(Asce) Be 1943-5592 0001770

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Coupled CFD–FEM Simulation Methodology

for Fire-Exposed Bridges


Zhi Liu1; Julio Cesar G. Silva2; Qiao Huang3; Yuji Hasemi4; Yili Huang5; and Zhaoyuan Guo6

Abstract: Evaluating the fire performance of bridges is necessary for damage mitigation as more fire-related accidents occurred on bridge struc-
tures. Current solutions adopting temperature curves can result in significant inaccuracy by neglecting the inhomogeneous characteristic of the
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

fire-induced thermal environment. This paper proposed a numerical methodology for analyzing the coupled thermomechanical response of var-
ious bridges exposed to fires. The computational fluid dynamics (CFD) approach was implemented to reproduce the fire condition more real-
istically by modeling the combustion process and fire-driven flow. Then, an interface was adopted to extract the thermal boundary from the fire
model. At last, the thermomechanical finite-element method (FEM) was coupled with the CFD model for determining the fire-induced response
of the global bridge, thermally and structurally. By incorporating the multiscale FEM, this methodology can be extended to various large-scale
bridges subjected to localized fires. The proposed approach was validated through a real fire experimental study on a steel column. To demon-
strate the application of this strategy, a complex case study was carried out. A long-span cable-stayed bridge was investigated considering its
girder segment was exposed to an under-deck tanker fire. Numerical results showed that the proposed method was able to capture the surround-
ing temperature field with strong thermal gradients and can predict not only the localized thermomechanical response of exposed segments but
also the global structural performance evolution for large-scale complex bridges. The under-deck fire can introduce a significant impact on the
entire cable-stayed bridge. Thereby, the multiscale FEM modeling strategy is required for the long-span bridges exposed to localized fires.
DOI: 10.1061/(ASCE)BE.1943-5592.0001770. © 2021 American Society of Civil Engineers.
Author keywords: Bridge; Fire; Computational fluid dynamics; Finite-element method; Thermomechanical behavior; Coupled simulation;
Multiscale modeling.

Introduction performance evolution originates from temperature-induced mate-


rial deterioration. The majority of existing studies simplified fire
Fire is one of the most dangerous hazards threatening bridge safety. conditions as spatially homogeneous temperatures varying over
This is particularly the case as more fuel-powered and fuel-carrying time, like the ISO834 or hydrocarbon fire (HC) profile (CEN 2002).
vehicles are shuttling on and under bridges. As per statistical anal- One primary problem with applying temperature curves is the
yses, more than 3% of counted bridge failures in the United States distortion of reconstruction for the fire-induced temperature field.
after 1951 were due to fires, which is close to earthquake-caused Extensive studies have demonstrated that a significant inaccuracy
bridge failures (Harik et al. 1990; Wardhana and Hadipriono can be introduced (Alos-Moya et al. 2014; Quiel et al. 2015;
2003; Garlock et al. 2012). To develop efficient mitigation strat- Buchanan and Abu 2017; Beneberu and Yazdani 2018). Prescrip-
egies against potential fire accidents, deeply understanding the tive curves were obtained based on the layered model established
fire performance of bridges is becoming more necessary. for indoor fires (CEN 2002), whereas bridge fires are usually char-
Predicting the response of fire-exposed bridges is twofold: ther- acterized by gradient temperature and high fire intensity.
mally and structurally. Determining the fire-induced surrounding Sporadic studies have implemented the computational fluid
temperature field is the foundation because the structural dynamics (CFD) approach to provide a more realistic fire basis.
Choi (2008) incorporated the CFD technique to calculate the spa-
1
Ph.D. Candidate, Dept. of Bridge & Tunnel Engineering, Southeast tial–temporal temperature distribution of reinforced concrete
Univ., Nanjing 211100, China. ORCID: https://orcid.org/0000-0001-8518 (RC) beams and gave an acceptable prediction for the structural
-3894. response. The experimental studies on composite beams con-
2
First Lieutenant, Engineering Corps of Brazilian Navy, Rio de Janeiro ducted by Alos-Moya et al. (2017) and their numerical valida-
22220-030, Brazil. ORCID: https://orcid.org/0000-0003-2798-9384. tions (Alos-Moya et al. 2019) showed that the CFD method is
3
Professor, Dept. of Bridge & Tunnel Engineering, Southeast Univ.,
capable of predicting the temperature distribution for fire-bridge
Nanjing 211100, China (corresponding author). Email: qhuanghit@126
.com analyses. However, the laborious coupling procedures in these
4
Professor Emeritus, Dept. of Architecture and Architectural Engineer- studies are merely limited to small bridges consisting of orthog-
ing, Waseda Univ., Tokyo 169-8050, Japan. onal geometries because of the lack of a methodology that can
5
Graduate Student, Dept. of Bridge and Tunnel Engineering, Southeast tackle complex structures, trapping the current understanding of
Univ., Nanjing 211100, China.
6
the bridge behavior in fires lingering at the application of simple
Jiangsu Provincial Transportation Engineering Construction Bureau, temperature models.
69 Shigu Rd., Nanjing 210004, China. This paper proposed a fire-thermomechanical simulation frame-
Note. This manuscript was submitted on October 28, 2020; approved on
work for analyzing the behavior of bridges exposed to fires. Fire
May 27, 2021; published online on August 2, 2021. Discussion period open
until January 2, 2022; separate discussions must be submitted for individual conditions were considered using the CFD technique to reproduce
papers. This paper is part of the Journal of Bridge Engineering, © ASCE, the spatially inhomogeneous temperatures. Then, the response was
ISSN 1084-0702. captured by performing a coupled thermomechanical finite-element

