that the input information was complete and accurate, the output of the simulations would have provided definitive responses to the first three objectives of the Investigation. However, the available information, as extensive as it was, was neither complete nor of assured precision. As a result, the Investigation Team took steps to ensure that the conclusions of the effort were credible explanations for how the buildings collapsed and the extent to which the casualties occurred. Chapter 6 144 NIST NCSTAR 1, WTC Investigation One principal step was the determination of those variables that most affected the outcome of the various computer simulations. Sensitivity studies and examination of components and subsystems were carried out for the modeling of the aircraft impact, the fires, and the structural response to impact damage and fires. For each of the most influential variables, a central or middle value and reasonable high and low values were identified. Further computations refined the selection of these values. The computations also were improved to include physical processes that could play a significant role in the structural degradation of the towers. The Investigation Team then defined three cases for each building by combining the middle, less severe, and more severe values of the influential variables. Upon a preliminary examination of the middle cases, it became clear that the towers would likely remain standing. The less severe cases were discarded after the aircraft impact results were compared to observed events. The middle cases (which became Case A for WTC 1 and Case C for WTC 2) were discarded after the structural response analysis of major subsystems were compared to observed events. The more severe case (which became Case B for WTC 1 and Case D for WTC 2) was used for the global analysis of each tower. Complete sets of simulations were then performed for Cases B and D. To the extent that the simulations deviated from the photographic evidence or eyewitness reports, the investigators adjusted the input, but only within the range of physical reality. Thus, for instance, the observed window breakage was an input to the fire simulations and the pulling forces on the perimeter columns by the sagging floors were adjusted within the range of values derived from the subsystem computations. The results were a simulation of the structural deterioration of each tower from the time of aircraft impact to the time at which the building became unstable, i.e., was poised for collapse. Cases B and D accomplished this in a manner that was consistent with the principal observables and the governing physics. 6.14.2 Results of Global Analysis of WTC 1 After the aircraft impact, gravity loads that were previously carried by severed columns were redistributed to other columns. The north wall lost about 7 percent of its loads after impact. Most of the load was transferred by the hat truss, and the rest was redistributed to the adjacent exterior walls by spandrels. Due to the impact damage and the tilting of the building to the north after impact, the south wall also lost gravity load, and about 7 percent was transferred by the hat truss. As a result, the east and west walls and the core gained the redistributed loads through the hat truss. Structural steel and concrete expand when heated. In the early stages of the fire, temperatures of structural members in the core rose, and the resulting thermal expansion of the core columns was greater than the thermal expansion of the (cooler) exterior walls. The floors also thermally expanded in the early stages of the fires. About 20 min after the aircraft impact, the difference in the thermal expansion between the core and exterior walls, which was resisted by the hat truss, caused the core columns’ loads to increase. As floor temperatures increased, the floors sagged and began to pull inward on the exterior wall. As the fires continued to heat areas of the core that were without insulation, the columns weakened and shortened and began to transfer their loads to the exterior walls through the hat truss until the south wall started to bow inward due to the inward pull of the sagging floors. At about 100 min, approximately 20 percent of the core loads had been transferred by the hat truss to the exterior walls due to weakening of Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 145 the core, the loads on the north and south walls had each increased by about 10 percent, and those on the east and west walls had about a 25 percent increase. The increased loads on the east and west walls were due to their relatively higher stiffness compared to the impact damaged north wall and bowed south walls. The inward bowing of the south wall caused failure of exterior column splices and spandrels, and these columns became unstable. The instability spread horizontally across the entire south face. The south wall, now unable to bear its gravity loads, redistributed these loads to the thermally weakened core through the hat truss and to the east and west walls through the spandrels. The building section above the impact zone began tilting to the south as the columns on the east and west walls rapidly became unable to carry the increased loads. This further increased the gravity loads on the core columns. The gravity loads could no longer be redistributed, nor could the remaining core and perimeter columns support the gravity loads from the floors above. Once the upper building section began to move downwards, the weakened structure in the impact and fire zone was not able to absorb the tremendous energy of the falling building section and global collapse ensued. 