Conference PaperPDF Available

Figures

Content may be subject to copyright.
SiF 2020 The 11th International Conference on Structures in Fire
The University of Queensland, Brisbane, Australia, November 30th December 2nd, 2020
THE COLLAPSE OF WORLD TRADE CENTER 7: REVISITED
Mhd Anwar Orabi
1
, Liming Jiang
2
, Asif Usmani
3
, Jose Torero4
ABSTRACT
In structural fire engineering literature, the catastrophic events of Sept. 11 2001 stand out as a major reason
and motivation for research on improving the understanding of structural behaviour in fire. These events
included the first complete collapse of a tall steel framed structure solely due to fire. World Trade Center 7
(WTC 7) was a 47-storey office building within the WTC complex that was set alight by debris from WTC
1 collapse, and in turn also collapsed 7 hours later. In the following years, detailed investigations were
carried out by the National Institute of Standards and Technology (NIST), Weidlinger Associates Inc.
(WAI), Guy Nordenson Associates (GNA), Arup, and the BRE Centre for Fire Safety Engineering. Each
of these teams analysed the fire and structure differently, and reached varying conclusions with regards to
the mechanisms responsible for initiating and propagating the collapse of WTC 7. This paper will give an
overview of how the various investigations were performed, what collapse hypotheses were made, and how
it is possible that the building emergency power system may have been responsible for failure.
Keywords: finite element modelling; world trade center; progressive collapse; case study
1 INTRODUCTION
During the events of Sept. 11 2001, WTC 1 and WTC 2 were each struck by a commercial airliner. With
all WTC buildings in relatively close proximity, the debris from the twin towers caused severe damage to
the surrounding structures many of which collapsed. WTC 7 was located north of WTC 1 and suffered
considerably little damage that did not directly jeopardise the its structural integrity. The debris, however,
set over ten floors alight within WTC 7 and diverted the attention of first respondents away from the
building [1]. After seven hours of fire burning on consecutive floors and traversing the building perimeter,
the structure finally gave in with what appeared to be a failure of the core where the penthouse on top of
the the building first sank into the building followed by the outside framing. As the first tall steel-framed
building to collapse solely due to fire, WTC 7 has gained a reputation as a major incentive for bettering the
community awareness of the risks fire poses to buildings. Despite this, the collapse of WTC 7 is not as
well-understood as what one may think considering its importance to the field. After years of investigation
by some of the world leading experts in both fire and structural engineering, there remains multiple possible
hypotheses without any consensus on how the building behaved and failed because of the fire. The authors
of this article believe that there is valuable scientific merit for the community in reviewing this case study
more closely and considering the big picture when the worst-case scenario really takes place. This paper
begins with introducing the WTC 7 structure, its unique backup energy systems, and the events that led to
its complete collapse. An overview of previous expert investigations and their results is then presented,
with focus on how they analysed the fire and the structure. Finally, a different hypothesis for the collapse
1
Mr., Deparment of Building Services Engineering, The Hong Kong Polytechnic University,
e-mail: anwar.orabi@connect.polyu.hk, ORCID: https://orcid.org/0000-0001-5083-3623
2
Dr., Deparment of Building Services Engineering, The Hong Kong Polytechnic University,
e-mail: liming.jiang@polyu.edu.hk, ORCID: https://orcid.org/0000-0001-8112-2330
3
Prof., Deparment of Building Services Engineering, The Hong Kong Polytechnic University,
e-mail: asif.usmani@polyu.edu.hk, ORCID: https://orcid.org/0000-0003-2454-5737
4 Prof., Deparment of Civil, Environmental and Geomatic Engineering, University College London,
e-mail: j.torero@ucl.ac.uk, ORCID: https://orcid.org/ 0000-0001-6265-9327
is proposed that explains the fate of the 12,000 gallons of diesel fuel that were unaccounted for after the
incident, and the role of the large transfer structures of the building in initiating and progressing the collapse.
