A Suggested Cause of The Fire-Induced Collapse of The World Trade Towers..Quintiere2002

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Fire Safety Journal 37 (2002) 707–716

Short Communication

A suggested cause of the fire-induced collapse of


the World Trade Towers
J.G. Quintierea,*, M. di Marzoa, R. Beckerb
a
Department of Fire Protection Engineering, University of Maryland, College Park, MD, 20742, USA
b
Faculty of Civil Engineering, Technion - Israel Institute of Technology, Israel
Received 24 May 2002; received in revised form 3 June 2002; accepted 5 June 2002

Abstract

An analysis is presented that calculates the temperature of the steel truss rods in the World
Trade Center towers subject to a fire based on the building ventilation factor. The CIB
correlation is used for the fire. Conduction analyses are made taking into account variable
properties for the steel and the insulation. A structural failure model is described based on
compression buckling of the truss rods due to a reduction in the Young’s modulus. The
computed times for the estimated failure or incipient collapse of the floors in both towers has
been computed as 105720 min for WTC 1 (north) and 5179 min for WTC 2 (south),
compared to the collapse times from the aircraft impact of 104 and 56 min, respectively. The
00 00
insulation thickness and the difference of 19.1 mm (34 ) and 38.1 mm (112 ) between the two
towers appear to have been the root cause of the collapses. r 2002 Elsevier Science Ltd. All
rights reserved.

1. Introduction

The fall of the World Trade towers has been initially reported and generally
accepted as due to the collisions by aircraft and its fuel. Subsequent analyses and
findings have dampened that cause and have pointed more at the fire involving the
building contents [1, p. 2, 2]. It is estimated that about 29 Mg of jet fuel was in each
aircraft and about 9.4 Mg of jet fuel was expended in the fireballs of the initial
impacts, and the remaining jet fuel may have burned for several minutes more on a
floor [2]. These calculations are based on standard formulas in the literature, and

*Corresponding author. Tel.: +1-301-405-3993; fax: +1-301-405-9383.


E-mail address: jimq@eng.umd.edu (J.G. Quintiere).

0379-7112/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 7 9 - 7 1 1 2 ( 0 2 ) 0 0 0 3 4 - 6
708 J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716

Nomenclature

A ceiling and wall surface area


Ao ventilation opening area
D diameter of the truss rod
E Young’s modulus
h overall heat transfer coefficient
hconv convective heat transfer coefficient
Ho compartment height
kins thermal conductivity of the insulation
L length of the truss rod
P critical buckling stress
r radial coordinate, see Eqn. (1)
t time
T temperature
Tgas gas temperature

Greek letters
a thermal diffusivity
s Stefan–Boltzmann constant.

easily and clearly show that the jet fuel was only responsible for the ignition of the
contents. The lack of clarity in the media is indicative of the lack of knowledge or
reliance on the science of fire in reaching a quick answer. The recent FEMA report
[1, pp. 236–38] could not determine the cause of the collapses. They cited a number
of issues: (1) aircraft damage but acknowledge that the two buildings would have
remained standing had they not received a fire load, (2) the jet fuel although it
burned out quickly ignited the contents and caused rapid fully developed floor fires,
(3) the type of steel truss floor system should be subject to more detailed evaluation.
This is what we have done. They also cited that fireproofing needs to adhere under
impact and fire, and connection performance needs to be better understood. In
00
particular, two 15.9 mm (58 ) and two 25.4 mm (100 ) diameter bolts to the exterior
00
columns, and two 15.9 mm (58 ) diameter bolts to the interior columns connected the
floor trusses. These connections have been questioned as weak links.
Instead, we have focused on the transverse and main trusses, and in particular on
the 27.7 mm (1.0900 ) diameter truss rods. Our hypothesis is that the truss rods are the
weak link because they have the lowest steel cross-sectional mass and the fire would
increase its temperature the fastest. Therefore, we have based our analysis on a link
between the temperature evolution and the buckling under restrained elongation.
Our hypothesis is illustrated in Fig. 1. We suggest that buckling of the transverse-
truss compressed diagonal rods, which were adjacent to the main-truss, occurred first
due to the combination of their restrained elongation and quick loss of buckling
resistance induced by their fast temperature rise (marked as St. A in Fig. 1). With the
J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716 709

