You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301346219

Uncertainty and Sensitivity Analysis of Reinforced Concrete Frame Structures


Subjected to Column Loss

Article in Journal of Performance of Constructed Facilities · February 2017


DOI: 10.1061/(ASCE)CF.1943-5509.0000930

CITATIONS READS

33 1,277

4 authors:

Xiaohui Yu Da-Gang Lu
Harbin Institute of Technology Harbin Institute of Technology
77 PUBLICATIONS 401 CITATIONS 206 PUBLICATIONS 1,106 CITATIONS

SEE PROFILE SEE PROFILE

Qian Kai Bing Li


Guilin Institute of Technology Nanyang Technological University
87 PUBLICATIONS 1,644 CITATIONS 194 PUBLICATIONS 4,406 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Life-cycle performance, reliability and design of structures View project

Seismic Fragility, Vulnerability and Risk of Civil Engineering Structures View project

All content following this page was uploaded by Qian Kai on 30 October 2018.

The user has requested enhancement of the downloaded file.


Uncertainty and Sensitivity Analysis of Reinforced
Concrete Frame Structures Subjected to Column Loss
Xiao Hui Yu, Ph.D. 1; Da Gang Lu, Ph.D. 2; Kai Qian, Ph.D., M.ASCE 3; and Bing Li, Ph.D., M.ASCE 4

Abstract: In this study, the variability inherent in the performance of RC frame structures to bridge over a column loss was investigated,
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

considering the uncertainties existing in gravity loads, material properties, and construction geometries. Uncertainty and sensitivity analysis
were conducted in tandem to provide the influence and significance of the uncertain parameters on RC frames to resist progressive collapse.
Quasi-static pushdown analysis was used to assess the residual load-resisting capacity of structures with an initialized damage. A set of
structural models were generated for uncertainty analysis by using the correlation-reduced Latin hypercube sampling method. Sensitivity
analysis was performed by independently varying parameters with one standard deviation away from their means. Two typical RC frame
structures with different span aspect ratios were taken as the study cases. The macromodeling technique, which was validated by the test
results, was used for analysis. The study results indicated that structural uncertainties have significant effects on the behavior of RC frame
buildings for mitigating progressive collapse caused by the loss of a ground column. Among the studied uncertain parameters, the gravity
loads and the properties of reinforcement have the most effects on the residual load-resisting capacities of damaged frames. DOI: 10.1061/
(ASCE)CF.1943-5509.0000930. © 2016 American Society of Civil Engineers.
Author keywords: Progressive collapse; Pushdown analysis; Uncertainty analysis; Sensitivity analysis; Column loss; RC frame building.

Introduction occurs, as illustrated by the past events, such as the collapse of


Murrah Federal Building in 1995 and World Trade Centers in 2001.
Progressive collapse is defined by ASCE/SEI 7 (ASCE/SEI 2010) With increasing prevalence of terrorism, the probability of struc-
as “the spread of an initial local failure from element to element, tures attacked with explosives or other violent means are increas-
which eventually results in the collapse of an entire structure or a ing, and designing buildings to resist progressive collapse became a
disproportionately large part of it.” Generally, the initial damages pressing and urgent need in engineering community.
that could trigger disproportionate collapse are often caused by an During the building design process, there are numerous sources
abnormal load (accident event), such as gas explosions, blast, of uncertainties, such as environment loads, structural strength,
vehicular collisions, tornados, or other extreme loads (Ellingwood the approximations in finite-element models, and limitations in
et al. 2007). These extreme loads, however, are not routinely con- supporting databases (Ellingwood et al. 2007). These unavoidable
sidered in normal design. Thus, the buildings may suffer severe uncertainties do affect the progressive collapse resistance of build-
damages or even progressive collapse when they are unfortunately ings and should be dealt with in a probabilistic manner (Ellingwood
subjected to such abnormal loads. Progressive collapse is a low- and Dusenberry 2005; Ellingwood 2006; Ellingwood et al. 2007).
probability event as it requires both the occurrence of an abnormal Recently, the importance of investigating the probabilistic behavior
load to trigger initial damages, and the structure lacks of robustness of buildings against progressive collapse has been emphasized
to absorb the energy released because of the initial damage (DoD by several researchers (Park and Kim 2010; Szyniszewski 2010;
2013). For instance, the full-scale test results showed that two Xu and Ellingwood 2011; Kim et al. 2011; Le and Xue 2014).
ground columns can be successfully removed from the steel frame However, to date, the studies regarding progressive collapse are
buildings without causing progressive collapse (Song and Sezen overwhelmingly deterministic (El-Tawil et al. 2013). Without an
2013; Morone and Sezen 2014; Song et al. 2014). However, cata- elaborate probabilistic analysis of progressive collapse, a reliable
strophic consequences can be triggered once progressive collapse judgment cannot be made on whether the existing deterministic
studies would really capture the collapse extent. This merits further
1
Assistant Professor, Key Lab of Structural Dynamic Behavior and
studies to fill the gap between the growing demand of suitable man-
Control of the Ministry of Education, Harbin Institute of Technology, agement of the risk of progressive collapse and the available lack
Harbin 150090, China. E-mail: xiaohui.yu@hit.edu.cn of information about the inherent probabilistic characteristic in the
2
Professor, Key Lab of Structural Dynamic Behavior and Control of the performance of the buildings to mitigate progressive collapse. To
Ministry of Education, Harbin Institute of Technology, Harbin 150090, do this, it is important to understand the influence and significance
China. E-mail: ludagang@hit.edu.cn of the uncertain parameters on the performance of structures sub-
3
Professor, College of Civil Engineering and Architecture, Guangxi jected to column loss. Unfortunately, there is no related information
Univ., Nanning 530004, China. E-mail: qiankai@gxu.edu.cn readily for RC buildings.
4
Associate Professor, School of Civil and Environmental Engineering, The objective of this study is to investigate the variability inher-
Nanyang Technological Univ., Singapore 639798 (corresponding author).
ent in the behavior of RC frame buildings to resist progressive
E-mail: cbli@ntu.edu.sg
Note. This manuscript was submitted on July 23, 2015; approved on collapse. To achieve this objective, uncertainty and sensitivity anal-
April 13, 2016; published online on July 5, 2016. Discussion period open ysis are run in tandem, in which the former has a great focus on
until December 5, 2016; separate discussions must be submitted for indi- uncertainty quantification and propagation and the latter is used to
vidual papers. This paper is part of the Journal of Performance of Con- identify the important uncertain parameters. The quasi-static push-
structed Facilities, © ASCE, ISSN 0887-3828. down analysis (Khandelwal and El-Tawil 2011; Qian et al. 2015b)

© ASCE 04016069-1 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


α ×(1.2DL+0.5LL) α ×(1.2DL+0.5LL)
1.2DL+0.5LL 1.2DL+0.5LL

Δ Δ

removed removed
(a) (b)
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Displacement-controlled pushdown procedure: (a) removal of an exterior column; (b) removal of an interior column

is adopted to analyze the residual capacity of RC frame buildings 3.0


Linear Nonlinear Losing capacity
subjected to a column loss. As an effective tool, the conventional
(deterministic) quasi-static pushdown method (called “pushdown” 2.5
hereafter) has been applied to RC frame buildings by numerical
2.0