2 WTC 7 AND THE SEPTEMBER 11 DISASTER
2.1 Structural layout
WTC 7 was a rather unique structure in a prime location in one of the most expensive real estate markets
in the world. Created as a later addition to the World Trade Center Complex, WTC 7 was erected atop an
existing electrical substation that supplied power to lower Manhattan and had to make most efficient use of
the trapezoidal plot it was located upon. The tower had 47 stories and was 186 m high making it the third
tallest building in the complex after WTC1 and 2 respectively. Its footprint was larger than either of the
twin towers with a trapezoidal plan with North, South, East and West dimensions of 100 m × 75 m × 43 m
× 45 m respectively as shown in Figure 1 which also highlights the eastern bay which would become the
focus of many of the collapse hypotheses. The first four floors from the ground contained the power
substation, the lobby of the building, as well as other functional space such as a cafeteria and a conference
room. In addition, the first floor also contained the pumps for the emergency power system. The fifth and
sixth floors were primarily mechanical spaces containing ventilation and other equipment. Eleven of the
fifteen power generators that comprised a critical component of the building emergency system discussed
in the next section were also located on the fifth floor along with a small diesel “day tank. Floors above
the sixth consisted mainly of both open-plan and traditional office spaces. The gravity system consisted of
wide flange beams and columns supporting a 140 mm thick concrete-on-steel-deck slab composite with the
floor framing using headed shear studs. Lateral load resistance was provided by moment resisting
connections along the perimeter frame and by two-storey high belt truss systems between the 22nd and 24th
floors and the 5th and 7th floors. Core bracing below the 7th floor resisted lateral loads transferred to it by
the floor diaphragm. Most floor framing connections consisted of various types of shear connections
including fin plate, knife, and multiple variants of seated connections [2]. SFRM (spray applied fire resistant
materials) type fire protection was applied to the structural components with a 3-hour rating for the columns
and 2-hour rating for the steel decking and beams, which exceeded the required fire protection of 2 and 1.5
hour-rating for the columns and beams respectively. The entire building was also covered with sprinklers
fed from the city mains from the ground to the 20th floor, and complemented with water tanks that serviced
the 21st floor and above. To transfer the load from the structure into the columns and foundation while
bridging over the Con Edison power substation below, a transfer system was used which consisted of three
trusses, eight cantilevered built-up girders, and three transfer girders all between floors 5 and 7 as shown
in Figure 2.
Figure 1. Typical floor layout of WTC 7
Figure 2. Transfer structure for WTC 7 [2]
2.2 Emergency fuel system
To ensure continuous operation of the primary functions of the building, an emergency power system was
installed. This system had a diesel fuel capacity of about 117 m3 which fed through several pump sets that
powered generators located on the 5th, 7th, 8th, and 9th floors. Separate from this system was the Salomon
Brothers emergency power system which had its own subgrade diesel tanks with a capacity of 45 m3, and
nine power generators all located on the 5th floor [3]. About 87 m3 of the fuel in the primary emergency
system storage tanks was recovered a few months after the collapse, which is within the expected residual
amount. None of the fuel within the Salomon Brothers system was recovered, however, and there were no
traces of leaked residue within the proximity of the then-damaged tanks. Since there was a pre-existing
275-gallon day tank on the 5th floor as part of the primary emergency power system, the Salomon Brothers
system had to be designed without a day tank. The system, therefore, was designed to supply fuel
continuously to the generators as long as they were operational. The rate at which fuel was pumped to the
generators was controlled by a regulator downstream from the generators. As long as the pressure
downstream from the generators is kept low, fuel would continuously circulate to power up the generators
which in turn powered up the pumps [3].
2.3 Fire and collapse
After WTC1, the Pentagon, and WTC2 were attacked, all 4,000 occupants of WTC 7 evacuated the
building. When WTC1 began to collapse at 10:28 AM, debris launched by the collapse damaged and
penetrated the south western face of WTC 7 severing the perimeter columns over floors 7 through 17 along
the building corner. Large areas of the Southern façade glass were also broken. Nonetheless, the debris
damage to the building and its structure did not pose an immediate threat to its stability. The water supply
to the sprinklers from the city mains was disrupted due to the collapse of WTC1 and 2, and the early
evacuation of building occupants made the building a low priority for the fire department which was
preoccupied with rescue and firefighting operation elsewhere. Multiple fires erupted as a result of the
burning debris that made it into the building, with visible fires over floors 7 through 9 and 11 through 13
burning uncontrolled for about 7 hours. The fires burned throughout the lower fire floors (floors 7, 8, and
9) spreading in a clockwise manner and travelled counter clockwise in the upper fire floors 11, 12, and 13
[4]. At 5:21 PM, a kink in the penthouse located on top of the building appeared followed immediately by
the penthouse plunging into the building suggesting an implosion followed immediately by the progressive
collapse of the entire structure beginning somewhere in the lower portion of the building.
3 PREVIOUS INVESTIGATIONS
The collapse of WTC 7 destroyed the electrical substation below it, for which the owners sued for damages.
This resulted in several expert teams investigating the collapse and each providing a plausible scenario by
which the building may have failed. Additionally, the National Institute of Standards and Technology
(NIST) were commissioned to do an independent study for public interest. In this section, the various
collapse scenarios are introduced and discussed starting with NIST, with particular attention paid to
numerical modelling approaches.
