Professional Documents
Culture Documents
Abstract: This paper presents a computational investigation of the robustness of a typical concrete deck–steel beam composite floor
system with simple shear connections in the event that a center column has been removed. The study provides insight into the behavior
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 06/18/14. Copyright ASCE. For personal use only; all rights reserved.
and failure modes of simple shear connections and composite floor systems comprised of such connections. Analyses of a connection
subassemblage indicate that loads are primarily resisted by cable action of the beams after column loss resulting in increasing tensile
forces in the beams and connections that could eventually precipitate failure. Simulation results show that the floor deck contributes
significantly to the floor system response through: diaphragm action to prevent the exterior columns from being pulled inward and
membrane action primarily through the reinforcement mesh and metal deck. The analyses indicate that the capacity of the analyzed floor
system under the column removal scenario is significantly less than the load specified by the General Services Administration’s current
progressive collapse guidelines. This suggests that applying these guidelines, the composite floor system studied would be vulnerable to
collapse if a center column is lost.
DOI: 10.1061/共ASCE兲0733-9445共2008兲134:11共1717兲
CE Database subject headings: Floors; Finite element method; Collapse; Composite materials; Connections.
Introduction of such systems to collapse depends upon the strength of the shear
beam-to-column connections. Astaneh-Asl et al. 共2001兲 con-
Steel moment resisting frame systems are commonly used in low- ducted an experimental study to explore the ability of a typical
to midrise construction in seismically active regions in the United composite floor system to resist progressive collapse. The shear
States. The lateral load resistance of such systems is often con- connections were a combination of single shear plate 共shear tab兲
centrated in a few “moment” frames, whereas the remainder of beam-to-beam and single-angle-web-with-seat angle beam-to-
the gravity system derives its stability by “leaning” on the perim- column connections. The tests indicated that the system could
eter frames. The gravity system is typically comprised of beams absorb removal of a middle perimeter column and successfully
connected to columns through simple shear connections in con- resist, without collapse, the dead and live loads and associated
junction with a concrete deck that acts compositely with the steel dynamic effects through catenary action in the floor system com-
beams. The motivation behind such systems stems from the need ponents. Using an analytical approach, Foley et al. 共2006兲 showed
to reduce construction costs by limiting expensive-to-construct that in the event of a center column loss, a composite floor system
moment resisting connections between beams and columns and with bolted clip angle gravity connections could be expected to
by locating them in a small number of moment resisting bays. carry the dead and service live loads, but without the full dynamic
Shear connections in the gravity system are generally modeled as load effects.
pinned and their moment resistance is typically ignored in routine The study reported in this paper utilizes computational struc-
design calculations. tural simulation to evaluate the robustness of concrete deck-steel
Few researchers have investigated the collapse resistance, or beam composite floor systems under the assumption that a center
structural robustness, of composite floor systems with shear con- column has been removed. The beam-to-column shear connection
nections in the event of a loss of a center column. The resistance considered is a single plate shear connection, which is a configu-
ration commonly used in the gravity portions of seismically and
1 nonseismically designed buildings. Two key areas are investi-
Research Structural Engineer, Building and Fire Research Labora-
gated: 共1兲 the behavior and failure modes of shear connections
tory, National Institute of Standards and Technology, Gaithersburg, MD
20899-8611. E-mail: fahim.sadek@nist.gov and 共2兲 the behavior and failure modes of the composite floor
2
Professor, Dept. of Civil and Environmental Engineering, Univ. of system; in particular, the relative role of the steel beams and their
Michigan, Ann Arbor, MI 48109. E-mail: eltawil@umich.edu connections, the metal deck, and the concrete slab including the
3
Senior Research Structural Engineer, Building and Fire Research wire mesh reinforcement in resisting collapse.
Laboratory, National Institute of Standards and Technology, Gaithers- The investigation is carried out using two different simulation
burg, MD 20899-8611. E-mail: hsl@nist.gov models. In the first, a beam–column subassembly is modeled
Note. Associate Editor: James S. Davidson. Discussion open until without the contribution of the floor deck under the notional re-
April 1, 2009. Separate discussions must be submitted for individual
moval of a center column. This analysis provides insight into the
papers. The manuscript for this paper was submitted for review and pos-
sible publication on August 16, 2007; approved on March 21, 2008. This behavior and failure of the shear connection and forms the basis
paper is part of the Journal of Structural Engineering, Vol. 134, No. 11, for a reduced model that is subsequently used in the analysis of
November 1, 2008. ©ASCE, ISSN 0733-9445/2008/11-1717–1725/ the floor system. In the second model, the floor system, including
$25.00. steel beams and their connections, columns, floor slab, and metal
2
W16x26
6.1 m
W14x22
Typical
4
W16x26
Typical
6
8mm
9.14 m 9.14 m 9.14 m 9.14 m 9.14 m
38.1mm 38.1mm
Fig. 1. Plan layout of ten-story building designed for Seismic Design
Category C 共Liang et al. 2008兲. Highlighted area indicates floor bays Fig. 2. Single-plate shear connection used in the gravity frames
considered in this paper.
floor portion that is considered for the analysis reported herein.