Load factor, α
simulation (Lu et al. 2011; Fascetti et al. 2015) and experimental
testing (Sasani et al. 2007; Yi et al. 2008; Qian and Li 2012b, a,
1.5
2013; Qian et al. 2015a). However, the deterministic approach is
not suitable for a probabilistic analysis. Thus, it is extended by cou- 1.0
pling a set of structural models with the random parameters that are
Pushdown curve
generated with a correlation-controlled Latin hypercube sampling 0.5 DC-I
(CLHS) technique. Seventeen parameters including gravity loads, DC-II
material properties, and construction geometries are selected for 0.0
0 200 400 600 800 1000 1200
uncertainty analysis in this paper.
Displacement, Δ /mm

Fig. 2. Typical pushdown curve and the considered damage criteria


Analysis Methodology
identified on it

Pushdown Analysis and Damage Criteria


Pushdown analysis starts by removing the critical members and
static procedure (NSP) of the alternative path method (APM); Ωn ×
then investigates the complete behavior of the damaged structure
ð1.2 × DL þ 0.5 × LLÞ (GSA 2013; DoD 2013). A maximum load
by increasing the gravity loads proportionally until system collapse.
factor that is less than Ωn implies that the structure may collapse as
Only the gravity loads on the damaged bays are raised, whereas
a result of a column loss.
the remaining parts of the damaged structure are subjected to nomi-
From the pushdown curve, two damage criteria (DC) are defined
nal gravity loads. The gravity load is the combination of dead
to describe the failure states of the frames. DC-I is defined as first
load (DL) and live load (LL), and the nominal combination is
yielding of steel rebar in RC beams, whereas DC-II is defined as
determined as 1.2 × DL þ 0.5 × LL (GSA 2013; DoD 2013). A
reaching the ultimate load-resisting capacity. Fig. 2 illustrates a typ-
displacement-controlled procedure is used by applying an incre-
mental vertical displacement Δ downward of the node, in which the ical pushdown curve with the aforementioned two damage criteria.
column is removed at first and the corresponding applied gravity DC-I marks the first plastic-hinge formulation in beams. Beyond
load in the bays that suffered damages is recorded, as shown DC-II, further increasing the displacement will decrease the load-
in Fig. 1. resisting capacity.
The recorded gravity load is represented as α × ð1.2 × DL þ
0.5 × LLÞ, where α = load factor calculated as the ratio between Uncertainty Analysis Method
the recorded gravity load corresponding to Δ and the nominal grav-
ity load. As suggested in DoD (2013), in normal service condition, For the uncertainty analysis, a total of N var ¼ 17 uncertain param-
the pressure applied should be 1.2 × DL þ 0.5 × LL, and the value eters including gravity loads, material properties, and reinforce-
of α equals 1.0. However, as a quasi-static pushdown loading re- ment details and dimensions are selected on the basis of the
gime (no dynamic effects are included) is used in this study, the available literature (Ellingwood et al. 1980; Mirza et al. 1979;
load-resisting capacity is varied with increasing the vertical dis- Mirza and MacGregor 1979b, a). To deal with the parameter un-
placement at the lost column. Thus, α is varied with Δ, and the certainties, the Latin hypercube sampling (LHS) technique is used.
load-displacement relationship in coordinates of Δ versus α is used The LHS technique uses stratification of the probability distribution
as the “pushdown curve.” Using α to represent the load-resisting functions of random variables and requires much less of simula-
capacity could help judge whether the building can bridge over the tions compared with the original type of Monte Carlo simulation
column loss more easily. In particular, the load factor α equal to the (MCS). However, the standard LHS may bring undesired spurious
dynamic increase factor (DIF) Ωn corresponds to the state in which correlation into the sample scheme and result in unfavorable effects
the applied vertical load reaches the load specified in the nonlinear on the statistical results. To avoid this, the CLHS technique is used.

© ASCE 04016069-2 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


For details on CLHS, please refer to Olsson and Sandberg (2002). variables are represented by three levels: the base value and the
It was demonstrated that the CLHS with 100 samples could provide upper and lower bounds corresponding to the means and 1 SD
better estimation than the standard LHS with 1,000 samples beyond and below the means, respectively. Pushdown analyses are
(Olsson and Sandberg 2002). Therefore, the CLHS with N sam ¼ performed for the structural samples by solely varying each param-
100 samples is used in this study. By varying the structural param- eter at its lower and upper bounds, whereas the other variables are
eters with the random samples, a population of N sam possible in- held constant at their base values. The absolute difference between
stances of the structure is created. Then, the random pushdown the responses corresponding to the parameter at its upper and
curves are collected by conducting pushdown analysis for each of lower bounds are used to measure the sensitivity of the parameter.
the structural samples. The load factors corresponding to DC-I and A special bar chart widely known as “tornado diagram” (Kim et al.
DC-II are identified on these curves. The random pushdown curves, 2011) is used to demonstrate the importance of each parameter.
together with the identified load factors, eventually form the data The importance of the parameter is determined by the width of
pool for statistics. Through summarizing the random pushdown the bar (swing). For each bar, the ends of the bar are calculated
curves, the variation of α-capacity with the growing Δ is calcu- according to the lower and upper bounds of the parameter. Thus,
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

lated, which is valuable to provide insight on the variability of the a total of 2 × N var þ 1 structural samples are generated for sensi-
whole pushdown process. The identified α-capacity samples cor- tivity analysis, where 2 × N var are conducted at the lower and
responding to different DC are fitted with suitable distributions to upper bounds of parameters, whereas the additional 1 is conducted
depict their probabilistic properties. The probabilistic performance at their means.
of the structure with a column removed can be reliably estimated,
assuming that a large enough number of structures have been
sampled. Analysis Procedures
Fig. 3 illustrates the methodologies and main steps in this study.
As shown in the figure, structural modeling and definition of
Sensitivity Analysis Method
uncertainty parameters are the basic and first steps. For each case
The objective of the sensitivity analysis is to investigate the per- with a specific initial damage, a group of N sam þ 2 × N var þ 1 ¼
formance variability of the damaged structure caused by varying 135 structural models must be generated. As two RC frame struc-
uncertain parameters. To achieve this goal, a simple method based tures with two different initial damages are examined, a total of
on the probability theory is adopted. In this method, the random 4 × 135 ¼ 540 of pushdown analyses were carried out.