3.1 NIST
Investigation by NIST commenced after WTC 7 debris was cleared, and so the primary investigation
material was videos of the incident and the information known about the structure as per the drawings
available [5,6]. To analyse the failure, NIST considered four major factors: 1. Structural damage from the
debris of WTC 1, 2. Fire propagation as seen in the videos, 3. Computational Fluid Dynamics modelling of
the fire calibrated with the videos, and 4. Nonlinear static and dynamic finite element models of the
building.
3.1.1 Fire
According to the NIST investigation, the most severe fires were those occurring on floors 7 through 9 and
11 through 13 [1]. Fires on other floors may not have had a significant effect because they only burned
briefly and are thus not visible on video. Floors 7 through 9 comprised mostly of open plan offices with a
fuel load of 20 kg/m2 mainly consisting of office equipment and paper. This fuel load was arrived at by
considering the fire spread rate from the videos and comparing with the load-dependent calculated rate.
The upper fire floors, 11 through 13, had more traditional offices separated by partitions and joined with
false ceilings. The paper load on these floors was higher and so the fuel load NIST arrived at was 32 kg/m2
[1]. With this data, NIST used Fire Dynamics Simulator (FDS) to produce a model of each of the fire floors
individually. The models for floors 7, 8, and 12 were calibrated based on the visibility of flame and glass
breakage in the videos. The models for the fire on the 9th floor was derived from the 8th floor model, and
the models for floors 11 and 13 were based on the 12th floor model due to lack of sufficient footage for
calibration of each individual model. Using FDS derived gas temperatures, NIST then calculated the
structural temperatures in the structural members and varied them by +/-10% to create three scenarios that
envelope the potential building fires and account for uncertainties in the modelling to some extent [4,7].
3.1.2 Structural analysis
The structural analysis of the building was performed in two stages: stage 1. Implicit pseudo-static analysis
of the bottom 16 floors of the building considering the heating effects, and 2. Dynamic explicit model of
the entire building to investigate debris impact, heating damage, and collapse propagation. The first model
was created in ANSYS and accounted for shear stud connection failure in the eastern part of the fire floors,
lateral torsional buckling of floor beams, and removal of slab sections that had reached large strains.
Connections were modelled as collections of springs in series or parallel depending on the type of
connection modelled, and shear studs were modelled as ‘break elements’ that would be removed once they
reached their capacity. The results of the first model analysis showed the same behaviour occurring but at
slightly different time due to the variance in the heating scenario temperatures. The thermo-mechanical
damage generally resulted from connection failure which led to loss of vertical support of several floor
beams. Lateral torsional buckling also occurred in some members considering loss of shear studs and thus
absence of lateral support. The analysis was stopped after 3.5 to 4 hours of heating, and results transferred
over to a dynamic explicit model built in LS-DYNA [1].
The second model was analysed in incremental steps, starting with gradual application of dead load to avoid
introducing undue vibrations into the structure. The impact and damage from the WTC1 debris was then
added instantaneously, and then the model was allowed to stabilise. The addition of the debris impact indeed
showed that the structure could redistribute its load-bearing capacity and avoid disproportionate damage or
collapse as was observed in reality. The temperature of the structure was then gradually increased up to the
levels reached in the ANSYS model to induce secondary effects, before finally applying material and
component damage instantaneously to the model. After the damage was applied, the model showed that the
structure would undergo progressive collapse.
3.1.3 Collapse hypothesis
The collapse of WTC 7 according to the NIST models and investigation primarily initiated, because of
connection failure. The connection between the internal column 79 and the girder linking it to perimeter
column 44 failed by girder walk-off during the heating phase. To the east of column 79, many of the floor
beams were lost to either loss-of-vertical-support or lateral torsional buckling. The severe heating and
restrained expansion of the floor slab caused what remained of the floor to lose its stiffness by concrete
crushing, and severely weakened the fire floors causing partial collapse which could not be impeded and
resulted in cascading floor failure that left column 79 unsupported over 9 floors causing its buckling failure
which caused the kink observed in the collapse video. The buckling of column 79 was followed by similar
failures of columns 80 and 81, and then the progressive collapse of the adjacent column lines. The apparent
implosion of the interior of the building was followed by perimeter failure starting at a location adjacent to
the entry point of the majority of the WTC1 debris damage in the south west of the structure. The perimeter
columns failed in quick succession and the complete failure of the building occurred [4].