The floor system consisted of 82.5 mm lightweight concrete
deck is analyzed to evaluate its response under the loss of a center 共17.3 kN/ m3兲 topping on 76.2 mm metal deck. The slab acts
column. compositely with the steel beams through shear studs. The steel
beams are connected to the columns using single plate simple
shear connections 共shear tab connections兲, which were designed
Description of Prototype Floor System according to the AISC LRFD Specifications. The shear tab is fillet
welded to the column and bolted to the beam web as shown in
The National Institute of Standards and Technology designed pro- Fig. 2.
totype steel framed buildings for the purpose of studying their The lightweight concrete topping slab has a nominal com-
response to an event that may cause progressive collapse 共Liang pressive strength of 20.7 MPa. ASTM A992 structural steel
et al. 2008兲. The buildings are ten-story office buildings with plan 共Fy = 344.8 MPa兲 was used in the beams and columns. For the
dimensions of 30.5 m ⫻ 45.7 m and utilize moment-resisting shear connections, 22.2 mm diameter ASTM A325 high strength
frames as the lateral load resisting system. Two seismic design bolts were used, whereas the 9.5 mm thick shear tabs were ASTM
categories were considered to study the effectiveness of seismic A36 steel 共Fy = 248.2 MPa兲. For the purpose of the analyses re-
design and detailing, including seismic connections, in resisting ported in this paper, the thickness of the light-gauge metal deck
progressive collapse. As a result, the buildings were designed for was assumed to be 20 gauge 共0.9 mm兲 and the concrete slab was
共1兲 Seismic Design Category C 共SDC C兲 共Atlanta兲, which resulted reinforced using a wire mesh 共welded wire fabric兲 152 mm by
in a design using intermediate moment frames as defined in the 152 mm, W1.4⫻ 1.4 steel.
American Institute of Steel Construction 共AISC兲 Seismic Provi-
sions 共2002兲, and 共2兲 Seismic Design Category D 共SDC D兲 共Se-
attle兲, which resulted in a design using special moment frames. Modeling of Shear Connections
The design loads on the buildings were determined based on
the International Building Code 共IBC兲 共ICC 2003兲. The material The performance of shear tab connections under seismic 共cyclic兲
design standards used in the design of members and their connec- loading was reported in Liu and Astaneh-Asl 共2004兲. In their
tions were those referenced in American Society of Civil Engi- study, the connection experienced primarily rotational demands.
neers 共ASCE兲 7-02 共2002兲 including the AISC Load and This study, however, focuses on the performance of the connec-
Resistance Factor Design 共LRFD兲 Specifications for Structural tion under the notional removal of a center column. As a result,
Steel Buildings 共AISC 1999兲 and the AISC Seismic Provisions the connection is subject to not only large rotations, but to axial
for Structural Steel Buildings 共2002兲. For typical floors, the dead forces as well due to the large deflections that the beam experi-
load consisted of the self-weight of the floor of 2.2 kN/ m2 and a ences as a result of column removal. As no experimental data are
super-imposed dead load of 1.44 kN/ m2; whereas the design live available to quantify the response of these connections under the
load was assumed to be 4.79 kN/ m2. For the roof, the self-weight column removal scenario, computational models with various lev-
of the slab was 2.2 kN/ m2, the superimposed dead load was els of complexity are used to understand the response character-
0.48 kN/ m2; and the design live load was 0.96 kN/ m2. The re- istics of the connections. The simulations presented in this study
duction in live loads was based on Section 1607.9.1 of IBC 共ICC are conducted using LS-DYNA; an explicit, general purpose,
2003兲. finite-element software package 共Hallquist 2006兲. In all analyses,
The lateral load resisting system, comprised of moment loads are applied at a slow rate to ensure a static response 共no
frames, was located only around the perimeter of the buildings. dynamic amplification兲.
