Engineering Structures 56 (2013) 664672

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Experimental and analytical progressive collapse assessment of a steel

frame building
Brian I. Song a, Halil Sezen b,
a
b

URS Corporation, Warrenville, IL, USA

Department of Civil and Environmental Engineering and Geodetic Science, The Ohio State University, Columbus, OH, USA

a r t i c l e

i n f o

Article history:
Received 19 August 2011
Revised 19 June 2012
Accepted 31 May 2013
Available online 2 July 2013
Keywords:
Progressive collapse
Steel buildings
Column failure
Load redistribution
Collapse experiment

a b s t r a c t
A eld experiment and numerical simulations were performed to investigate the progressive collapse
potential of an existing steel frame building. Four rst-story columns were physically removed from
the building to understand the subsequent load redistribution within the building. Experimental data
from the eld tests were used to compare and verify the computational models and simulations. Due
to the scarcity of data from full-scale tests, the experimental data produced during this research is a valuable addition to the state of knowledge on progressive collapse of buildings. The progressive collapse
design guidelines typically recommend simplied analysis procedures involving instantaneous removal
of specied critical columns in a building. This paper investigates the effectiveness of such commonly
used progressive collapse evaluation and design methodologies through numerical simulation and experimental data.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Progressive collapse is generally dened as small or local structural failure resulting in damage and failure of the adjoining members and, in turn, causing total collapse of the building or a
disproportionately large part of it. Progressive collapse of building
structures is initiated by loss of one or more vertical load carrying
members, usually columns. After one or more columns fail, an
alternative load path is needed to transfer the load to other structural elements. If the neighboring elements are not designed to resist the redistributed loads, failure will happen with further load
redistribution until equilibrium is reached, resulting in partial or
total collapse of the structure.
Progressive collapse is triggered by abnormal loading that
causes local failure of one or more columns if the building lacks
sufcient ductility, continuity and/or redundancy. The local or
complete collapse may cause signicant casualties and damage
disproportionate to the initial failure. A notable example is partial
collapse of the Ronan Point apartment building in London. An accidental gas explosion in a corner kitchen on the 18th oor initiated
progressive collapse of the 24-story building in 1968. This event
triggered extensive progressive collapse research and led to development of design guidelines for the prevention of progressive
collapse [13].

Corresponding author. Tel.: +1 614 292 1338.

E-mail addresses: inhyoks@gmail.com (B.I. Song), sezen.1@osu.edu (H. Sezen).
0141-0296/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.05.050

The World Trade Center 7 (WTC 7) in New York City was a 47story ofce building adjacent to the WTC towers (WTC 1 and 2)
that collapsed following the terrorist attacks of September 11,
2001. WTC 7 collapsed several hours after the collapse of twin
WTC towers. The NIST report [11] concluded that: An initial local
failure occurred at the lower oors (below oor 13) of the building
due to re and/or debris induced structural damage of a critical
column (the initiating event) which supported a large span oor
bay with an area of about 2000 square feet. Vertical progression
of the initial local failure occurred up to the east penthouse, as
the large oor bays were unable to redistribute the loads, bringing
down the interior structure below the east penthouse. Horizontal
progression of the failure across the lower oors triggered by
damage due to the vertical failure, resulting in a disproportionate
collapse of the entire structure. The FEMA 403 [6] study emphasized the signicance of res on the collapse. This is a good example of disproportionate collapse caused by debris and/or re
induced failure of a column or columns in a tall steel building. In
this research, several columns were sequentially removed from a
building, which can resemble the initial debris damage and gradual
and intensifying re damage or a various other loads.
Failure of one or more columns in a building and the resulting
progressive collapse may be a result of a variety of events with different loading rates, pressures or magnitudes. The magnitude and
probability of natural and man-made hazards are usually difcult
to predict. Therefore, most of the current progressive collapse
design guidelines are threat-independent and do not intend to
prevent such local damage, e.g., ACI 318 [1]. Rather, their purpose

