Analysis of Sequential Failures Leading to the Collapse of WTC 7, Part I: Structural

Response to Fire

Therese McAllister1, Omer Erbay2, Andrew Sarawit3, Mehdi Zarghamee4, Robert MacNeill5,

John Gross6, Steven Kirkpatrick7

1 Research Structural Engineer, PhD, PE, National Institute of Standards and Technology, Gaithersburg, MD, 20899-8611,

therese.mcallister@nist.gov, 301-975-6078, fax 301-869-6275

2 Senior Staff, PhD, Simpson Gumpertz & Heger, Inc., Waltham, MA, 02453

3 Senior Staff, PhD, Simpson Gumpertz & Heger, Inc., Waltham, MA, 02453

4 Senior Principal, PhD, Simpson Gumpertz & Heger, Inc., Waltham, MA, 02453

5 Principal Engineer, Applied Research Associates, Mountain View, CA 94043

6 Research Structural Engineer, PhD, PE, National Institute of Standards and Technology, Gaithersburg, MD, 20899-8611

7 Principal Engineer, PhD, Applied Research Associates, Mountain View, CA 94043

Abstract

This paper is the first of two papers that present the structural analysis approach used to model

the sequence of fire-induced damage and failures leading to the global collapse of WTC 7. This

paper presents the analysis methods developed and results obtained during the investigation

conducted by National Institute of Standards and Technology (NIST) to determine the cause of

collapse of World Trade Center 7 (WTC 7). The structural analysis required a two-phase

approach to address both the gradual response of the structure to fire before collapse initiation

(approximately 6 hrs) and the rapid response of the structure during the collapse process

(approximately 15 s). This paper describes the first phase, a pseudo-static (implicit) analysis that

simulated the response of structural elements to fires that spread and grew over several hours.

1
The analysis 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. The second paper describes the dynamic (explicit) analysis that used

the damage predicted in the first phase as initial conditions, and simulated the progression of

structural failures that resulted in the global collapse of the building.

Keywords: World Trade Center, WTC 7, fire-induced damage, structural analysis, failure, global

collapse

1.0 Introduction

World Trade Center 7 (WTC 7) was structurally damaged by falling debris during the collapse of

WTC 1 at 10:28:22 a.m. (eastern daylight time, EDT) on September 11, 2001. The damage

included severed exterior columns on the lower floors. The collapse of WTC 1 also resulted in

initiation of fires on at least 10 floors of WTC 7, extensive window breakage on the south face,

and loss of city water that supplied the automatic sprinkler system in the lower floors. After

nearly 6 hrs of burning, WTC 7 collapsed at 5:20:52 p.m. EDT.

The National Institute of Standards and Technology (NIST) conducted an investigation into the

collapse of the WTC buildings (NIST 2008, McAllister et al. 2008). A specific objective of the

WTC investigation was to determine why and how WTC 7 collapsed. A series of detailed

analyses were performed, consisting of : 1) a fire dynamics simulation to model the spread and

2
growth of the fires with time, 2) a thermal analysis to predict the temporal and spatial

distribution of temperature (temperature time-histories at every node) in the structure, and 3) a

two-phased structural analysis approach consisting of (a) a finite element analysis to simulate the

response of the structure to the fire-induced temperature histories that led to collapse initiation

(Part I), and (b) a dynamic finite element analysis to simulate the sequence of subsequent

structural failures that led to the collapse of the building (Part II).

In the following sections, the key aspects of the pseudo-static model development are presented.

Additionally, the response of the WTC 7 steel framed and composite floor structural system to

fire effects is discussed.

1.1 Modeling Approach

A single software application and solution method was not available to address both the gradual

response of the structure to fire before collapse initiation (up to 6 hrs) and the rapid response of

the structure during the collapse process (approximately 10 s). Therefore, a two-phased

structural analysis approach was adopted for the analysis of WTC 7. In the first phase, a

structural model was developed in ANSYS1 (2007) and used to determine the pseudo-static

structural response to spatially and temporally varying fire-induced temperatures, and to predict

the local structural failures that occurred up to collapse initiation. In the second phase, a

companion structural model was developed in LS-DYNA (2007) and was used to simulate

dynamic structural response in the seconds just prior to the total collapse of the building and the

1
Certain commercial software or materials are identified to describe a procedure or concept adequately. Such
identification is not intended to imply recommendation, endorsement, or implication by NIST that the software or
materials are necessarily the best available for the purpose.

