You are on page 1of 15

Engineering Science and Technology, an International Journal 35 (2022) 101193

Contents lists available at ScienceDirect

Engineering Science and Technology,


an International Journal
journal homepage: www.elsevier.com/locate/jestch

Progressive collapse risk of steel framed building considering column


buckling
Hussein Elsanadedy a,1,⇑, Halil Sezen b, Husain Abbas a, Tarek Almusallam a, Yousef Al-Salloum a
a
Chair of Research and Studies in Strengthening and Rehabilitation of Structures, Department of Civil Engineering, College of Engineering, King Saud University, Riyadh 11421,
Saudi Arabia
b
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 a b s t r a c t

Article history: Steel buildings are generally susceptible to the risk of progressive collapse in case one or a few load-
Received 21 October 2021 carrying members are lost in an extreme event such as blast, impact, or fire. Thus, to avoid the catas-
Revised 25 May 2022 trophic collapse of steel structures, it is imperative to investigate and learn from the performance of
Accepted 1 June 2022
existing steel buildings for progressive collapse potential. The Ohio Union building, a multi-story steel-
Available online 11 June 2022
framed building that existed on the campus of Ohio State University, was tested earlier before its demo-
lition by one of the co-authors for progressive collapse assessment under the successive removal of four
Keywords:
columns located at the first-story level. The aim of this study is to conduct NLD (nonlinear dynamic) anal-
Progressive collapse
Steel building
ysis for the building considering buckling of columns to numerically assess its potential for progressive
Column buckling collapse and then compare it with test observations. The calibrated finite element (FE) model was then
NLD analysis extended to incorporate more hypothetical sequential and simultaneous column-loss scenarios. As one
LS analysis of the most widely accepted documents used among practicing engineers for numerical assessment of
progressive collapse potential of buildings, the 2003 GSA (General Services Administration) guidelines
were also applied. The simplified LS (linear static) analysis approach of the GSA guidelines was used
for assessing the progressive collapse risk of the building, and the results were then compared with both
test observations and NLD analysis. New dynamic increase factors (DIFs) are recommended for both
force- and deformation-controlled actions to be used with the LS analysis.
Ó 2022 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction loss of column(s) under extreme load cases. Yet, the results of
structural analysis obtained by adopting these procedures are
In fact, progressive collapse occurs when major vertical load- different.
carrying elements fail leading to the residual structure being overly The SAP2000 software was employed by Marjanishvili and
loaded and consequent significant changes in its boundary condi- Agnew [10] to model the progressive collapse of multi-story build-
tions and large demands on elements neighboring the lost mem- ings using nonlinear dynamic (NLD), linear dynamic (LD), nonlin-
ber(s). Failure of elements could propagate as the load is ear static (NLS), and linear static (LS) analyses. A 20-story
redistributed. This process could continue until the structure building was numerically investigated by Fu [11] using the NLD
resumes equilibrium. Examples of real progressive collapse inci- approach, and it was concluded that columns, which are directly
dents for existing buildings include a) the gas explosion of the linked to the lost column in the same floor, should be designed
Ronan Point Building in the UK in 1968 [1,2], b) bombing of the for axial load of 2(D + 0.25L), where D and L are the dead and live
Murrah Federal Building in the US in 1995 [3], and c) collapse of loads, respectively. In another research, the progressive collapse
the World Trade Centre in the US in 2001 [4–7]. risks of seismically designed 10-story steel-braced buildings were
The alternate load path method has been adopted in various numerically assessed by Khandelwal et al. [12] using the alternate
codes, e.g. [8,9], for the quantification of damage owing to sudden load path method. The structural response was predicted with the
help of discrete springs and 2D macro-models of beam-column.
Powel [13] compared different analysis procedures and concluded
⇑ Corresponding author. that adopting a dynamic increase factor (DIF) of 2 in the static anal-
E-mail address: helsanadedy@ksu.edu.sa (H. Elsanadedy). ysis would be conservative. Tsai and Lin [14] reported that the NLS
Peer review under responsibility of Karabuk University. analysis procedure, adopted for the progressive collapse simulation
1
On leave from Helwan University, Cairo, Egypt.

https://doi.org/10.1016/j.jestch.2022.101193
2215-0986/Ó 2022 Karabuk University. Publishing services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

in RC (reinforced concrete) buildings, is more conservative than the decrease the beam displacement at the removed column location.
NLD analysis. A simplified method was suggested by Vlassis et al. Alrubaidi et al. [31] studied experimentally and numerically the
[15] and Izzuddin et al. [16] for evaluating the progressive collapse progressive collapse resistance of regularly used intermediate
risk of steel frames using the NLS approach. Another simplified moment resisting steel frame connections under the event of col-
approach was suggested by Naji and Irani [17] to predict the pro- umn removal. Among all studied connections, the connection
gressive collapse potential of steel-framed buildings using the designed according to the Turkish-Earthquake Code [32] had the
load–displacement and capacity curves for a fixed beam. Elsanad- highest progressive collapse resistance. Rodríguez et al. [33] pre-
edy et al. [18] carried out FE modeling to study the likelihood of sented the results of sensitivity and fragility analysis for the pro-
existing steel-framed building to progressive collapse due to differ- gressive collapse potential of steel moment-resistant frames.
ent blast scenarios. The developed models were validated with the Tornado diagrams were derived to quantify the sensitivity of the
help of available test results of steel tube subjected to blast loads. progressive collapse to design parameters, and fragility models
Talebi et al. [19] used different cross-sections of buckling were developed for assessing the probability of failure as a function
restrained brace systems (BRBs) in structural frames to mitigate of a specified intensity measure. The results revealed the impact of
their progressive collapse against different fire scenarios. While the random variables/studied parameters on the fragility/progres-
using BRBs, the results revealed the role of higher stiffness, sym- sive collapse potential of steel moment-resistant framed buildings.
metry in behavior, and redistribution of load to adjoining elements Recently, Elsanadedy et al. [34] examined experimentally and
without the buckling of the braces in elevating the collapse tem- numerically the progressive collapse risk of different types of sim-
perature. In later research [20], the authors investigated the effect ple shear steel beam-column connections under column-removal
of the inverted-V system of BRBs on the progressive collapse of a 4- events. It was found that the one with double angle and bolted
story steel-framed structure against a column loss caused by fire. web-cleats had the highest resistance against progressive collapse.
The results revealed an enhancement in global collapse duration A large number of numerical and experimental studies involv-
due to the higher stiffness provided by BRBs. In addition, as com- ing few structural members have been performed to assess the
pared to the ordinary concentric brace systems, the frames with applicability of seismic design guidelines for progressive collapse
BRBs had higher resistance to progressive collapse. As an alterna- mitigation. Nevertheless, experimental investigations to validate
tive, different schemes of strengthening connections in steel- the results of the studies at the system level are scanty, which is
framed structures were tried experimentally and numerically by primarily due to the fact that such studies are costly to undertake.
Alrubaidi et al. [21]. The schemes were successful at mitigating The field tests conducted by the second author [35–37] on existing
the risk of progressive collapse of steel frames to a great extent. steel framed buildings involve the sequential removal of four col-
Shan and Li [22] studied numerically the progressive collapse umns of the first story. Song and Sezen [35] carried out full-scale
behavior of steel frames triggered by different cases of fire expo- field tests for column removal scenario on the five-story (B + 4)
sure. The frames had partial infill walls. The analysis showed that Ohio Union building built in 1950 on the campus of the Ohio State
higher-grade steel, higher number, and fewer openings in walls University. Song et al. [36] performed tests on the 3-story building
raised the temperature, thereby causing collapse. The authors used in Northbrook, Illinois. During the field tests on these buildings,
an equivalent strut model to propose a design procedure for reduc- there was no or very limited live load. The results of these field
ing the progressive collapse risk caused by exposure to fire. Guo tests reveal that the steel-framed buildings were safe even after
et al. [23] studied experimentally the progressive collapse of nine the loss of four perimeter columns, which highlights the robust-
beam-column assemblages having welded unreinforced flange- ness of these buildings.
bolted web (WUF-B) connections under a column-loss event trig- In this research, NLD analysis considering column buckling was
gered by fire exposure. The test results were compared with the carried out for the Ohio Union steel building tested by Song and
control specimen tested at ambient. The catenary mechanism Sezen [35] to numerically assess its potential for progressive col-
was observed at a displacement of more than half of its maximum lapse and then compare it with test observations. The calibrated
value. A loss of progressive collapse resistance was noted for fire FE model was then further extended to incorporate more hypo-
exposed frames compared to the control specimen. thetical sequential and simultaneous column-loss scenarios. More-
Hadidi et al. [24] clarified that steel-framed buildings that are over, the simplified LS analysis approach of the 2003 GSA
designed only for seismic requirements as per AISC-LRFD [25] are guidelines [9] was used for the progressive collapse assessment
vulnerable to progressive collapse with regard to the 2002 UFC of the Ohio Union building, and the results were then compared
requirements [8]. Chen et al. [26] utilized the energy principle to with both test observations and NLD analysis.
investigate the progressive collapse robustness of steel-framed
buildings. Zoghi and Mirtaheri [27] conducted NLD analysis to
assess the progressive collapse potential of seismically designed 2. Building description
steel-framed building accounting for the effect of infill walls. It
was demonstrated that modeling the RC slabs as well as the infill Fig. 1 shows the five-story moment-resistant framed Ohio
walls could improve the stability and catenary action of the build- Union building, tested earlier by the second author [35] for pro-
ing. Qiao et al. [28] proposed a simplified approach to investigate gressive collapse assessment. The building had a rectangular foot-
the progressive collapse resistance of steel-framed buildings by print with nine rows of three columns each. Table 1 summarizes
idealizing them as single-story substructures. Naji [29] suggested the sections used in beams and columns. Fig. 2 depicts the longitu-
a simple approach to assess the progressive collapse risk of steel dinal frame elevation and the ground floor framing plan [38]. The
beam-column assemblies due to the middle column loss event. list of steel sections presented in Table 1 shows I-beams that were
The developed approach considers the impact of the span/depth common during the period of construction of the building in the
ratio of the beam on the progressive collapse performance of the 1950s [39]. The building being old and out of service, it was
assemblies. In a study by Ebrahimi et al. [30], FE models were planned to be demolished, but before its demolition, the building
developed to estimate the progressive collapse risk of 6-story was used for the progressive collapse study through the successive
steel-framed buildings in the event of column loss. Two upgrading removal of four columns at the first-story level in the order as indi-
schemes for column-removal scenarios were investigated. They cated in Fig. 1. The time duration of column removal was kept
involved the use of braces and cables. The study indicates that short to represent the instantaneous column removal requirement
the use of bracing members and cables could substantially of the GSA design guidelines [9]. It should be noted that prior to the
2
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