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

is to provide a level of resistance against disproportionate collapse

and to increase the overall structural integrity. Design guidelines
typically require minimum level of redundancy, strength, ductility
and element continuity. The codes typically prescribe simplied
analysis procedures requiring instantaneous removal of certain
critical columns in a building, e.g., GSA [8]. In this paper, effectiveness of such commonly used progressive collapse evaluation and
design methodologies is investigated through numerical simulations and experimental testing of the building.
A large number of numerical studies have been conducted to
evaluate the effectiveness and consistency of the current progressive collapse design guidelines. However, very limited experimental research has been performed to validate the results of these
computational studies and to verify the methodologies prescribed
in the guidelines. This is mainly because it is difcult to construct
and test full-scale building specimens and such large-scale testing
is discouragingly expensive. In this study, an existing steel frame
building, Ohio Union building, was tested by physically removing
four rst-story columns. The building was instrumented and the
experiment was conducted prior to its scheduled demolition. The
building was also modeled and analyzed using the computer program, SAP 2000 [15], following the requirements of the current
progressive collapse evaluation and design guidelines. The results
from static and dynamic analysis of the building were compared
with the experimental data.
2. Progressive collapse guidelines
American Society of Civil Engineers (ASCE 7, [3]), General Services Administration [8], Department of Defense (Unied Facilities
Criteria, [4], and National Institute of Standards and Technology
[12] have developed criteria and guidelines to evaluate, design
and improve structural integrity and progressive collapse resistance of existing and new buildings. ASCE 7 [3] provides design
load combinations including abnormal loads and associated
probabilities. It also presents general direct and indirect design approaches to ensure structural integrity following local damage to a
primary load-carrying member. In this paper, the collapse resistance of the test building is evaluated using the load combinations
recommended by the ASCE 7 standard and GSA guidelines [8].
General Services Administration [8] provides guidelines for
evaluation of existing buildings and design of new buildings
against progressive collapse. A simplied threat independent
methodology is recommended for buildings with fairly regular
plans and up to ten stories above ground. A linear elastic static
analysis of the building is required after the instantaneous removal
of a rst story column located near the middle of longitudinal and
transverse perimeter frame or at the corner of the building. Progressive collapse and possible subsequent failure of elements are
investigated using the calculated demand-to-capacity ratio (DCR)
for each structural element. DCR is dened as the ratio of the force
(moment, shear, or axial force) calculated after the instantaneous
loss of a column and the corresponding capacity of the member.
In this study, the test building was analyzed using the load combinations specied by the GSA and the corresponding DCRs were calculated. The acceptance criteria provided by the GSA was then used
to assess the potential for progressive collapse.
3. Building experiment
The Ohio Union building, shown in Fig. 1, was located on the
Ohio State University campus. The four-story moment frame building was constructed in 1950. The building included a rectangular
oor plan with three columns on each transverse axis and nine
columns along the longitudinal axes. Column and beam section

665

properties and the longitudinal test frame geometry are shown

in Table 1 and Fig. 2, respectively. In Table 1, the rst and last numbers are the depth (in inch units) and nominal weight (lb/ft) of the
columns or beams, respectively (1 in. = 25.4 mm, 1 ft = 305 mm,
and 1 lb = 4.448 N). The letters WF and B are wide-ange (WF)
shaped I-beam and light I-beam, respectively, which were commonly used in the 1950s [2].
Before the buildings demolition, four rst-story columns were
removed in the following order: (1) two columns near the middle
of the longitudinal perimeter frame, (2) column in the building corner, and (3) column next to the corner column. As shown in Figs. 1
and 3, four of the nine exterior columns were rst torched near the
top and bottom. Only a small portion of the ange was left intact
when the cross sections were cut. The middle column segment between the torched sections was then pulled out by a bulldozer
using a steel cable (Fig. 3).
The columns were removed within a very short time period
representing an instantaneous column removal as recommended
in the design guidelines. As shown in Fig. 4, 15 strain gauges were
installed on the columns and beams closely linked to the removed
columns to monitor the redistribution of gravity loads using the
change in strains measured during the removal of columns. During
the column removal process, a portable data acquisition system
and a scanner connected to a laptop computer recorded the strains.
No signicant visible damage was observed in the building even
after the four columns were removed. Detailed description of the
test building, instrumentation, experimental procedure and recorded data can be found in Song [16].
During the eld experiment, strains in members neighboring
the removed columns were measured as each column was torched
and removed. In this study, universal general purpose strain
gauges with a resistance of 120 0.3% Ohms were used. All strain
values dropped to negative values after each column was torched
or removed, and then stabilized after a certain amount of time.
These negative strain values indicate that the structural members
contracted and compressed when the neighboring columns were
torched. Most of the measured strain values dropped more when
the columns were torched than when they were removed during
the experiment. The largest drop of strain values was observed
when the last column was torched.