Sampling of random variables for


uncertainty analysis
(Nsam samples) Development of
Selection of Nvar deterministic
random variables Sampling of random variables for Structural model
sensitivity analysis
(2×Nvar+1 samples)

Development of random
structural models
(Nsam+2×Nvar+1 models)

Pushdown analysis
(Nsam+2×Nvar+1 cycles)

Uncertainty analysis Sensitivity analysis


(Using Nsam pushdown curves) (Using 2×Nvar+1 pushdown curves)

Pushdown uncertainty Pushdown sensitivity

DC uncertainty DC sensitivity

Fig. 3. Analysis steps of uncertainty and sensitivity evaluation under pushdown analysis

© ASCE 04016069-3 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


Subassembly for test
(Specimen P2)
Subassembly for test

7@3300

7@3300
(Specimen P1)

3600

3600
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

4@8400 4@6000

(a) (b)

4#9

2#10
4#9 4#10

2#10
#4 hoops@140
720

#4 hoops@140
560

600
#4 hoops@120
4#9 4#9
4#10
400 600
Section A-A Section B-B Section C-C
(c)

Fig. 4. Frame elevation and the typical reinforcement details of the case study frame buildings (also shown is frame subassembly used in validation
study) (mm): (a) elevation of LF; (b) elevation of TF; (c) reinforcement arrangement

Case Study Buildings and Analytical Modeling compressive strength of concrete (f c ¼ 27.6 MPa); Young modu-
lus of concrete (Ec ¼ 24.3 GPa); yield strength of reinforcement
Characteristics of the Case Study Buildings (fy ¼ 413.8 MPa); and Young modulus of reinforcement (Es ¼
200 GPa). All the design characteristics of the case buildings are
Two RC moment-resisting frames with different span aspect ratios summarized in Table 1.
were selected as the case study buildings designed according to
ACI 318-08 (ACI 2008). The design dead loads distributed on the
floor and superimposed because of the ceiling were assumed as 5.2
and 1.0 kN=m2 , respectively. The design live load was assumed to Table 1. Design Characteristics of the Case Buildings
be 3.0 kN=m2 . Seismic forces were also considered in the design Characteristics Descriptions
with an assumption of a low-seismicity zone having a D-class site
Dead loads distributed on 5.2 kN=m2
and a stiff soil profile. The design spectral response acceleration the floor
parameter at 1 s is 0.34 g. Dead loads superimposed 1.0 kN=m2
Fig. 4 shows the frame elevation and the typical reinforcement due to ceiling
in beams and columns, in which TF and LF are used to represent Live load 1.0 kN=m2
the frames with span length of 6,000 and 8,400 mm, respectively. Site condition D-class site stiff soil profile
An identical elevation is used for both TF and LF. The height of Spectral response 0.34 g at 1 s
the first floor is 3,600 mm, and that of floors above is 3,300 mm. acceleration parameter
The frames have the same column section (see section C-C), Bay width 6,000 mm for TF
8,400 mm for LF
whereas LF has a larger beam section (see section A-A) than TF Story height 3,600 mm for the first floor
(see section B-B) because of its larger beam span. Two substruc- 3,300 mm for the floors above
tures taken from TF and LF have been tested at a quarter scale by Dimensions and reinforcement Section C-C for the columns
Qian et al. (2015a) to examine the load-carrying mechanism of details (Fig. 4) Section A-A for the beams of LF
RC buildings in resisting progressive collapse. The test data pro- Section B-B for the beams of TF
vide the opportunity to validate the structural modeling. The sub- Concrete properties f c ¼ 27.6 MPa (4 ksi)
structures from LF and TF are labeled as P1 and P2 (also shown Ec ¼ 24.3 GPa (3,600 ksi)
in Fig. 4), respectively, to be consistent with Qian et al. (2015a). Steel properties f y ¼ 413.8 MPa (60 ksi)
Es ¼ 200 GPa (29,000 ksi)
The nominal properties of the materials used in design are

© ASCE 04016069-4 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


Analytical Modeling membrane action developed in slabs could significantly reduce
the likelihood of progressive collapse (Qian and Li 2012b; Qian
As mentioned previously, 540 cycles of pushdown analyses are
et al. 2015a).
conducted in this study. It is considerably computation intensive;
therefore, it is not feasible to generate models on the basis of a high-
fidelity finite-element (FE) modeling approach. Recently, a macro- Validation of Analytical Modeling
model strategy based on the open-source software OpenSees had
The critical responses of the beam-column subassemblies (P1 and
been proven to achieve similar accuracy to the high-fidelity FE
P2) measured from the quarter-scale test models were used for
models (Bao et al. 2008; Kazemi-Moghaddam and Sasani 2015;
validation. In these tests, the middle column beneath the beam
Li and Sasani 2015; Fascetti et al. 2015). Compared with the con-
was preremoved, and a pushdown test was carried out by impos-
tinuum FE model, the macromodel could cut down the computation
ing an increasing displacement at the top of the middle column.
burden significantly. Thus, the macromodel approach based on
The edge columns of the specimen were designed to fix on the
OpenSees, which is similar to that used in Fascetti et al. (2015),
strong legs to equivalently simulate the fixed boundary conditions.
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

was adopted to develop the structural model.


To create a symmetric failure mode for the test specimens, a steel
In the macromodel, several displacement-based fiber elements
assembly was installed below the hydraulic jack to ensure concen-
were used to discretize the beams and columns, each of which
tric loading. The center-to-center spans of specimens P1 and P2
includes five integration points. The corotational transformation
are 2,100 and 1,500 mm, respectively. Moreover, the cross-
(Felippa 2000) was used to exactly account for the second-order sectional dimensions of specimens P1 and P2 are 180 × 9 and
effect (P − δ effect). Each RC section was discretized into the 140 × 90 mm, respectively. Fig. 6 shows the dimensions and
fibers of the steel reinforcement bars and the confined and uncon- reinforcement details of the test specimens, in which R6 denotes
fined concrete. The well-known nonlinear constitutive law by Kent the plain round bar with a 6-mm diameter, and T10 denotes the
and Park (1971) was used to define the uniaxial nonlinear behav- deformed bar with a 10-mm diameter. Because of space limitation,
ior of concrete material (it is the “Concrete02 Material” on the the details of the tests are not provided. However, it can be found
OpenSees library). The stress-strain envelop properties of the con- in Qian et al. (2015a). Fig. 7 shows the macromodel scheme of the
fined concrete was determined by Mander’s model (Mander et al. substructures with the pin supports at the outer sides. The simu-
1988). The material used to define the steel reinforcing bars is “hys- lated force versus displacement responses are compared with those
teretic material,” which allows the definition of trilinear behavior. from the experiments, as shown in Fig. 8. It can be seen that the
The first branch is the elastic stage, the second one is the hardening simulation results agree well with the test results, proving the reli-
stage with a hardening ratio of 0.01, and the final branch is defined ability of the adopted macromodeling technique. The sharp drops
with a negative slope to account for bar fracture. A displacement- shown in the load-displacement curves of the test results are attrib-
controlled integrator scheme was used to perform pushdown uted to the fracture of rebars at the interfaces between the mid-
analysis until the structure loses stability, whereas the nonlinear column and beams. As the fracture of the rebar is not always
equations involved in the solution of the problems were dealt with simultaneous, the drop of the resistance in P1 and P2 has some
the Krylov-Newton algorithm (Scott and Fenves 2010). kind of differences.
For initiation of progressive collapse, one ground exterior col-
umn or interior column was notionally removed, as shown in Fig. 5.
For simplicity, TF-E, TF-I, LF-E, and LF-I are used hereafter to Uncertainty Consideration
represent the cases of removing exterior and interior columns for
TF and LF, respectively. In this model, only catenary and compres- Seventeen uncertain parameters were selected as random variables,
sive arch actions are included, whereas the effects of slab are not which can be categorized into four groups: (1) gravity loads, in-
considered. The available test data have shown that the tensile cluding dead loads of DLf and DLr on floor and roof, respectively,