© ASCE 04021074-1 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


method (FEM) simulation by applying the thermal exposure computational cost can result because the structural FEM model
extracted from the fire model using an interface. The proposed of the entire bridge is usually required to consider the local thermal
framework was validated based on a real fire experiment on a evolution and the global redistribution of internal force. Extending
steel column. Combining with the multiscale FEM approach, large- the coupled FDS–FEM analyses to large-span complex bridges is
span bridges subjected to localized fires are investigatable also con- now prevented by these dilemmas, which will be addressed by
sidering the interaction between the exposed portion and other the programmatic and effective simulation methodology proposed
components. For demonstration, a cable-stayed bridge was studied in this paper.
using the proposed approach. Not only the localized thermome-
chanical response was determined but also the entire bridge perfor-
mance was captured, providing a dependable foundation for future CFD–FEM Coupled Simulation Methodology
safety evaluations of fire-exposed bridges.
This paper proposed a methodology for evaluating the fire perfor-
mance of various bridge structures by coupling the CFD-based fire
Previous Approaches model and FEM-based thermomechanical model, as illustrated in
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Three main steps were included: replicating the fire scenario
Numerically evaluating the behavior of fire-exposed bridges is considering environmental variables using a CFD code, extracting
basically a thermomechanical FEM simulation based on the fire- and transferring the thermal boundary from the fire model to the
induced surrounding temperature field, which is the emphasis of FEM model of exposed members, and performing the coupled ther-
how precisely a fire scenario can be represented. For compart- momechanical FEM simulation.
ment fires, the one-zone model was developed assuming a homo- For complex or long-span bridges, a cautious numerical plan
geneous distribution of temperature, density, internal energy, was required to reduce the computational cost because these simu-
and pressure of the gas, and then, a two-zone model was further lations, especially the CFD, are time-consuming iteration pro-
proposed by defining two thermal layers assuming the combus- cesses. It is critical to determine how many structural members
tion products accumulate in the form of layers (CEN 2002). Var- are required in each step and to what extent should their geometric
ious temperature curves are the illustration of the one-zone details be modeled. Depending on the relative size of the fire-
model or the upper layer in the two-zone model. Following affected region to the global scale, the partial and entire modeling
these compartment-aimed assumptions, numerous numerical of a bridge are alternatives in the first step, and a detailed discreti-
studies have been contributed for bridges to predict their resist- zation and multiscale FEM modeling of the entire bridge are alter-
ance before fire, failure process during fire, and residual strength natives in the third step.
after fire (Aziz and Kodur 2013; Kodur et al. 2013; Aziz et al.
2015).
CFD-Based Fire Modeling
As an advanced solution, the CFD approach involves the mod-
eling for the characteristics of the fire source, the geometry of com- The aim of the CFD-based modeling for the fire condition is to
ponents, and fire-driven fluid heat flow. The open-source program reproduce the thermal circumstance. The dimensions of the re-
FDS (version 6.7.5) has been extensively utilized for replicating the quired computational domain in which the simulation will be
fire scenario in infrastructures. Because adopting prescriptive tem- carried out mainly depend on how the fire-driven heat flow
perature curves for bridges can yield very different thermal re- will progress. The necessary domain must enclose not only the
sponses comparing to the FDS-based fire modeling (Alos-Moya flame but also the space conveying heat transfer to consider a
et al. 2014), coupling the FDS model with the FEM model has in- comprehensive fire exposure. On the other hand, the domain
creased as the major concern in the recent decade. A critical review should be as narrow as possible to minimize the computational
of the studies is summarized in Table 1. expenditure.
The crux of coupling two models is the passed quantities. The Because thermal conduction and radiation highly depend on the
heat flux was a direct quantity for thermal analysis, like the study configurations of bridges, their real geometry should be considered
(Choi et al. 2012) for concrete members. The surface temperature in the CFD simulation. In this methodology, all components of
and gas temperature were adopted in Choi et al. (2012), Gong small-scale bridges in fire conditions were modeled. For large-scale
and Agrawal (2015, 2016), and Cui et al. (2020). The better solu- bridges subjected to localized fires, the computational domain en-
tion is increasing as the adiabatic surface temperature (Wickström closed the exposed portion and its adjacent members, which will
et al. 2007), whose utilization is able to avoid the transfer of the be discretized using high-fidelity elements in the following FEM
heat flux, surface temperature, and convective heat-transfer coeffi- analyses.
cient simultaneously. Its successful application has been reported The open-source program FDS developed by the National In-
on general beam bridges (Alos-Moya et al. 2014; Peris-Sayol stitute of Standards and Technology (NIST 2020) was adopted to
et al. 2015; Wu et al. 2020). simulate fire behavior. It is a large-eddy-simulation-based code
The exclusive way through which quantities were passed in designed to model the fire-driven fluid flow by solving numeri-
these studies was to define measuring devices on needed locations cally the Navier–Stokes equations. Validated by various exper-
before the FDS simulation. The device approach can be acceptable iments, FDS is capable of predicting fire-related quantities,
by small-scale bridges with regular geometries. However, placing such as the gas-phase temperature field, heat flux, and smoke
devices priorly will be very laborious and even become impractica- movement.
ble for irregular bridges. Although endeavors have been contrib- Before the FDS simulation, the thermal FEM model of the mem-
uted to unconventional bridge structures, their fire performance bers that will be considered in the CFD simulation was developed
cannot be estimated precisely by using insufficient quantities to provide geometric information. The FDS model is limited to a
based on FDS models. computational domain made up of rectilinear volumes. Inclined ge-
Also, for localized fires imposing thermal effects on a limited ometries are not supported at present in FDS because it solves
portion, the behavior evolution of large-scale bridges involves equations using the finite-difference method. This paper scripted
the interaction between all structural members. A dramatic a code to orthogonalize irregular geometries into FDS-compatible