3.2 Arup & Guy Nordenson and Associates
Arup and Guy Nordenson took a different approach to the analysis of the WTC 7 collapse. First, Arup
performed an FDS and FE studies to predict the collapse initiation event. Then, GNA performed a series of
nonlinear studies to study the progression of collapse from its initiation point.
3.2.1 Fire
Similar to NIST, Guy Nordenson Associates (GNA) performed a photographic study of the debris damage
and fire in WTC 7 [8] and Arup performed a large scale FDS study of the 12th floor, which was one of the
six floors (7th-9th, 11th-13th) where persistent potentially structurally damaging fires occurred. Arup’s FDS
model was calibrated with the observed fire behaviour on the day, and the window breakage was
programmed following the photographs and videos. The fuel load used was 36.6 kg/m2 which correlated
well with the values provided in the NFPA Handbook for private and government office fuel loads [9,10].
No vertical fire spread was allowed, mechanical ventilation was assumed to have been off, and the building
core was not breached and remained free from fire throughout the simulation [11]. The simulation results
were in very good agreement with the observed fire in the north and east faces of the building, but less so
with that in the west face. However, as Arup had determined that the critical fire space was to the east of
the core, the FDS results were deemed acceptable [11]. In order to use the FDS simulation for structural
analysis, an idealized but consistent heating regimen of 1 hour heating at 800 °C and 1 hour cooling at 20
°C was implemented [9]. This approach was based on the ceiling-level temperatures at multiple monitoring
points throughout the area east of the core, and despite being simple, it accounted for the traveling nature
of the fire, preheating, and conservation of energy. Considering the applied fire protection, particularly the
assumption that the space between the ribbed floor slab and girders was left unfilled with SFRM, heat
transfer analysis was performed at the University of Edinburgh. The scenario of unfilled ‘flutes’ presented
with a unique temperature distribution in the floor girder where the top flange was the hottest, contrary to
standard thermal exposure in fire where top flanges would be the coldest. Similarly, heat transfer analysis
was performed considering filled flutes, and using an additional heating scenario where the ambient
temperatures would peak at 700 °C instead of 800 °C.
3.2.2 Structural analysis
The structural model used by Arup was built in Abaqus/Explicit [12] and represented the east half of a
typical floor, with symmetry boundary conditions representing the other side. The ribbed slab was
abstracted with uniform-depth shell elements including the steel decking and using Concrete Damaged
Plasticity [13,14] to consider material nonlinearity. While most of the steel framing was idealised using
beam elements, columns 44 and 79 as well as girders 79-80, 76-79, and 44-79 were modelled with a higher
grid resolution using shell elements [15]. Likewise, both connections of girder 44-79 and all connections
into column 79 were modelled in full detail using shell elements to represent the various plates, tie elements
to model the bolts, and rigid links to represent the welds. Shear studs were only modelled in the north east
corner of the floor, and assumed to be rigid elsewhere. Their force-slip relationship was temperature
dependent, and ‘broke’ beyond a slip limit of 6 mm as proposed by [16]. Details of the model as built by
Arup can be more clearly seen in Figure 3 [11]. Both mass and time scaling were used to ensure an efficient
solution, while inertial forces were monitored to maintain reasonably accurate results [15]. The Arup
analysis predicted that failure could indeed commence in the vicinity of column 79 in either heating or
cooling and only if the flutes between the girder and slab were unfilled. The expansion of the beams in the
north east, as well as the breakage of the studs over the girder, results in the girder severing its connection
bolts and being pulled off of its seat. Given this initiating event assessed by Arup, GNA investigated
collapse propagation using a full-scale elastic model built in SAP 2000 [17]. The model was supported by
multiple nonlinear sub-models that would manually inform the progression of the analysis such as removing
particular columns if they were prone to buckling. None of the models used by GNA were thermo-
mechanically loaded and instead the analysis focused on collapse progression rather than assessment of
thermal damage which was assumed to have been adequately addressed by the prerequisite Arup analysis.
Figure 3. The Arup model and its details as built in Abaqus [11]
3.2.3 Collapse hypothesis
The Arup-GNA investigation concluded that several factors and defects contributed to the collapse of WTC
7. While the debris from WTC 1 contributed to the collapse by starting the fires, its actual impact and
structural damage was largely inconsequential. Despite the fires burning uncontrollably for 7 hours
throughout the structure, their effect seriously jeopardised the building only after reaching the eastern face.