4. Analysis procedures and results

Numerical simulations of the test building were performed
using the computer program SAP2000 [15] to investigate the
progressive collapse performance of the building. At the time of
testing, the frames carried only dead loads due to weight of walls,
slabs, beams, and columns. In the linear static analysis, the dead
loads were multiplied by 2.0 as recommended in the GSA guidelines [8]. The live load was assumed to be zero in all analyses
because the test building was not occupied, and most of the partitions, furniture and other non-structural loads were removed from
the building. To calculate the dead load of the walls, densities of
glass and brick were assumed to be 2579 kg/m3 and 1920 kg/m3,
respectively. Properties of frame members were obtained from
the original structural drawings and design notes. Yield strength
of all frame members of the Ohio Union building was assumed to
be 345 MPa (50 ksi), as specied in the original design drawings.
Details of the modeling and analysis assumptions and results are
reported in Song [16] and Song et al. [17,18].
Two-dimensional (2-D) as well as three-dimensional (3-D)
models of the building were developed to analyze and compare
the progressive collapse response. Fig. 5 shows 2-D and 3-D
SAP2000 models of the Ohio Union building with frame member
numbers. As in the actual building experiment, four circled

666

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

Fig. 1. (a) Building before demolition, (b) four rst-story columns exposed, (c) columns removed, and (d) building during the gradual demolition process.

Table 1
Column and beam sections of the Ohio Union building.
Column section

Beam section

Column number

Column type

Beam number

Beam type

C1
C2
C3
C4
C5
C6
C7
C8
C9

10
12
12
10
10
10
10
10
10

B1
B2
B3
B4
B5
B6
B7
B8
B9

24
21
16
21
18
14
14
24
18

WF
WF
WF
WF
WF
WF
WF
WF
WF

72
133
120
100
89
54
112
60
33

B 76
B 68
B 58
WF 62
WF 50
B 17.2
B 22
WF 76
WF 45

columns were sequentially removed in the following order: columns 27, 22, 2, and 7.
Linear static, nonlinear static, linear dynamic, and nonlinear
dynamic analysis methods, in order of increasing complexity, can
be used to analyze a structure to investigate its structural behavior.
Researchers investigated the advantage and disadvantage of each
of these procedures for progressive collapse analysis [14]. A complex analysis is desired to obtain more realistic results representing the actual nonlinear and dynamic response of the structure
during the progressive collapse. However, both GSA and DOD
guidelines recommend the simplest method, linear static, for the
progressive collapse analysis since this method is cost-effective
and easy to perform. One of the objectives of this paper is to

Fig. 2. Longitudinal frame elevation including beam and columns sections (see Table 1).

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

667

Fig. 3. Before and after removal of middle part of a column.

Fig. 4. Plan view of strain gauge placement in Ohio Union building with columns and beam labeled (15 strain gauges are shown in the circles).

compare the simplest and most complicated analysis procedures

(i.e., linear static and nonlinear dynamic procedures) for the evaluation of progressive collapse potential of the test building.
4.1. 2-D linear static analysis
Linear static analysis is a simple and commonly used method to
investigate progressive collapse potential of a building [3], and [8].
Fig. 6 shows the elastic moment diagrams after the removal of each
column from the Ohio Union building. When the rst two columns
were removed, the largest bending moments were localized and
typically occurred in the members above or immediately next to
the removed columns. The maximum moments signicantly increased and spread within the frame when three and four columns
were removed.
Demand-to-capacity ratios (DCR) were calculated for each
frame member, and the building response was evaluated by

comparing the calculated DCR values based on the recommendations of GSA guidelines. DCR for moment is dened as the ratio
of the maximum moment demand Mmax of the beam or column
calculated from linear elastic analysis to its expected ultimate
moment capacity Mp, which is calculated as the product of plastic
section modulus and yield strength. In Mp calculations for columns,
the effect of the axial load is neglected because the column axial
loads were relatively small and did not signicantly affect the
moment capacity of the cross section.

DCR

M max
Mp

Fig. 7 shows the moment diagram and the corresponding

maximum DCR values at the end of each beam and top of each
column after four columns were removed from the frame. The
columns in the top story had higher DCR values, indicating that
after removal of columns additional loads were transferred

668

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

Fig. 5. (a) Two-dimensional SAP2000 model with frame member numbers and (b) three-dimensional SAP2000 model of the Ohio Union building (circled columns are
removed in the order shown).