© ASCE 04021074-2 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Table 1. Literature studying the fire performance of bridges using the CFD-FEM approach
Title Authors (year) Summary
Integrated fire dynamic and Choi et al. (2012) A fire-heat-structural simulation approach was proposed for analyzing the behavior of
thermomechanical modeling of a bridge structures subjected to fires. The solid-phase temperature at the external surface of steel
under fire members predicted by the FDS-based fire model was approximated using a fourth-order
polynomial form over time and was considered as the component temperature. Inside steel
members, the temperature was assumed to be linearly distributed. For concrete obstructions,
the surface heat fluxes calculated by FDS were transferred to the ABAQUS-based FEM
model to undertake a thermomechanical simulation. For validation, the authors simulated a
three-point loading test of an RC beam under the ISO834 condition. The predicted midspan
and quarter-span deflection was 28% and 31% lower than the experimental values,
respectively, at the end of the test. The behavior of the I-80/880 interchange in Oakland,
California, collapsed due to fire was then reproduced using the proposed approach
Analysis of a bridge failure due to fire Alos-Moya et al. The authors simulated the fire-induced failure process of the I-65 overpass in Birmingham,
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

using computational fluid dynamics and (2014) Alabama which was a three-span continuous beam bridge composed of seven steel I-beams
finite-element models and a concrete slab. The adiabatic surface temperature proposed by Wickström et al. (2007)
was collected from the developed FDS fire model by manually adding measuring devices
and applied to the thermomechanical ABAQUS model of steel beams. For comparison,
simulations were performed using standard and HC temperature curves. The thermal and
structural results were very different from those calculated by coupled CFD–FEM method.
Further parametric studies were carried out to estimate the influence of fire intensity, FEM
discretization, and axial boundary conditions of bearings
Analysis of the influence of geometric, Peris-Sayol et al. A simulation was conducted for a simply-supported bridge with a span of 12.2 m. The
modeling, and environmental parameters (2015) adiabatic surface temperature was measured in FDS and applied as the thermal boundary to
on the fire response of steel bridges the thermomechanical FEM model of the entire steel portion. Parametric studies were
subjected to realistic fire scenarios carried out to analyze the impact caused by different horizontal restraints, count of
considered components in FEM modeling, vertical clearance, wind, and bridge
configuration, and their failure patterns were also compared and discussed
Numerical simulation of fire damage to a Gong and Agrawal The authors performed a coupled simulation to replicate the impact caused by a vehicle fire
long-span truss bridge (2015) to Ed Koch Queensboro Bridge, which is a long-span truss bridge. The fire scenario was
modeled in FDS and the surface temperature was extracted and then applied to the exposed
portion for the coupled thermomechanical simulation. Based on the fire test of a beam
component exposed to the ISO834 fire, the approach validation was performed using a
thermomechanical simulation. The predicted out-of-plane deformation of stringers
presented a correlation with the observed phenomenon
Safety of cable-supported bridges during Gong and Agrawal To investigate the stability of stiffening girders in cable-supported bridges subjected to fire,
fire hazards (2016) series coupled simulations on isolated orthotropic box girder segments were performed. The
over-deck and under-deck fire scenarios were replicated in FDS and surface temperatures
were collected then passed to the thermomechanical FEM model of these segments. For
conservation, the inside heat convection and radiation were not considered. Different axial
compressive loads, 0%, 30%, and 60% of the yield capacity for decks, were applied at the
two ends of segments representing anchored suspension, self-anchored suspension, and
cable-stayed bridges, respectively. Cables or suspenders were considered as rigid vertical
constraints and two ends of segments were constrained in the longitudinal and transverse
directions. The bulking failure behavior of decks was analyzed by examining and
thermal-induced axial force for various fire intensities and distances from the fire source to
the deck floor. Simulation results indicated that the magnitude of axial force is critical for the
potential buckling failure of cable-supported bridges during fire scenarios. The authors
clarify the necessity of conducting more investigation on the behavior of cable-supported
bridges exposed to fire
A localized fire model for predicting the Wu et al. (2020) Series coupled FDS–ABAQUS simulations were conducted for a general concrete box
surface temperature of box girder bridges girder with inclined geometries. The bridge was subjected to various localized tanker truck
subjected to tanker truck fire fires. Inclined webs were orthogonalized into rectangles overlaid by measuring devices to
transfer the adiabatic surface temperatures. The authors proposed a modified SFPE fire
model for predicting the flame length and distribution of heat flux of box girder bridges in
localized fires
Stability assessment of a suspension bridge Cui et al. (2020) The stability of a steel pylon near a tanker fire in a suspension bridge was investigated by the
considering the tanker fire nearby steel– authors. The gas temperature calculated by FDS was applied to the FEM model of the entire
pylon bridge. Results showed that the fire source defined by burning time and firepower can
decrease the stability coefficient of the exposed pylon significantly