Fig. 1. Diagram of suggested sequence of events triggering progressive collapse (truss drawings taken
from FEMA report [1, pp. 2–8 & 2–9])
710 J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716

buckling of these diagonals, the compressive forces in the transverse-truss middle


diagonals were suddenly increased. As they had a similarly reduced buckling stiffness
they buckled soon after (marked as St. B), and the whole transverse-truss failed,
sagging downwards (marked as St. C). Sagging of the floor deck followed,
developing some membrane action in the steel deck, and possibly disconnecting the
concrete within the floor span from the main-truss’s upper bulk of concrete due to
combined large tensile and shear stresses (marked as St. D). Due to the ductility and
continuous nature of the deck itself and the continuous nature of the transverse-truss
bottom chord, the floor probably did not fall down immediately, but rather stayed
hanging from the main-truss bottom chord for some period. Subsequently, the
buckled transverse-trusses and steel/concrete deck fell down on the floor below, thus
aggravating loading conditions and enhancing its failure.
Simultaneously the fire challenged the main-trusses as well. Again, the thin
diagonal rods heated much faster than the other members. However, only part of
them was in compression and with various loading ratios. As the ‘‘truss’’ was not
statically determinate (its upper and lower chords were highly continuous and did
not include hinged joints at all), buckling of some compressed diagonals alone
(marked as St. A1 in Fig. 1) would not necessarily cause immediate collapse;
however, it would be a main trigger to further deterioration of capacity, leading
eventually to total failure. Our hypothesis is that the whole collapse mechanism of
this truss was as follows: Despite the slotted holes at the supports, the main-truss
upper chord was too restrained and could not elongate freely. A horizontal force was
induced at the supports, which were highly above the neutral line of the main truss’s
bending as a whole. This induced a high compressive force into the upper chord, and
a superimposed bending moment on the truss as a whole, increasing its tendency to
either strongly deflect downwards (with possible buckling of the upper chord in the
vertical plane), or buckle in a transverse mode. However, since the upper chord had a
much larger stiffness in the horizontal direction, transverse buckling did not occur.
With most of the compressed diagonals already buckled, the ‘‘beam’’ or ‘‘truss’’
behavior did not exist any longer, and the upper chord acted as a ‘‘cable’’, from
which all the other members are hanging downwards.
This increased the pressure on the inner edge of the end seats while increasing the
tension in the bolts, actually bending and shearing the bolts due to the enlarged
rotations. Eventually, the bolts failed as well, possibly rupturing the seat
connections, and total collapse of the floor ensued, placing load on the floor below.
With several floors simultaneously involved in fire, this would cause several floors to
load a floor below. The connections at the columns could not tolerate the load and
progressive collapse ensued even to floors that were not on fire. The progressive
collapse mechanism is well reported and accepted, but the initiating event is not
established. We offer the truss failure as described.
The following analyses were prompted by the FEMA report [1, pp. 2–12] that
00
stated that the average thickness of the fireproofing on the trusses was 19.1 mm (34 ),
00
and were in the process of being upgraded to 38.1 mm (112 ). On September 11, 2001,
WTC 1 (north) had the higher level in the impact zone (floors 94–98), and WTC 2
(south) only had the higher level on floor 78 with impact over 78–84. The insulation
J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716 711

was sprayed mineral fiber, reportedly Cafco DC/F similar to Isolatek Blaze-Sheild II
having a value of thermal conductivity of 0.043 W/m K at 241C. The DC/F is slightly
higher at 0.046 W/m K.