duration of 100 s (rate of displacement increase = 1 mm/s). The


analysis was run until column failure. Axial load versus mid-
height lateral displacement was generated for the column and
the buckling load was then determined. This model was verified
using the published data in the literature for steel columns failing
in buckling under pure axial compression. A total of seven columns
were used for model verification, as seen in Table 2. These columns
included two specimens tested by Rasmussen and Hancock [42]
and five columns tested by Li et al. [43]. Experimental and numer-
ical results for all 7 columns are enlisted in Table 2, which shows
the accuracy of the predicted buckling load with errors ranging
from 0.6% to 5.8%.
The validated local FE model was used to compute the buckling
load of basement columns C1 to C5 of the Ohio Union building, as
seen in Table 3. It is worth noting that columns were divided into
Fig. 1. Ohio Union building after removing four first-story columns. beam elements of size 250 mm. A mesh sensitivity study was con-
ducted for column C1 as a case study. Two other models were cre-
ated for column C1 with an element size of 125 mm and 50 mm in
Table 1 the second and third models, respectively. It was identified that as
Member sections of the Ohio Union building.
the mesh size decreased from 250 mm to 50 mm, the buckling load
Column ID Column sec. Beam ID Beam sec. of column C1 increased by only 5%. Accordingly, it was decided to
C1 WF10  72 B1 B24  74 use a mesh size of 250 mm in all columns in order to have a global
C2 WF12  133 B2 B20  68 building model with a reasonable number of elements and hence
C3 WF12  120 B3 B16  58 reduce the computational time. In addition to the local FE model-
C4 WF10  100 B4 WF21  62
ing, the equations of the ANSI/AISC 360-16 [25] were used to cal-
C5 WF10  89 B5 WF18  50
C6 WF10  54 B6 B14  17.2 culate the buckling load of columns C1 to C5, which were 13% to
C7 WF10  112 B7 B14  22 20% less than those predicted by the FE model. This is because
C8 WF10  60 B8 WF24  76 the equations of the ANSI/AISC 360-16 [25] contained an approxi-
C9 WF10  33 B9 WF10  21 mation for the column strength curve that provides a transition
between elastic buckling and yielding. Moreover, the formulas of
removal of columns, almost all loads including live load and inte- the ANSI/AISC 360-16 [25] implemented safety factors to account
rior partitions were removed from the floors and the building for accidental eccentricity and initial imperfection of steel col-
was only exposed to the own weight of all structural elements plus umns. Axial load versus mid-height lateral displacement of the five
the weight of exterior non-structural masonry walls. Even though basement columns C1 to C5 is presented in Fig. 6, from which the
four columns were removed at the first-story level, the building did maximum load of each curve depicts the critical buckling load
not collapse, as seen in Fig. 1. Full details of the building, field test identified in Table 3.
procedure, recorded observations, and test measurements are
given in Song [38].
4. NLD FE model of the building
3. Local FE model for column buckling
Nonlinear dynamic (NLD) FE modeling of the Ohio Union build-
ing was carried out using the LS-DYNA package [40]. The FE model
A local FE model was developed for the single steel column,
was developed with the help of FEMB Pre-Processor 28.0.
using the commercial FE package LS-DYNA [40], to compute its
critical buckling load, as seen in Fig. 3. In this model, 2-node
Hughes-Liu beam elements – with cross-section integration and 4.1. Finite element discretization
having compression, tension, bending, and torsion capabilities
were used to model the columns with the boundary conditions Fig. 7 illustrates the FE model of the whole building where beam
shown in Fig. 3. The constitutive model employed for steel beam and shell elements were utilized. All columns and beams were rep-
elements was type 24 in the material model library of LS-DYNA resented by 2-node Hughes-Liu beam elements. Four-node quadri-
software [40]. This is a piecewise linear plasticity model in which lateral shell elements having Belytschko-Tsay formulation [44]
an elasto-plastic material with arbitrary stress versus strain curve were used for modeling the RC floor slabs. The global model has
and arbitrary strain-rate dependency can be defined. In this model, a total of 94,803 elements (35,523 beam elements and 59,280 shell
the stress–strain behavior may be treated by a bilinear curve by elements). It should be noted that the exterior masonry walls were
defining the tangent modulus (see Fig. 4), or alternatively, a curve not physically included in the model in order not to increase the
of effective stress versus effective plastic strain may be defined. solution time considerably; however, their weights were calcu-
Failure in this material model is controlled by effective plastic lated and input as line loads on exterior steel beams. Based on
strain. In this study, a bilinear stress–strain curve was utilized, the available information, the beam-column joints of the Ohio
and the input material model parameters included yield stress, Union building were welded unreinforced flange (WUF) moment
Poisson’s ratio, Young’s and tangent moduli, and plastic strain to connections. They were assumed to have higher capacity and duc-
failure. An initial imperfection of L/1000 was considered at the col- tility than the connected components (beams and columns), and
umn mid-height as recommended by Marques et al. [41], where L failure of beams and columns was assumed to occur before the
is the column height. A displacement-controlled loading was connection failure. Therefore, the connections were assumed to
applied to the column by a prescribed displacement–time history be fully rigid, and they were not explicitly modeled. This assump-
curve assigned on the top node, as seen in Fig. 5. It is identified tion may be valid as the prime goal of the study was to focus on the
from the figure that a target downward displacement (in the neg- global behavior of the building rather than the microscopic perfor-
ative Z-direction) of 100 mm was applied on the top node in a total mance of the beam-column connections.
3
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Fig. 2. Building frames including beam and column sections (see Table 1): (a) longitudinal frame elevation; (b) ground floor framing plan.

!1=p
e_
y¼1þ ð1Þ
C

where e_ is the rate of strain (in s1), and p and C are the model
parameters (taken as 1.6 and 250, respectively). The material model
used for RC slabs is the Concrete Eurocode EC2 (type 172). The
model is good for simulating the beam and shell behavior. The
model simulates discrete reinforcement rebars in concrete mem-
bers with smeared reinforcement. This model can represent con-
crete crushing in compression, concrete cracking in tension, and
yielding of steel rebars. The input parameters for the Concrete Euro-
code EC2 model include concrete strengths in compression and ten-
sion, tension stiffening modulus of concrete, Young’s modulus of
steel rebars, Poisson’s ratio of steel rebars, yield stress of steel
rebars, and ratios of steel rebars in the x and y directions. For the
nonlinear dynamic analysis used in this study, the damping ratio
was assumed to be 1%, which is reasonable for steel buildings
[38,45].
Fig. 3. Local FE model for prediction of buckling load of steel columns.