Gravity frames 共not part of the lateral-force-resisting system兲
were used for the interior of the buildings. As a result, the design
High Fidelity Connection Model
of all interior floors and columns was the same for the SDC C and
SDC D buildings. A plan view of the Building in the SDC C zone The beam selected for analysis is the shorter-span beam 共6.1 m
is shown in Fig. 1. The highlighted area in Fig. 1 indicates the span兲 oriented in the north-south direction 共Fig. 1兲. The beam is a
Fig. 4. 共Color兲 共a兲 Applied vertical load versus center column vertical displacement for detailed and reduced model analyses; 共b兲 horizontal
reaction versus center column vertical displacement for detailed and reduced model analyses
Rigid Panel
Beam
dbg si Spring Element
Beam Web
Nonlinear springs
Column
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 06/18/14. Copyright ASCE. For personal use only; all rights reserved.
(a) Reduced Component Connection (RCC) Model (b) Reduced Coarse Shell Connection
(RCSC) Model
occur, and the contribution from shear deformation of the shear tional elastic rotational capacity of these connections is estimated
tab is likely small, the beam is connected to the column using a as 0.02 rad. This gives a total estimated rotational capacity for
rigid constraint equation in the vertical direction. such connections as
The force–deformation response for the nonlinear springs used
in the connection model is shown in Fig. 6共a兲. The rotational t,max = 0.17 − 0.00014dbg 共3兲
stiffness, ks, of the connection without any contribution from the
Therefore, the deformation limit for a bolt spring is
floor slab is given in FEMA 355D 共2000兲 as
where si = distance of each bolt from the center of the bolt group
关Fig. 5共a兲兴. Reduced Coarse Shell Connection Model
The yield and ultimate forces of each spring, Fy,i and Fu,i, Another reduced model using a coarse shell element mesh is used
respectively, are calculated based on the governing failure mode to represent the beam, the column, and their connection 关Fig.
or limit state of the connection under horizontal 共axial兲 loading. 5共b兲兴. The element size is about 38 mm. The connection is repre-
These can be estimated based on the following considerations sented using one row of shell elements with a thickness equal to
with a resistance factor = 1: that of the beam web. The stress-strain characteristics of these
• The bolt yield and ultimate capacity in shear with threads shell elements are derived based on the bolt spring parameters in
excluded; Eqs. 共1兲–共4兲. The model is supplemented with a spring element as
• The ultimate strength of transversely loaded fillet welds along shown in Fig. 5共b兲 to ensure accurate shear rigidity and strength.
the side of the shear plate based on the AISC Specification The spring has an elastic-perfectly plastic behavior with an ulti-
共2005a,b, Sec. J2.4兲. In this case, Fy,i = Fu,i; mate load calculated as the difference between the actual shear
• The block shear yield and ultimate capacity of the web or capacity of the connection and the shear capacity based on ana-
shear tab based on the AISC Specification 共2005a,b, Section
J4.3兲; and
• The tear-out 共bearing at bolt holes兲 yield and ultimate capacity
Force Force
through the beam web or through the shear tab based on the COMPRESSION COMPRESSION
Load (kN)
Comparison between Model Results and Discussion 60 Model
Results of the three models are shown in Fig. 4. The plots indicate 40
good agreement between the detailed and reduced model results.
The analyses confirm that the loads under a column removal sce- 20
nario are primarily resisted by cable action, which results in al-
0
most uniform axial tensile forces in the beam. These tensile forces
0 10 20 30 40 50
increase until the connection can no longer sustain this axial load,
after which the connection could fail in one of the failure modes Displacement (mm)
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 06/18/14. Copyright ASCE. For personal use only; all rights reserved.
discussed previously. The beam 共between the bolts and the pin
Fig. 7. Comparison between model and test data for tear-out failure
support at the other end兲 remains in the elastic range.
The rotation at peak load is 0.088 rad, substantially less that
the rotation capacity of this connection based on seismic testing, The force–deformation characteristics of the spring are shown in
which is 0.14 rad. The reason for this discrepancy in rotations is Fig. 6共b兲. The spring stiffness, kC, and strength f c are estimated
due to differences in the mechanisms by which the applied loads according to FEMA-355D 共2000兲
are resisted. Under seismic loading, the beam rotates about a cen-
ter of rotation, moving some bolts toward the column and others Ecbfcts
kC = 共5兲
away from the column. Ductility demands in this case are mainly lt
dependent on the distance of each bolt from the center of rotation.