α×(1.2DL+0.5LL) 1.2DL+0.5LL α×(1.2DL+0.5LL)


1.2DL+0.5LL 1.2DL+0.5LL

(a) (b)

Fig. 5. Location of a removed column: (a) removal of an exterior column; (b) removal of an interior column

© ASCE 04016069-5 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Dimensions and reinforcement details of the test substructures (mm): (a) construction geometry of P1; (b) reinforcement arrangement of P1;
(c) construction geometry of P2; (d) reinforcement arrangement of P2

Displa cement-based fiber element 80


Nodes to joint adjuncct elements 70
Load 60

Force/kN
50
Column removal
40
Stress

Ultimate 30
height
Yield 20
Strain
width

y 10
Fracture Experiment
Simulation
0
0 50 100 150 200 250 300 350 400
Compression Steel bar
z Fiber section (a) Displacement/mm

80
Stress

Tension
70
Maximum
60
Strain
Force/kN

Crushing 50

40
Maximum Concrete
30

Fig. 7. Macromodeling schemes used in OpenSees and the application 20


to the simulation of the tested substructures 10 Experiment
Simulation
0
0 50 100 150 200 250 300
and live load on the floor LL; (2) properties of concrete material, (b) Displacement/mm
including compressive strength fc , tensile strength ft , and elastic
Fig. 8. Comparison of experiment and simulation results: (a) P1; (b) P2
modulus Ec ; (3) properties of reinforcement materials, including
yield strengths f y and fyt for longitudinal reinforcement and trans-
verse stirrups, respectively, ultimate strength f u for longitudinal
reinforcement, and elastic modulus Es ; and (4) construction geom- live load (Ellingwood et al. 1980). The sustained live load, also
etries, including span length L, beam width Bb , beam height Hb , called “arbitrary-point-in-time live load,” remains relatively con-
concrete cover t, longitudinal reinforcement areas As;Beam and stant within a particular occupancy containing, for example, the
As;Column in beams and columns, respectively, and transverse stirrup weight of furniture and heavy equipment, whereas the extraordi-
area Ast . The size of the column section was not included, as col- nary live load, also called “intermittent live load” (JCSS 2001),
umns contribute less than beams to the performance to resist commonly has small duration, arising from gathering of people,
progressive collapse. Table 2 summarizes the statistical properties for example, a crowded room during special events. In this study,
of the considered variables and the references used to quantify each only the variability in the sustained live load is accounted for be-
of the variables. The total live loads on a floor area can be divided cause the loss of a column can occur anytime during the life of the
into two components: the sustained live load and the extraordinary structure.

© ASCE 04016069-6 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


Table 2. Summary of the Considered Uncertainty Parameters
Uncertainty Coefficient of
Category parameters Mean (or mediana) variation (or SD) Distributions References
Gravity loads DLf 1.05 × designed ¼ 5.50 kN=m2 0.10 Normal Ellingwood et al. (1980)
DLr 1.05 × designed ¼ 1.05 kN=m2
LL 12 psf ¼ 0.575 kN=m2 0.4 Beta Ellingwood et al. (1980)
Concrete fc 3,390 psi = 23.4 MPa 0.18 Normal Ellingwood et al. (1980)
properties ft 339 psi = 2.34 MPa 0.18 Normal
Ec 1.0 × designed ¼ 24,264 MPa 0.077 Normal Mirza et al. (1979)
Reinforcement fy 71.0 ksi = 490 MPa 0.093 Beta Mirza and
properties f y;t 48.8 ksi = 337 MPa 0.107 Beta MacGregor (1979b)
fu 110.8 ksi = 764.1 MPa 0.08 Beta
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

Es 29,000 ksi = 200,000 MPa 0.033 Normal


Construction Bb Bb þ 0.11 in: ¼ 323b or 403 mmc 0.144 in: ≈ 4 mmd Normal Mirza and
geometries Hb H b − 1=8 in: ¼ 557b or 717 mmc 1=4in: ≈ 7 mmd Normal MacGregor (1979a)
L 1.0 × designed ¼ 6,000 mmb or 8,400 mmc 11=16 in: ≈ 18 mmd Normal
As;beam 1.01 × designed ¼ 647.70 mm2 0.04 Modified lognormale
As;Column 1.01 × designed ¼ 799.6 mm2
As;t 1.01 × designed ¼ 128.0 mm2
t 0.8 in. = 21 mm 0.145 × 21 ≈ 3 mmd Normal Lee and Mosalam (2004)
a
Lognormal variables.
b
Dimensions in TF.
c
Dimensions in LF.
d
SD.
e
The accepted modification constant is 0.91, which indicates that logðAs =designed − 0.91Þ is normally distributed.

Generally, it is not easy to obtain accurate and complete corre- Δ and assigning them with the percentiles of 100 × ð0.5=N sam Þ;
lation relationships among different random variables because of 100 × ð1.0=N sam Þ; : : : ; and 100 × ½ðN sam − 0.5Þ=N sam . The dis-
the lack of knowledge and data. Therefore, only the correlation be- persion, β αjΔ , of the conditional α-capacity on a given Δ is then
tween f c and Ec was considered by a coefficient of 0.8 (Lee and determined by ðα84% 16%
Δ − αΔ Þ=2 and shown in Fig. 10, in which
16% 84%
Mosalam 2004), and the other parameters were assumed as inde- αΔ and αΔ are 16 and 84% fractile α-capacities, respectively.
pendent of each other. This partial correlation was involved into the The aforementioned method to summarize random pushdown
samples by Nataf transformation (Liu and Der Kiureghian 1986). curves is inspired by that used in incremental dynamic analysis
The spatial variability of material and dimensional properties is not (IDA) for statistical analysis of IDA curves (Vamvatsikos and
accounted for during random sampling with an implicit assumption Cornell 2002). It can be seen that the variation of pushdown re-
that all the random parameters are spatially fully correlated over the sponse is very small initially, whereas it is promoted quickly with
same structure. the increase of the vertical displacement. At a large displacement
stage, the value of β αjΔ can even go beyond 0.4. Such a large level
of variability reveals that it is unreasonable and unsafe to neglect
Uncertainty Analysis the effect of structural uncertainties for progressive collapse analy-
sis because it will underestimate the failure potential conditioned
Pushdown Uncertainty on large deformation. At the descending part of the pushdown
curves, some structural samples have lost their load-resisting capac-
By using 100 samples from the parameter distributions, the struc- ity. As a consequence, the fractile curves in this stage could include
tural model set is developed. Random pushdown curves are then reasonable error because of reduction of available data for statistics.
generated by performing analyses for each structural model from This is the main reason why β αjΔ does not increase significantly
the model set, as shown in Fig. 9. For pushdown analysis, a unified after the peak resistance.
target displacement, i.e., Δ ¼ 1,800 mm, is applied to the node To draw more reliable conclusions from the random pushdown
beyond the miss column for random structural models. Because of curves, the statistical pushdown curves in terms of the mean and 1
the variability of structural models, some structural models lose SD above and below the mean are also estimated and compared
their load-resisting capacity before reaching the target displace- with the fractile pushdown curves as shown in Fig. 11. The 16,
ment, whereas others survive. Moreover, a significant variation in 50, and 84% fractile curves are practically coincident with the
the α-capacity conditioned on a given Δ is apparent. For example, mean - SD, mean and mean þ SD curves, respectively. It is illus-
as shown in Fig. 9(d), the maximum values of α-capacity are vary- trated that structural uncertainty does not cause a biased predic-
ing within the interval between 1.5 and 3.0. Similar observations tion of mean response. Thus, a normal distribution becomes a
are also made in the other cases. These observations reveal the natural option to depict the probabilistic property of pushdown
significant effects of structural uncertainties on the load-resisting performance.
capacity. Thus, it is very important to examine the random nature
in pushdown curves and bring in the probabilistic view in pro-
DC Uncertainty
gressive collapse studies.
The 16, 50, and 84% fractile pushdown curves are derived (also After the pushdown curves are studied, the next step is to
plotted in Fig. 9) by sorting the conditioned α-capacity on a given confirm the effects of structural uncertainty on the specific DC.