6.14.3 Results of Global Analysis of WTC 2 Before aircraft impact, the load distribution across the exterior walls and core was symmetric with respect to the centerline of each exterior wall. After aircraft impact, the exterior column loads on the south side of the east and west walls and on the east side of south wall increased. This was due to the leaning of the building core towards the southeast. After aircraft impact, the core carried 6 percent less load. The north wall load reduced by 6 percent and the east face load increased by 24 percent. The south and west walls carried 2 percent to 3 percent more load. In contrast to the fires in WTC 1, which generally progressed from the north side of the building to the south side over approximately 1 hour, the fires in WTC 2 were located on the east side of the core and floors from the time of impact until the building collapsed, with the fires spreading somewhat from south to north. With insulation dislodged over much of the same area, the structural temperatures became elevated in the core, floors, and exterior walls at similar times. During the early stages of the fires, columns with dislodged insulation elongated due to thermal expansion. As the structural temperatures continued to rise, the columns thermally weakened and consequently shortened. Thermal expansion of the floors also occurred early in the fires, but as floor temperatures increased, the floors sagged and began to pull inward on the exterior columns. The south exterior wall displaced downward following the aircraft impact, but did not displace further until the east wall became unstable 43 min later. The inward bowing of the east wall, due to the inward pull of the sagging floors, caused failure of exterior column splices and spandrels and resulted in the east wall columns becoming unstable. The instability progressed horizontally across the entire east face. The east wall, now unable to bear its gravity loads, redistributed them to the thermally weakened core through the hat truss and to the east and west walls through the spandrels. The building section above the impact zone began tilting to the east and south as column instability progressed rapidly from the east wall along the adjacent north and south walls, and increased the gravity load on the weakened east core columns. The gravity loads could no longer be redistributed, nor could the remaining core and perimeter columns support the gravity loads from the floors above. As with WTC 1, once the upper building section began to move downwards, the weakened structure in the impact Chapter 6 146 NIST NCSTAR 1, WTC Investigation and fire zone was not able to absorb the tremendous energy of the falling building section and global collapse ensued. 6.14.4 Events Following Collapse Initiation Failure of the south wall in WTC 1 and east wall in WTC 2 caused the portion of the building above to tilt in the direction of the failed wall. The tilting was accompanied by a downward movement. The story immediately below the stories in which the columns failed was not able to arrest this initial movement as evidenced by videos from several vantage points. The structure below the level of collapse initiation offered minimal resistance to the falling building mass at and above the impact zone. The potential energy released by the downward movement of the large building mass far exceeded the capacity of the intact structure below to absorb that through energy of deformation. Since the stories below the level of collapse initiation provided little resistance to the tremendous energy released by the falling building mass, the building section above came down essentially in free fall, as seen in videos. As the stories below sequentially failed, the falling mass increased, further increasing the demand on the floors below, which were unable to arrest the moving mass. The falling mass of the building compressed the air ahead of it, much like the action of a piston, forcing material, such as smoke and debris, out the windows as seen in several videos. NIST found no corroborating evidence for alternative hypotheses suggesting that the WTC towers were brought down by controlled demolition using explosives planted prior to September 11, 2001. NIST also did not find any evidence that missiles were fired at or hit the towers. Instead, photographs and videos from several angles clearly show that the collapse initiated at the fire and impact floors and that the collapse progressed from the initiating floors downward, until the dust clouds obscured the view. 6.14.5 Structural Response of the WTC Towers to Fire without Impact or Thermal Insulation Damage To complete the assessment of the relative roles of aircraft impact and ensuing fires, NIST examined whether an intense, but conventional, fire, occurring without the aircraft impact, could have led to the collapse of a WTC tower, were the tower in the same condition as it was on September 10, 2001. NIST used the observations, information, and analyses developed during the Investigation to enable the formulation of probable limits to the damage from such a fire. Since a complete analysis beyond the actual collapse times of the towers was not conducted, the findings in this section represent NIST’s best technical judgment based on the available observations, information, and analyses: • Ignition on a single floor by a small bomb or other explosion. If arson were involved, there might have been multiple small fires ignited on a few floors. • Air supply determined by the building ventilation system. • Moderate fire growth rate. In the case of arson, several gallons of an accelerant might have been applied to the building combustibles, igniting the equivalent of several workstations. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 147 • Water supply to the sprinklers and standpipes maliciously compromised. • Intact structural insulation and interior walls. The four cases described in this chapter represented fires that were far more severe than this: • About 10,000 gallons of jet fuel were sprayed into multiple stories, quickly and simultaneously igniting hundreds of workstations and other combustibles. • The aircraft and subsequent fireballs created large open areas in the building exterior through which air could flow to support the fires. • The impact and debris removed the insulation from a large number of structural elements that were then subjected to the heat from the fires. Additional findings from the Investigation showed that: • Both the results of the multiple workstation experiments and the simulations of the WTC fires showed that the combustibles in a given location, if undisturbed by the aircraft impact, would have been almost fully burned out in about 20 min. • In the simulations of Cases A through D, none of the columns and trusses for which thermal insulation was intact reached temperatures at which significant loss of strength occurred. Thermal analyses showed that steel temperatures in areas where the insulation remained intact rarely exceeded 400 °C in WTC 1 and 500 °C in WTC 2. • In WTC 1, if fires had been allowed to continue past the time of building collapse, complete burnout would likely have occurred within a short time since the fires had already traversed around the entire floor and most of the combustibles would already have been consumed (see Figure 6–38). During the extended period from collapse to burnout, the steel temperatures would likely not have increased very much. The installed insulation in the fire-affected floors of this building had been upgraded to an average thickness of 2.5 in. • In a fire simulation of WTC 2, that was extended for 2 hours beyond Case D and with all windows broken during this period, the temperatures in the truss steel on the west side of the building (where the insulation was undamaged) increased for about 40 min before falling off rapidly as the combustibles were consumed. Results for a typical floor (floor 81) showed that temperatures of 700 °C to 760 °C were reached over approximately 15 percent of the west floor area for less than 10 min. Approximately 60 percent of the floor steel had temperatures between 600 °C and 700 °C for about 15 min. Approximately 70 percent of the floor steel had temperatures that exceeded 500 °C for about 45 min. At these temperatures, the floors would be expected to sag and then recover a portion of the sag as the steel began to cool. Based on results for Cases C and D, the temperatures of the insulated exterior and core columns would not have increased to the point where significant loss of strength or stiffness would occur during these additional 2 hours. With intact, cool core columns, any inward bowing of the west exterior wall that might occur would be readily supported by the adjacent exterior walls and core columns. Chapter 6 148 NIST NCSTAR 1, WTC Investigation • Both WTC 1 and WTC 2 were stable after the aircraft impact, standing for 102 min and 56 min, respectively. The global analyses with structural impact damage showed that both towers had considerable reserve capacity. This was confirmed by analysis of the post-impact vibration of WTC 2, the more severely damaged building, where the damaged tower oscillated with a peak amplitude that was between 30 percent and 40 percent of the sway under hurricane force winds for which the towers were designed and at periods nearly equal to the first two translation and torsion mode periods calculated for the undamaged structure. • Computer simulations, supported by the results of large-scale fire tests and furnace testing of floor subsystems, showed that insulated structural steel, when coated with the average installed insulation thickness of ¾ in., would not have reached high temperatures (i.e., greater than 650 °C) from nearby fires for a longer time than the burnout time of the combustibles (approximately 20 min for 4 lb/ft2 of combusted material). Simulations also showed that variations in thickness resulting from normal application, even with occasional gaps in coverage, would not have changed this result. • Inward bowing of the exterior walls in both WTC 1 and WTC 2 was observed only on the face with the long-span floor system. In WTC 1, this was found to be the case even though equally extensive fires were observed on all faces. In WTC 2, fires were not observed on the long-span west face and were less intense on the short-span faces than on the east face. • Inward bowing was a necessary but not sufficient condition to initiate collapse. In both WTC 1 and WTC 2, significant weakening of the core due to aircraft impact damage and thermal effects was also necessary to initiate building collapse. • The tower structures