4.2. Constitutive models 4.3. Boundary conditions and load application

The material model with Piecewise Linear Plasticity (type 24) For representing the actual boundary conditions of the base-
was utilized for steel beams and columns. The magnification factor ment columns of the Ohio Union building, pin-type supports were
of the yield strength of the Cowper-Symonds was used for incorpo- assumed for the bottommost nodes of the basement columns. The
rating the strain-rate effect, as given in the following formula: column-removal scenario used in the numerical model followed
4
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Fig. 4. Effective stress versus effective plastic strain for bilinear curve of material model type 24 [40].

At the time of column removal, the elements representing the


removed column are instantaneously deleted from the model via
the ‘‘restart” option of the software, which necessitates the recal-
culation of the stiffness matrix for keeping the structural model
in its deformed/stressed condition. The analysis was then repeated
five times. In the initial run, all elements of the building were
included and the analysis was set to stop at 3 s. In the second
run, the ‘‘restart” option was used in which column 1 of the first
story was instantaneously deleted and this run was set to stop at
5 s. The third run was also a ‘‘restart” analysis in which column 2
was deleted and this run was set to end at 7 s. In the last two runs,
the ‘‘restart” option was employed in which column 3 (in the
fourth run) or column 4 (in the fifth run) was deleted from the
model and the run was set to terminate at 9 s (for the fourth
run) or 15 s (for the fifth run). The termination time of 15 s was uti-
lized for capturing the full dynamic behavior of the building after
the loss of the identified columns.
Fig. 5. Prescribed vertical displacement–time history curve for top column node.

4.4. Discussion of results

exactly the sequence used in the progressive collapse field tests Fig. 8 depicts the deformed shape of the building at the end of
conducted for the Ohio Union building in which column 1 of the analysis time for scenario S1. A steady-state maximum vertical
Fig. 7 was removed first, followed by column 2, then column 3, displacement of about 234 mm was predicted for the floor slabs in
and finally column 4 [35,38]. This scenario is denoted here as S1, the location of the first floor where column 3 is removed. Due to
as seen in Table 4. Similar to the field test, the building was only the unavailability of actual field measurement for the vertical dis-
subjected to its self-weight, as identified in Table 4. placement of the first-floor level at the location of column 3
In the analysis, the gravity load was input as a ramp function, removal, the predicted value of 234 mm could not be validated.
and it was held constant when it reached its peak value at 2 s. This The relationship between time and axial force is shown in Fig. 9
was to minimize the oscillatory behavior of the building due to the for critical basement columns C1 to C5 under missing column sce-
dynamic nature of the gravity load input in the time domain. Col- nario S1. It is clear that the peak axial forces in the five basement
umn removal scenario S1 involved, in sequence, removal of col- columns were considerably less than their axial capacity (critical
umns 1, 2, 3, and 4 at times 3, 5, 7, and 9 s, respectively. buckling load). Table 5 shows the key results of the NLD analysis

Table 2
Buckling load of columns tested by other researchers.

Researcher Buckling load


Column ID Exp (kN) FE (kN) Error (%)
Rasmussen and Hancock [42] I1000C 2092 2079 0.6
I1650C 1751 1736 0.9
Li et al. [43] H-30–1 8493 8976 5.7
H-30–2 8994 9072 0.9
H-50–1 7207 7288 1.1
H-50–2 7125 7293 2.4
H-70–2 3690 3906 5.8

5
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Table 3
Prediction of axial load capacity of basement columns C1 to C5 (effective length factor = 1; yield stress of steel = 345 MPa)*.

Column ID Column slenderness parameter (c) fcr (MPa) Pu-AISC (kN) Pu-FE (kN) Pu-AISC/Pu-FE
C1 0.90 245 3348 4162 0.80
C2 0.74 274 6917 7962 0.87
C3 0.75 273 6217 7236 0.86
C4 0.88 249 4724 5690 0.83
C5 0.89 248 4183 5056 0.83

*fcr = flexural buckling stress as per Section E3 of ANSI/AISC 360-16 [25]; Pu-AISC = ultimate load as per ANSI/AISC 360-16 [25]; Pu-FE = ultimate load as per FE analysis.

Fig. 6. Axial load versus mid-height lateral displacement for basement columns of Fig. 8. Deformed shape of building at the end of analysis time for scenario S1
Ohio Union building. (Units: mm, s).

of the building due to missing column scenario S1. It is also evident


that in all basement columns C1 to C5, peak axial loads after col-
umn loss scenario S1 were in the range of 24% to 72% of the critical
buckling load, whereas the residual axial load values ranged from
21% to 56% of the axial load capacity. Based on the NLD analysis,
this scenario did not result in the failure of any of the building col-
umns, and as a result, progressive collapse could not be triggered
as confirmed by the field tests conducted at Ohio State University
for the same building (see Fig. 1) [35,38]. Accordingly, the FE model
is considered validated based on the global behavior of the build-
ing obtained from test observations. The validated numerical
model of the building was further extended to investigate more
critical column-loss scenarios in which more loads are applied
and/or more critical first-story columns are removed.
As seen in Table 4, more hypothetical column-removal scenar-
Fig. 7. Global FE model of the whole building using LS-DYNA software.
ios (S2 to S8) were investigated for which the load combination fol-
lowed the nonlinear dynamic (NLD) analysis of the 2003 GSA
guidelines [9]. The gravity load on the building was increased to
include the superimposed dead load (SIDL) and 25% of the full
Table 4
design live load (L). As per the 2003 GSA guidelines [9], the full
Missing first-story column scenarios used for nonlinear dynamic FE analysis.
design live load was reduced in case of progressive collapse assess-
Scenario Sequence of column removal* Applied load** ment, admitting that the full live load application is less probable
Multiple column removal scenarios during an extreme event. In this building, the superimposed dead
S1 Column 1, then Column 2, then Column 3, SW load (including HVAC, mechanical and electrical, and flooring
then Column 4 materials attached to the slabs) was taken as 3.0 kN/m2, and the
S2 Column 1, then Column 2, then Column 3, SW + SIDL + 0.25L
then Column 4
full design live load was assumed as 4.0 kN/m2 for typical floors
S3 Columns 3, 6 and 7 simultaneously SW + SIDL + 0.25L and 1.0 kN/m2 for the roof. The parametric study shown in Table 4
S4 Columns 3, 4 and 5 simultaneously SW + SIDL + 0.25L includes four different multiple column loss scenarios (S2, S3, S4,
S5 Columns 3, 4 and 6 simultaneously SW + SIDL + 0.25L and S5) in addition to three single-column removal scenarios (S6,
Single column removal scenarios S7, and S8). It should be noted that the single-column loss scenar-
S6 Column 2 only SW + SIDL + 0.25L ios were studied for comparison with the LS approach of the 2003
S7 Column 6 only SW + SIDL + 0.25L
GSA guidelines in which single-column loss scenarios are used for
S8 Column 3 only SW + SIDL + 0.25L
progressive collapse assessment. These guidelines require the
*See Fig. 7 for reference. removal of a single exterior column located: (a) close to the middle
**SW = self-weight; SIDL = superimposed dead load; L = live load.

6
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

of the long direction (Column 2 (Scenario S6)), (b) close to the mid-
dle of the short direction (Column 6 (Scenario S7)), or (c) on the
corner (Column 3 (Scenario S8)). It should be noted that since sce-
nario S2 involved sequential removal of four columns at different
times, the analysis was set to terminate at 15 s from its onset.
However, for other multiple column loss scenarios S3 to S5 in
which three columns were simultaneously removed or for single-
column loss scenarios S6 to S8, the analysis was set to terminate
at 10 s.
The deformed shape of the building at the end of the analysis is
shown in Fig. 10(a) and 10(b) for scenarios S2 and S3, respectively.
The deformed shape at different times till collapse is displayed in
Figs. 11 and 12, for scenarios S4 and S5, respectively. For single-
column loss scenarios S6 to S8, the deformed shape of the building
is depicted in Fig. 13(a) to 13(c), respectively, at the end of the
analysis. The key results of the NLD FE analysis due to different
missing column scenarios are also enlisted in Table 5. It is clear
Fig. 9. Time vs axial force for critical basement columns under missing column
that for all studied scenarios, except S4 and S5, progressive col-
scenario S1. lapse potential was not identified and failure was not predicted

Table 5
Key results of NLD analysis due to different missing column scenarios.