© ASCE 04016069-7 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


3.5 3.5
3 3
2.5 2.5

Load factor, α

Load factor, α
2 2

1.5 1.5

1 1
Random pushdown curves Random pushdown curves
0.5 16% and 84% fractile curves 0.5 16% and 84% fractile curves
50% fractile curve 50% fractile curve
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
(a) Displacement, Δ/mm (b) Displacement, Δ/mm
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

3 4

2.5
3
Load factor, α

Load factor, α
2

1.5 2

1
Random pushdown curves 1 Random pushdown curves
0.5 16% and 84% fractile curves 16% and 84% fractile curves
50% fractile curve 50% fractile curve
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
(c) Displacement, Δ/mm (d) Displacement, Δ/mm

Fig. 9. Random pushdown curves shown against the 16, 50, and 84% fractile curves of α given Δ: (a) TF-E; (b) TF-I; (c) LF-E; (d) LF-I

0.5 capacity; thus, it is unreasonable to ignore the effects of the


TF-E structural uncertainties for design buildings to resist progressive
TF-I collapse.
0.4
LF-E On the basis of the probability distributions of the α-capacity
LF-I corresponding to DC-II, a further step is taken to re-examine the
βα | Δ

0.3 DIF under a probabilistic perspective. For RC frame structures,


the DIF is determined by Ωn ¼ 1.04 þ 0.45=ðθpra =θy þ 0.48Þ,
0.2 where θpra and θy = plastic and elastic rotation angles of the
elements, components, and connections (GSA 2013; DoD 2013).
0.1 According to this equation, the DIFs are calculated as 1.41 and 1.43
for TF and LF, respectively. According to the fitted normal distri-
butions of DC-II, the failure probabilities are then calculated as 3.0,
0.0
0 300 600 900 1200 1500 1800 3.0, 16, and 1% for the cases of TF-E, TF-I, LF-E, and LF-I, re-
Displacement, Δ/mm spectively. Except for LF-E, the remaining three cases achieved
very limited failure probabilities. Thus, it can be concluded that the
Fig. 10. Comparison of β αjΔ for different scenarios of column removal TF-E, TF-I, and LF-I are very likely to survive from a sudden re-
moval of a first-story column. For the case of LF-E, the failure
probability is not within the accepted level.

The α-capacities for DC-I and DC-II are determined from each of
the random pushdown curves, as shown in Fig. 12. The correspond- Sensitivity Analysis
ing empirical failure probabilities are then calculated and idealized
by a continuous normal distribution, as shown in Fig. 13. It is ap-
parent that the idealized normal distributions fit the empirical prob- Pushdown Sensitivity
abilities well. Table 3 provides the fitted distribution parameters. Pushdown sensitivity analyses are performed by testing each un-
It is found that the median (50%) α-capacities from statistics are certainty parameter independently with 1 SD above or below its
almost identical with the means from fitting. This is expected be- mean. The sensitivity is then evaluated from the pushdown curves,
cause the mean and 50% fractile pushdown curves are quite close as shown in Fig. 14. The curve with the mean parameters (base-case
to each other (Fig. 11). As indicated in Table 3, DC–II has a mean curve) always appears as a solid line, whereas the curves corre-
α-capacity that is approximately 1.5–2.0 times larger than that of sponding to upgraded and degraded parameter are distributed
DC-I. As DC-I and DC-II are defined as the elastic limit and the around its sides. Because of limitation in paper length, only the
plastic limit, respectively, the observation implies that the mean of results of case TF-I are presented. Similar observations were ob-
the safety margin ratio ranges from 1.5 to 2.0 for the studied frames. tained for remaining cases.
The SDs for DC-I and DC-II are approximately 0.2 and 0.4, respec- As shown in the figure, the parameters of DLf , LL, f y , fu , Hb ,
tively. This illustrates the significant variability in load-resisting and As;Beam have relatively large effects, whereas the rest of the

© ASCE 04016069-8 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


3 3

2.5 2.5

Load factor, α

Load factor, α
2 2

1.5 1.5

1 Mean curve 1 Mean curve


50% fractile curve 50% fractile curve
0.5 Mean±SD 0.5 Mean±SD
16% and 84% fractile curves 16% and 84% fractile curves
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
Displacement, Δ/mm Displacement, Δ/mm
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

3 3.5

2.5 3

2.5
Load factor, α

Load factor, α
2
2
1.5
1.5
1 Mean curve Mean curve
50% fractile curve 1 50% fractile curve
0.5 Mean±SD 0.5 Mean±SD
16% and 84% fractile curves 16% and 84% fractile curves
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
(c) Displacement, Δ/mm (d) Displacement, Δ/mm

Fig. 11. Comparison of statistical pushdown curves based on the lognormal assumption and the normal assumption: (a) TF-E; (b) TF-I; (c) LF-E;
(d) LF-I

3.5 3.5
3 3

2.5 2.5
Load factor, α
Load factor, α

2 2

1.5 1.5
1 1
Random pushdown curve Random pushdown curve
0.5 DC-I 0.5 DC-I
DC-II DC-II
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
(a) Displacement, Δ/mm (b) Displacement, Δ/mm

3 4

2.5
3
Load factor, α
Load factor, α

1.5 2

1
Random pushdown curve 1 Random pushdown curve
0.5 DC-I DC-I
DC-II DC-II
0 0
0 300 600 900 1200 1500 1800 0 300 600 900 1200 1500 1800
(c) Displacement, Δ/mm (d) Displacement, Δ/mm