had significant capacity to redistribute loads (a) from bowed walls to adjacent exterior walls with short-span floors via the arch action of spandrels, and (b) between the core and exterior walls via the hat truss and, to a lesser extent, the floors. In evaluating how the undamaged towers would have performed in an intense, conventional fire, NIST considered the following factors individually and in combination: • The temperatures that would be reached in structural steel components with intact insulation. • The extent of the area over which high temperatures (e.g., greater than 600 °C where significant thermal weakening of the steel occurs) would be reached at any given time. • The duration over which the high temperatures would be sustained concurrently in any given area. • The length of the floor span (long or short) where high temperatures would be reached. • The number of floors with areas where high temperatures would be sustained concurrently in the long-span direction. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 149 • The potential for inward bowing of exterior walls (i.e., magnitude and extent of bowing over the width of the face and the number of floors involved) due to thermally induced floor sagging of long-span floors and associated inward pull forces. • The capacity of the structure to redistribute loads (e.g., via the spandrels, hat truss, and floors) if the thermal conditions were sufficiently intense to cause inward bowing of the exterior walls. In addition, NIST considered the following known facts: • WTC 1 did not collapse during the major fire in 1975, which engulfed a large area (about one-fourth of the floor area or 9,000 ft2) on the southeast quadrant of the 11th floor. At the time, office spaces in the towers were not sprinklered. The fire caused minimal damage to the floor system with the ½ in. specified insulation thickness applied on the trusses (four trusses were slightly distorted), and at no time was the load-carrying capacity compromised for the floor system or the structure as a whole. • Four standard fire resistance tests of floor assemblies like those in the WTC towers conducted as part of this Investigation showed that (a) it took about 90 min of sustained heating in the furnace for temperatures to exceed 600 °C on steel truss members with either ½ in. or ¾ in. insulation thickness, and (b) in no case was the load-carrying capacity compromised by heating of the floor system for 2 hours at furnace temperatures, with applied loads exceeding those on September 11 by a factor of two. From these findings, factors, and observed performance, NIST concluded: • In the absence of structural and insulation damage, a conventional fire substantially similar to or less intense than the fires encountered on September 11, 2001, likely would not have led to the collapse of a WTC tower. • The condition of the insulation prior to aircraft impact, which was found to be mostly intact, and the insulation thickness on the WTC floor system contributed to, but did not play a governing role, in initiating collapse of the towers. • The towers likely would not have collapsed under the combined effects of aircraft impact and the subsequent multi-floor fires encountered on September 11 if the thermal insulation had not been widely dislodged or had been only minimally dislodged by aircraft impact. These findings apply to fires that are substantially similar to or less intense than those encountered on September 11, 2001. They do not apply to a standard fire or an assumed fire exposure which has (a) uniform high temperatures over an entire floor or most of a floor (note that the WTC floors were extremely large) and concurrently over multiple floors, (b) high temperatures that are sustained indefinitely or for long periods of time (greater than about 20 min at any location), and (c) combusted fire loads that are significantly greater than those considered in the analyses. They also do not apply if the capacity of the undamaged structure to redistribute loads via the spandrels, hat truss, and floors is not accounted for adequately in a full 3-dimensional simulation model of the structure. Chapter 6 150 NIST NCSTAR 1, WTC Investigation 6.14.6 Probable WTC 1 Collapse Sequence Aircraft Impact Damage • The aircraft impact severed a number of exterior columns on the north wall from the 93rd to the 98th floors, and the wall section above the impact zone moved downward. • After breaching the building’s perimeter, the aircraft continued to penetrate into the building, severing floor framing and core columns at the north side of the core. Core columns were also damaged toward the center of the core. Insulation was damaged from the impact area to the south perimeter wall, primarily through the middle one-third to one-half of the core width. Finally, the aircraft debris removed a single exterior panel at the center of the south wall between the 94th and 96th floors. • The impact damage to the exterior walls and to the core resulted in redistribution of severed column loads, mostly to the columns adjacent to the impact zones. The hat truss resisted the downward movement of the north wall. • Loads on the damaged core columns were redistributed mostly to adjacent intact core columns and to a lesser extent to the north perimeter columns through the core floor systems and the hat truss. • As a result of the aircraft impact damage, the north and south walls each carried about 7 percent