3
sequential failures that occurred during the collapse. Figure 1 shows the sequence of analyses

conducted as part of the NIST investigation of the collapse of WTC7.

The pseudo-static model accounted for temperature-dependent material degradation and

component failure mechanisms. Failure criteria were developed to identify when a structural

component was no longer contributing to the structural system, and was impeding structural

convergence. When sufficient damage had occurred such that the structural system appeared to

be approaching instability, the fire-induced damage from the 16-story pseudo-static model was

input into the 47-story dynamic model as initial conditions for the phase two analyses.

1.2 Structural System

The 47-story building was 186 m (610 ft) high and constructed over a pre-existing electrical

substation owned by Consolidated Edison. Above Floor 7, the building had typical steel framing

for high-rise construction. The floor systems had steel beams acting compositely with normal

weight concrete slabs on a 76 mm (3 in.) metal deck, with a total floor thickness of 140 mm (5.5

in). Figure 2 shows the floor framing and column numbering system for typical tenant floors.

Shear connections were used at all interior floor framing connections (i.e., beams to girders and

girders to columns). The shear connections were constructed with a single shear plate (also

referred to as a fin connection), double angles, and seated connections. Single shear plate and

double angle connections connected interior beams to girders and girders to interior columns.

Seated connections connected the floor beams and girders to the exterior columns, the north side

of Column 79, and the south side of Column 81. Moment connections were used in the exterior

4
framing and portions of the core framing at Floors 5 and 7 as part of the lateral load resisting

system.

2.0 16-Story Pseudo-Static Model for Structural Response to Fire-Induced Temperatures

The 16-story pseudo-static finite element model used for calculating the nonlinear response of

the structural system to fire-induced temperatures, included the following features:

o Floors from ground level to Floor 16 (Figure 3), as sustained fires were observed on

Floors 7 to 9 and Floors 11 to 13 (McAllister et al. 2008).

o Representations of columns, beams, girders, composite floor slab, and connections.

o Geometric and material nonlinearities, including temperature-dependent material

properties of steel and concrete, thermal expansion, stiffness and strength degradation,

and creep of steel at high temperatures. Figure 4 shows temperature-dependent stress-

strain relationships used for ASTM A 572 Grade 50 steel and for normal weight concrete.

o Evolving temperature states, input as a temperature time history for each node at 30 min.

intervals, during the time period of analysis.

o Connection models that captured failure of connection components using “break

elements” (Zarghamee et al., 2005, McAllister et al., 2008, and Sarawit et al., 2010),

including bolt shear, plate tear-out, or beam walk-off from its seat.

o Failure criteria for component failure mechanisms, including connection component

failures, failure of shear studs within a composite floor system, buckling instability of

beams and girders, and crushing and cracking failures of concrete floor slabs.

5
Material and geometric nonlinearities were assigned to the structure between Floors 8 and 14,

while Floor 7 and below and Floors 15 and 16 were modeled linearly with sub-structuring

(super-elements in ANSYS) to reduce the size of the model. The structural loads from the

portion of the building above Floor 16 were applied to the Floor 16 columns to represent the

gravity loading on columns in the lower 16 stories of the building. The sub-structuring and

exclusion of the building portion above Floor 16 allowed the analysis to remain tractable while

including details of the framing and floor connections at the lower floors where fires were

observed.

Beams, girders, and columns were modeled with a 3-D linear finite strain beam element that is

well suited for large rotation and/or large strain nonlinear solutions, namely BEAM188 (ANSYS

2007). Typically, the columns were meshed with 0.6 m (2 ft) long elements, and beams were

meshed with 0.9 m (3 ft) long elements. The floor slab was modeled with a 4-node finite strain

shell element that is well suited for large rotation and/or large strain nonlinear solutions, namely

SHELL181 (ANSYS 2007). Typical mesh size for the floor slab was 0.9 m x 0.9 m (3 ft x 3 ft).

Temperature-dependent inelastic material properties were used for beam and shell elements.