Fig. 12. Random pushdown curves conditioned on different scenarios of column removal together with the identified DC: (a) TF-E; (b) TF-I;
(c) LF-E; (d) LF-I

© ASCE 04016069-9 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


1 1

0.8 0.8

Failure probability
Failure probability
0.6 0.6

0.4 0.4
DC-I: Empirical DC-I: Empirical
0.2 DC-II: Empirical 0.2 DC-II: Empirical
DC-I: Fitted DC-I: Fitted
DC-II: Fitted DC-II: Fitted
0 0
0.5 1 1.5 2 2.5 3 3.5 0.5 1 1.5 2 2.5 3 3.5
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

(a) Load factor, α (b) Load factor, α

1 1

0.8 0.8

Failure probability
Failure probability

0.6 0.6

0.4 0.4
DC-I: Empirical DC-I: Empirical
DC-II: Empirical DC-II: Empirical
0.2 0.2
DC-I: Fitted DC-I: Fitted
DC-II: Fitted DC-II: Fitted
0 0
0.5 1 1.5 2 2.5 3 3.5 0.5 1 1.5 2 2.5 3 3.5 4
(c) Load factor, α (d) Load factor, α

Fig. 13. Empirical probability function of DS and the fitted normal distributions: (a) TF-E; (b) TF-I; (c) LF-E; (d) LF-I

parameters show very limited influences as these three curves are the means could increase the ultimate deformation of structure
almost identical. As expected, the gravity loads have impressive significantly, which agrees with the results from Bao et al. (2008)
effects on pushdown results as shown in Figs. 14(a and c). The dead and Qian and Li (2013) well. In their studies, increasing the amount
load on the roof, DLr , shows negligible influence [Fig. 14(b)] as it of beam transverse reinforcement ratio will considerably increase
accounts for a very limited percentage of the total dead loads. the ultimate deformation capacity of the RC frames to resist pro-
Compared with DLf , the live load on the floor, LL, shows less gressive collapse.
influence [Fig. 14(b)] because only the variation in the sustained
load component is considered. It is predictable that f y has signifi-
cant effects initially. However, its effects decrease with increasing DC Sensitivity
displacement at the stage until peak α-capacity, in which catenary The sensitivity of each uncertain parameter on specific DC is
action becomes the main load-resisting mechanism, and are con- investigated by the tornado diagrams, as shown in Figs. 15 and 16.
trolled by the ultimate strength of the reinforcement fu . As shown As illustrated in the figures, the parameters of DLf and f y dem-
in Fig. 14(j), the effects of f u further confirmed the prediction. onstrated their importance on DC-I, whereas DLf and f u have the
Moreover, the height H b and the reinforcement area As;Beam of most influences for DC-II. Furthermore, the effect of As;Beam
beams also show considerable influences. on both DC is also considerable. It is expected that DLf is the
Some of the remaining parameters apart from the aforemen- most influential parameter because of its close association with
tioned show significant clear effects on the ultimate deformation α-capacity. Because DLr and LL only account for a small part of
capacity of the frames. For instance, as shown in Figs. 14(h and q), the total gravity loads, they show less significance than DLf . As
increasing the stirrup-related parameters fyt and Ast to 1 SD above DC-I is defined by the first yielding of steel bars in beams, fy
shows significant effect. f u becomes significantly influential for
DC-II because it contributes largely to the peak load-resisting
Table 3. Moments from Fitting the Empirical Probabilistic Distributions of capacity. fc shows considerable effect on DC-I placed in line to
DC-I and DC-II As;Beam because f c is a key factor for the compressive arch action
DC-I DC-II that is the main load-carrying mechanism corresponding to DC-I.
When it comes to DC-II, however, fc has a narrow swing because
Scenarios Median Mean SD Median Mean SD
the catenary action becomes the dominant load-carrying mecha-
TF-E 1.30 1.30 0.22 2.09 2.07 0.36 nism where concrete plays a negligible role.
TF-I 1.05 1.06 0.16 2.17 2.13 0.37 The tornado diagrams corresponding to different DC and differ-
LF-E 1.25 1.24 0.18 1.87 1.81 0.40 ent studied frames also appear different. In particular, the tornado
LF-I 1.39 1.40 0.21 2.59 2.55 0.46
diagrams seem more symmetric for DC-I compared with that for

© ASCE 04016069-10 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


3 2.5 2.5
2 2

Load factor, α
Load factor, α

Load factor, α
2
1.5 1.5

Mean-std 1 Mean-std 1 Mean-std


1
Mean 0.5 Mean 0.5 Mean
Mean+std Mean+std Mean+std
0 0 0
0 300 600 900 1200 0 300 600 900 1200 0 300 600 900 1200
(a) Displacement, Δ/mm (b) Displacement, Δ/mm (c) Displacement, Δ/mm

2.5 2.5 2.5


2 2 2
Load factor, α

Load factor, α

Load factor, α
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

1.5 1.5 1.5


1 Mean-std 1 Mean-std 1 Mean-std
0.5 Mean 0.5 Mean 0.5 Mean
Mean+std Mean+std Mean+std
0 0 0
0 300 600 900 1200 0 300 600 900 1200 0 300 600 900 1200
(d) Displacement, Δ/mm (e) Displacement, Δ/mm (f) Displacement, Δ/mm

2.5 2.5 2.5


2 2 2

Load factor, α
Load factor, α

Load factor, α

1.5 1.5 1.5


1 Mean-std 1 Mean-std 1 Mean-std
0.5 Mean 0.5 Mean 0.5 Mean
Mean+std Mean+std Mean+std
0 0 0
0 300 600 900 1200 0 300 600 900 1200 0 300 600 900 1200
(g) Displacement, Δ/mm (h) Displacement, Δ/mm (i) Displacement, Δ/mm

2.5 2.5 2.5


2 2 2
Load factor, α

Load factor, α
Load factor, α

1.5 1.5 1.5


1 Mean-std 1 Mean-std 1 Mean-std
0.5 Mean 0.5 Mean 0.5 Mean
Mean+std Mean+std Mean+std
0 0 0
0 300 600 900 1200 0 300 600 900 1200 0 300 600 900 1200
(j) Displacement, Δ/mm (k) Displacement, Δ/mm (l) Displacement, Δ/mm

2.5 2.5 2.5


2 2 2
Load factor, α

Load factor, α
Load factor, α

1.5 1.5 1.5


1 Mean-std 1 Mean-std 1 Mean-std
0.5 Mean Mean Mean
0.5 0.5
Mean+std Mean+std Mean+std
0 0 0
0 300 600 900 1200 0 300 600 900 1200 0 300 600 900 1200
(m) Displacement, Δ/mm (n) Displacement, Δ/mm (o) Displacement, Δ/mm

2.5 2.5
2 2
Load factor, α
Load factor, α

1.5 1.5
1 Mean-std 1 Mean-std
0.5 Mean 0.5 Mean
Mean+std Mean+std
0 0
0 300 600 900 1200 0 300 600 900 1200
(p) Displacement, Δ/mm (q) Displacement, Δ/mm