less gravity load after impact, and the east and west walls each carried about 7 percent more load. The core carried about 1 percent more gravity load after impact. Thermal Weakening of the Structure • Under the high temperatures and stresses in the core area, the remaining core columns with damaged insulation were thermally weakened and shortened, causing the columns on the floors above to move downward. The hat truss resisted the core column shortening and redistributed loads to the perimeter walls. The north and south walls’ loads increased by about 10 percent, and the east and west walls’ loads increased by about 25 percent, while the core’s loads decreased by about 20 percent. • The long-span sections of the 95th to 99th floors on the south side weakened with increasing temperatures and began to sag. Early on, the floors on the north side had sagged and then contracted as the fires moved to the south and the floors cooled. As the fires intensified on the south side, the floors there sagged, and the floor connections weakened. About 20 percent of the connections on the south side of the 97th and 98th floors failed. • The sagging floors with intact floor connections pulled inward on the south perimeter columns, causing them to bow inward. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 151 Collapse Initiation • The bowed south wall columns buckled and were unable to carry the gravity loads. Those loads shifted to the adjacent columns via the spandrels, but those columns quickly became overloaded as well. In rapid sequence, this instability spread all the way to the east and west walls. • The section of the building above the impact zone (near the 98th floor), acting as a rigid block, tilted at least 8 degrees to the south. • The downward movement of this structural block was more than the damaged structure could resist, and global collapse began. 6.14.7 Probable WTC 2 Collapse Sequence Aircraft Impact Damage • The aircraft impact severed a number of exterior columns on the south wall from the 78th floor to the 84th floor, and the wall section above the impact zone moved downward. • After breaching the building’s perimeter, the aircraft continued to penetrate into the building, severing floor framing and core columns at the southeast corner of the core. Insulation was damaged from the impact area through the east half of the core to the north and east perimeter walls. The floor truss seat connections over about one-fourth to one-half of the east side of the core were severed on the 80th and 81st floors and over about one-third of the east perimeter wall on the 83rd floor. The debris severed four columns near the east corner of the north wall between the 80th and 82nd floors. • The impact damage to the perimeter walls and to the core resulted in redistribution of severed column loads, mostly to the columns adjacent to the impact zones. The impact damage to the core columns resulted in redistribution of severed column loads, mainly to other intact core columns and the east exterior wall. The hat truss resisted the downward movement of the south wall. • As a result of the aircraft impact damage, the core carried about 6 percent less gravity load. The north wall carried about 10 percent less, the east face carried about 24 percent more, and the west and south faces carried about 3 percent and two percent more, respectively. • The core was then leaning slightly toward the south and east perimeter walls. The perimeter walls restrained the tendency of the core to lean via the hat truss and the intact floors. Thermal Weakening of the Structure • Under the high temperatures and stresses in the core area, the remaining core columns with damaged insulation were thermally weakened and shortened, causing the columns on the floors above to move downward. Chapter 6 152 NIST NCSTAR 1, WTC Investigation • At this point, the east wall carried about 5 percent more of the gravity loads, and the core carried about 2 percent less. The other three walls carried between 0 percent and 3 percent less. • The long-span floors on the east side of the 79th to 83rd floors weakened with increasing temperatures and began to sag. About one-third of the remaining floor connections to the east perimeter wall on the 83rd floor failed. • Those sagging floors whose seats were still intact pulled inward on the east perimeter columns, causing them to bow inward. The inward bowing increased with time. Collapse Initiation • As in WTC 1, the bowed columns buckled and became unable to carry the gravity loads. Those loads shifted to the adjacent columns via the spandrels, but those columns quickly became overloaded. In rapid sequence, this instability spread all across the east wall. • Loads were transferred from the failing east wall to the weakened core through the hat truss and to the north and south walls through the spandrels. The instability of the east face spread rapidly along the north and south walls. • The building section above the impact zone (near the 82nd floor) tilted 7 degrees to 8 degrees to the east and 3 degrees to 4 degrees to the south prior to significant downward movement of the upper building section. The tilt to the south did not increase any further as the upper building section began to fall, but the tilt to the east was seen to increase to 20 degrees until dust clouds obscured the view. • The downward movement of this structural block was more than the damaged structure could resist, and the global collapse began. 