There, the high loads borne by column 79, the asymmetry of floor framing into girder 44-79, and most
importantly the un-filled flutes (no fire protection for the top flange) caused the girder to be pulled off of
its seat [18]. From that initiating event, the analysis performed by GNA showed that a sequence of
potentially stoppable events would lead to global collapse. Just after the failure of girder 44-79, the north
eastern floor slab would collapse over multiple floors causing column 79 to also lose its west-facing
connection and thus buckling. The failure of column 79 and its supported floor area leaves column 80
unbraced and its buckling then causes the kink observed in the penthouse as seen in the collapse videos.
The falling debris then damages the transfer trusses and results in the failure of the next column line
followed by pulling of the remaining core and causing the failure of interior of the building. Finally, the
perimeter frame begins to buckle resulting in total collapse of the building.
3.3 Weidlinger & Associates and Hughes Associates
WAI and Hughes Associates were tasked to investigate the collapse of WTC 7 on behest of the defendants
in the litigation over the power substation. Two reports were released based on their investigation, one
regarding the fire and another dealing with the structural response. Unfortunately, the former, produced by
Hughes Associates, was never made public and so little about it can be said.
3.3.1 Fire
Despite the WAI report adhering to the theme of thermo-mechanical response to thermal action presented
in the Hughes Associates report, some important insight about the fire is given. First, it is assumed that the
debris damage not only started the fires over multiple floors, but also inhibited the fire prevention measures
in the building. It is assumed that most compartmentation was lost and that the sprinklers were destroyed
due to the debris impact. It is also noted that the presence of widespread fires on affected floors resulted in
a severely reduced ability to arrest potential progressive collapse by weakening the floor capacity to resist
the local impact of a falling floor.
3.3.2 Structural analysis
Similar to Arup and NIST, WAI also used an explicit dynamic model to study the collapse of WTC 7. The
analysis performed by WAI relied on one full building model, and two sub-models representing the eastern
bay of one and two floors respectively. The sub-models were used to study the thermo-mechanical response
of the east bay, while the full building model was used to study the collapse propagation. All models were
built using WAI’s inhouse nonlinear FE software ‘FLEX’, and included a high level of detail. All shear
studs, welds, connections and and slab-beam contact were modelled explicitly in the sub models. Moreover,
all beams, girders and columns were modelled using shell elements. The global model was less complex
with connections built as consolidations of uniaxial spring based on connection capacity estimation using
the component method, and beam elements representing the steel framing.
Mechanical load was applied to the models by increasing the gravity and the density incrementally until
full quasi static load was applied. Thermal loading was also a quasi-static phase employing time scaling
and incurring few dynamic effects. The thermal loading continued until failure initiation, at which point
dynamic analysis proceeds until re-stabilisation or total collapse.
3.3.3 Collapse hypothesis
WAI analysis also showed that the collapse initiated in the eastern part of the building. However, the
initialising event did not occur in the vicinity of column 79 but in the proximity of column 80. Uncontrolled
heating in the eastern bay had led to floor deflections in the order of 1/15 of the span, which resulted in
high tensile forces in the floor beams. At the same time, the uncontrolled heating had resulted in high
connection temperatures of which the nearest to column 80 would sever and lead to zipping off of adjacent
connections. With this, the eastern portion of the floor would collapse impacting a similarly weakened floor
resulting in progressive failure of the east section of the interior floor of WTC 7. Columns 79, 80, and 81
then unsupported over multiple floors would buckle and cause the kink in the penthouse. The collapsing
debris severely damaged the transfer structure on the fifth and sixth floors, and also resulted in westwards
internal collapse propagation. The unbraced perimeter columns then buckled and the total failure of the
structure occurred [19].
4 MECHANICAL ROOM FIRE
As previously mentioned, the Salomon Brothers emergency power system was designed as a pressurised
loop where fuel would continuously be pumped from the storage tanks as long as the generators were
operational. It is possible that once a power outage was detected on 11 September, the Salomon Brothers
power system would have started. Had there been a penetration in the double-walled fuel line anywhere
away from the first few generators on the fifth floor, a low-pressure signal in the valve rig at the end of the
system may have signalled to the pumps to continuously push fuel into the system [3]. There was no
indication of any automatic leak-detection system in the Salomon Brothers system, and it is therefore
possible for all available fuel to have been available at the fifth floor to feed a potential fire there. NIST did
consider this scenario in their analysis but dismissed it based on three reasons:
1. A pool fire would have raised the temperature beyond the generators operational limit thus cutting power
supply to the fuel pumps.
2. The fire would not have heated the structure to critical levels.
3. A critical fire in that location would have resulted in significant exhausted smoke visible from the outside
of the structure but none were seen.