Fig. 6. Moment diagrams: (a) after one column was removed, (b) after two columns were removed, (c) after three columns were removed, and (d) after four columns were
removed.

upward as well as to the adjacent spans. Smaller cross section

used in the top story columns is another reason for the higher
DCR values observed in the top story. As shown in Fig. 7, the
maximum DCR value of 2.83 was calculated in Column 10 in
the top story. The maximum calculated beam DCR value was
0.94 in beam 63 in the third oor level.

Fig. 8 shows DCR values for each frame member for all column
removal cases. Frame member numbers up to 45 are columns, and
beams are numbered from 46 to 85 (Fig. 5a). After the rst column
was removed, DCR values for all columns and beams were below
0.5. The DCR values after the loss of second column was similar
to those of third column loss, all of which were less than 1.5. The

669

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

2.83

0.34

2.25

0.32
0.21

1.56

0.16

0.22

0.09

0.44

0.17

0.63

1.54
0.91

0.24

0.07

0.12
0.34

0.45

0.14
0.18

0.41

0.40

0.17
0.26

0.56

0.48

0.07

0.86
0.55

0.35

0.41

1.44

1.21

0.45
0.51

0.37
0.42

0.00

0.94

0.58

0.91

0.47

0.45

2.24
0.83

0.56
0.89

2.14

0.38

2.38

1.10
0.82

0.40

0.27

0.42

0.32

0.15

0.80

0.49

0.77

0.39

0.11

0.03

Fig. 7. Moment diagram and corresponding DCR values after the loss of four columns in the Ohio Union building.

Fig. 8. Change in DCR values of each frame member for all cases.

DCR values for columns were remarkably increased after the fourth
column was lost. Columns were impacted more than beams when
all four columns were removed from the frame. As acceptance criteria, the maximum DCR limits specied in GSA [8] are 2.0 and 3.0
for columns and beams for the test building, respectively. After all
four columns were removed, no beams and ve columns (i.e., columns 8, 9, 10, 20 and 25) exceeded the DCR criteria. The change in
DCR values for beams was not signicant compared with that of
columns. The DCR values of beams were always less than 1.0. This
is probably due to potential redistribution of loads to the adjacent
beams in the analyzed frame.
After four columns were removed, the building was more
susceptible to progressive collapse. This was also reected in the
maximum displacements calculated from linear static analysis.
As columns were sequentially removed, the maximum vertical displacements were calculated as 11.40, 11.54, 30.73, and 17.93 cm at
the joints immediate above the rst (column 27), second (column
22), third (column 2) and forth (column 7) removed columns,
respectively.

phenomenon and therefore nonlinear dynamic analysis is more

realistic and accurate than linear static analysis.
In nonlinear dynamic analysis, a major load bearing structural
element is removed dynamically and the structural material is allowed to undergo nonlinear behavior. Fig. 9 illustrates the replacement of a removed column by equivalent loads in nonlinear
dynamic analysis. First, the building is modeled with its dead load
assigned. After the internal (equivalent) forces in a given column
are determined from static analysis, the column is replaced with
its equivalent forces to simulate the instantaneous removal of
the column. As shown in Fig. 10, the equivalent load is rst assigned with a uniform time history function. This corresponds to
the initial case where the column is still in place and carrying
the dead load. Then the column is suddenly removed using a step
function. The sum of a uniform time history function and the column loss function represents the column loss. This is referred to
as a time-history analysis where the response of the structure is
calculated during and after the removal of column(s) as a function
of time.
In this study, both geometric and material nonlinear behaviors
were considered in the nonlinear dynamic analysis. Material properties such as yield strength, ultimate strength, and ductility were
important parameters to design a building model. P-Delta effect
was considered as a geometric nonlinearity. Also, several dynamic
and nonlinear parameters including time step, damping ratio, and
plastic hinges was dened before performing nonlinear time history analysis.
The vertical displacements of the joints above each removed
column were calculated during and after removal of each column
in the rst story. Fig. 11 shows the vertical displacement history
of Joint 1, 2, 3, and 4 above the rst (column 27), second (column
22), third (column 2), and fourth (column7) removed columns,
respectively (Fig. 5a) after the removal of fourth column. The columns were removed at time of 0 s and negative values indicate

4.2. 2-D nonlinear dynamic analysis

Progressive collapse is a dynamic event involving vibration of
building elements and resulting in internal dynamic forces affected
by inertia and damping. Progressive collapse is inherently a
nonlinear event in which structural elements are stressed beyond
their elastic limit to failure. Nonlinear dynamic procedure reects
the dynamic and nonlinear aspects of the progressive collapse

Fig. 9. Column removal load representation for nonlinear dynamic analysis.

