Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Procedia Engineering 00 (2017) 000–000
Available online at www.sciencedirect.com
Procedia Engineering 00 (2017) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Engineering 199 (2017) 1170–1174

X International Conference on Structural Dynamics, EURODYN 2017

X International Conference on Structural Dynamics, EURODYN 2017
Reliability-Based Dynamic Analysis of Progressive Collapse of
Reliability-Based Dynamic Analysis of Progressive Collapse of
Highway Bridges
Highway Bridges
Feng Miao and Michel Ghosn *
Feng
Department of Civil Engineering, Miao
The and Michel
City College of New Ghosn * New York, NY 1003, USA
York /CUNY,
Department of Civil Engineering, The City College of New York /CUNY, New York, NY 1003, USA

Abstract
Abstract
Structural systems optimized to meet member design criteria specified in design standards may not provide sufficient levels
of structural
Structuralrobustness to withstand
systems optimized possible
to meet local design
member failurescriteria
that could be caused
specified by unforeseen
in design standardsextreme
may notevents.
provideIn sufficient
fact, the sudden
levels
failure of anrobustness
of structural element may result in possible
to withstand the failure offailures
local anotherthatcreating
could abechain
caused reaction that might
by unforeseen progress
extreme throughout
events. thesudden
In fact, the entire
structure
failure oforana element
major portion of it leading
may result to catastrophic
in the failure of anothercollapse.
creating Concerns about possible
a chain reaction threats
that might to major
progress bridges instigated
throughout the entire
interest inordeveloping
structure criteria
a major portion of for evaluating
it leading their robustness.
to catastrophic collapse.Bridge-specific
Concerns about criteria are needed
possible threats toas major
existing procedures
bridges for
instigated
buildingsinare
interest not suitable
developing for bridges
criteria because oftheir
for evaluating differences in their
robustness. structural configurations
Bridge-specific criteria areand their dead
needed and transient
as existing loads.for
procedures
This paper
buildings describes
are not suitableafor
methodology
bridges becausefor assessing the capability
of differences of bridgesconfigurations
in their structural to resist progressive collapse
and their dead andwhile accounting
transient loads.for
the This
uncertainties in the aapplied
paper describes loads and
methodology the load the
for assessing carrying capacities
capability of the
of bridges members
to resist as wellcollapse
progressive as the while
systemaccounting
of damaged for
structures.
the The reliability
uncertainties analysisloads
in the applied methodology is illustrated
and the load carryingusing models
capacities ofofthea steel box girder
members as wellbridge
as theandsystem
a steel of
truss bridge
damaged
subjected toThe
structures. different initial
reliability damage
analysis scenarios. The
methodology paper outlines
is illustrated how the
using models of results
a steel from reliability-based
box girder bridge and astructural dynamic
steel truss bridge
analyses can
subjected to be implemented
different initial to develop
damage structuralThe
scenarios. robustness criteria that
paper outlines howwould lead tofrom
the results consistent levels of safety.
reliability-based structural dynamic
analyses can be implemented to develop structural robustness criteria that would lead to consistent levels of safety.
© 2017 The Authors. Published by Elsevier Ltd.
© 2017 The Authors. Published by Elsevier Ltd.
© 2017 The under
Peer-review Authors. Published by
responsibility of Elsevier Ltd. committee of EURODYN 2017.
the organizing
Peer-review under responsibility of the organizing committee of EURODYN 2017.
Peer-review under responsibility of the organizing committee of EURODYN 2017.
Keywords: Progressive Collapse; Structural Redundancy; Structural Robustness; Highway Bridges; Structural Reliability
Keywords: Progressive Collapse; Structural Redundancy; Structural Robustness; Highway Bridges; Structural Reliability