obstructions snapping to the underlying grids by identifying the [dimensions, heat release rate per unit area (HRRPUA), fuel
closest gridding planes. The FDS model can thus be established type, and so on], material properties of the structure, ventilation,
by incorporating other information, such as the fire source and so on.

© ASCE 04021074-3 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Simulation methodology.

Data Mapping from the CFD Model to the FEM Model where α = absorptivity. According to the Stefan–Boltzmann law,
the emitted radiative thermal energy is defined as follows:
Heat propagates from flames and hot gases to structure surfaces
in the ways of radiation and convection, making up the total q̇′′r,emi = εσ(Tsurf )4 (4)
heat flux, q̇′′tot
where ɛ = surface emissivity; and σ = Stefan–Boltzmann constant.
q̇′′tot = q̇′′rad + q̇′′conv (1) Based on Kirchhoff’s assumption that emissivity and absorptivity
are equal, Eq. (2) becomes
which is shown in Fig. 2(a). q̇′′rad = ε[q̇′′inc − σ(Tsurf )4 ] (5)
The radiative heat flux, q̇′′rad , is the difference between the radi-
ative flux absorbed (q̇′′r,abs ) and emitted (q̇′′r,emi ): Newton’s law of cooling states that the convective heat flux,
q̇′′conv , is proportional to the temperature difference between gas
q̇′′rad = q̇′′r,abs − q̇′′r,emi (2) (Tgas) and surface (Tsurf). Thereby, Eq. (1) can be written as follows:

The absorbed radiative flux can be considered as part of the inci- q̇′′tot = ε[q̇′′inc − σ(Tsurf )4 ] + h(Tgas − Tsurf ) (6)
dent radiative heat flux (q̇′′inc ):
where h = convective heat-transfer coefficient. Supposing a per-
fectly insulated surface, whose temperature is the adiabatic surface
q̇′′r,abs = αq̇′′inc (3) temperature, TAST, and the emissivity keeps unchanged. When sub-
jected to the same heating conditions, the total heat flux of the ideal

© ASCE 04021074-4 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 2. Illustration of (a) heat transfer from the fire source to exposed surfaces; and (b) example (TAST) of quantity composition for thermal exposure
and interface from the FDS to the FEM model, for a generic fire-exposed structure.