670

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

Fig. 10. Time history function for column loss simulation.

downward displacements. As shown in Fig. 11, the joints above the

four removed columns settled at the permanent displacements of
6.05, 6.12, 17.93, and 9.98 cm, respectively. The maximum
transient vertical displacements calculated from 2-D nonlinear dynamic analysis were 7.11, 7.24, 20.47, and 11.33 cm at Joint 1, 2, 3,
and 4, respectively.
4.3. Comparison of results from 2-D and 3-D analyses
A 3-D model of the Ohio Union building was developed, and
progressive collapse analysis was performed using this model.
Fig. 12 shows a comparison of DCR values for moments determined
from 2-D and 3-D models after four columns were removed. In 2-D
linear static analysis, columns were more impacted than beams.
Five columns exceeded the DCR criteria of 2.0 [8], but none for
the beams after four columns were removed. The DCR values of
all beams were less than 1.0, and the maximum DCR value observed in beams was 0.94. However, DCR values calculated from
3-D linear static analysis showed an opposite trend compared to
2-D results. Beams were more inuenced by the column loss. The
maximum DCR value of beams was 1.49 while that of columns
was 0.96. The reason that beams had higher DCR values than
columns in the 3-D linear static analysis was possibly due to the
larger deformation and participation of beams in the transverse
direction. It was found that beams, especially in the top story, were

Fig. 11. Displacement of joints above each removed column after all columns were
removed.

Fig. 12. Comparison of DCR values determined from 2-D and 3-D linear static
analysis after removal of four columns.

signicantly deformed in the transverse direction after each column removal. 2-D linear static analysis may lead to limited and
underestimated demands for beams.
More interestingly, it was observed that DCR values calculated
from the 3-D linear static analysis were smaller than those from
2-D linear static analysis for columns and most beams. As shown
in Fig. 12, all members had DCR values of less than 1.5, and satised GSA acceptance criteria of 2.0 for columns and 3.0 for beams.
This could be mainly due to contribution of transverse beams. The
transverse beams can distribute loads to the connected columns
and beams in the transverse direction, leading to a decrease of
force demands in structural members.
Table 2 shows the comparison of maximum vertical displacements calculated from 2-D and 3-D analyses. 3-D models showed
lower maximum displacements than 2-D models for both linear
static and nonlinear dynamic analysis. Similar to the DCR results,
the transverse beams connected to the interior columns and the
beams increased the overall resistance of structure, leading to
smaller deformations in the 3-D model. As shown in Table 2, Linear
static analysis resulted in higher maximum vertical displacements
than nonlinear dynamic analysis in both 2-D and 3-D models. For
example, the maximum vertical displacement calculated from
2-D linear static analysis was 30.73 cm at Joint 3 while that from

671

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

Table 2
Comparison of vertical displacement (cm) after all columns removal.
Joints above removed columns

2-D model

3-D model

Linear static analysis

Joint
Joint
Joint
Joint

1
2
3
4

11.40
11.53
30.73
17.93

Nonlinear dynamic analysis

Maximum

Permanent

7.11
7.24
20.47
11.33

6.05
6.12
17.93
9.98

Linear static analysis

7.85
7.95
10.03
7.37

Nonlinear dynamic analysis

Maximum

Permanent

4.22
4.27
5.64
3.68

3.66
3.71
5.08
3.38

Table 3
Plastic hinge rotations (h, degree) at the location where each column was removed after all columns removal.
Removed columns

2-D nonlinear dynamic analysis ()

3-D nonlinear dynamic analysis ()