1. Introduction
1. Introduction
Structural systems optimized to meet member design criteria as specified in current design standards may not
Structural
provide systems
sufficient optimized
levels to meet
of robustness member design
to withstand criteria
a possible as failure
local specified in current
following design standards
an unforeseen extrememay not
event.
provide sufficient
In fact, local failurelevels ofstructural
of one robustnesselement
to withstand a possible
may result local failure
in the failure following
of another an aunforeseen
creating extreme
chain reaction event.
of failures
In
thatfact, local failure
progress of onethe
throughout structural element
structure may
leading to result in of
a level thedamage
failure ofdisproportionate
another creatingtoa chain reaction
the initial of failures
damage or to
that progress throughout the structure leading to a level of damage disproportionate to the initial damage or to

* Corresponding author. Tel.: +1-212-650-8002; fax: +1-212-650-6965.

* E-mail address:author.
Corresponding ghosn@ccny.cunny.edu
Tel.: +1-212-650-8002; fax: +1-212-650-6965.
E-mail address: ghosn@ccny.cunny.edu
1877-7058 © 2017 The Authors. Published by Elsevier Ltd.
Peer-review©under
1877-7058 2017responsibility
The Authors. of the organizing
Published committee
by Elsevier Ltd. of EURODYN 2017.
Peer-review under responsibility of the organizing committee of EURODYN 2017.

1877-7058 © 2017 The Authors. Published by Elsevier Ltd.

Peer-review under responsibility of the organizing committee of EURODYN 2017.
10.1016/j.proeng.2017.09.300
 Feng Miao et al. / Procedia Engineering 199 (2017) 1170–1174 1171
2 Feng Miao and Michel Ghosn/ Procedia Engineering 00 (2017) 000–000

catastrophic collapse [1, 2]. Progressive collapse occurs when a sudden local change in structural geometry due to
the loss of load-carrying members results in dynamic forces exceeding the bearing capacities of the surrounding
elements [3]. Engineers need to evaluate the potential of progressive collapse for all high risk structures including
highway bridges which form critical links in our infrastructure network as they are essential catalysts for economic
growth and the wellbeing of the communities they serve. Existing guidelines for analyzing buildings for the
potential of disproportionate collapse may not be suitable for bridges because of the differences in the topologies
and configurations of the types of structural systems and their permanent and transient loads [1]. Furthermore,
structural analysis guidelines and criteria should account for the large uncertainties associated with estimating the
applied loads and the capacity of structural systems to resist collapse following an initial local failure [4, 5, 6, 7].
Probabilistically, the progressive collapse process can be represented by the following equation [4]:
P(C ) =∑ ∑ H
P(C | D) P(D | H) P(H)
D
(1)

where P (C ) is the probability of system collapse, P ( H ) is the probability of occurrence and intensity of hazard H;
P ( D / H ) is the probability of local structural damage scenario, D, given the occurrence of the damage-initiating
hazard H, and P (C / D ) is the probability of structural collapse given an initial damage scenario D. The
term P (C / D ) is often used to assess structural robustness [5].
The objectives of this paper are to: 1) describe a structural reliability analysis methodology proposed by Miao
and Ghosn [7] to estimate P (C / D ) which is the probability of structural collapse of a highway bridge given a
specified initial damage scenario D; and 2) calibrate appropriate deterministic analysis procedures that can be
implemented in bridge engineering practice to verify that a particular highway bridge provides sufficient levels of
structural robustness. The methodology is illustrated using as examples a steel box girder bridge and a truss bridge
subjected to different initial damage scenarios [7].

2. Analysis Models for Example Bridges

2.1 Progressive Collapse Analysis

In this study, the progressive collapse analysis process is modelled by instantaneously applying ramped up
reaction loads equal and opposite to those that were originally applied on the damaged element before damage took
place. Thus, an impulsive dynamic load is applied to the structure that is held in its initial, undamaged position by
the reaction loads. Fig. 1 shows a schematic representation outlining the dynamic analysis process [3, 7].