surface is zero: The general approach for data transfer is the device method. It
requires the definition of measuring devices before the FDS simu-
ε[q̇′′inc − σ(TAST )4 ] + h(Tgas − TAST ) = 0 (7) lation to record quantities representing the fire environment, like
TAST. As a direct approach, the device method has been exclusively
Combining Eqs. (6) and (7), the total heat flux transferred to the utilized in bridge fire engineering. However, for complex bridges
real structural surface can be written as a function of TAST: with irregular geometries, the device method is not applicable be-
cause inclined geometries will be transformed to orthogonal ob-
q̇′′tot = εσ[(TAST )4 − (Tsurf )4 ] + h(TAST − Tsurf ) (8) structions aligning to the meshing grids in FDS. This makes it
impossible to set devices on the obstructions in advance. Also, de-
Eq. (8) shows that the heat flux to an exposed surface can be deter-
vices are manually arranged, which is impractical for long-span
mined by TAST, which can be calculated based on the measured
bridges with crisscross panels being considered.
temperature using a plate thermometer during experiments. This
To collect the thermal boundary from the FDS simulation, the
concept proposed by Wickström et al. (2007) provides a simple sol-
fds2ftmi developed by Silva et al. (2016) was used as the interface
ution of using one quantity for transferring the heat flux from the
in this framework. It was developed based on fds2ascii, a Fortran
fire model to advanced three-dimensional (3D) solid-phase pro-
program, to map the data from FDS to ANSYS (2020 R2). The
grams rather than using the one-dimensional (1D) heat-transfer
fds2ftmi collects from ANSYS the geometric information of ex-
model in FDS.
posed surfaces including the coordinates and normal directions.
To correct the different profiles of orthogonalized obstructions
Based on each keypoint position, it searches into the boundary
by FDS and the actual surface in the FEM model, another treatment
files (.bf) of FDS results, iterating over time and orientation to tran-
was introduced for inclined geometries. Fig. 2(b) exemplifies the
scribe the variables called the thermal exposure, TAST and h, into
vector composition of TAST in a two-dimensional (2D) space.
ASCII text files. For application convenience, the code outputs
Eq. (7) shows that TAST is orientation-dependent because it is re-
these variables into a file in ANSYS APDL format. The applicabil-
strained to radiative incident energy and the convective heat-
ity of the fds2ftmi code to transcribe thermal exposure from FDS to
transfer coefficient, which are both orientation-dependent. For a ge-
thermal FEM model has been validated by experiments and com-
neric 3D exposed surface with a normal vector of n, vAST , and vh
plex structures exposed to fire such as spheres and inclined panels
can be summed by the contributions in the directions of three coor-
(Silva et al. 2014, Zhang et al. 2016). More details about fds2ftmi
dinate axes:
can be found in Silva et al. (2016).
vAST = TAST,x · i + TAST,y · j + TAST,z · k
(9)
vh = hx · i + hy · j + hz · k Thermomechanical FEM Simulation

The thermal exposure can be considered as the projections of vAST Two approaches are available in the coupled thermomechanical
and vh onto n: FEM analysis. One is the direct coupling method. It handles the
problem by using a coupled-field element type containing all neces-
vAST · n sary degrees of freedom. The direct coupling method is advanta-
TAST,n = |vAST | · |n| · cos (vAST , n) = |vAST |
|vAST | · |n| geous in strongly coupled problems and requires less user
(10)
vh · n intervention. However, it only applies to limited types of bilinear
hn = |vh | · |n| · cos (vh , n) = |vh | solid elements, which will introduce a tremendous computational
|vh | · |n|
need for large-scale thin-walled steel structures. The other is the se-
Given an initial surface temperature, the total heat flux imposed quential coupling method. By applying results from one analysis as
on a generic structural surface can be calculated by using Eqs. (8) loads of another analysis, two or more analyses in different fields are
and (10), providing a thermal load to the following FEM performed sequentially and separately. This method is extensively
simulations. used for engineering structures because it is geometry-friendly and

© ASCE 04021074-5 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


time-frugal. The methodology proposed in this paper implemented
the sequential coupling method for the FEM simulation.
To minimize the computational cost, only the members ther-
mally affected by the fire were considered in the thermal FEM
model. The heat flux was evaluated automatically using Eq. (8), ac-
cording to TAST and h at each time step. On the exposed surface
sides, both of TAST and h were applied for calculating radiative
and convective heat fluxes, respectively.
It is important to note that FDS is also capable of calculating the
structure temperature using the 1D thermal conduction model.
However, the in-plane heat conduction is neglected. Studies like
Silva et al. (2016) have verified that using the 1D heat-transfer
model in FDS can introduce a considerable inaccuracy. The tem-
perature of the exposed portion can be overestimated, and the
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

neighboring region is usually underestimated in its temperature.