This assessment was based on FDS results assuming a pool fire occurring in the vicinity of column 79 just
outside the mechanical room located on the 5th floor. It is possible that a breach in the fuel line and masonry
wall of the mechanical room would have led to a pool fire inside the mechanical room, or a fire may have
started just outside the mechanical room and propagated into it through damage in the partition. Both of
these scenarios would have heated the large transfer trusses in the mechanical room resulting in potentially
critical damage to the structural integrity of the building. Such a diesel fuel fire would likely generate large
amounts of smoke that, if not all had exhausted through other openings in the building or its core, would
have been visible outside the east face of the structure. As noted by Mowrer [9], petroleum-based smoke
was indeed observed in the videos and witnessed by first responders after 3:30 PM just outside the north
east corner of the building as seen in Figure 4 [8].
Figure 4: Smoke outside the north east corner of WTC7 between 4:15 and 4:38 PM [8]
4.1 Fire scenario
Assuming a pool fire in the mechanical room, an FDS analysis was run by the BRE Center for Fire
Engineering [20]. Several ventilation conditions were assessed considering the opening of the mechanical
room to the plenum east of the building, and the results were compared to the Thomas curve [21] since the
fire had reached steady state burning within 300 600 seconds in all cases. All FDS simulations gave
predictions below the Thomas curve indicating that the analyses were conservative and well-below
experimental data and reality. By considering energy balance while neglecting re-radiation and conduction,
and assuming a linear growth phase for the fire it was found that the column and truss-member temperatures
would reach steady state with a small delay but are otherwise the same as the gas-phase temperatures.
Considering the fire protection, a 20% lag in heat uptake may occur but would be inconsequential assuming
that for the columns it was thermal degradation and so maximum attained temperatures that were most
critical [20]. Similarly, a CFD analysis of a pool fire in the vicinity of column 79 was performed and it was
found that the column temperatures would be very similar to the mechanical room fire scenario.
4.2 Structural analysis
The approach that the BRE Centre had taken towards the structural analysis of the collapse was unique as
it had paid more attention to the global structural behaviour and focused on load redistribution rather than
on localised failures of individual connections and elements. A 15-storey model was built in Abaqus using
beam-column elements to represent the steel framing, shell elements to represent the concrete slab. Vertical
load corresponding to the column forces from the upper floors was applied at the top of the columns which
were allowed to translate vertically but not planarly. All elements were connected to one another rigidly;
no connections were allowed to fail and composite action was maintained throughout the analysis. The
rationale for this modelling approach is because the investigators believed that adding all the details entails
making many assumptions most of which would complicate the problem and obscure what may been a
global limit state. Unlike the other models, this particular approach did not aim to recreate the exact events
but rather give insight into the global behaviour of the structure and unlock the many possible ways the
buildings may have potentially failed.
An elimination analysis was performed where different collapse scenarios were modelled and checked for
consistency with the observed collapse. Heating of only truss 1, heating of only truss 2, heating of both
trusses, and heating of only column 79 were all tested. In each of the scenarios, load transfer played a
critical role in overall deformation of the structure, and it was identified that it was most likely that a
combination of heating both trusses were responsible for the collapse as will be discussed in the next
section. This analysis also showed that heating of column 79 just outside the mechanical room would likely
result in a similar global collapse.
4.3 Collapse hypothesis
A fire in the mechanical room would heat columns 77 and 80 as well as the diagonals of truss 2 which
connects them. As these columns lose their capacity, their load is transferred to columns 76, 79, and 81.
Likewise, the exposed and partially heated truss 1 is also losing capacity due to heating thus transferring
its own capacity to the core and to column 79. Given that column 79 is likely to have been heated due to a
diesel fire in its location, and that it also is the most heavily loaded column, it is likely that it would have
failed. From this, it was clear that failure of any one of truss 1, truss 2, or column 79 would overload the
other two and lead to their failure. It was, therefore, that combined heating of truss 1 and truss 2, and
potential heating of column 79 was the primary factor for global collapse of WTC 7. The initiating failure
could have been either failure of column 79 or of truss 2, because a collapse event beginning at either of
these two points had the potential to manifest the kink observed in the penthouse and cause its sinking as
was seen in the videos.
5 CONCLUSIONS
The collapse of WTC 7 remains to this day a unique event in the history of structural fire engineering.
Despite occurring nearly twenty years ago, there is no real way to know what exactly caused its failure.