2. Analysis

There are three parts to this analysis: (1) The fire, (2) The heat transfer to the steel,
and (3) The structural failure. The fire depends principally on the ventilation, and
secondly on the fuel type. The fuel loading (weight or energy available on the floor)
will determine the duration of the fire. The heat transfer to the steel principally
depends on the insulation, and the variation of thermal conductivity is crucial in
such an analysis. The gas-phase heat transfer conductance is very high and the
surface temperature of the insulation can usually be taken as the gas temperature;
however, we did not make such approximations. Instead, we strove to calculate the
heat transfer as accurately as possible, and the fire conditions were allowed to vary
over a plausible range. The time of structural failure of the main-truss rods depends
on the temperature-induced decreasing modulus of elasticity lowering the resistance
to buckling. The time of structural failure of the transverse-truss rods depends on the
combined effects of increasing compressive stresses stemming from temperature rise
under elongation restraint, and the decreasing modulus of elasticity lowering the
resistance to buckling.

2.1. Fire temperature

The fire temperature during fully developed conditions is computed from the
correlations reported from the CIB test analyses [3]. These tests are of scales up to
1.5 m high and for wood cribs covering the floor. Fuels of higher heats of
combustion and larger scale can produce higher temperatures, but this is only
suggested theoretically [4].
The gas temperatures are based on Ao ranging from 124 and 267 m2 for WTC 1 [1]
(p. 2–23), Ho equal to 3 m, and A based on a floor of 63.5 m by 63.5 m and a utility
core of 26.5 m by 41.8 m that is about 4100 m2. With these values, the group
1=2
A=Ao Ho can be estimated between 15 and 33 m–1/2. Fig. 2 provides a gas
temperature range between 8001C and 10001C.
The corresponding floor burning rates in wood are 14.5–18.7 kg/s or about
250 MW per floor. For typical floor loadings of 30 kg/m2, the fire duration per floor
would be about 80–100 min. There is variation in this duration since the burning rate
is based on a wood crib configuration; other configurations would vary. But the
results suggest, based on the actual event that the duration was not much greater
than the WTC 1 time to collapse. This would imply that had the steel structure
survived longer, the fire duration time would have been less than the failure time, and
no complete failure would have occurred. The loading and burning rate are key
issues here and more information on the demography of the WTC furnishings, and
research on different fuel types are needed.
712 J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716

1200

1000
T (°C)
800

600

400
0 10 20 30 40 50
1/2 -1/2
A/AoHo (m )
Fig. 2. Temperature range of exposure in WTC 1 & 2 [3].

Hence, a fire temperature of 9001C was used in our calculations based on the mean
in Fig. 2, and will commence and stay constant throughout. This is justified based on
the rapidly developing fire from the jet fuel, and on the estimate that the fire duration
on a floor is longer than the collapse time.

2.2. Heat transfer to the steel

The heat transfer analysis is a standard numerical calculation of the heat


conduction equation for a composite cylinder of homogeneous insulation and steel.
The governing equation is given as:
 
qT a q qT
¼ r : ð1Þ
qt r qr qr
The gas-phase boundary condition uses a constant convective heat transfer
coefficient of 20 W/m2 K and blackbody conditions for the radiant heat transfer.
The latter is justified because of the large scale of the fire, and the convective
component is a typical value for fire conditions and is negligible compared to
radiation. The combined heat transfer coefficient is expressed as:
2
h ¼ hconv þ sðTgas þ T 2 ÞðTgas þ TÞ: ð2Þ
The heat transfer coefficient is in W/m2 K and the temperatures in the equation are in
degrees Kelvin. These values used for the overall heat transfer coefficient have small
effect on the final results since the insulation provides the significant resistance to
heat flow. At the insulation-steel interface, we assume that there is no contact-
resistance; we impose continuity of the temperature and conservation of energy
across the boundary.
The properties of both the steel and insulation vary considerably with temperature
and cannot be neglected. The manufacturer of the insulation states the conductivity
is 0.046 W/m K at normal temperature, and the literature indicates an approximate
J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716 713

linear increase [5,6]. We used:

kins ¼ 0:046 þ 0:00024 ðT  20Þ ð3Þ

The thermal conductivity is expressed in W/m K and the temperature in this


expression is in degree centigrade. The thermal diffusivity of the insulation is
considered constant and equal to 3.1  107 m2/s [5]. The steel exhibits a strong
variation both in thermal conductivity as well as in volumetric specific heat. We have
carefully fitted the curves provided by Lie [7].
00
Note that for WTC 1 (north) we assign the insulation to be 38.1 mm (112 ) in
300
thickness while for the WTC 2 (south), the insulation thickness is 19.1 mm (4 ) [1]
(pp. 2–12). Eq. (1) is integrated with the Crank Nicholson implicit scheme starting
from a uniform initial condition at 201C throughout the steel and insulation. The
temperature results are shown in Fig. 3.
It must be realized that there is an incipient delay time that must be added to the
results in Fig. 3 since the ceiling membrane would retard the heating. It can be
argued that the impact of the aircraft destroyed the ceiling partition, and this is
certainly true in places. But where it has been destroyed, the fuel on the floor is also
likely pushed away. The aircraft had a weight at impact of about 170 tons and each
tower weighed about 750,000 tons; or the impact is like a sparrow hitting a 125 kg
person [8]. Moreover the engines and landing gear are the dominant destructive
missiles. The aircraft aluminum skin and the fuel would quickly dissipate its
momentum. Hence, we believe the fuel from the furnishings and contents would be
pushed to areas where there would be sustained fire, with no fire likely in the
immediate impact area. These fuel piles would increase the fire duration in those
areas, and first attack the ceiling membrane.
The acoustical ceiling tiles initially installed in the World Trade Center towers
were of one type (Armstrong World Industries Inc.) and were not specified to be fire
resistant. If not immediately destroyed, this membrane would likely retard the direct
heating of the insulated truss by about 10 min, an estimation based on experience.

1000

800

600
T (°C)

400
WTC-2
200
WTC-1
0
0 20 40 60 80 100 120
t (min)
Fig. 3. Temperature of the steel truss rods for WTC 1 & 2.
714 J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716

Any reconstruction of the incidence needs to account for this membrane effect more
completely.
It has been speculated that the impact knocked off the protective insulation. We
do not believe that this occurred to any extent in the fire region since a calculation of
the bare steel chord would suggest failure in 10–15 min. Again, in the impact area
this may have happened, and testing needs to be done to assess the robustness of the
insulation to impact.

3. The structure

The truss rods can be analyzed in tension and in compression at the mid-section.
Compression can be shown to be the critical condition. The rods were A36 steel
with a yield stress of 253 Mpa (36 ksi) and an ultimate stress of 422 MPa (60 ksi)
[1, p. B2–3]. The rods are nominally 0.889 m (3500 ) long. The critical buckling stress of
the rod in compression is given as:
 
pD 2
P¼E ð4Þ
4L

For E equal to 210,900 MPa (30,000 ksi), the critical buckling stress is 126 MPa
(17.9 ksi). Since this is less than the yield stress, the member fails in buckling. With
the design load stress as 126 MPa (17.9 ksi), if the overall safety factor is 2–3, then the
actual load just before the fire may have induced a stress of about 42.2–63.3 MPa
(6–9 ksi). With the rising of the rod temperature, it wishes to elongate, but the
continuity of the upper and lower truss chords restrains this elongation. The
reduction in E due to temperature to render this stress, 53711 MPa (7.571.5 ksi),
into a critical buckling stress must be about 70,300–105,500 MPa (10,000–15,000 ksi)
[7]. The corresponding critical failure temperature is about 630–7701C, accordingly.
With this temperature bound we estimate the time to failure for the towers to be
given in Table 1 using a fixed fire temperature of 9001C.
The time to failure suggested in the table complemented by the addition of 10 min
to account for the suspended ceiling membrane gives times close to the actual failure

Table 1
Time for critical events (minutes)