2.2 Box Girder Bridge Characteristics

A grillage model is established for a 36.6 m-long and 7.2 m wide steel box-girder bridge having the cross
section of Fig. 2.a. The 200 mm concrete deck is reinforced both ways with two layers of rebars. The bridge’s
longitudinal members are designed following the current AASHTO specifications [8] to yield member reliability
index values βmember=3.5. To simulate damage, a 15.2 cm fracture is assumed to cut through the entire width of the
bottom flange and propagate through the two webs of a box whose capacity is reduced to that of the slab only.

Internal forces Static forces Dynamic forces

P P
P
P P

Dead Load+Live load Dead Load+Live load

(a) (b) (c)

Force function

0 T0 T1
Tr
t

Fig. 1. Progressive collapse analysis process

1172 Feng Miao et al. / Procedia Engineering 199 (2017) 1170–1174
Feng Miao and Michel Ghosn/ Procedia Engineering 00 (2017) 000–000 3

Table 1. Composite steel box-girder bridge member properties

Inertia Torsional Const. Nominal Moment Probability dist.

Bias COV
I, (cm4) J, (cm4) Capacity (kN-m) type
Composite long. box beams 3.4x106 7.2x105 8400 1.12 10% Lognormal
Cracked longitudinal beams 1.3x105 2.6x105 172 1.14 13% Lognormal
Transverse box beams 1.1x107 3.9x107 19815 1.12 10% Lognormal
Transverse slab beams 2.6x105 5.1x105 343 1.14 13% Lognormal

The nonlinear behavior of the members is represented by moment-curvature relationships. The maximum
curvature is obtained based on section dimensions and material properties. Comparisons with results of field tests
indicate that the plastic hinge length Lp is equal to the depth of the section [9]. Table 1 lists the nominal member
properties and the associated statistical information for the probabilistic model. The longitudinal member properties
are assumed to be statistically independent of those of the transverse members. The dead loads, that nominally add
up to 18.03 kN/m per longitudinal beam, are assumed to follow normal probability distributions with a bias of 1.05
and a COV of 10%. The live load model is obtained from weigh-in-motion data and projected to estimate the
maximum load for different bridge service periods. The live load model is presented in terms of the effects of a
standard semi-trailer truck having a total weight 320kN as shown in Fig. 3. The static load is augmented by a
dynamic amplification IM=1.09 and the combined effect, including epistemic uncertainties, has a COV=18% [11].

Fig. 2. Configurations of box-girder bridge cross section (on the left) and truss bridge profile (on the right)

2.3 Truss Bridge Characteristics

The 31.7 m through-truss bridge is supported by two identical parallel trusses having the configuration shown in
Fig. 2. The concrete deck is 178 mm thick and 10.36 m wide. The truss members are first designed to produce a
reliability index βmember=3.5 as implied in traditional design standards [8]. The structural model includes the main
struts and the connections, which consist of the bolts and gusset plates. Table 2 gives a listing of the truss Grade 36
(248 MPa) members’ cross sectional areas. The bias for the main member capacities is set at 1.10 and the COV is
11% assuming lognormal distributions. The nominal dead weights are obtained as 31.52 kN/m assuming they follow
a normal probability distribution associated with a bias = 1.05 and a COV=10%. The static live load model for
different service lives, normalized in terms of the effect of a typical semi-trailer 3-S2 standard truck [10], is shown
in Fig. 3. As is the case with the box girder bridge, the static live load on the truss is augmented by a dynamic
amplification of 1.09 and the combined live load is associated with a COV=18% [11].

1.0 1.0

0.8 0.8
Cumulative Probability

Cumulative Probability

original
One year original
Two years
0.6 Five years
0.6 One year
Two years
Ten years Five years
0.4 Seventy-Five years 0.4

0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Normalized 3S2 Trucks Normalized 3S2 Trucks

Fig. 3. Live Load Models for 36.6 m bridge (on left) and 31.7 m bridge (on right)
 Feng Miao et al. / Procedia Engineering 199 (2017) 1170–1174 1173
4 Feng Miao and Michel Ghosn/ Procedia Engineering 00 (2017) 000–000