Multiple expert teams had investigated the disaster and tried to recreate the various events that led to the
progressive failure of the structure. With most debris gone, the expert teams had to rely on photos, videos,
and numerical modelling to piece together their different hypotheses. FDS and explicit FEM played a
crucial role in each of the investigative teams’ approaches, which shows a trend towards more
computationally intensive methods for forensic analysis of fire-attacked structures. By reviewing the
various modelling approaches and hypotheses of the investigators some important insights about the case
can be gained:
1. There is consensus amongst the investigators regarding two main aspects of the collapse of WTC7:
fire was the primary cause of the failure, and the failure initiated in the east side of the building
somewhere in the lower 14 floors.
2. FDS simulations of affected floors 7-9 and 11-13 calibrated against videos produce comparable
temperatures in the structural components. Even the small variations in temperature, however, may
result in different outcomes depending on the nature of the structural model.
3. Despite the tremendous efforts of the various investigative teams, there is no consensus regarding
the exact initiating mechanism. Arup’s analysis showed that girder 44-79 would be pulled off of its
seat, opposite to the prediction by NIST which show the girders pushed off of their seats. WAI’s
analysis points towards a floor beam connection failing followed by entire floor collapsing. Finally,
the UoE mechanical room fire scenario predicts that the failure may have occurred due to a
completely different fire caused by a breach of the emergency power system.
4. Numerical models are only idealisations of real events, and thus by nature they would be dependent
on the decisions made by the analyst. There are many discrepancies between the different numerical
models and outcomes for WTC 7, and each may be sensitive to different factors. There still remains
a gap in the knowledge on the most suitable approach to modelling large structural systems in fire,
and what is the right level of abstraction to capture the most important aspects of the behaviour
without complicating the models beyond reasonable computational cost and pushing it outside its
predictive capacity.
REFERENCES
1. NIST, NIST NCSTAR 1A: Final Report on the Collapse of the World Trade Center Building 7, 2008.
2. NIST, NIST NCSTAR 1-1: Design , Construction , and Maintenance of Structural and Life Safety Systems,
2005.
3. NIST, NIST NCSTAR 1-1J: Documentation of the Fuel System for Emergency Power in World Trade Center
7, 2005.
4. NIST, NIST NCSTAR 1-9: Structural Fire Response and Probable Collapse Sequence of the World Trade
Centre Building 7, 2008.
5. Frankel Steel Limited, Erection Drawings: 7 World Trade Center, (1985).
6. Office of Irwin G. Cantor, Structural Design Drawings: 7 World Trade Center, (1985).
7. T. McAllister, R. MacNeill, O. Erbay, A. Sarawit, M. Zarghamee, S. Kirkpatrick, J. Gross, Analysis of
structural response of WTC 7 to fire and sequential failures leading to collapse, J. Struct. Eng. (United States).
138 (2012) 109117. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000398.
8. Guy Nordenson and Associates, Photographic Analysis: Volume A Photographic Timeline, New York, 2009.
9. F. Mowrer, Expert Report by Frederick Mowrer, 2010.
10. National Fire Protection Association, Fire Protection Handbook, 20th edition, 2008.
11. Ove Arup & Partners Consulting Engineers PC, World Trade Center 7 Floor 12 Fire Effects Simulation, 2010.
12. Simulia, Abaqus, (2018).
13. J. Lee, G.L. Fenves, Plastic-damage model for cyclic loading of concrete structures, J. Eng. Mech. 124 (1998)
892900. https://doi.org/10.1061/(ASCE)0733-9399(1998)124:8(892).
14. Dassault Systèmes, Abaqus Analysis Users Manual, (2012).
15. Arup, JA-3240: Master Assumptions List, 2009.
16. Z. Huang, I.W. Burgess, R.J. Plank, The influence of shear connectors on the behaviour of composite steel-
framed buildings in fire, J. Constr. Steel Res. 51 (1999) 219237. https://doi.org/10.1016/S0143-
974X(99)00028-0.
17. Guy Nordenson and Associates, WTC7 Global Collapse Analysis Report and Summary of Findings, 2010.
18. Arup, WTC -7 Structural Fire Analysis Report (Runs 1-4), 2010.
19. Weidlinger Associates Inc., WTC 7 Collapse Analysis and Assessment Report, 2010.
20. BRE Centre for Fire Safety Engineering, Analysis of the Impact of a Fire in the Mechanical Room (5th & 6th
Floor) of the World Trade Center 7 Building, 2010.
21. P.. Thomas, A.J.M. Hseselden, Fully Developed Fires in Single Compartments: new correlations of burning
rates, 1973.