Failure of floor framing connections and shear studs was modeled on Floors 8 to 14, as indicated

in Figure 5. A single-floor fire simulation performed prior to the analysis of the full structure

showed no connection damage west of Columns 73 though 76 contributing to the failures on the

east side of the structure, where the collapse was observed to initiate. Outside the indicated area,

structural damage, such as buckling of the steel frame and crushing and cracking of the concrete

slab, was modeled, but connection failures were not modeled. Connections failures were also

6
not modeled in the exterior moment frame, as no failures were observed there prior to the onset

of global collapse, or at column splices, as the purpose of the ANSYS model was to accumulate

local failures up to the point of buckling in a column.

Connection models, including models for shear studs, were constructed with a combination of

rigid beams, contact elements, control elements, spring elements, and user-defined “break

elements”, which modeled component failure (Zarghamee et al. 2005, McAllister et al. 2008, and

Sarawit et al. 2010). A control element is unidirectional and can turn on/off during an analysis to

connect or disconnect parts of the model during analysis. A break element is a multi-degree of

freedom elastic spring with the capability of disconnecting once its capacity is reached. Break

elements were developed initially for the collapse analysis of the WTC 1 and WTC 2 towers

(Zarghamee et al. 2005) to simulate complex modes of failure in connections using relatively

few degrees of freedom. An analytical model for a fin connection is shown in Figure 6

(McAllister et al. 2008).

The force and moment capacity of a break element were defined with temperature-dependent

properties by modifying the room temperature properties with a temperature-dependent reduction

factor. Different tensile and compressive capacities were assigned to connections where

appropriate. Connections with multiple failure modes required several break elements connected

in series and/or parallel as determined by the logical sequence of partial failures prior to the total

failure of the connection. The inclusion of contact elements in the connection models allowed

for slip and construction clearances (gaps) to be taken into account and thus insured different

responses to tensile and compressive loads.

7
Temperature data were obtained from Fire Dynamics Simulator (FDS) analyses of the fires and

the analyses of heat transfer from the heated gases to the structural components (McAllister et.al.

2008). Temperature data were input to the pseudo-static analyses at 30 min intervals for the fires

observed on Floors 7 to 9 and Floors 11 to 13 for a total of a 6.0 h time period. Linear variation

of temperatures was assumed for structural analysis between the discrete temperatures provided.

Three different thermal cases were used in the heat transfer analyses and pseudo-static analyses.

Case A used temperature data obtained from the FDS simulation of the observed fires. Cases B

and C increased and decreased, respectively, the Case A gas temperature by 10 percent. These

cases were within the range of realistic and reasonable fires in WTC 7 on September 11, 2001,

and were judged to be within the range of uncertainty for the observed fires (McAllister et.al.

2008). This range modified the severity of the fire in three ways. It increased the local heat

transfer to the structural members, it increased the area over which the structural members were

heated, and it increased the chance of hot zone (zone near the fire and the heated upper gas layer)

overlap on adjacent floors. A 10 percent increase or decrease in gas temperatures resulted in a

roughly 30 percent increase or decrease in the heat flux to structural members.

The model required the double precision version of ANSYS 11.0 for a 64-bit operating system.

The model had approximately 93,000 nodes and 101,000 elements. Inclusion of user-defined

elements prohibited parallel processing, so a 64-bit workstation with a quad-core, 3.0 GHz

processor, and 64 GB of random access memory (RAM) was used. The analysis time was

8
approximately 6 months for a single series of temperature states that simulated up to 4 h of

heating.

2.1 Load Application Sequence

The loads applied to the model include the dead load, 25 percent of the design live load on all

floors, and nodal temperature histories due to fire. Gravity loads were applied to the model in

stages that simulated the sequence of construction, where the floor slab participates in the floor

stiffness after the concrete has hardened.

Structural elements heat slowly relative to the rapidly fluctuating gas temperatures in a fire.

Component temperatures, which were determined by the fire dynamic and thermal analyses

(McAllister et al. 2008), were applied incrementally to represent the temperature state of the

structural system in 30 min time steps. The temperatures were assumed to vary linearly between

consecutive steps. The temperature time histories of several floor framing and slab components

on the east side of Floor 13 are shown in Figure 7.

2.2 Treatment of Failed Elements

Within the ANSYS model, failure criteria were developed for addressing failed components.