Under the column removal scenario, the cable force that develops f C = 0.85f ⬘c bfcts 共6兲
in the beam is transmitted through the bolts, forcing all bolts to
move away from the column. Ductility demands in this case are where Ec = Young’s modulus for concrete; bfc = column flange
dictated by the geometry of the subassemblage, i.e., by the need width; ts = concrete slab thickness; lt = spring length; and f ⬘c
for the beam and its connections to stretch to accommodate = specified concrete strength.
the increased chord length associated with the new deformed The extra moment capacity provided by composite action of
configuration. the slab is lost when the plastic rotation 共in radians兲 reaches a
The General Services Administration’s 共2003兲 progressive col- critical value 共FEMA-355D 2000兲
lapse guidelines recommend a rotation of 0.015– 0.025 rad as an c,max = 0.029 − 0.0000079dbg 共7兲
acceptance criterion for partially restrained connections. The De-
partment of Defense Guidelines 共UFC 2005兲 use a rotation range Therefore, the deformation limit for the spring is
of 0.9–2° 共⬍0.035 rad兲 based on the governing limit state and the
⌬C,max = dtC,max 共8兲
level of protection as acceptance criteria. These acceptance crite-
ria appear to be quite conservative when compared to the rota- where dt = distance between the center of the bolt group and the
tional capacity calculated for the beam configuration and span center line of the slab. At this displacement, the compression
considered in this study. capacity of the spring is assumed to drop to zero.
As shown in Fig. 8, individual bolt failures are manifested as
sudden drops in the moment capacity of the connection, with the
Validation connection losing all resistance when all three bolts fail. Fig. 8
Two levels of validation exercises are conducted to provide con- also shows that the effect of binding on negative bending strength
fidence in the proposed models. In the first, the proposed tear-out is significant. The model underestimates negative moment capac-
model results are compared to test results for Specimen 41 in Rex
and Easterling 共2003兲, which also failed by tear out. The compari-
1.5
Moment / Maximum Positive Moment
count for the possibility that the concrete slab might bear against
the column. Framing and Metal Deck
This model includes the columns, beams, and their connections
along with the metal deck, which is connected to the top flange of
Discussion of Simulation Results
the beams via puddle welds as previously explained. The concrete
Pushdown analysis is conducted using the developed model, slab and its reinforcement are not included in this model. The
where the unsupported center column is statically pushed down applied load versus the vertical displacement of the center 共re-
under displacement control until the system collapses. The devel- moved兲 column is shown in Fig. 11. The model fails when the
oped model is exercised as follows, by successively including the shell elements representing the shear connections reach the failure
contribution of various system components to gain insight into strain and are eroded. The metal deck experiences large displace-
how these components influence system response: ments and some localized yielding, but remains intact and at-
tached to the floor beams. At peak load, the perimeter columns
Framing Only facing the removed column are pulled slightly inward, but the
This simulation includes only the columns, beams, and their con- displacements are an order of magnitude smaller than those ob-
nections. The metal deck, concrete slab, slab reinforcement, and tained from the analysis with the framing only without con-
shear studs are removed from the model. The applied load versus straints. Fig. 11 indicates that the vertical loading capacity of the
the vertical displacement of the center 共removed兲 column is floor framing with the metal deck is larger than that without the
shown in Fig. 11. The system fails when the shell elements rep- metal deck. This shows that the metal deck is capable of provid-
resenting the shear connections reach the failure strain and are ing significant membrane action 共mainly one-way membrane ac-
sequentially eroded. The analysis indicates that due to the large tion in the direction of the flutes兲, and thus contributing to the
axial tensile forces generated in the affected beams, the perimeter floor system strength.
columns facing the removed column are pulled inward 共Fig. 12兲.
In particular, the column to the north of the removed column with Detailed Floor Model (Framing, Metal Deck, and Concrete
the minor axis facing the sagging beam experienced excessive Slab)
deformations and yielding. Although the stability of this column This model includes all the elements of the floor system. The
might be questionable in this scenario due to the large displace- applied load versus the vertical displacement of the center 共re-
ments of the columns and the subsequent second order P – ⌬ ef- moved兲 column is also shown in Fig. 11, which indicates that the
fects, this behavior helps reduce the demand on the shear capacity of the floor framing with the concrete slab is larger than
connections, and as a result the failure of the floor model occurs that without the slab. This demonstrates that the concrete deck is
at a larger vertical displacement when compared to the failure capable of providing significant membrane action, and thus con-
displacement obtained from the analysis of the beam–column tributes to the floor system strength. The simulations show that
subassembly 关compare, e.g., Fig. 11 to Fig. 4共a兲兴. the concrete slab is, however, severely damaged 共cracked兲 due to
The analysis is repeated after adding a constraint to prevent the the large tensile strains imposed on it as a result of the membrane
horizontal displacement of the columns at the floor level. This action. Fig. 13 shows the damage contours in the concrete slab at
constraint is intended to represents the large in-plane stiffness of a vertical displacement of about 635 mm, where damage indices
the floor slab, which is not included in this simulation. The results of 0 and 2 signify no cracking and severe cracking, respectively.