Article
Full-text available
The utilisation of composite floor systems in modern construction allows for building large floor areas with efficient use of material and labour. The efficiency of this flooring system in carrying ambient loads as well as the thermal thinness of its steel framing makes it hypothetically sensitive to fire loads as a small exposure time may result in a rapid increase in temperature. However, a series of fire events and then large-scale experiments at the turn of the century demonstrated that composite floors could display surprising robustness in the face of high temperatures. Soon after, the total collapse of the World Trade Center towers 1, 2 and 7 showed the opposite. For the two decades since then, a large amount of experimental and theoretical work has been performed on this topic. This review aims to formulate an understanding of the thermo-mechanical behaviour of composite floors in fire, and to leverage that understanding to design more efficient and more robust buildings. We first surveyed the literature, and then summarized the composite floors being studied into three categories: isolated composite slab panels, composite floor assemblies, and large-scale tests. Then, theoretical design and analysis methods are discussed in light of the reviewed literature. Finally, this review coalesces the knowledge generated from the literature and lists the remaining challenges in the area, and set out a set of recommendations for designing composite floor systems for fire.
Article
Full-text available
The collapse of several tall composite buildings over the last two decades has shown that the performance of tall, composite and complex buildings in fire is a necessary design consideration that ought to go beyond simple code compliance. To this end, several advancements in the field of numerical simulation of both the fire and the thermomechanical response of structures have been made. In isolation, the practical benefit of these advancements is limited, and their true potential is only unlocked when the results of those numerical simulations are integrated. This paper starts by showcasing recent developments in the thermal and thermomechanical analysis of structures using OpenSees. Integration of these developments into a unified simulation environment combining fire simulation, heat transfer, and mechanical analysis is then introduced. Finally, a demonstration example based on the large compartment Cardington test is used to showcase the necessity and efficiency of the developed simulation environment for thermomechanical simulation of composite structures in fire.
Article
This paper presents the structural analysis approach used and results obtained during the investigation conducted by the National Institute of Standards and Technology (NIST) to model the sequence of fire-induced damage and failures leading to the global collapse of World Trade Center 7 (WTC 7). The structural analysis required a two-phased approach to address both the gradual response of the structure to fire before collapse initiation (approximately 4 h) and the rapid response of the structure during the collapse process (approximately 15 s). This paper emphasizes the first phase, a pseudostatic (implicit) analysis that simulated the response of structural elements to fires that spread and grew over several hours and presents key aspects of the second phase, a dynamic (explicit) analysis that used the first-phase damage as initial conditions and simulated the progression of structural failures that resulted in global collapse. The analyses accounted for (1) geometric nonlinearities; (2) temperature-dependent nonlinear materials behavior for both members and connections (including thermal expansion, degradation of stiffness, yield and ultimate strength, and creep); and (3) sequential failure of structural framing and connections. Analysis uncertainty was addressed by determining rational bounds on the complex set of input conditions and by running several multiphase analyses within those bounds. The structural response from each analysis was compared to the observed collapse behavior. This approach allowed evaluation of fire-induced damage, sequential component failures, and progression of component and subsystem failures through global collapse of WTC 7. DOI: 10.1061/(ASCE)ST.1943-541X.0000398. (c) 2012 American Society of Civil Engineers.
Article
A new plastic-damage model for concrete subjected to cyclic loading is developed using the concepts of fracture-energy-based damage and stiffness degradation in continuum damage mechanics. Two damage variables, one for tensile damage and the other for compressive damage, and a yield function with multiplehardening variables are introduced to account for different damage states. The uniaxial strength functions are factored into two parts, corresponding to the effective stress and the degradation of elastic stiffness. The constitutive relations for elastoplastic responses are decoupled from the degradation damage response, which provides advantages in the numerical implementation. In the present model, the strength function for the effective stress is used to control the evolution of the yield surface, so that calibration with experimental results is convenient. A simple and thermodynamically consistent scalar degradation model is introduced to simulate the effect of damage on elastic stiffness and its recovery during crack opening and closing. The performance of the plastic-damage model is demonstrated with several numerical examples of simulating monotonically and cyclically loaded concrete specimens.
National Fire Protection Association, Fire Protection Handbook
  • F Mowrer
F. Mowrer, Expert Report by Frederick Mowrer, 2010. 10. National Fire Protection Association, Fire Protection Handbook, 20th edition, 2008.
The influence of shear connectors on the behaviour of composite steelframed buildings in fire
  • Z Huang
  • I W Burgess
  • R J Plank
Z. Huang, I.W. Burgess, R.J. Plank, The influence of shear connectors on the behaviour of composite steelframed buildings in fire, J. Constr. Steel Res. 51 (1999) 219-237. https://doi.org/10.1016/S0143-974X(99)00028-0.