Scenario Variation of axial load of critical basement columns Peak vertical disp. of 1st floor (mm) Potential
for
Col. Axial load Service load Peak load after Residual value Col. failure Value at Value after Residual
progressive
ID capacity column-loss after column- service column- value
collapse
scenario loss scenario load loss after column-
scenario loss scenario
(PN) (kN) (Ps) (kN) Ps/PN (%) (Pp) (kN) PP/PN (%) (Pt) (kN) Pt/PN (%)
S1 C1 4162 1377 33 2991 72 2313 56 No 10 285 234 No
C2 7962 1191 15 2490 31 2014 25 No
C3 7236 1319 18 2873 40 1996 28 No
C4 5690 1163 20 2722 48 1979 35 No
C5 5056 1116 22 1194 24 1075 21 No
S2 C1 4162 1892 45 3539 85 3064 74 No 18 473 440 No
C2 7962 2166 27 3619 45 3020 38 No
C3 7236 2439 34 4581 63 3270 45 No
C4 5690 2210 39 3757 66 2872 50 No
C5 5056 2159 43 2215 44 2040 40 No
S3 C1 4162 1892 45 1969 47 1844 44 No 18 452 394 No
C2 7962 2166 27 5817 73 4491 56 No
C3 7236 2439 34 2710 37 2479 34 No
C4 5690 2210 39 2309 41 2185 38 No
C5 5056 2159 43 2323 46 2169 43 No
S4 C1 4162 1892 45 4153 100 4 0 Yes 18 1 1 Yes
C2 7962 2166 27 7938 100 129 2 Yes
C3 7236 2439 34 7207 100 177 2 Yes
C4 5690 2210 39 5661 100 0 0 Yes
C5 5056 2159 43 5041 100 0 0 Yes
S5 C1 4162 1892 45 4149 100 238 6 Yes 18 1 1 Yes
C2 7962 2166 27 7954 100 306 4 Yes
C3 7236 2439 34 7222 100 7 0 Yes
C4 5690 2210 39 5673 100 0 0 Yes
C5 5056 2159 43 5051 100 19 0 Yes
S6 C1 4162 1892 45 2695 65 1885 45 No 18 105 60 No
C2 7962 2166 27 3584 45 2261 28 No
C3 7236 2439 34 4127 57 2741 38 No
C4 5690 2210 39 2274 40 2133 37 No
C5 5056 2159 43 2219 44 2155 43 No
S7 C1 4162 1892 45 1978 48 1859 45 No 18 115 65 No
C2 7962 2166 27 4342 55 2967 37 No
C3 7236 2439 34 2550 35 2430 34 No
C4 5690 2210 39 2550 45 2432 43 No
C5 5056 2159 43 2235 44 2163 43 No
S8 C1 4162 1892 45 2965 71 2326 56 No 18 135 84 No
C2 7962 2166 27 3099 39 2223 28 No
C3 7236 2439 34 2490 34 2432 34 No
C4 5690 2210 39 2490 44 2431 43 No
C5 5056 2159 43 2490 49 2431 48 No

7
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Fig. 10. Deformed shape of building at the end of analysis time for: (a) scenario S2; (b) scenario S3 (Units: mm, s).

Fig. 11. Deformed shape of building for scenario S4 at: (a) T = 3.1 s; (b) T = 3.5 s; (c) T = 4.0 s; (d) T = 5.0 s (Units: mm, s).

in any of the columns. Even though scenarios S2 and S3 involved For scenario S4 which involved the simultaneous removal of
multiple column losses, the progressive collapse was not predicted columns 3, 4, and 5 of the first story, progressive collapse potential
and the peak axial loads of the critical basement columns were in was predicted due to the large cantilever portion of the first floor of
the range of 44% to 85% of the axial load capacity for scenario S2, the building (size: 20.72 m in the longitudinal direction and 7.62 m
and 37% to 73% of the buckling load for scenario S3 (Table 5). This in the transverse direction) carrying all dead plus 25% of the full
also clarifies that frames of the Ohio Union building can redis- design live load. As seen in Fig. 14(a), basement columns C1, C2,
tribute the gravity load (self-weight + superimposed dead and C3 reached their buckling load at times between 3 s and 4 s,
load + 25% of full design live load) once multiple columns are however, columns C4 and C5 had their peak axial load at times
simultaneously or sequentially removed so that a cantilever part between 4 s and 5 s. Fig. 11 shows that some of the columns
of not more than 13.0 m span in any direction of the floor is started to buckle at 3.5 s and more columns buckled at 4 s. The
formed. For larger cantilever portions, the progressive collapse building finally collapsed at 5 s due to the buckling of all columns
was triggered as noted in scenarios S4 and S5. (see Fig. 11(d)).

8
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Fig. 12. Deformed shape of building for scenario S5 at: (a) T = 3.1 s; (b) T = 3.5 s; (c) T = 4.0 s; (d) T = 7.0 s (Units: mm, s).

Fig. 13. Deformed shape of building at the end of analysis time for: (a) scenario S6; (b) scenario S7; (c) scenario S8 (Units: mm, s).

In column loss scenario S5 which included simultaneous change of axial load for all critical columns under missing column
removal of first-story columns 3, 4, and 6, the progressive collapse scenario S5 is shown in Fig. 14(b). From the figure, it is noted that
was also predicted for the whole building, as seen in Fig. 12. This even though columns were removed at time 3 s, columns C1, C3,
was due to the large cantilever portion of the exterior panel (Size: and C4 reached their peak axial load at later times of approxi-
15.24 m in the transverse direction and 13.0 m in the longitudinal mately 6 s, 6.7 s, and 5 s, respectively. Whereas, basement columns
direction) carrying all dead plus 25% of the live load. The dynamic C2 and C5 had their peak axial load at earlier times of 3.3 s and
9
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Fig. 14. Time vs axial force for critical basement columns under column-loss scenarios: (a) S4; (b) S5.

3.5 s, respectively. The collapse of the whole building occurred at deflection is compared with the allowable limit of 1/360 of the
7 s, as seen in the deformed shape in Fig. 12(d), due to the buckling beam span as set by the ANSI/AISC 360-16 code [25], and it is also
of all columns. It is worth noting that for multiple column loss sce- compared with the limiting value Dlimit,GSA based on the acceptance
narios S4 and S5, similar findings were reported by Elsanadedy criteria for nonlinear analysis recommended in the 2003 GSA guide-
et al. [18] for NLD analysis of an existing steel-framed building lines [9]. The limiting deflection Dlimit,GSA is taken as the minimum
under two different blast threat scenarios. The blast-generated of 0.21Lb and 20Dv,max-LS, where Lb is the beam span and Dv,max-LS is
waves caused failure of multiple ground-story columns so that a the peak vertical deflection of the beam at the removed column
large cantilever portion of the first floor (>15 m in one direction) location as obtained from the linear static (LS) analysis of the build-
was formed. This, in turn, led to the progressive collapse of the ing under the same loading combination used in the NLD analysis. It
steel-framed building. should be noted that the Dv,max-LS values were obtained from the LS
For single missing column scenarios S6 to S8 (identified earlier), analysis of the Ohio Union building that will be discussed later in
Table 6 lists the key results of the NLD analysis for critical first- Sec. 5. The maximum connection rotation was compared with the
floor beams and beam-column connections attached to the acceptance limit of 0.025 rad, as recommended by the 2003 GSA
removed column. These results are for the sustained building after guidelines [9] for fully restrained steel beam-column connections.
the column-removal scenarios, and they are presented for the crit- It is evident from Table 6 that the location of the removed column
ical beams and connections in the long and short sides of the build- has a significant impact on the building deformation, and out of the
ing, as seen in Table 6. For critical beams, the key results are three missing column locations, removal of the corner column in
presented in terms of maximum vertical deflection and peak bend- scenario S8 was the most critical with regard to increasing the
ing moment. For critical beam-column connections, the maximum deformation of the building after column loss. The maximum rota-
rotation is listed. As per both ASCE 41-13 [46] and UFC guidelines tion of the critical beam-column connections after the removal of
[47], the connection rotation was taken as the chord rotation of the the corner column in scenario S8 was 17% to 50% more than that
beam member, and it was computed from for exterior column missing scenarios S6 and S7. As seen from
Table 6, the maximum vertical beam deflection for the three scenar-
Dv ;max
hmax ¼ ð2Þ ios S6 to S8 considerably exceeded the limit of the ANSI/AISC 360–
Lb
16 code [25], which indicates damage to the nonstructural elements
where Dv,max is the maximum vertical deflection at the removed such as claddings and plaster ceiling. However, the maximum beam
column location and Lb is the span of the beam connected to the deflections were significantly less than the acceptance limit of the
removed column. The key results presented in Table 6 are compared 2003 GSA guidelines. In addition, for all single missing column sce-
with the thresholds of codes and guidelines. The maximum beam narios, the peak bending moment in the critical beams was less

Table 6
Key NLD analysis results of critical beams and connections in the sustained building due to single missing column scenarios*.