of this analysis are also shown in Fig. 11. Similar to the previous Two observations are evident from Fig. 13: 共1兲 the damage
analysis, the model fails when the connections fail. The vertical 共cracking兲 contours are arranged in a diamond pattern around the
displacement at failure, however, is more consistent with the dis- central column and 共2兲 the beams remain straight and the slab is
placement at which the shear connection fails when a single kinked at the beam locations, i.e., yield lines form there.
beam–column subassembly is analyzed 关compare Fig. 11 to Fig. It is evident from the analysis results that, although membrane
4共a兲兴. action is generated by the reinforcement in the slab, the damaged
cluding dynamic effects兲 specified by the GSA guidelines of structural steel framing systems.” Rep. No. MU-CEEN-SE-06-01,
9.67 kN/ m2. Thus, applying the GSA design criteria, this particu- Marquette Univ., Milwaukee, Wis.
lar floor system would be vulnerable to collapse under the speci- General Services Administration 共GSA兲. 共2003兲. “Progressive collapse
fied column removal. Further research must be conducted in order analysis and design guidelines for new federal office buildings and
to generalize this conclusion. Research is also needed to develop major modernization projects.” GSA, Washington, D.C.
design oriented models for evaluating the robustness of composite Hallquist, J. 共2006兲. LS-DYNA keyword user’s manual, Version 971, Liv-
floor systems. ermore Software Technology Corporation, Livermore, Calif.
Hawkins, N. M., and Mitchell, D. 共1979兲. “Progressive collapse of flat
plate structures.” ACI J., 76共7兲, 775–808.
Acknowledgments International Code Council 共ICC兲. 共2003兲. International building code,
Falls Church, Va.
This research was supported by the National Institute of Stan- Liang, X., Shen, Q., and Gosh, S. K. 共2008兲. “Assessing ability of seismic
dards and Technology 共NIST兲. The writers acknowledge the in- structural systems to withstand progressive collapse: Design of steel
valuable comments and through review by John L. Gross of NIST frame buildings.” Rep., submitted to the Building and Fire Research
and Kurt Gustafson of AISC. The second writer acknowledges the Laboratory, National Institute of Standards and Technology, Gaithers-
discussions and helpful insight of Professor Kapil Khandelwal at burg, Md.
Notre Dame University, South Bend, India. Certain commercial Liu, J., and Astaneh-Asl, A. 共2004兲. “Moment rotation parameters for
products, software, or materials are identified in this paper to composite shear tab connections.” J. Struct. Eng., 130共9兲, pp. 1371–
describe a procedure or concept adequately. Such identification is 1380.
Rambo-Roddenberry, M. D. 共2002兲. “Behavior and strength of welded
not intended to imply recommendation, endorsement, or implica-
stud shear connectors.” Ph.D. dissertation, Dept. of Civil Engineering,
tion by NIST that the products, software, or materials are neces-
Virginia Polytechnic Institute and State Univ., Blacksburg, Va.
sarily the best available for the purpose. Rex, C. O., and Easterling, W. S. 共2003兲. “Behavior and modeling of a
bolt bearing on a single plate.” J. Struct. Eng., 129共6兲, 792–800.
Sadek, F. 共2005兲. “Federal building and fire safety investigation of the
References World Trade Center disaster: Baseline structural performance and air-
craft impact damage analysis of the World Trade Center towers.”
AISC. 共1999兲. Load and resistance factor design specifications for struc- NIST NCSTAR 1-2, National Institute of Standards and Technology.
tural steel buildings, Chicago. Gaithersburg, Md.
AISC. 共2002兲. Seismic provisions for structural steel buildings, Chicago. Unified Facilities Criteria 共UFC兲. 共2005兲. “Design of buildings to resist
AISC. 共2005a兲. Seismic provisions for structural steel buildings, Chicago. progressive collapse.” UFC 4-023-03, Dept. of Defense, Washington,
AISC. 共2005b兲. Specifications for structural steel buildings, Chicago. D.C.