Members that were no longer contributing structurally to the response of the building were

numerically softened or removed. This approach significantly improved computational

efficiency and eliminated many convergence problems. When an element failed according to the

criteria presented below, either the element stiffness was reduced to a lower level or the element

was removed (the element remained in the ANSYS model but contributed a near-zero stiffness

9
value). When element stiffness was reduced, the mass was preserved, and when the member was

removed, the mass of the member was lost. Failed beams and girders were removed, and failed

slab sections had their stiffness reduced.

Shear Stud Failure. Shear stud failure in composite floor systems occurs when the concrete slab

crushes or cracks around the shear stud or the shear stud separates from the steel framing. The

shear strength of studs depends on rib geometry, slab thickness, concrete strength, steel strength,

stud location relative to the steel deck ribs, and loading direction. Using the results from two

sources, Rambo-Rodenberry (2002) and AISC (2005), the average of the strong axis and weak

axis strength of shear studs is 19.5 kip, which was used for shear studs in all load directions.

Rambo-Rodenberry also measured a relative movement between concrete and metal deck at

failure of about 0.2 in., which is indicative of the limited ductility of the shear stud connection.

Lateral-Torsional Buckling of Beams and Girders. When lateral support of the top (compression)

flange was lost due to failure of the shear studs, floor beams could laterally displace and buckle

in the lateral-torsional mode. The girders did not have shear stud connections to the concrete

slab. Therefore, girders with one-sided floor framing were also subject to lateral-torsional

buckling. Buckled members were removed when the web rotation was large enough that the

member would become unstable under its own gravity loading. In other words, it was assumed

that if a beam or girder web rotated so that there was a relative lateral displacement between the

top and bottom flange of more half of the flange width, the beam or girder would be unable to

support the gravity load and was removed.

10
Loss of Vertical Support for Beams and Girders. Under elevated temperatures, a beam could

lose vertical support at its ends through connection failure, including walking-off the beam

support seat (referred to here as walk-off failure). Walk-off failure is due to either 1) movement

along the axis of the beam due to sagging of beams or girders during the cooling phase of the

beam, or 2) lateral displacement of the beam resulting from thermal expansion of the orthogonal

beams. Gravity loads in a beam are transferred to the bearing seat from the bottom flange of the

beam near the web. Therefore, when the web was no longer supported by the bearing seat, the

beam was assumed to have lost support, as the flexural stiffness of the bottom flange was

assumed to be insufficient to transfer the gravity loads. Under such a condition, the beam was

removed. While axial walk-off was possible, the calculated connection failure mode was lateral

walk-off.

Cracking and Crushing of the Concrete Slab. Local temperature effects and failure of

supporting beams and girders caused slab elements to undergo large compressive and tensile

strains. The ANSYS material model of a concrete slab, reinforced with welded-wire-mesh, was

based on an isotropic plasticity formulation without strength degradation at high strains and,

hence, it could not represent cracking under tensile strains or crushing under compressive strains.

To address this issue, slab elements that reached the tensile or compressive failure strains were

removed. Any slab element with a principal tensile strain equal to or greater than 0.0015 at mid-

depth of the section was assumed to be fully cracked in tension. (Concrete starts to micro-crack

at a tensile strain of about 1.4 x 10-4; beyond this strain level, the concrete will soften, until at a

strain level of about 1.5 x 10-3 concrete no longer supports tension and steel reinforcement is

11
yielded.) Similarly, any slab element with a principal in-plane compressive strain equal to or

more than 0.004 was considered fully crushed in compression.

When these strains were reached, the slab was softened, i.e., the concrete elastic modulus was

numerically reduced. The reduced elastic modulus was equivalent to the in-plane stiffness that

would be provided by the welded-wire-mesh in the slab. Softening of a section of the slab, and

not removing it from the analysis, allowed the weight and the live loads that were applied to the

slab elements to be transferred to the remaining beams, girders, or spandrel beams.

3. Structural Response to Fire

Three different thermal response cases (Cases A, B, and C) were used as input to the pseudo-

static analysis. It became apparent as the pseudo-static analyses progressed that the connection,

beam, and girder failures occurred essentially at the same locations and with similar failure

mechanisms, but shifted in time between the three cases. Case B failures occurred at the earliest

time, followed by Case A, and then Case C. Therefore, only Case B results are described below.

As the temperature histories were applied, the following structural responses led to failures in a

number of structural components, which were softened or removed.