Scenario Location of Long side of the building Short side of the building
missing
Beams Beam-column Beams Beam-column
column
connections connections
Dv,max Dlimit, Dlimit, Mmax Mp hmax hlimit, Dv,max Dlimit, Dlimit, Mmax Mp hmax hlimit,
(mm) AISC GSA (kN.m) (kN. (rad) GSA (mm) AISC GSA (kN.m) (kN. (rad) GSA
(mm) (mm) m) (rad) (mm) (mm) m) (rad)
S6 Exterior in the 105 21 1621 769 830 0.014 0.025 105 21 1600 1064 1100 0.014 0.025
long side
S7 Exterior in the 115 18 1365 726 1107 0.018 0.025 115 21 1600 786 1100 0.015 0.025
short side
S8 Corner 135 18 1365 782 1107 0.021 0.025 135 21 1600 976 1100 0.018 0.025

*Dv,max = maximum vertical displacement of beams connected to the removed column; Dlimit,AISC = allowable floor deflection as per ANSI/AISC 360-16 [25]; Dlimit,GSA =-
acceptance limit of vertical beam displacement as set in Table 2.1 of the 2003 GSA guidelines [9]; Mmax = peak bending moment in beams connected to the removed column;
Mp = plastic moment of beam section as per ANSI/AISC 360-16 [25], and it is calculated as fyZx, where fy = yield stress of steel = 345 MPa and Zx = plastic modulus of beam
section about the x-axis; hmax = maximum rotation of beam-column connection at the removed column location; hlimit,GSA = acceptance limit of connection rotation as set in
Table 2.1 of the 2003 GSA guidelines [9].

10
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

than the plastic moment calculated as per the ANSI/AISC 360-16 Table 7
code [25], which reveals the nondevelopment of plastic hinges in Missing first-story column scenarios used for LS analysis.

the critical beams connected to the removed columns. Moreover, Scenario Sequence of column removal* Applied load**
the maximum rotations of critical beam-column connections were Multiple column removal scenarios
less than the acceptance limit of the 2003 GSA guidelines [9]. Con- S1 Columns 1, 2, 3, and 4 simultaneously 2SW
clusively, from the perspective of the 2003 GSA guidelines, the S2 Columns 1, 2, 3 and 4 simultaneously 2(SW + SIDL + 0.25L)
building is deemed structurally adequate and satisfies the progres- S3 Columns 3, 6 and 7 simultaneously 2(SW + SIDL + 0.25L)
S4 Columns 3, 4 and 5 simultaneously 2(SW + SIDL + 0.25L)
sive collapse requirements.
S5 Columns 3, 4 and 6 simultaneously 2(SW + SIDL + 0.25L)
Based on the NLD analysis conducted in this study, it can be
Single column removal scenarios
concluded that single-column missing scenarios in steel-framed
S6 Column 2 only 2(SW + SIDL + 0.25L)
buildings may not cause their progressive collapse, and in build- S7 Column 6 only 2(SW + SIDL + 0.25L)
ings with typical framing layout, removing the corner column is S8 Column 3 only 2(SW + SIDL + 0.25L)
more critical than removing exterior columns with regard to
**SW = self-weight; SIDL = superimposed dead load; L = live load.
increasing the building deformation after column loss. This conclu-
sion is validated in a previous study of Chen et al. [26], who con-
ducted NLS and NLD analyses on a multistory steel-framed by taking a load factor of two for the load combination, the GSA
building under four different single missing ground-story column guidelines consider – in a simple way – the dynamic effect that hap-
scenarios using the load combination of the 2003 GSA guidelines pens once a vertical supporting member is suddenly lost from the
[9]. In their study, the column missing events involved three sce- building. Following the LS analysis, a demand-to-capacity ratio
narios for exterior columns in addition to one scenario for the cor- (DCR) is calculated for different steel frame members, using:
ner column. None of these scenarios triggered the progressive
collapse of the building, and with respect to the building deforma- Q UD
DCR ¼ ð4Þ
tion, the most critical scenario was the removal of the corner col- Q CE
umn. For the missing corner column scenario, the peak vertical
displacement of the first floor was about 162 mm in the NLD anal- where QUD is the demand force for acting loads; and QCE is the cor-
ysis. The corresponding maximum rotation in the critical beam- responding unfactored capacity. For calculating QCE, the resistance
column connection was about 0.02 rad, which is approximately (under capacity) factor is not used; however, the nominal strength
the same as that predicted in this study for the Ohio Union building of the material could be increased by 5% to 10% to represent the
under missing corner column scenario S8 (=0.021 rad as seen in actual material strength (5% adopted in the current study as per
Table 6). Ref. [18]). The collapse of a member is assumed to occur when
the DCR value exceeds 2.0, which conforms to the 2003 GSA
guidelines.
5. Linear static (LS) analysis
Demand-to-capacity ratios (DCR values) were computed for the
eight column-loss scenarios, outlined previously. Based on the first
In addition to the NLD analysis procedure discussed previously,
iteration of the LS analysis, listed in Table 8 are demand-to-
the simpler LS analysis method of the 2003 GSA guidelines [9] was
capacity ratios (DCR values) for columns and beams. For beams,
employed for progressive collapse assessment of the Ohio Union
the ratios are based on the demands in flexure and shear, whereas,
building, and then a comparison was made between the two meth-
for columns, these are based on the interaction of axial force and
ods. In the LS analysis, the building was simulated with the help of
bending moment. As clarified from Table 8, the GSA acceptance cri-
ETABS software [48]. Steel columns and beams were simulated
teria of DCR  2.0 are satisfied for almost all beams and columns of
with frame elements, whereas shell elements were utilized to rep-
the single-column removal scenarios S6, S7, and S8.
resent the RC slabs. Similar to the NLD model of LS-DYNA, masonry
As per the GSA guidelines, only two beams failed in flexure for
walls were not incorporated in the model. Table 7 enlists the miss-
scenario S6; whereas, for scenario S7, failure was predicted for two
ing first-story column scenarios used for the LS analysis. It included
columns and two beams in flexure. For scenario S8, two beams in
five multiple column loss scenarios (S1 to S5) in addition to three
flexure and one column failed. However, for scenarios S6, S7, and
single-column removal scenarios (S6, S7, and S8) as seen in Table 7.
S8, the DCR values are more than 1.0 for 13 columns (at various
Unlike the NLD analysis, in which columns in scenarios S1 and
stories), which, indicates a low progressive collapse risk [49]. In
S2 were sequentially removed, columns in the LS analysis were
the second iteration of the LS analysis in which failed columns
simultaneously removed. Even though multiple column-loss sce-
and beams were removed from the ETABS model, no further failure
narios were not required in the GSA guidelines for progressive col-
was predicted for any of the members of the building. This con-
lapse assessment, they were studied in the LS analysis for
firms the low progressive collapse risk of the building. For
comparison with the NLD analysis. However, the three single-
multiple-column loss scenarios, DCR values exceed 2.0 for 8 to 18
column loss events are in line with the recommendations of the
columns and exceed 1.0 for 30 to 73 columns as seen in Table 8,
GSA guidelines for the progressive collapse assessment of multi-
which indicates a high potential for progressive collapse. This
story buildings. These guidelines require the removal of an exterior
was confirmed in the second cycle of analysis for scenarios S1 to
column located: (a) close to the middle of the long direction (Col-
S5 in which all failed beams and columns were physically removed
umn 2 (Scenario S6)), (b) close to the middle of the short direction
from the model. In this iteration, many more beams and columns
(Column 6 (Scenario S7)), or (c) on the corner (Column 3 (Scenario
were found to have DCR values exceeding 2.0 and the building is
S8)). As per the 2003 GSA guidelines [9], the load employed in the
thus considered to have a high risk of progressive collapse and
LS analysis was given by:
no further iterations were required.

2SW for scenario S1 In brief, as per the 2003 GSA guidelines, the steel-framed build-
Load ¼ ð3Þ ing shows low progressive collapse risk if a single column is lost as
2ðSW þ SIDL þ 0:25LÞ for scenarios S2 to S8
per scenarios S6, S7, and S8. Nevertheless, in case multiple column
where SW = self-weight; SIDL = superimposed dead load; and loss is invoked as per scenarios S1 to S5, the building may have a
L = live load. As mentioned before, the live load is reduced by 75% high risk of progressive collapse. This conclusion contradicts the
acknowledging that the full live load is less likely. In the meantime, results of the actual progressive collapse field experiments
11
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Table 8
DCR values for steel frame elements (based on LS analysis results after 1st iteration).