Also, some studies adopted the gas temperature to transfer the
heat from the FDS model to the FEM model. However, applying
the gas temperature can neither predict the surface temperature as
demonstrated by Zhang et al. (2013). Therefore, this methodology (a) (b)
implemented the FEM method to perform the 3D heat-transfer
analysis for predicting the temperature more precisely. Fig. 3. Experimental configuration (in cm) of Case 3 by Kamikawa
Determining the structural response should involve all the compo- et al. (2006): (a) front view; and (b) side view.
nents interacting with the exposed portion. In this paper, small bridges
were entirely considered in the structural FEM models. For large-
scale bridges, the entire model was established to consider the inter-
action between all components although the localized fire is imposed Numerical Results
on a few components. In this task, the multiscale FEM was adopted The fire scenario was replicated using FDS within the computational
to simultaneously provide both global and local mechanical charac- domain of 0.75 × 0.45 × 1.8 m in x-, y-, and z-directions, respectively.
teristics. The structural FEM model of thermally affected members The domain was discretized with 180,000 cubic cells in the dimen-
was discretized identically with the thermal model. Other members sion of 1.5 cm. The thermal and mechanical material model pre-
were modeled in simple configurations and connected to the well- scribed in CEN (2005a) was adopted in the FDS model and
discretized portion to ensure structural interaction. The necessity re- following FEM simulations. Fig. 4 visualizes the predicted fire be-
garding the entire modeling will be demonstrated by a complex case havior and its comparison with the observed flame. Note that the
study on a cable-stayed bridge, whose fire-induced response pre- flame oscillated drastically due to the turbulent combustion. Because
dicted by using the proposed approach was different from previous of the insulation clothes employed in experiments, the simulated tur-
studies. The time-dependent nodal temperatures obtained from the bulence slightly differs from the observed behavior.
thermal FEM simulation were imported as the thermal body load The thermomechanical response of the specimen was simulated in
for conducting the structural simulation considering both the me- ANSYS. Two-layered shell elements were utilized to discretize the
chanical and thermal loads. In this way, the structural behavior of FEM model with a side length of 1 cm. To reproduce the boundary
the entire bridge subjected to a localized fire was calculated. condition, the model bottom was fully constrained and the column
cap was restricted laterally. A constant force of 186 kN was applied
to the nodes of the square section at the specimen top.
Experimental Validation Fig. 5 compares the thermomechanical response predicted by
the coupled models to the measured values. The thermal response
includes the temperature evolutions at the centroids of the front
Test Description
panel, side panel, and back panel, and the vertical deformation at
The experimental study carried out by Kamikawa et al. (2006) was the column cap representing the structural response. The simulated
validated using the methodology proposed in this paper. In the seri- temperature evolution agrees well with the tested specimen whose
ous tests, a burner was placed upward adjacent to the base of steel temperature increased to steady at about 20 min and decreased
columns applied with different boundary conditions. Case 3 was re- quickly at 60 min when the burner was extinguished. Regarding
produced because the tested specimen was applied with a constant the structural performance, the vertical deformations at the column
axial force, corresponding to the loading characteristics of the stiff- top were very close before 20 min and began to recover after
ening girder exposed to a localized fire in the cable-stayed bridge as 60 min. The slight difference is mainly attributed to the insulation
investigated below. clothes and instrumental deviations, which cannot be considered
In Case 3, as shown in Fig. 3, the 1.6-m-tall column specimen precisely in the FEM model. Overall, the proposed methodology
had an STKR400-type section with a side length of 0.1 m and a is capable of reproducing the fire-thermomechanical response of
thickness of 3.2 mm. A propane burner with a footprint of 30 cm the specimen in Case 3 (Kamikawa et al. 2006).
by 30 cm and a height of 25 cm was used as the fire source. During
the test, the column bottom was fixed to the ground, and a 186-kN
vertical load was maintained at the specimen cap when the burner Complex Case Study: Cable-Stayed Bridge
released a steady 52.5-kW fire lasting 60 min. There was also a
60-min cooling phase after heating. The behavior measurement in- To demonstrate the application of the proposed methodology in
volved the body temperature at four surfaces and the vertical defor- predicting the fire-induced response for various bridges, a cable-
mation at the column cap. stayed bridge exposed to localized under-deck tanker fire was

© ASCE 04021074-6 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


numerically investigated. This is a more complex case than general steel plate system is composed of a roof, two webs, a flat floor,
beam bridges because cable-stayed bridges usually involve criss- two tilt floors, trapezoidal shape stiffeners, and diaphragms. The
cross panels that cannot be handled by current methods. supporting system includes vertical bearing at the transitional
pier, auxiliary pier, and pylon crossbeam without any longitudinal
or lateral constraints.
Prototype Bridge
The prototype cable-stayed bridge has a span of 63 + 257 + 648 +
Material Model
257 + 63 m, as shown in Fig. 6. The bridge deck consists of 89 seg-
ments with a height of 3.2 m and a width of 37.2 m, which are sup- Steel is temperature-sensitive in that its thermal properties vary and
ported by two-plane stayed cables every 12 m in the auxiliary span mechanical properties also deteriorate significantly in a fire envi-
and 15 m in the main span and side span. The Q345 orthotropic ronment. Standards such as CEN (2005a) and AISC (2005) have
prescribed these material properties at high temperatures. To be
more representative, this study adopted the material property mod-
els at high temperatures specified in CEN (2005b) for the steel
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

decks, which are shown in Fig. 7.

Fire Scenario
Fire behaviors are complex and different fire sources can be very
distinct. The engineering approach quantifies the burning intensity
as the firepower, equaling to the product of the HRRPUA and the
fire area. The fire development can be represented by the HRRPUA
curve, which is divided into the growth phase, steady phase, and
decay phase (Karlsson and Quintiere 2000). The t-squared function

Q̇ = αt 2 (11)

is widely used to describe the initial growth stage because the fire is
nearly accelerating. In the steady phase, the fire is fuel-bed-controlled,
which will reach a maximum burning rate with sufficient oxygen sup-
port. Fig. 8 compares the HRRPUA curves adopted in existing studies
during the first 6 min. Some of these fires were more intensive than
Fig. 4. Comparison between simulated and observed fire behaviors in the ultrafast fire classified in Karlsson and Quintiere (2000), which
Case 3 by Kamikawa et al. (2006). is characterized by α = 0.19 kW/s2. Previous studies adopted the
HRRPUAMAX ranging from 1,600 to 4,500 kW/m2. This study

Fig. 5 Comparison of the thermomechanical behavior of the specimen in Case 3 by Kamikawa et al. (2006).