3.1 Thermal Effects on Shear Studs

Failure of shear stud connections occurred on the east side of the building where the floor slab

was continuous and the floor beams framed into girders on one side only (see Figure 2).

Generally, the steel framing heated more quickly than the concrete slab. When a concrete slab

section was heated, it was restrained by cooler adjacent slab sections. When the floor beams

12
were heated, their thermal expansion was first restrained by the shear studs and, after the shear

stud connections had failed, by weak axis flexure of the girder. Both factors led to differential

thermal expansion between the concrete slab and the floor framing.

The capacity of 28 shear studs on a floor beam in the northeast corner was estimated at about

546 kip, which is less than the force produced in a fully restrained floor beam with an average

temperature increase of 100 oC. Therefore, shear stud connection failures were expected to occur

early in the heating process. Shear stud connection failures also can occur when concrete

crushes due to compression forces in the slab, causing the shear studs to be ineffective.

Review of literature did not find much data that documented shear stud connection failure in

composite floors subjected to fire. One reason for the lack of such failures is that most

experimental tests of composite floor sections (a beam and slab section) have the same boundary

conditions applied to the beam and floor slab, which results in similar rates of thermal expansion

between the beam and slab.

3.2 Thermal Effects on Columns

None of the columns reached temperatures over 300 °C and, therefore, none of the columns

failed or buckled due to fire-induced thermal weakening (significant thermal weakening occurs

at temperatures greater than about 500 °C). The interior columns were not thermally restrained,

so they did not develop additional compressive loads due to thermal expansion. The exterior

columns had some restraint to thermal expansion, due to the moment frame construction of the

exterior framing. However, the exterior columns tended to have lower temperatures than the

13
interior columns, as they were only heated on one side and the heat dissipated to the outside.

Thermal expansion of either interior or exterior columns had little or no effect on the failure

mechanisms that occurred in the floor systems.

3.3 Thermal Effects on Floor Beams, Girders, and Concrete Slab

The calculated temperatures in the steel floor framing of Floor 13 at 3.0 h, 3.5 h, and 4.0 h are

shown in Figure 8 (a, b, c). These temperatures are largely due to heating from the fire on the

floor below. The corresponding calculated failures of floor beams and girders of Floor 13, based

on the described failure criteria, at 3.0 h, 3.5 h, and 4.0 h are shown in Figure 8 (d, e, f). (For

temperatures and damages to other floors, see McAllister et al. 2008). Although cooling

occurred for many structural members in other parts of the structure, critical components that

contributed to collapse initiation had not begun cooling between 3.0 h and 4.0 h of heating (see

Figure 7).

At temperatures less than approximately 400 °C (when averaged over the beam length),

restrained thermal expansion effects caused two types of failures in the floor beams and girders:

1) lateral-torsional buckling of beams and girders, and 2) failure of connections.

Localized elevated temperatures in the concrete slab resulted in compressive failure through

crushing, and loss of composite action with the floor beams due to failure of concrete at shear

stud connections. This failure mechanism usually occurred at locations where fires were burning

and slab temperatures were much higher in a localized area. Slab tensile failures occurred when

14
beams and girders failed at their connections and the slab experienced reverse curvature over

adjacent intact members.

3.4 Thermal Effects on Connections

Thermal expansion of beams and girders caused connection failures in the form of 1) bolt shear,

2) failure of welds, and 3) walk-off of the seated connections after bolts had sheared off. Shear

failure of bolts or failure of welds at the beam and girder web in header connections resulted in

loss of horizontal and vertical support to the beam and girders.

In seated connections, the shear failure of bolts at the bearing seat and at the top clip or plate

caused loss of horizontal support, but not vertical support. Loss of vertical support occurred

when the beam or girder walked off the bearing seat or the bearing seat weld failed. Girder

walk-off occurred when the beams that framed into girders from one side thermally expanded

and the resulting axial compression forces pushed the girder laterally from one side, sheared the

bolts at the seated connection, and then continued to push the girder until it walked off the

bearing seat. Other factors that contributed to this failure were an absence of shear studs on the

girders, which would have provided lateral restraint, and the one-sided framing of the east

girders by the floor beams, which allowed the floor beams to push laterally on the girder when

thermally expanding.