Scenario Beams Columns Potential for progressive


collapse
Flexure Shear P-M interaction
S1 DCRmax = 2.815 DCRmax = 0.615 DCRmax = 2.256 High
DCR > 1.0 for 43 beams DCR > 1.0 for 30 columns
DCR > 2.0 for 3 beams DCR > 2.0 for 8 columns
S2 DCRmax = 3.747 DCRmax = 0.806 DCRmax = 2.934 High
DCR > 1.0 for 125 beams DCR > 1.0 for 55 columns
DCR > 2.0 for 12 beams DCR > 2.0 for 18 columns
S3 DCRmax = 7.266 DCRmax = 1.161 DCRmax = 2.799 High
DCR > 1.0 for 201 beams DCR > 1.0 for 72 columns
DCR > 2.0 for 15 beams DCR > 2.0 for 9 columns
S4 DCRmax = 2.83 DCRmax = 0.724 DCRmax = 2.9 High
DCR > 1.0 for 63 beams DCR > 1.0 for 50 columns
DCR > 2.0 for 12 beams DCR > 2.0 for 13 columns
S5 DCRmax = 8.182 DCRmax = 1.248 DCRmax = 5.078 High
DCR > 1.0 for 143 beams DCR > 1.0 for 73 columns
DCR > 2.0 for 17 beams DCR > 2.0 for 17 columns
S6 DCRmax = 2.337 DCRmax = 0.517 DCRmax = 1.479 Low
DCR > 1.0 for 20 beams DCR > 1.0 for 13 columns
DCR > 2.0 for 2 beams DCR > 2.0 for zero columns
S7 DCRmax = 3.486 DCRmax = 0.497 DCRmax = 2.14 Low
DCR > 1.0 for 9 beams DCR > 1.0 for 13 columns
DCR > 2.0 for 2 beams DCR > 2.0 for 2 columns
S8 DCRmax = 3.836 DCRmax = 0.632 DCRmax = 2.135 Low
DCR > 1.0 for 16 beams DCR > 1.0 for 13 columns
DCR > 2.0 for 2 beams DCR > 2.0 for one column

conducted for the building before its demolition as reported else- mLIF ¼ 4:3  0:00327d ð6Þ
where [35,38]. In the field tests, the progressive collapse was not
triggered due to the sequential column removal scenario S1 (see where d = beam depth (in mm). It should be noted that in the above
Fig. 1); nevertheless, a high risk of progressive collapse was pre- equation, the mLIF factor is the smallest of any primary beam that is
dicted from the LS analysis. This is a major difference from the directly connected to the removed column. For nonlinear static
NLD analysis in which progressive collapse potential was high for analysis of steel frames, the UFC guidelines [47] use the following
scenarios that only involved simultaneous removal of columns so equation to compute the DIF for both force- and deformation-
that a large cantilever portion of the floor slab (having a span in controlled actions.
either direction exceeding 15.0 m) is formed (scenarios S4 and
0:76
S5). For all other scenarios, as per the NLD analysis, the progressive DIF ¼ 1:08 þ ð7Þ
hpra =hy þ 0:83
collapse was not anticipated and failure was not predicted in any of
the columns as detailed previously. In conclusion, the LS analysis where hpra is the plastic rotation angle, which depends on the type
may overestimate the progressive collapse risk of steel buildings and flexibility of the beam-column connection as given by the
in case of multiple-column removal scenarios. This overestimation acceptance criteria of the guidelines, and hy is the yield rotation that
in the LS analysis results may be attributed to the factor of 2.0 used corresponds to the effective stiffness of the beam members. Tsai
in Eq. (3) to represent the dynamic effect owing to instantaneous and Lin [14] proposed a force-based definition in which the DIF is
column removal. expressed as the ratio between the static resisting force and
dynamic resisting force to the same displacement demand. The
nonlinear static approach led to a conservative estimate for the col-
6. Assessment of dynamic increase factor (DIF)
lapse resistance if a DIF of 2.0 was used. Ruth et al. [50] concluded
that a load factor of 1.5 can capture the dynamic effects when static
Several formulations are available in the literature for estimat-
analyses are adopted, as the load factor of 2.0 may overestimate the
ing the dynamic increase factor (DIF) used in the static analysis. As
dynamic effects. Rezvani et al. [51] also concluded that the DIF for
mentioned previously, the 2003 GSA guidelines [9] used a DIF of
steel moment-resisting frames susceptible to progressive collapse
2.0 to simply account for the dynamic effect that happens once a
under column removal event is at most 1.13. Brunesi and Parisi
vertical supporting member is suddenly lost from the building. In
[52] carried out nonlinear regression analysis to derive simplified
the UFC guidelines [47], all actions in structural members are clas-
formulas for prediction of the DIF to be used in the pushdown anal-
sified as either deformation-controlled or force-controlled actions.
ysis of both RC framed buildings designed only to gravity loads, as
Examples of deformation- and force-controlled actions, respec-
per EC2 [53], and RC earthquake-resistant buildings designed to
tively, include bending moment in beams and axial force in col-
comply with EC8 [54]. For example, Brunesi and Parisi [52] pro-
umns. For linear static analysis, the UFC guidelines use a DIF of
posed the following equation to predict the DIF for 3D modeling
2.0 for all force-controlled actions; however, for deformation-
of EC2-conforming RC buildings.
controlled actions of steel frames, the DIF is estimated from

DIF ¼ 0:9mLIF þ 1:1 ð5Þ DIF ¼ 15:68D0:62


v ð8Þ

where mLIF is a factor calculated from the following formula for where Dv is the vertical displacement (in mm) of the beam-column
welded unreinforced flange (WUF) moment connections. joint located on top of the removed column.
12
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

In this study, the LS analysis was repeated for the case of gravity Peak axial load of columns from NLD analysis
DIF FCA ¼ ð9Þ
loads without dynamic amplification factor and the results were Peak axial load of columns from LS analysis
compared with those of the NLD analysis to come up with more
The DIFFCA values are listed in Table 9 for columns of each sce-
representative values for the dynamic increase factor (DIF) for
nario. As illustrated, these values range from 1.02 to 1.67 with an
the cases of force- and deformation-controlled actions. The peak
average value (m) of 1.22 and a standard deviation (r) of 0.21.
axial load of basement columns was selected as an example for
Thus, a more reasonable value for DIFFCA may be recommended
force-controlled actions, whereas the peak bending moment in
as 1.5, which lies in between (m + r) and the maximum value,
critical first-floor beams – connected to the removed columns –
as listed in Table 9. For assessment of the dynamic increase factor
was chosen as an example for deformation-controlled action. For
for deformation-controlled actions (DIFDCA), the peak bending
scenarios S6 to S8, peak axial load was calculated for critical base-
moments in the critical first-floor beams were obtained from the
ment columns C1 to C5 after the single column removal scenario,
FE analysis at four different locations after the single column
as shown in Table 9 for both analysis types. As seen in Table 9,
removal scenarios S6 to S8, as seen in Table 10. These locations
the dynamic increase factor for force-controlled actions (DIFFCA)
included positive and negative moments in the critical beams of
was estimated for each column from:

Table 9
Estimation of dynamic increase factor for force-controlled actions (DIFFCA) in cases of single missing column scenarios.

Scenario Applied load in model* Column designation Peak axial load after column-loss scenario (kN) DIFFCA
NLD analysis LS analysis
(LS-DYNA model) (ETABS model)
S6 SW + SIDL + 0.25L C1 2695 1897 1.42
C2 3584 2151 1.67
C3 4127 2919 1.41
C4 2274 2166 1.05
C5 2219 2136 1.04
S7 SW + SIDL + 0.25L C1 1978 1881 1.05
C2 4342 2857 1.52
C3 2550 2437 1.05
C4 2550 2193 1.16
C5 2235 2143 1.04
S8 SW + SIDL + 0.25L C1 2965 2384 1.24
C2 3099 2287 1.35
C3 2490 2437 1.02
C4 2490 2193 1.14
C5 2490 2144 1.16
Mean 1.22
Standard deviation 0.21
Minimum 1.02
Maximum 1.67

*SW = self-weight; SIDL = superimposed dead load; L = live load.

Table 10
Estimation of dynamic increase factor for deformation-controlled actions (DIFDCA) in cases of single missing column scenarios.