6.14.8 Accuracy of the Probable Collapse Sequences Independent assessment of the validity of the key steps in the collapse of the towers was a challenging task. Some of the photographic information had been used to direct the simulations. For example, the timing of the appearance of broken windows was an input to the fire growth modeling. However, there were significant observables that were usable as corroborating evidence, as shown in Tables 6–10 and 6–11. Some of these were used to establish the quality of the individual simulations of the aircraft impact and the fire growth, as described in Sections 6.9 and 6.10. While the agreement between observations and simulation was not exact, the differences were within the uncertainties in the input information. The generally successful comparisons lent credibility to the overall reconstruction of the disaster. There remained a small, but important number of observations against which the structural collapse sequences could be judged. The comparisons are for Cases B and D impact damage and temperature histories, for which the better agreement was obtained. Reconstruction of the Collapses NIST NCSTAR 1, WTC Investigation 153 Table 6–10. Comparison of global structural model predictions and observations for WTC 1, Case B. Observation Simulation Following the aircraft impact, the tower still stood. The tower remained upright with significant reserve capacity. The south perimeter wall was first observed to have bowed inward at 10:23 a.m. The bowing appeared over nearly the entire south face of the 94th to 100th floors. The maximum bowing was 55 in. on the 97th floor. (The central area in available images was obscured by smoke.) The inward bowing of the south wall at 10:28 a.m. It extended from the 94th to the 100th floor, with a maximum of about 43 in. As the structural collapse began, the building section above the impact and fire zone tilted at least 8 degrees to the south with no discernable east or west component in the tilt. Dust clouds obscured the view as the building section began to fall downward. The south side bowed and weakened. The analysis stopped as the initiation of global instability was imminent. The time to collapse initiation was 102 min from the aircraft impact. There was significant weakening of the south wall and the core columns. Instability was imminent at 100 min. Table 6–11. Comparison of global structural model predictions and observations for WTC 2, Case D. Observation Simulation Following the aircraft impact, the tower still stood. The tower remained upright with significant reserve capacity. The east perimeter wall was first observed to have bowed inward approximately 10 in. at floor 80 at 9:21 a.m. The bowing extended across most of the east face between the 78th and 83rd floors. The inward bowing of the east wall had a maximum value of about 9.5 in. at 9:23 a.m. The bowing extended from the 78th floor to the 83rd floor. The building section above the impact and fire area tilted to the east and south as the structural collapse initiated. The angle was approximately 3 degrees to 4 degrees to the south and 7 degrees to 8 degrees to the east prior to significant downward movement of the upper building section. The tilt to the south did not increase as the upper building section began to fall, but the tilt to the east rose to approximately 25 degrees before dust clouds obscured the view. At point of instability, there was tilting to the south and east. The time to collapse initiation was 56 min after the aircraft impact. The analysis predicted global instability after 43 min. The agreement between the observations and the simulations is reasonably good, supporting the validity of the probable collapse sequences. The exact times to collapse initiation were sensitive to the factors that controlled the inward bowing of the exterior columns. The sequence of events leading to collapse initiation was not sensitive to these factors. Chapter 6 154 NIST NCSTAR 1, WTC Investigation 6.14.9 Factors that Affected Building Performance on September 11, 2001 • The unusually dense spacing of perimeter columns, coupled with deep spandrels, resulted in a robust building that was able to fragment the aircraft upon impact and redistribute loads from severed perimeter columns to adjacent, intact columns. • The wind loads used for the WTC towers, which governed the design of the framed-tube system, significantly exceeded the requirements of the building codes of the era and were consistent with the independent NIST estimates that were based on current state-of-the-art considerations. • The robustness of the perimeter framed-tube system and the large lateral dimension of the towers helped the buildings withstand the impact of the aircraft. • The composite floor system enabled the floors to redistribute loads from places of aircraft impact damage to other locations, avoiding larger scale collapse upon impact. • The hat truss resisted the significant weakening of the core by redistributing loads form the damaged columns to intact columns. As a result of these factors, the buildings would likely not have collapsed under the combined effects of the aircraft damage and subsequent fires if the insulation had not been widely dislodged. The thickness and the condition of the insulation prior to aircraft impact did not play a governing role in the initiation of building collapse. NIST NCSTAR 1, WTC Investigation 155