Fig. 14. Sensitivity of the pushdown curves to the considered uncertain parameters: (a) DLf ; (b) DLrf ; (c) LL; (d) f c ; (e) f t ; (f) Ec ; (g) f y ; (h) f yt ;
(i) Es ; (j) f u ; (k) Bb ; (l) Hb ; (m) L; (n) t; (o) As;Beam ; (p) As;Column ; (q) Ast

© ASCE 04016069-11 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


DLf DLf
fy fy
As,Beam As,Beam
LL fc
Hb LL
fc t
t Hb
Ec L
L Bb
Es DLr
Ast As,Col
fyt Es
Bb fyt
DL Ast
As,Colr Mean variables ft Mean variables
ft Ec
fu fu

1.2 1.3 1.4 1.5 0.9 1.0 1.1 1.2 1.3


Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

Load factor, α Load factor, α


(a) (b)

DLf DLf
fy fy
As,Beam As,Beam
LL fc
fc t
t LL
Hb Hb
Ec Ec
L L
fyt fyt
Es Es
As,Col As,Col
DLr DLr
Bb Bb
ft
Mean variables
Mean variables ft
fu fu
Ast Ast
1.1 1.2 1.3 1.4 1.5 1.2 1.3 1.4 1.5 1.6 1.7
Load factor, α Load factor, α
(c) (d)

Fig. 15. Tornado diagrams for DC-I: (a) TF-E; (b) TF-I; (c) LF-E; (d) LF-I

DC-II. Moreover, especially for DC-II, the bars in the tornado performance of RC frame structures subjected to a column loss
diagrams for LF are more skewed compared with that for TF. have been examined by means of uncertainty and sensitivity anal-
The skewness of bars does not affect the significance of the uncer- ysis, respectively. The following general conclusions have been
tain parameters because the significance is only decided by the drawn:
length of bars. Nevertheless, the skewness of bars reflects different 1. The dispersion of random pushdown curves could go beyond
influences of the uncertain parameters on the performance of the 0.4. The standard deviations of α-capacity corresponding to
studied frames under different missing-column scenarios. On the DC-I and DC-II were approximately 0.2 and 0.4, respectively.
basis of the obtained probabilistic distributions for DC-I and The considerable variability on their responses reveals the sig-
DC-II (Table 3), for the same studied frame, the variability inherent nificant effects of structural uncertainty on the performance of
in DC-I is smaller than that existing in DC-II, thereby leading to a RC frame structures against progressive collapse. A normal dis-
more symmetric tornado diagram. LF is found to be more signifi- tribution was proven suitable to depict the probabilistic charac-
cantly influenced by uncertainties than TF because, conditioned on teristics of the behavior of the damaged structure with an initial
the same DC, the fitted SD values for LF are larger than that for TF loss of column.
(Table 3). This may partially contribute to the asymmetry observed 2. The DIF defined for the case study frames according to the
in the tornado diagrams for LF. To sum up, in addition to the un- guidelines (GSA 2013; DoD 2013) were less than the mean
certainty parameters, the damage state, the structural design, and α-capacities of DC-II. Moreover, from the fitted normal distri-
the missing-column scenarios totally impose influence on the re- bution of DC-II, the conditional failure probability are com-
sults of sensitivity analysis. The exact relationship between them puted as 3.0, 3.0, 16, and 1% for the cases of TF-E, TF-I,
deserves further studies and is beyond the scope of this study. LF-E, and LF-I, respectively. The limited failure probabilities
imply that the studied frames could survive the sudden loss of
a column with high likelihood.
Conclusions 3. The parameters of dead load in floor (DLf ), yield and ultimate
strengths of steel bars (f y and f u ), and reinforcement area of
The influence and significance of the inherent randomness of grav- beams (As;Beam ) show apparent effects on the residual capacity
ity loads, material properties, and construction geometries on the of structures in resisting progressive collapse.

© ASCE 04016069-12 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


DLf DLf
fu fu
As,Beam As,Beam
fy fy
Hb LL
LL Hb
t t
L L
fc fc
DLr DLr
fyt fyt
As,Col Es
Ast Ast
Bb Bb Mean variables
Es As,Col
Mean variables
Ec ft
ft Ec
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.0 2.1 2.2 2.3 2.4 2.5 2.6
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

Load factor, α Load factor, α


(a) (b)

DLf DLf
fu fu
Bb L
As,Beam Bb
As,Beam
LL
L fy
fy t
Es DLr
t fyt
Hb LL
fc Hb
Ast fc
DLr Ec
As,Col Es
fyt
Mean variables
Ast
Ec Mean variables As,Col
ft ft
1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.0 2.5 3.0 3.5
Load factor, α Load factor, α
(c) (d)

Fig. 16. Tornado diagrams for DC-II: (a) TF-E; (b) TF-I; (c) LF-E; (d) LF-I

Acknowledgments Ellingwood, B. R. (2006). “Mitigating risk from abnormal loads and


progressive collapse.” J. Perform. Constr. Facil., 10.1061/(ASCE)0887
The financial support received from the Scientific Research -3828(2006)20:4(315), 315–323.
Foundation for the Returned Overseas Chinese Scholars, State Ellingwood, B. R., and Dusenberry, D. O. (2005). “Building design
Education Ministry, the National Science Foundation of China for abnormal loads and progressive collapse.” Comput.-Aided Civ.
(Grant Nos. 51408155, 51378162, 51568004), the China Post- Infrastruct. Eng., 20(3), 194–205.
doctoral Science Foundation (2014M551251), the Heilongjiang Ellingwood, B. R., Smilowitz, R., Dusenberry, D. O., Duthinh, D., Lew,
Postdoctoral Science Foundation (LBH-Z14114), and the Funda- H. S., and Carino, N. J. (2007). “Best practices for reducing the
potential for progressive collapse in buildings.” NISTIR 7396, U.S.
mental Research Funds for the Central Universities (HIT. NSRIF.
Dept. of Commerce, National Institute of Standards and Technology,
2015099) are gratefully appreciated. Gaithersburg, MD.
El-Tawil, S., Li, H., and Kunnath, S. (2013). “Computational simulation of
gravity-induced progressive collapse of steel-frame buildings: Current
References trends and future research needs.” J. Struct. Eng., 10.1061/(ASCE)ST
ACI (American Concrete Institute). (2008). “Building code requirements .1943-541X.0000897, A2513001.
for structural concrete and commentary.” ACI 318-08, Farmington Fascetti, A., Kunnath, S. K., and Nisticò, N. (2015). “Robustness evalu-
Hills, MI. ation of RC frame buildings to progressive collapse.” Eng. Struct., 86,
ASCE/SEI (Structural Engineering Institute). (2010). “Minimum design 242–249.
loads for buildings and other structures.” ASCE/SEI 7, Reston, VA. Felippa, C. A. (2000). “A systematic approach to the element-independent
Bao, Y., Kunnath, S. K., El-Tawil, S., and Lew, H. S. (2008). “Macromodel- corotational dynamics of finite elements.” Center for Aerospace
based simulation of progressive collapse: RC frame structures.” J. Struct. Structures, College of Engineering, Univ. of Colorado, Boulder, CO.
Eng., 10.1061/(ASCE)0733-9445(2008)134:7(1079), 1079–1091. GSA (General Services Administration). (2013). “Alternate path analysis &
DoD (Department of Defense). (2013). “Design of building to resist design guidelines for progressive collapse resistance.” Washington, DC.
progressive collapse.” UFC 4-023-03, Washington, DC. JCSS (Joint Committee on Structural Safety). (2001). “Probabilistic model
Ellingwood, B., Galambos, T. V., MacGregor, J. G., and Cornell, C. A. code. Part 2: Live load.” 〈http://www.jcss.byg.dtu.dk/Publications/
(1980). “Development of a probability based load criterion for Probabilistic_Model_Code〉 (May 2011).
American National Standard A58—Building code requirement for Kazemi-Moghaddam, A., and Sasani, M. (2015). “Progressive collapse
minimum design loads in buildings and other structures.” National evaluation of Murrah Federal Building following sudden loss of column
Bureau of Standards, Dept. of Commerce, Washington, DC. G20.” Eng. Struct., 89, 162–171.