On Floors 10, 11, and 12, tensile failure of knife connections caused the failure of the

connections between Columns 76 and 79. The tension was caused by the expansion of the

girders between Columns 76 and 79 on Floor 13.

15
3.5 Failures Leading to Collapse Initiation

After 4.0 h of heating, the floor framing and slabs at Floors 8 to 14 were weakened by fires and

Columns 79, 80, and 81 had lost lateral support at several floors due to the failure of connections,

floor beams, and girders. The fire-induced failures of the floor framing resulted in increased

unsupported column lengths for Columns 79, 80, and 81. Figure 10 indicates failed floor beams

or girders in the pseudo-static analysis after 4.0 h, where failed members (by either buckling or

end connection failure) are shown as red and intact members as blue. Figure 99hows that at this

point in time, Column 79 was laterally unsupported at three floors in the east-west and north-

south directions, Columns 80 was laterally unsupported at one floor in the east-west and north-

south directions, and Column 81 was laterally unsupported at one floor in the north-south

direction.

The pseudo-static analysis performed by the pseudo-static model provided the response of the

structure to fire, but did not include the effect of the dynamics of failure and transfer of

momentum and kinetic energy of the failing components to the remaining parts of the structure.

To calculate the dynamic response of the structure and determine whether the failures resulted in

total collapse, a separate analysis using dynamic model was performed and the sequence of

failures were determined starting the analysis with the temperatures and the resulting cumulative

damages existing at different points in time.

16
4.0 Summary

To address both the gradual response of the structure to fire before collapse initiation and the

rapid response of the structure during the collapse process, a two-phased structural analysis

approach was adopted for the analysis of WTC 7. In the first phase (this paper), a 16-story

pseudo-static finite element model was used to simulate the pseudo-static response of the

structure to the fire-induced temperatures time histories and the resulting structural failures. In

the second phase (Part II), a 47-story dynamic finite element model was used to simulate the

sequence of subsequent structural failures and dynamics of the collapse process.

The 16-story pseudo-static model was developed to determine the structural response of WTC 7

to elevated temperatures resulting from fires on Floors 7 to 9 and Floors 11 to 13 that spread and

grew. The finite-element model represented columns, beams, girders, composite floor slab, and

connections. The model employed geometric and material nonlinearities, including temperature-

dependent material properties of steel and concrete, thermal expansion, stiffness and strength

degradation, and creep of steel at high temperatures. Temperature time histories for each node

were input at 30 min. intervals. Detailed connection models captured failure of connection

components, including bolt shear, plate tear-out, or beam walk-off from its seat. Failure criteria

to identify component failure mechanisms, including connection component failures, failure of

shear studs within a composite floor system, buckling instability of beams and girders, and

crushing and cracking failures of concrete floor slabs were developed and used to remove failed

components.

17
Three different thermal response cases (Cases A, B, and C) were used as input to the pseudo-

static analysis, where the temperature data from simulations of observed fires (Case A) were

increased (Case B) and decreased (Case C) by 10 percent. This range modified the severity of the

fire in three ways: the local heat transfer to the structural members increased, the area over which

the structural members were heated increased, and the chance of hot zone overlap on adjacent

floors increased. The pseudo-static analyses had similar failure mechanisms and patterns, but

were shifted in time between the three cases. Case B failures occurred at the earliest time, and

were used in the dynamic analyses.

The results of the pseudo-static analysis simulated a sequence of structural failure of beams and

girders by lateral torsional buckling and connection failures that led to collapse initiation in the

dynamic analysis. Collapse initiation was a result of fire-induced failures of connections, beams,

and girders in the lower floors that resulted in loss of lateral support for Column 79 over several

floors. The damage from falling debris and the failures predicted from the pseudo-static analysis

of the structural response to temperature histories over several hours were input into the dynamic

model as initial conditions to simulate the progression of structural failures and global collapse

of the building, as is discussed in Part II of this paper.

5.0 Recommendations for Future Research

The computational and time demands of the analyses pointed out that the following capabilities

would improve future finite element analyses of structural response to fire:

18
 - Beam elements that can be use for heat transfer and structural analysis. At present, solid

elements are used for heat transfer analyses and temperature data is manually applied to

beam element nodes.

- A computationally efficient concrete material model for shell elements that includes

temperature dependent properties and can simulate cracking and crushing.