Table 2. Truss member cross sections

No. Area (cm2) No. Area (cm2) No. Area (cm2) No. Area (cm2) No. Area (cm2) No. Area (cm2)
1 90.32 7 96.3 13 113.48 19 15.87 25 11.87 31 17.10
2 96.32 8 90.32 14 104.77 20 1.23 26 23.74 32 3.55
3 111.10 9 68.97 15 72.32 21 15.87 27 23.74 33 3.55
4 121.61 10 72.32 16 68.97 22 5.61 28 11.87 34 17.10
5 121.61 11 104.77 17 26.32 23 26.32 29 41.16 35 12.77
6 111.10 12 113.48 18 5.61 24 41.16 30 12.77

3. Reliability Analysis

A probabilistic dynamic analysis of progressive collapse is performed assuming that each bridge has been
subjected to an extreme event that led to the sudden damage to one critical member. No analysis of the hazard that
causes the damage is performed because the interest is in determining the conditional probability of collapse
P (C / D ) . A sensitivity analysis is perfumed to study how the system reliabilities of the damaged structures relate to
member reliabilities. The reliability analysis is performed using a variation on the Subset Simulation method [7, 12].
The reliability results are compared to those of deterministic incremental static load analyses of the different bridge
systems after the removal of the damaged members to obtain the associated Incremental Load factors, ILF. If
properly calibrated, bridge engineers could use results of deterministic analyses to infer the system reliabilities
implicit in bridge structures and estimate their levels of structural robustness [5, 6, 7].

3.1 Box-girder Bridge

The dynamic probabilistic progressive collapse analysis is executed assuming that a fracture takes place at the
mid-span of one box when the maximum 1-month live load is on the bridge. For the original design that produced
βmember=3.5, the damaged system has a reliability βprogressive =2.46. The nonlinear incremental analysis showed that
collapse occurs at an Incremental Load Factor ILF = 1.80 . If the capacity of each longitudinal beam is multiplied by
0.8, the reliability index becomes βprogressive=1.33 and ILF = 1.53 . By repeating the same procedure for different
member capacities, the relationship between ILF and the reliability index βprogressive is established as shown in Fig. 4.
The results in Fig. 4 can be used by code writers to help select the appropriate incremental progressive analysis
criteria that would meet a target reliability level. For example, if a reliability index βprogressive=2.0 is required, then
an engineer can simply execute a nonlinear incremental analysis of a fractured bridge and ensure that the damaged
bridge will be able to carry its dead weight and 1.75 times the 3S-2 live load to be considered sufficiently robust. An
ILF=1.54 would correspond to a reliability index βprogressive=0.80. These reliability indexes are those related to the
conditional probability of collapse given the fatigue fracture of the one box of the two-box bridge P(C D) . A
reliability index βprogressive=0.80 would lead to a conditional reliability P(C D) =21.19% if the bridge members are
designed to meet a reliability index for fatigue fracture βmember=2.0 or P(D H ) P ( H ) =2.28% as intended by the
AASHTO [8] specifications, then the unconditional probability of collapse will be 0.48% for a reliability index
equal to 2.59. This value is similar to that allowed for the rating of existing bridges in the AASHTO MBE [10].

Fig. 4. Reliability index βprogressive versus Incremental Load Factor ILF for box-girder (on the left) and truss bridge (on the right)
1174 Feng Miao et al. / Procedia Engineering 199 (2017) 1170–1174
Feng Miao and Michel Ghosn/ Procedia Engineering 00 (2017) 000–000 5