© ASCE 04016069-13 J. Perform. Constr. Facil.

J. Perform. Constr. Facil., -1--1


Kent, D. C., and Park, R. (1971). “Flexural members with confined Park, J., and Kim, J. K. (2010). “Fragility analysis of steel moment frames
concrete.” J. Struct. Div., 97(7), 1969–1990. with various seismic connection subjected to sudden loss of a column.”
Khandelwal, K., and El-Tawil, S. (2011). “Pushdown resistance as a Eng. Struct., 32(6), 1547–1555.
measure of robustness in progressive collapse analysis.” Eng. Struct., Qian, K., and Li, B. (2012a). “Dynamic performance of RC beam-column
33(9), 2653–2661. substructures under the scenario of the loss of a corner column-
Kim, J., Park, J. H., and Lee, T. H. (2011). “Sensitivity analysis of steel experimental results.” Eng. Struct., 42, 154–167.
buildings subjected to column loss.” Eng. Struct., 33(2), 421–432. Qian, K., and Li, B. (2012b). “Slab effects on response of reinforced
Le, J. L., and Xue, B. (2014). “Probabilistic analysis of reinforced con- concrete substructures after loss of corner column.” ACI Struct. J.,
crete frame structures against progressive collapse.” Eng. Struct., 76, 109(6), 845–856.
313–323. Qian, K., and Li, B. (2013). “Performance of three-dimensional reinforced
Lee, T. H., and Mosalam, K. M. (2004). “Probabilistic fiber element concrete beam-column substructures under loss of a corner column
modeling of reinforced concrete structures.” Comput. Struct., 82(27), scenario.” J. Struct. Eng., 10.1061/(ASCE)ST.1943-541X.0000630,
2285–2299. 584–594.
Li, M., and Sasani, M. (2015). “Integrity and progressive collapse resis- Qian, K., Li, B., and Ma, J. X. (2015a). “Load-carrying mechanism to resist
Downloaded from ascelibrary.org by GUANGXI UNIVERSITY on 01/05/17. Copyright ASCE. For personal use only; all rights reserved.

tance of RC structures with ordinary and special moment frames.” progressive collapse of RC buildings.” J. Struct. Eng., 10.1061/(ASCE)
Eng. Struct., 95, 71–79. ST.1943-541X.0001046, 04014107.
Liu, P. L., and Der Kiureghian, A. (1986). “Multivariate distribution models Qian, K., Li, B., and Zhang, Z. W. (2015b). “Testing and simulation of
with prescribed marginals and covariances.” Probab. Eng. Mech., 1(2), 3D effects on progressive collapse resistance of RC buildings.” Mag.
Concr. Res., 67(4), 163–178.
105–112.
Sasani, M., Bazan, M., and Sagiroglu, S. (2007). “Experimental and ana-
Lu, D. G., Cui, S. S., Song, P. Y., and Chen, Z. H. (2011). “Robustness
lytical progressive collapse evaluation of actual reinforced concrete
assessment for progressive collapse of framed structures using push-
structure.” ACI Struct. J., 104(6), 731–739.
down analysis methods.” Int. J. Reliab. Saf., 6(1–3), 15–37.
Scott, M. H., and Fenves, G. L. (2010). “Krylov subspace accelerated
Mander, J. B., Priestley, M. J. N., and Park, R. (1988). “Theoretical stress-
Newton algorithm: Application to dynamic progressive collapse simu-
strain model for confined concrete.” J. Struct. Eng., 10.1061/(ASCE)
lation of frames.” J. Struct. Eng., 10.1061/(ASCE)ST.1943-541X
0733-9445(1988)114:8(1804), 1804–1826. .0000143, 473–480.
Mirza, S. A., Hatzinikolas, M., and MacGregor, J. G. (1979). “Statistical Song, B. I., Giriunas, K. A., and Sezen, H. (2014). “Progressive collapse
descriptions of strength of concrete.” J. Struct. Div., 105(ST6), testing and analysis of a steel frame building.” J. Constr. Steel Res., 94,
1021–1037. 76–83.
Mirza, S. A., and MacGregor, J. G. (1979a). “Variability in dimensions Song, B. I., and Sezen, H. (2013). “Experimental and analytical progres-
of reinforced concrete members.” J. Struct. Div., 105(ST4), sive collapse assessment of a steel frame building.” Eng. Struct., 56,
751–766. 664–672.
Mirza, S. A., and MacGregor, J. G. (1979b). “Variability of mechanical Szyniszewski, S. (2010). “Effects of random imperfections on progressive
properties of reinforcing bars.” J. Struct. Div., 105(ST5), collapse propagation.” ASCE Structures Congress, ASCE, Reston, VA,
921–937. 3572–3577.
Morone, D. J., and Sezen, H. (2014). “Simplified collapse analysis Vamvatsikos, D., and Cornell, C. A. (2002). “Incremental dynamic
using data from building experiment.” ACI Struct. J., 111(4), analysis.” Earthquake Eng. Struct. Dyn., 31(3), 491–514.
925–934. Xu, G., and Ellingwood, B. R. (2011). “Probabilistic robustness assess-
Olsson, A. M. J., and Sandberg, G. E. (2002). “Latin hypercube sampling ment of pre-Northridge steel moment resisting frames.” J. Struct. Eng.,
for stochastic finite element analysis.” J. Eng. Mech., 10.1061/(ASCE) 10.1061/(ASCE)ST.1943-541X.0000403, 925–934.
0733-9399(2002)128:1(121), 121–125. Yi, W. J., He, Q. F., Xiao, Y., and Kunnath, K. (2008). “Experimental study
OpenSees [Computer software]. Pacific Earthquake Engineering Research on progressive collapse-resistant behavior of reinforced concrete frame
Center, Univ. of California, Berkeley, CA. structures.” ACI Struct. J., 105(4), 433–439.

© ASCE 04016069-14 J. Perform. Constr. Facil.

View publication stats J. Perform. Constr. Facil., -1--1

You might also like