- Parallel processing capabilities for elements that model failure (i.e., the user defined

connections with „break‟ elements).

6.0 References

ANSYS. (2007). ANSYS Mechanical Release 11.0, ANSYS Inc., Southpointe, 275 Technology

Drive, Canonsburg, PA, 15317.

Frankel (1985), Frankel Steel Limited, Fabrication Shop Drawings, 7 World Trade Center.

LS-DYNA (2007). “Keyword User‟s Manual”, Livermore Software Technology Corporation,

Version 971, May.

Kirkpatrick, S. W., R. T. Bocchieri, F. Sadek, R. A. MacNeill, S. Holmes, B. D. Peterson, R. W.

Cilke, C. Navarro. (2005). Federal Building and Fire Safety Investigation of the World Trade

Center Disaster: Analysis of Aircraft Impacts into the World Trade Center Towers, NIST

NCSTAR 1-2B. National Institute of Standards and Technology. Gaithersburg, MD, September.

19
McAllister, T. P., R. G. Gann, J. D. Averill, J. L. Gross, W. L. Grosshandler, J. R. Lawson, K. B.

McGrattan, H. E. Nelson, W. M. Pitts, K. R. Prasad, F. H. Sadek. (2008). Federal Building and

Fire Safety Investigation of the World Trade Center Disaster: Structural Fire Response and

Probable Collapse Sequence of World Trade Center Building 7. NIST NCSTAR 1-9. National

Institute of Standards and Technology. Gaithersburg, MD, November.

MacNeill, R., S. Kirkpatrick, B. Peterson, and R. Bocchieri, (2008). Federal Building and Fire

Safety Investigation of the World Trade Center Disaster: Global Structural Analysis of the

Response of World Trade Center Building 7 to Fires and Debris Impact Damage. NIST

NCSTAR 1-9A. National Institute of Standards and Technology, Gaithersburg, MD, November.

NIST NCSTAR 1A. (2008). Federal Building and Fire Safety Investigation of the World Trade

Center Disaster: Final Report on the Collapse of the World Trade Center Building 7, NIST,

Gaithersburg, MD, Nov.

Rambo-Roddenberry, M.D., (2002). “Behavior and Strength of Welded Stud Shear Connectors”.

Ph.D. thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA, April 8.

Sarawit, A.T., O.O. Erbay, Y. Kitane, M.S. Zarghamee, T.P. McAllister, J.L. Gross, (2010).

“Modeling Structural Response to Fire with Break Elements”. Submitted to the ASCE Journal of

Structural Engineering for publication.

Zarghamee, M. S., S. Bolourchi, D. W. Eggers, F. W. Kan, Y. Kitane, A. A. Liepins, M.

Mudlock, W. I. Naguib, R. P. Ojdrovic, A. T. Sarawit, P. R Barrett, J. L. Gross, and T. P.

20
McAllister. (2005). Federal Building and Fire Safety Investigation of the World Trade Center

Disaster: Component, Connection, and Subsystem Structural Analysis. NIST NCSTAR 1-6C.

National Institute of Standards and Technology. Gaithersburg, MD, September.

21
Figure 1. WTC 7 analysis sequence (McAllister et. al. 2008).

22
 N

Figure 2. Floors 8 through 45 framing plan with column numbering.

23
 1
ELEMENTS
MAY 29 2008
14:59:09

FL16

FL14 Upper
FL13 Superelement
FL12
FL11
FL10
FL9
FL8
FL7

Z
Lower
Y
X Superelement
FL1

N
WTC 7 16-Story Model

Figure 3. 16-story pseudo-static model representing lower 16 floors of the WTC 7 building.
The nonlinear part of the model subjected to thermal loading is shown in blue
(McAllister et. al. 2008).

24
 5
100

600 30

80 4

500 25

Stress (MPa)
Stress (ksi)

60 400 20

Stress (ksi)

Stress (MPa)
300 15
40 2

200 10
20 1
100 5

0 0 0 0
0 0.05 0.1 0.15 0.2 0 0.0005 0.001 0.0015 0.002 0.0025 0.003
Strain (in./in.) (mm/mm)
Strain (in./in.) (mm/mm)
T = 20 C T = 400 C T = 450 C
T = 20 C T = 100 C T = 200 C
T = 500 C T = 550 C T = 600 C
T = 650 C T = 700 C T = 750 C T = 300 C T = 400 C T = 500 C
T = 800 C T = 850 C T = 600 C T = 700 C T = 800 C

a) ASTM A572 Steel b) 4000 psi Normal Concrete

Figure 4. Multilinear isotropic hardening (MISO) model temperature-dependent stress-
strain relationships for ASTM A572 grade 50 steel and 4000 psi concrete slab.