Joint 1
Joint 2
Joint 3

0.53
0.54
1.80

0.31
0.32
0.50

the 2-D nonlinear dynamic analysis was 20.47 cm. It seems that
the impact factor of 2 (i.e., dead loads multiplied by 2) in linear static analysis led to very conservative results. Marjanishvili reported
that a more complicated analysis method such as nonlinear dynamic analysis may result in less severe structural response, due
to more accurate estimates of load distribution and less stringent
evaluation criteria.
Table 3 shows plastic hinge rotations at the location where
columns were removed after four columns removal. Plastic hinge
rotation was chosen as acceptance criteria for the nonlinear
dynamic analysis [8], which also possibly evaluates whether the
moment connection of the frame is strong enough to survive the
excessive moment rotation of the joints. Plastic hinge rotation
angle for beam members on each side of the removed column
can be measured between horizontal line and tangent to maximum
deected shape, which is dened by Eq. (2).

h tan1



dmax
L

where h is the maximum hinge rotation, dn the maximum displacement of columns at the location where the column is removed, and
L is the beam length or column spacing in the longitudinal direction.
As shown in Table 3, hinge rotations calculated from the 3-D
model were smaller than those from the 2-D model, because of
lower maximum displacement values in 3-D nonlinear dynamic
analysis. The maximum plastic hinge rotation was only 1.80 at
the hinge above third removed column (Column 2) in linear static
analysis. For both 2-D and 3-D nonlinear dynamic procedure, the
values of plastic hinge rotation were much smaller than 12 of
GSA [8] criteria, indicating that the Ohio Union building was not
susceptible to progressive collapse. Considering that no signicant
deformations were observed during eld testing, GSA criteria for

plastic deformations or hinge rotations may be more realistic than

the GSA criteria for force demands or DCR values.
4.4. Comparison of calculated and measured strains
Table 4 shows changes in strain (De) obtained from the eld
test, compared with those calculated from 2-D and 3-D models.
During the eld test, strain values changed as each column was
torched and removed. De (Field Test) reported in Table 4 are the
changes in strain values recorded by the strain gauges in the eld
after the last column torching. De (Computational Model) is the
changes in strain values after the last column removal. De is calculated by considering the combined effect of axial load and a bending moment, both of which were determined from the SAP2000
analysis. Details of calculations and assumptions are reported in
Song [16].
Total of fteen (15) strain gauges were used in this experiment
(see Fig. 4). Table 4 compares selected strain measurements and
analytical model results. Six strain gauges were selected since
strain gauges 1, 3, 7, 10 and 12 are attached to the same columns
of the selected gauges 2, 4, 8, 9 and 11, respectively, and strain
gauges 5, 6, 13 and 14 were attached above the removed columns.
The strain gauges on the same column showed very similar strain
measurements. The strain gauge 15, attached on Beam 67, was
selected from the experimental study to compare the results from
2-D and 3-D models because it was the only strain gauge left in the
perimeter frame in the 2-D model after the four columns were
removed (location of strain gauges are shown in Fig. 4). Strain
gauges 2, 4, 8, 9 and 11 were attached on the interior columns.
As shown in Table 4, for strain gauge 15 attached on Beam 67,
De calculated from the 3-D model was closer to the experimental
result than that from the 2-D model. 3-D model can account for
redistribution of the buildings weight to both exterior and interior

Table 4
Comparison of change in strain (De) obtained from the eld test after last column torching with that calculated from 2-D and 3-D analyses after all columns removal (% difference
was indicated in parentheses).
Strain gauge

Field test

2 (Column)
4 (Column)
8 (Column)
9 (Column)
11 (Column)
15 (Beam)

55  106
37  106
29  106
28  106
33  106
37  106

2-D model

3-D model

Linear static analysis

Nonlinear dynamic analysis

Linear static analysis

Nonlinear dynamic analysis

118  106 (219%)

64  106 (73%)

165  106 (200%)

121  106 (227%)
34  106 (17%)
104  106 (271%)
49  106 (48%)
53  106 (43%)

32  106
17  106
4  106
20  106
7  106
46  106

(42%)
(54%)
(86%)
(29%)
(79%)
(24%)