WTC 1 WTC 2

Time for the steel to reach 6301C 75 32


Time for the steel to reach 7701C 114 49
Time to failure for the suspended ceiling 10 10
Time to failure range 85–124 42–59
Computed average time to failure 105 51
Actual failure time 104 56
J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716 715

events in both buildings. We find this result extremely telling and we believe forms
the root cause initiating the collapses.
For structural steel elements in standard fire testing, the failure criterion ranges
from 550–6001C. Therefore, we have also made computations for the truss rod to
reach 6001C for fire temperatures of 800–10001C. The times to reach these
temperatures are roughly 75712 min for WTC 1 and 3276 for WTC 2. Hence, no
matter how one estimates failure, the difference between the collapse times of the two
buildings is of 4477 min compared to the actual collapse time difference of 48 min.
This strongly correlates with the insulation thickness.
Should one accept that indeed the generalized failure of the truss rods is the
initiating event, then it follows that the consequent failure of the connections could
be the result of the floor sagging and pulling on the connections downward thus
exerting a shearing action that could fail them. From some circumstantial evidence
shown in the FEMA report we see failure of the connections, as they were pulled
apart rather than failure of weak elements of the connections such as bolts. Further,
the images of the collapse of the WTC 1 (north) seem to show a burst of flame and
smoke outward just before the collapse. This, in our view corroborates the idea of a
generalized floor failure. The collapse of a floor would cause the smoke and fire to be
pushed out of the building from the space below as video images seems to indicate.
Upon loading one or two additional floors on the floor below, the whole structure
would start failing as more and more floors fails and increased instabilities in the
vertical unconstrained elements arise.

4. Conclusions

It appears that the insulation thickness on the truss rods was deficient and caused
the heating of the steel that led to weakening and collapse. The variation of 19.1 mm
00 00
(34 ) and 38.1 mm (112 ) thickness between the towers needs to be investigated further,
and understood in light of fire safety design principles. The logical questions include:
What was the basis of the original fire safety design and specification? Our estimate
00
of the truss rod failure by temperature in the ASTM E 119 test for 38.1 mm (112 )
thick insulation is about 67–79 min corresponding to failure at 5001C and 6001C,
respectively. Our estimate of fire duration was about 80–100 min. If these times are
correct, it suggested that the fire safety design criteria need assessment. It has been
implied that a two-hour criterion should have been used for this floor assembly.
Certainly, some factor of safety needs to be incorporated in relating fire duration
time with test ratings or computed failure times.
The truss rod failure is proposed as the initiating event. An important implication
of this proposed mechanism points to the dominant effect of the insulation applied
to the truss members. If one follows this theory, the reduced amount of insulation in
the WTC 2 could have resulted in a premature collapse of that tower by almost one
hour compared to WTC 1.
716 J.G. Quintiere et al. / Fire Safety Journal 37 (2002) 707–716

5. Recommendations

We have used available and presumably accurate information in this analysis. It


requires checking and a more complete analysis. We have pointed to a failure
mechanism that is consistent with calculations. This needs to be examined by more
complete testing of the structure, and more extensive calculations. The heating by the
fire needs more study. A scale model rendition might be a way to produce data more
accurately, and a basis for computer validation.

Acknowledgements

We appreciated checks on the early thermal calculations by Ali Rangwala. We


acknowledge useful discussions with colleagues: C.C. Fu, J. Milke and F. Mowrer,
and N. Shultz of VTEC Laboratories, Inc.

References

[1] World trade center building performance study. Federal Emergency Management Agency, FEMA
2002.p. 403.
[2] Torero JL, Quintiere JG, Steinhaus T. Fire safety in high-rise buildings, lessons learned from the WTC,
Jahresfachtagung der Vereingung zur Forderrung des Deutschen Brandschutzez e. V., Dresden,
Germany, 2002.
[3] Thomas PH, Heselden AJM. Fully developed fires in single compartments—a co-operative research
program of the Conseil International du Batiment. FR Note No. 923, Fire Research Station, UK,
1972.
[4] Quintiere JG. Fire behavior in building compartments. Proceedings of the 29th International
Symposium on Combustion. Sapporo, Japan: The Combustion Institute, accepted for presentation.
[5] Kreith F, Bohn MS. Principles of Heat Transfer. 6th ed. US: Brooks/Cole, Thomson Learning, 2001.
p. 44.
[6] Reed RJ. North American Combustion Handbook. 3rd ed., vol. 1, Cleveland, OH: North America
Manufacturing Co., 1986. p. 120.
[7] Lie TT. Structural Fire Protection. New York: American Society of Civil Engineers, 1992.
[8] Sometime Lofty Towers. Browntrout Publishers, 2001.

You might also like