25
 Col. 76

N
Col. 78

Note: Black colored dots show the location of shear studs and
connections modeled with break elements.

Figure 5. Area of the floor where connections modeled in detail using break elements
(McAllister et. al. 2008).

26
 Coped beam at beam-
to-girder connections

Local z
Break element to model Beam element to model bolt
tear-out, (USER104) stiffness (BEAM4)

Break element to model bolt

shear, (USER103, K1=0) Beam element to model shear
plate or clip angles (BEAM188)
A A
Break element to model
block/weld shear
(USER102)

Local x
Beam element Elevation View of Fin
to model girder
(BEAM188) Connections
Beam element
to model floor beam
(BEAM188) Rigid beams Fillet weld
(BEAM4)

A325 7/8 in. bolt

Shear Plate

Beam-to-girder “fin” connection Section A-A

Figure 6. Analytical model for a fin connection (based on fabrication shop drawings,
Frankel 1985) (McAllister et. al. 2008).

27
 1000
Column (Average)
Floor Girder (Average)
800 Slab Over Floor Girder (Top Surface)
Slab over Floor Girder (Bottom Surface)
Temperature (F)

Girder-to-Column Connection
600

400

200

0
0 1 2 3 4 5 6
Time (h)

a) Floor framing and concrete slab north of Column 79 including floor

girder, girder-to-column connection, slab over floor girder and Column 79.

1000
Column (Average)
Floor Beam (Average)
800 Slab Over Floor Beam (Top Surface)
Slab over Floor Beam (Bottom Surface)
Temperature (F)

Beam-to-Girder Connection
600

400

200

0
0 1 2 3 4 5 6
Time (h)

b) Floor framing and concrete slab south of Column 81 including floor

beam, beam-to-girder connection, slab over floor beam and Column 81.

Figure 7. Temperature time-history of selected floor framing and slab at Floor 13.

28
Figure 8. Floor 13 framing temperatures at (a) 3.0 h, (b) 3.5 h, and (c) 4.0 h. (Temperature
scale range: 0 ºC to 675 ºC; temperatures > 675 ºC are grey.) ANSYS floor 13 beam and
girder failures (by buckling or end connection damage) at (d) 3.0 h, (e) 3.5 h, and (f) 4.0 h
(McAllister et. al. 2008).

29
Figure 9. Pseudo-static analysis beam and girder failures around Columns 79, 80, and 81 at
4.0 hr (McAllister et. al. 2008).

30
List of Figures

Figure 1. WTC 7 analysis sequence.

Figure 2. Floors 8 through 45 framing plan with column numbering.

Figure 3. ANSYS 16-story model representing lower 16 floors of the WTC 7 building. The
nonlinear part of the model subjected to thermal loading is shown in blue (McAllister et. al.
2008).

Figure 4. ANSYS multilinear isotropic hardening (MISO) model temperature-dependent stress-

strain relationships for ASTM A572 grade 50 steel and 4000 psi concrete slab.

Figure 5. Area of the floor where connections modeled in detail using break elements
(McAllister et. al. 2008).

Figure 6. ANSYS analytical model for a fin connection (based on fabrication shop drawings,
Frankel 1985) (McAllister et. al. 2008).

Figure 7. Axial force in girder between Column 76 and 79 at different floor elevations.

Figure 8. ANSYS floor 13 framing temperatures at (a) 3.0 h, (b) 3.5 h, and (c) 4.0 h.
(Temperature scale range: 0 ºC to 675 ºC; temperatures > 675 ºC are grey.) ANSYS floor 13
beam and girder failures (by buckling or end connection damage) at (d) 3.0 h, (e) 3.5 h, and (f)
4.0 h (McAllister et. al. 2008).

Figure 9. ANSYS beam and girder failures around Columns 79, 80, and 81 at 4.0 hr (McAllister
et. al. 2008).

31