Scenario Applied load in model* Location of maximum moment in critical beams Maximum beam moment after DIFDCA
column-loss scenario (kN.m)
NLD LS
S6 SW + SIDL + 0.25L Positive moment in the long side 611 357 1.71
Negative moment in the long side 769 547 1.41
Positive moment in the short side 394 203 1.94
Negative moment in the short side 1064 659 1.61
S7 SW + SIDL + 0.25L Positive moment in the long side 359 216 1.66
Negative moment in the long side 726 516 1.41
Positive moment in the short side 786 495 1.59
Negative moment in the short side 360 246 1.46
S8 SW + SIDL + 0.25L Positive moment in the long side 380 228 1.67
Negative moment in the long side 782 620 1.26
Positive moment in the short side 256 132 1.94
Negative moment in the short side 976 669 1.46
Mean 1.59
Standard deviation 0.21
Minimum 1.26
Maximum 1.94

*SW = self-weight; SIDL = superimposed dead load; L = live load.

13
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

Table 11
Comparison of dynamic increase factors proposed in this study with available literature.

Scenario Dynamic increase factor for force-controlled actions (DIFFCA) Dynamic increase factor for deformation-controlled actions (DIFDCA)
Proposed in this 2003 GSA UFC guidelines Ruth et al. Proposed in this 2003 GSA UFC guidelines Ruth et al.
study guidelines [9] [47] [50] study guidelines [9] [47] [50]
S6 1.5 2 2 1.5 1.8 2 3.18 1.5
S7 1.5 2 2 1.5 1.8 2 3.17 1.5
S8 1.5 2 2 1.5 1.8 2 3.17 1.5

the long and short sides of the building. As illustrated in Table 10, 4. Based on the NLD analysis, the Ohio Union building has a high
the dynamic increase factor for deformation-controlled actions progressive collapse risk under dead load plus 25% of the live
(DIFDCA) was calculated from: load in scenarios involving simultaneous removal of three col-
umns so that a large cantilever portion of the floor slab (having
Peak bending moment of beams from NLD analysis
DIF DCA ¼ a span in either direction exceeding 15.0 m) is formed.
Peak bending moment of beams from LS analysis
5. It should be noted that in the multiple column loss scenarios S4
ð10Þ and S5 of the NLD analysis, the failure load of basement col-
For every single column missing event, Table 10 lists the DIFDCA umns C1 to C5 were almost the same as their critical buckling
values for critical beams of each scenario. As seen from the table, loads estimated from the FE analysis considering the case of
these values vary from 1.26 to 1.94, with an average value (m) of pure axial compression. This means that eccentricity introduced
1.59 and a standard deviation (r) of 0.21. Therefore, a more conve- in the columns from their framing with beams was insignificant
nient value of 1.80 is suggested in this study for DIFDCA. As seen to cause a noticeable reduction in the critical buckling load.
from Table 10, the recommended value lies in between (m + r) 6. The simplified LS analysis of the 2003 GSA guidelines overesti-
and the maximum value. mated the risk of progressive collapse for the Ohio Union build-
For single missing column scenarios S6 to S8, Table 11 depicts a ing under the sequential column loss scenario followed in the
comparison between the DIF values proposed in this study and field experiments. The LS analysis anticipated high progressive
those available in the literature. It is identified that both the collapse potential contrary to the field tests in which progres-
2003 GSA and UFC guidelines overestimated the dynamic effects sive collapse did not occur and none of the columns failed.
for both force- and deformation-controlled actions, and hence 7. Based on a comparison between the NLD and LS analyses car-
the LS analysis of the two guidelines may give a conservative esti- ried out for the Ohio Union building under different single col-
mation for the collapse resistance. The recommended value for umn loss scenarios, it was found out that the dynamic increase
force-controlled actions comes in line with the work of Ruth factors (DIF) suggested by both the 2003 GSA and UFC guideli-
et al. [50], who concluded that a load factor of 1.5 can capture nes as gravity load multipliers (accounting for the dynamic
the dynamic effects when static analyses are adopted. However, effect due to instantaneous column loss) may give a conserva-
the DIF value of Ruth et al. [50] underestimated the dynamic effects tive estimation for the collapse resistance and thus they can
for deformation-controlled actions. be reduced. Dynamic increase factors of 1.5 and 1.8 are recom-
mended in this study for force- and deformation-controlled
actions, respectively.
7. Conclusions

NLD and LS analyses were conducted on a multi-story steel-


framed building that existed on the campus of Ohio State Univer- Declaration of Competing Interest
sity to study its potential for progressive collapse under multiple
and single column-loss scenarios. In both analyses, the load combi- The authors declare that they have no known competing finan-
nations of the 2003 GSA guidelines were used. The results of the cial interests or personal relationships that could have appeared
two analyses were compared and new dynamic increase factors to influence the work reported in this paper.
were recommended for both force- and deformation-controlled
actions. The prime conclusions of this research are as follows: Acknowledgments

1. The simplified nonlinear FE model employed in this study using The authors are grateful to the Deanship of Scientific Research,
beam elements was found accurate in predicting the critical King Saud University, for funding through Vice Deanship of Scien-
buckling load of axially loaded steel columns tested in literature tific Research Chairs. The second author’s research visits to the
with errors of prediction ranging from 0.6% to 5.8%. King Saud University in 2017 and 2018 was funded by a Fulbright
2. Compared with the simplified nonlinear FE model, the equa- grant through the U.S. Department of State and King Saud Univer-
tions of the ANSI/AISC 360-16 used to calculate the critical sity. This support is greatly acknowledged.
buckling load of axially loaded steel columns were found con-
servative by giving peak loads that were 13–20% less than those References
predicted by the FE model.
[1] W. McGuire, Prevention of progressive collapse, Proc. Regional Conf. on Tall
3. The nonlinear dynamic analysis (NLD) procedure followed in Buildings Bangkok, Thailand, 1974.
this study was proven to be accurate in predicting the progres- [2] E.V. Leyendecker, B.R. Ellingwood, Design methods for reducing the risk of
sive collapse risk of the Ohio Union steel building due to the progressive collapse in buildings, Building Science Series No. 98; National
Bureau of Standards, Washington, D.C., USA, 1977.
sequential column removal scenario under its self-weight only.
[3] W.G. Corley, P.F.M. Sr, M.A. Sozen, C.H. Thornton, The Oklahoma City bombing:
As per both the progressive collapse field tests and the NLD Summary and recommendations for multihazard mitigation, J. Perfor. Constr.
analysis, the building did not progressively collapse and none Facil. 12 (3) (1998) 100–112.
[4] Z.P. Bazant, Y. Zhou, Why did the world trade center collapse?—Simple
of its columns failed after four exterior first-story columns were
analysis, J. Eng. Mech. 128 (1) (2002) 2–6.
sequentially removed.

14
H. Elsanadedy, H. Sezen, H. Abbas et al. Engineering Science and Technology, an International Journal 35 (2022) 101193