© ASCE 04021074-7 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


(a)
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

(b)

Fig. 6. Profile of prototype cable-stayed bridge: (a) elevation view; and (b) section view.

(a) (b)

(c) (d)

Fig. 7. Temperature-dependent material properties of Q345 steel: (a) conductivity; (b) specific heat; (c) thermal elongation; and (d) stress–strain
relationships.

© ASCE 04021074-8 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


steady state that the flame profile oscillated promptly due to the fire-
driven plume.
Fig. 10 shows the results of FDS-predicted thermal exposure,
whose TAST is the adiabatic surface temperature defined in
Eq. (3) and h is the convective heat-transfer coefficient. Unlike pre-
scriptive curves, the FDS-based fire model reconstructs an inhomo-
geneous thermal exposure to the segments above the fire source.
The evolutions of TAST and h at the floor centroid show that
TAST reached an oscillated steady state after 1 min when the
HRRPUA became constant. h also oscillated but was not steady be-
cause it depends on the velocity field, which changed drastically
with the plume.
Figs. 10(b and d) show the distributions of the averaged TAST
and h of the exposed floor, respectively. It is shown that TAST
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

above the flame was nearly 1,200°C, which was much higher
than that at other locations, and the thermal gradient was also the
largest in the central region. h was about 12 W/(m2 · K) at the
floor centroid, slightly lower than other locations. Overall, the ther-
Fig. 8. HRRPUA curves adopted in this paper and existing studies. mal exposures were symmetric about the transverse direction, and
the slight longitudinal asymmetry of thermal exposures was caused
by the deck slope and vector compositions.

Thermal Response of Fire-Affected Deck Segments


The thermal FEM model of the fire-affected portion involved the same
segments as those considered in the FDS model, including main
plates, ribs, and diaphragms. The four-noded 3D layered thermal
shell element, SHELL131, was used for discretizing the steel compo-
nents. It can tackle transient-state and steady-state thermal problems.
To impose these loads, the surface effect element SURF152 was
used. They were overlaid on SHELL131 elements sharing their
nodes with one extra node beyond. Adopting elements of
SHELL131 and SURF152 in ANSYS to simulate the thermal perfor-
mance of fire-exposed structures has been verified by adequate studies
Fig. 9. FDS model and steady flame behavior.
(Kodur et al. 2013; Zhang et al. 2020). The thermal FEM model had
116,381 shell elements and 232,762 surface effect elements in total to
consider the 3D thermal conduction and radiation.
assumed an HRRPUAMAX of 2,800 kW/m2 reached in 60 s following On the outside of the exposed floor, TAST was prescribed to the
the t-squared function. The combustion lasted 30 min in total. extra nodes and h was applied on the element surfaces for the cal-
This paper considered a tanker fire that occurred 18 m below culations of radiative and convective heat fluxes, respectively. The
the floor of the #26 deck segment. The fire source was close to roof panel of thermally influenced girder segments was constrained
the south pylon that might be caused by ship collisions with the to 20°C because of the considerable specific heat of the overlaid
pylon or terrorist attacks. The dimensions of the tanker were 15, pavement exposed to the ambient environment. Applying the ther-
6, and 3 m in the lateral, longitudinal, and vertical directions of mal boundaries obtained through fds2ftmi from FDS to the FEM
the cable-stayed bridge, respectively. According to the HRRPUA model, the transient-state thermal response of the fire-affected
curve defined as Eq. (5), the steady firepower was 252 MW. deck segments was determined with a maximum time step of 5 s
Note that this fire scenario was designed to demonstrate the simu- considering radiation and convection. The layered nodal tempera-
lation process instead of reproducing a real fire accident. ture results were the basis of the subsequent analysis of global
structural behavior.
Fig. 11 shows the temperature distribution of the exposed deck
Simulating the Bridge Response
floor at different exposure times. The temperature of the floor was
Fire Behavior significantly influenced by the fire source, which was 18 m below.
In the FDS model shown in Fig. 9, five deck segments were mod- The central region of the deck floor heated up first and then the tem-
eled: #26 segment, which was above the fire source, and two seg- perature propagated outward. At the fire extinguishment at 30 min,
ments on both ends. The calculation was performed within the a considerable district above the flame reached over 700°C.
computational parallelepiped of 44.4 m (X ) × 24.6 m (Y ) × 75.6 m The temperature distribution of the floor has value valleys along
(Z) to consider the heat transfer between the fire, deck, and environ- with diaphragms because diaphragms can absorb heat from the ex-
ment. The domain was discretized using uniform 0.7 m grids, con- posed floor.
taining 276,480 control volumes in total. The boundary results of Fig. 12 compares the evolutions of steel temperatures at six
obstructions were recorded every 10 s. locations on the deck floor. The distances between adjacent mea-
The fire behavior during steady combustion is shown in Fig. 9. suring points were 3 and 7 m in lateral and longitudinal directions,
The orthotropic deck including the slope was transformed into respectively. It is shown that the points closer to the fire center had
stair-stepping obstructions aligning with meshing grids. It is higher temperatures and elevation rates. Centroid P1, which is clos-
shown that with a steady power of 252 MW, the flame reached est to the fire, had the highest temperature reaching 436°C at
the deck floor 18 m above. Note that the fire was in a dynamic 10 min and up to 748°C at 30 min because it is exposed to the