3.2 Truss Bridge

Originally, all the truss members are designed to achieve a member reliability βmember=3.5. Subsequently the
designs are adjusted to produce higher reliabilities. Several damage scenarios are investigated. The first scenario,
D1, assumes that member 29 in Fig. 2 is dynamically removed. The second and third scenarios (D2 and D3) assume
that members 23 or 34 are suddenly removed, respectively. The results of the system analyses in Fig. 4 show the
curves for each damage scenario. On the whole, the three curves lie within a reasonably narrow band but slightly
shifted to the right of the curve of the box girder bridge. The variations are due to the differences in the type of
damages considered, the load path patterns, and live to dead load ratios of the two structural configurations. The
results indicate that if a target reliability index βprogressive=1.50 is required, the incremental analysis of the damaged
bridge should lead to a load carrying capacity equal to the bridge’s dead weight plus somewhere between 1.67 and
1.83 times the 3S-2 live load. If the average value ILF=1.75 is used, then the reliability index βprogressive will fall
between 1.25 and 1.75. If a reliability index βprogressive=0.80 is required, then the incremental load factor on the 3S2
truck should be between 1.45 and 1.63 with an average value ILF=1.55 which is very similar to the value obtained
for the box girder bridge for the same target reliability.

4. Conclusions

This paper described a procedure to perform a probabilistic analysis of progressive collapse should a member be
suddenly damaged due to the occurrence of an external hazard.
The probabilistic analysis procedure was illustrated using two typical bridge configurations: A two-box steel
girder bridge and a steel truss bridge. The box girder bridge is assumed to be susceptible to fatigue fracture at the
midspan of one box and the truss is assumed to lose the load carrying capacity of a main strut. These damage
scenarios are selected because these two bridge types have been generally designated as non-redundant, fracture-
critical configurations that are expected to collapse under their own weights if one of their members is fractured.
Because performing reliability analyses are beyond the day-to-day practice of bridge engineers, a methodology is
presented that can be used to calibrate reliability-based criteria to help bridge engineers perform traditional
incremental analyses to assess the ability of a bridge system to mitigate disproportionate collapse.
The results indicate that if a push down analysis of bridges in their damaged configurations are able to sustain
their dead loads and about 1.83 times the weight of two side-by-side standard semi-trailer trucks each carrying 320
kN, then they should be able to resist progressive collapse with a reliability index on the order of βprogressive=1.5. A
reliability index βprogressive=0.80 is achieved when the incremental load factor ILF is equal to 1.55.
Future research should compare the results obtained in this paper to those from the analysis of additional bridge
configurations and various damage scenarios that could include simultaneous failures in several members.

References

[1] GSA, Progressive collapse analysis and design guidelines for new federal office buildings and major modernization projects, Washington DC,
2013.
[2] U. Starossek, Progressive collapse of structures, Thomas Telford Publishing, London, 2009.
[3] S.M. Marjanishvili, Progressive analysis procedure for progressive collapse, J. Perform. Constr. Facil.,18(2), 2004, pp.79-85.
[4] G. Xu, B.R. Ellingwood, Probabilistic robustness assessment of pre-Northridge steel moment resisting frames, J. of Structural Engineering,
137(9), 2011, pp. 925-34.
[5] M. Ghosn, J. Yang, Bridge system safety and redundancy, NCHRP Report 776, National Academy Press, Washington DC, 2014.
[6] M. Ghosn, F. Moses, Redundancy in highway bridge superstructures, NCHRP Report 406, The National Academies, Washington DC, 1998.
[7] F. Miao, M. Ghosn, Reliability-based progressive collapse analysis of highway bridges, Structural Safety (63) 20016, pp. 33-46.
[8] AASHTO LRFD Bridge Design Specifications, Association of State Highway and Transportation Officials, Washington DC, 2014.
[9] J. Kim, E.B. Williamson, Finite-element modeling of twin steel box-girder bridges for redundancy evaluation, J. Bridge Eng., 20(10), 2015.
[10] AASHTO MBE, Manual for bridge evaluation, Association of State Highway and Transportation Officials, Washington DC, 2011.
[11] M. Ghosn, B. Sivakumar, F. Miao, Development of state-specific load and resistance factor rating method, J. Bridge Eng., 18(5), 2013, pp.
351-61.
[12] S.K. Au, J.L. Beck. Estimation of small failure probabilities in high dimensions by subset simulation. Probabilistic Engineering Mechanics,
(16) 2001, pp. 263-277.