[5] Z.P. Bazant, M. Verdure, Mechanics of progressive collapse: Learning from [30] A.H. Ebrahimi, M.E. Jamkhaneh, M.S. Amiri, 3D finite-element analysis of steel
World Trade Center and building demolitions, J. Eng. Mech. 133 (3) (2007) moment frames including long-span entrance by strengthening steel cables
308–319. and diagonal concentrically braced frames under progressive collapse, Pract.
[6] Z.P. Bazant, J.L. Le, F.R. Greening, D.B. Benson, What did and did not cause Period. Struct. Des. Constr. 23 (4) (2018) 04018025.
collapse of World Trade Center twin towers in New York?, J Eng. Mech. 134 [31] M. Alrubaidi, H. Elsanadedy, H. Abbas, T. Almusallam, Y. Al-Salloum,
(10) (2008) 892–906. Investigation of different steel intermediate moment frame connections
[7] K.A. Seffen, Progressive collapse of the world trade center: simple analysis, J. under column-loss scenario, Thin-Walled Struct. 154 (2020) 106875.
Eng. Mech. 134 (2) (2008) 125–132. [32] TEC-2007, Specification for buildings to be built in disaster areas, Ministry of
[8] US Department of Defense (DoD), Unified facilities criteria (UFC), DoD Public Works and Settlement, Ankara, Turkey, 2007.
Minimum antiterrorism standards for buildings, Washington, DC: [33] D. Rodríguez, E. Brunesi, R. Nascimbene, Fragility and sensitivity analysis of
Department of Defense, UFC 4-010-01, US Army Corps of Eng.; p. 31, 2002. steel frames with bolted-angle connections under progressive collapse, Eng.
[9] General Services Administration (GSA), Progressive collapse analysis and Struct. 228 (2021) 111508.
design guidelines for new federal office buildings and major modernization [34] H. Elsanadedy, M. Alrubaidi, H. Abbas, T. Almusallam, Y. Al-Salloum,
projects, Washington, DC: Office of Chief Architect, 2003. Progressive collapse risk of 2D and 3D steel-frame assemblies having shear
[10] S. Marjanishvili, E. Agnew, Comparison of various procedures for progressive connections, J. Constr. Steel Res. 179 (2021) 106533.
collapse analysis, J. Perfor. Constr. Facil. 20 (4) (2006) 365–374. [35] B.I. Song, H. Sezen, Experimental and analytical progressive collapse
[11] F. Fu, Progressive collapse analysis of high-rise building with 3-D finite assessment of a steel frame building, Eng. Struct. 56 (2013) 664–672.
element modeling method, J. Constr. Steel Res. 65 (6) (2009) 1269–1278. [36] B.I. Song, K.A. Giriunas, H. Sezen, Progressive collapse testing and analysis of a
[12] K. Khandelwal, S. El-Tawil, F. Sadek, Progressive collapse analysis of steel frame building, J. Constr. Steel Res. 94 (2014) 76–83.
seismically designed steel braced frames, J. Constr. Steel Res. 65 (3) (2009) [37] E. Akah, C. Wood, K. Ii, H. Sezen, Experimental investigation of a steel-framed
699–708. building for disproportionate collapse, Eng. J-Amer. Inst. Steel Constr. 55 (3)
[13] G. Powel, Progressive collapse: Case studies using nonlinear analysis, Proc. of (2018) 143–159.
Metropolis and Beyond-Structures Congress ASCE, Reston, VA, 2005. [38] B.I. Song, Experimental and Analytical Assessment on the Progressive Potential
[14] M.H. Tsai, B.H. Lin, Investigation of progressive collapse resistance and of Existing Buildings, Master’s Thesis, The Ohio State University, Ohio, 2010, p.
inelastic response for an earthquake-resistant RC building subjected to 125.
column failure, Eng. Struct. 30 (12) (2008) 3619–3628. [39] AISC, Manual of steel construction, (6th Ed.), American Institute of Steel
[15] A.G. Vlassis, B.A. Izzuddin, A.Y. Elghazouli, D.A. Nethercot, Progressive collapse Construction, Chicago, IL, USA, 1969.
of multi-storey buildings due to sudden column loss—Part II: Application, Eng. [40] Livermore Software Technology Corporation (LSTC), LS-DYNA keyword user’s
Struct. 30 (5) (2008) 1424–1438. manual – Vol. 1-3, Version R11, LSTC, Livermore, CA, 2018.
[16] B.A. Izzuddin, A.G. Vlassis, A.Y. Elghazouli, D.A. Nethercot, Progressive collapse [41] L. Marques, A. Taras, L.S. da Silva, R. Greiner, C. Rebelo, Development of a
of multi-storey buildings due to sudden column loss—Part I: Simplified consistent buckling design procedure for tapered columns, J. Constr. Steel Res.
assessment framework, Eng. Struct. 30 (5) (2008) 1308–1318. 72 (2012) 61–74.
[17] A. Naji, F. Irani, Progressive collapse analysis of steel frames: Simplified [42] K.J.R. Rasmussen, G.J. Hancock, Tests of high strength steel columns, J. Constr.
procedure and explicit expression for dynamic increase factor, Int. J Steel Steel Res. 34 (1) (1995) 27–52.
Struct. 12 (4) (2012) 537–549. [43] T.J. Li, G.Q. Li, S.L. Chan, Y.B. Wang, Behavior of Q690 high-strength steel
[18] H.M. Elsanadedy, T.H. Almusallam, Y.R. Alharbi, Y.A. Al-Salloum, H. Abbas, columns: Part 1: Experimental investigation, J. Constr. Steel Res. 123 (2016)
Progressive collapse potential of a typical steel building due to blast attacks, J. 18–30.
Constr. Steel Res. 101 (2014) 143–157. [44] T. Belytschko, J.I. Lin, T. Chen-Shyh, Explicit algorithms for the nonlinear
[19] E. Talebi, M.M. Tahir, F. Zahmatkesh, A.B. Kueh, Comparative study on the dynamics of shells, Comp. Methods Appl. Mech. Eng. 42 (2) (1984) 225–251.
behaviour of buckling restrained braced frames at fire, J. Constr. Steel Res. 102 [45] J.D. Stevenson, Structural damping values as a function of dynamic response
(2014) 1–12. stress and deformation levels, Nucl. Eng. Des. 60 (2) (1980) 211–237.
[20] E. Talebi, M.M. Tahir, F. Zahmatkesh, A.B. Kueh, A.M. Said, Fire resistance of a [46] American Society of Civil Engineers (ASCE), Seismic rehabilitation of existing
damaged building employing buckling restrained braced system, Adv. Steel buildings, Standard No. ASCE/SEI 41-13, Reston, Virginia, USA, 2013.
Constr. 14 (1) (2018) 1–21. [47] Department of Defense (DoD), Unified Facilities Criteria (UFC): UFC 4-023-03 -
[21] M. Alrubaidi, H. Abbas, H. Elsanadedy, T. Almusallam, R. Iqbal, Y. Al-Salloum, Design of buildings to resist progressive collapse, Washington DC, USA, Change
Experimental and FE study on strengthened steel beam-column joints for 3, 2016.
progressive collapse robustness under column-loss event, Eng. Struct. 258 [48] ETABS Version 17.0.1, Extended 3D analysis of building systems, Computers
(2022) 114103. and Structures, Inc. (CSI), Berkeley, CA, USA, 2018.
[22] S. Shan, S. Li, Fire-induced progressive collapse mechanisms of steel frames [49] A. Marchis, M. Botez, A.M. Ioani, Vulnerability to progressive collapse of
with partial infill walls, Structures 25 (2020) 347–359. seismically designed reinforced concrete framed structures in Romania, In
[23] Z. Guo, Z. Xu, H. Lin, Z. Zheng, H. Su, J. Chen, Y. Chen, Experimental tests of Proc. of the 15th World Conf. on Earthquake Engineering (WCEE), Lisbon,
post-fire beam-column assemblies with WUF-B connections against Portugal, 2012.
progressive collapse, J. Constr. Steel Res. 192 (2022) 107249. [50] P. Ruth, K.A. Marchand, E.B. Williamson, Static equivalency in progressive
[24] A. Hadidi, R. Jasour, A. Rafiee, On the progressive collapse resistant optimal collapse alternate path analysis: reducing conservatism while retaining
seismic design of steel frames, Struct. Eng. Mech. 60 (5) (2016) 761–779. structural integrity, J. Perform. Constr. Facil. 20 (4) (2006) 349–364.
[25] ANSI/AISC 360-16, Specifications for Structural Steel Buildings, American [51] F.H. Rezvani, A.E. Jeffers, B. Asgarian, H.R. Ronagh, Effect of column loss
Institute of Steel Construction, Chicago, IL, USA, 2016. location on structural response of a generic steel moment resisting frame,
[26] C.H. Chen, Y.F. Zhu, Y. Yao, Y. Huang, Progressive collapse analysis of steel Steel Comp. Struct. 25 (2) (2017) 217–229.
frame structure based on the energy principle, Steel Comp. Struct. 21 (3) [52] E. Brunesi, F. Parisi, Progressive collapse fragility models of European
(2016) 553–571. reinforced concrete framed buildings based on pushdown analysis, Eng.
[27] M.A. Zoghi, M. Mirtaheri, Progressive collapse analysis of steel building Struct. 152 (2017) 579–596.
considering effects of infill panels, Struct. Eng. Mech. 59 (1) (2016) 59–82. [53] CEN, European Committee for Standardization. EN 1992-1-1. Eurocode 2
[28] H. Qiao, Y. Yang, J. Zhang, Progressive collapse analysis of multistory moment (EC2): Design of concrete structures – Part 1. 1: General rules and rules for
frames with varying mechanisms, J. Perform. Constr. Facil. 32 (4) (2018) buildings. Brussels; 2004.
04018043. [54] CEN, European Committee for Standardization. EN 1998-1-5. Eurocode 8
[29] A. Naji, Progressive collapse analysis of steel moment frames: an energy-based (EC8): Design of structures for earthquake resistance – Part 1. 5: Specific rules
method and explicit expressions for capacity curves, J. Perform. Constr. Facil. for concrete buildings. Brussels, 2004.
33 (2) (2019) 04019008.

15

You might also like