672

B.I. Song, H. Sezen / Engineering Structures 56 (2013) 664672

were reached during this study based on the evaluation of experimental data and structural analysis of the test building.
The measured strain data compared relatively well with the
analysis results. In particular, 3-D model was more accurate than
the 2-D model, because 3-D models can avoid overly conservative
solutions as well as account for 3-D effects such as contribution of
transverse beams to overall resistance of the frame. The 3-D model
had lower DCR values and vertical displacements than 2-D model,
which was possibly due to inclusion of transverse beams in the 3-D
model. The 3-D model is believed to be more realistic than 2-D
model for the progressive collapse analysis.
The strain values calculated from the nonlinear dynamic analysis were smaller than those from the linear static analysis, and
were closer to the measured strains. Also, linear static analysis
showed higher DCR values and vertical displacements than nonlinear dynamic analysis for both 2-D and 3-D models. The amplication factor of 2 required for the dead load in linear static analysis
may lead to very conservative analysis results.
For future research, it would be better to consider the actual
material properties and connections of the building in the analytical models in order to obtain more reliable results.
Acknowledgements
This research was partially funded by the National Science
Foundation (CMMI 0745140), American Institute of Steel Construction, and URS Corporation; this is gratefully acknowledged. The
authors would like to thank SMOOT Construction, Loewendick
Demolishing Contractors, and the Ohio State University for providing access to the test building and help with the experiment.
References

Fig. 13. Comparison of calculated strains and strain measured by (a) Strain Gauge 4
and (b) Strain Gauge 15 after all columns were removed.

columns and beams while only exterior members were considered

in the 2-D model. All of De values calculated from the 3-D models
were very comparable to the measured strains.
Fig. 13 compares strain values (De) measured in the eld and
calculated from linear static and nonlinear dynamic analyses after
all columns were removed. For the interior columns (i.e., Strain
gauge 4) and the beam (i.e., Strain gauge 15), the measured strains
were closer to the De values calculated from the nonlinear dynamic analysis. The strain increments (De) calculated from the linear static analysis were much larger than the measured values. It
should be noted that the linear static analysis was performed by
amplifying the gravity (dead) loads by a factor of 2.0 following
the recommendations of GSA [8] while the unfactored dead load
was used in dynamic analysis. If the unfactored dead loads were
used in the linear static analysis, the calculated strain would be reduced by half to approximately 0.00006, which is still larger than
the maximum measured strain.
5. Conclusions
Progressive collapse performance of an existing steel frame
building was evaluated by physically removing four rst-story columns from the building and by performing linear static and nonlinear dynamic analysis of the building. The following conclusions

[1] ACI 318-11. Building code requirements for structural concrete and
commentary. Farmington Hills, MI: American Concrete Institute (ACI); 2011.
[2] AISC. Manual of steel construction. 6th ed. American Institute of Steel
Construction (AISC); 1969.
[3] ASCE. Minimum design loads for buildings and other structures. Reston, VA:
American Society of Civil Engineers (ASCE); 2005.
[4] DOD. Design of buildings to resist progressive collapse. Unied Facilities
Criteria (UFC) 4-023-03. Department of Defense (DOD); 2005.
[6] FEMA 403. World trade center building performance study: data collection,
preliminary observations and recommendation. Report: FEMA 403.
Washington, DC: Federal Emergency Management Agency (FEMA); 2002.
[8] GSA. Progressive collapse analysis and design guidelines for new federal ofce
buildings and major modernization projects. Washington, DC: General
Services Administration (GSA); 2003.
[11] NIST SP 1000-5. Progress report on the federal building and re safety
investigation of the world trade center. Gaithersburg, MD: National Institute
of Standards and Technology; 2004.
[12] NIST. The collapse of the world trade center towers. Final Report.
Gaithersburg, MD: National Institute of Standards and Technology (NIST);
2005.
[13] NIST. Best practices for reducing the potential for progressive collapse in
buildings. Gaithersburg, MD: National Institute of Standards and Technology
(NIST); 2006.
[14] Powell G. Progressive collapse: case study using nonlinear analysis. ASCE
structures congress and forensic engineering symposium. New York, NY; April
2024, 2005.
[15] SAP2000. SAP 2000 advanced structural analysis program. Version 12.
Berkeley, CA, USA: Computers and Structures, Inc. (CSI); 2009.
[16] Song BI. Experimental and analytical assessment on the progressive potential
of existing buildings. Masters thesis. The Ohio State University; 2010. p. 125.
[17] Song B, Sezen H. Evaluation of an existing steel frame building against
progressive collapse. In: ASCE structures conference, Austin, Texas; April 30
May 2, 2009.
[18] Song B, Sezen H, Giriunas K. Experimental and analytical assessment on
progressive collapse potential of actual steel frame buildings. In: ASCE
structures conference and North American steel construction conference,
Orlando, Florida; May 1215, 2010.