© ASCE 04021074-9 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


(a) (b)
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

(c) (d)

Fig. 10. Thermal exposures: (a) TAST evolution at the centroid of the flat floor; (b) distribution of averaged TAST at the floor; (c) h evolution at the
centroid of the flat floor; and (d) distribution of averaged h at the floor.

Fig. 12. Evolution of surface temperatures at representative points of


exposed floor.

most heat-carrying gases. At further locations, temperatures in-


creased due to the lesser fire exposure. The temperature lags that
P2 behind P1, P4 behind P3, and P6 behind P5 decreased as
their longitudinal distances from the fire source increased. This
Fig. 11. Isotherms of the floor in fire-affected deck segment at different
demonstrates that the thermal gradient also decreased on further
exposure times.
surfaces. Note that the temperatures at P1 and P2 fluctuated because

© ASCE 04021074-10 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Multiscale structural FEM model of the entire bridge.

the nodal temperatures were forced to follow the oscillations of within the central region reduced significantly to below 19 MPa.
TAST and h, which were applied as thermal boundaries. This stress reduction is caused by the decrease of the ultimate stress
led by the temperature elevation of steel, as shown in Figs. 7(d) and
Structural Response of the Entire Bridge 12. On the other hand, the stress increment of the floor is due to the
The mechanical response of the prototype bridge was calculated unbalanced expansion of the roof and floor in decks. The upper roof
from the multiscale structural FEM model of the entire bridge, as keeps at ambient temperature because no direct fire exposure was
shown in Fig. 13. The fire-affected deck segments were discretized imposed, whereas the temperature at the lower floor increased a
with four-node shell elements using the SHELL181 element type in lot. Confined by the tension force of stayed cables, the mean stress
ANSYS, and the fish-spine modeling technique was adopted for of the floor was enhanced to 158 MPa due to the expansion, and the
other segments using BEAM188. On the interfaces of beam ele- roof stress was partially released because of its temperature lag be-
ments and shell elements, the nodes were coupled with constraint hind the floor.
equations to transfer interactive actions. Pylons were also discre-
tized with beam elements. Stayed cables were modeled using
LINK180 elements to merely consider the axial tension force. Global Response
They were connected to pylons and decks using rigid beams to
The results of the deflection and internal force of the stiffening
transfer cable actions. These element types are commonly utilized
decks at four representative times are presented in Fig. 16, in
in global bridge analyses with good accuracy and profitable run-
which the deck segments modeled with shell elements are shaded.
ning time. There were 118,418 elements in total in the multiscale
Note that the internal force of the decks modeled using beam ele-
FEM model of the entire bridge. The structural performance was
ments was captured at the element midpoints, and the internal
determined by applying the nodal temperatures obtained previously
as the body load taking into account the geometric nonlinearity and force of the segments discretized with shell elements was calculated
the nonlinear material model shown in Fig. 7. by composing the force components to the section centroids that
were the same as that of beam elements.
The five shell-shaped deck segments above the fire source de-
Local Response formed most drastically due to the stiffness reduction caused by
The fire-induced structural impact was directly imposed on the high temperatures, as shown in Fig. 7. The maximum vertical dis-
girder segments through the material transformation. Figs. 14 and placement appeared above the fire centerline and was 21.0, 32.3,
15 show the Mises stress of the roof and floor above the flame at and 38.5 cm at 10, 20, and 30 min, respectively. Apart from that,
various exposure times, respectively. The deck floor experienced many girder segments had a considerable compensatory deforma-
two stress evolution stages. The first stage is from the completion tion: the decks in the south side span and midspan were lifted
state of the bridge to 10 min. The bottom stress before the fire and the decks in the north side span deflected downwards.
was evenly distributed at about 57 MPa. After the 5-min fire expo- The axial force in girders was almost unaffected by the fire. This
sure, the central stress increased up to 227 MPa, and the peripheral is because there are no constraints in the longitudinal direction and
stress also increased. The averaged stress within the studied portion no obvious secondary axial force was introduced by the short-
was almost doubled to 110 MPa at 10 min. exposed length, which was 1/17 of the full span. This observation
During the second stage from 10 to 30 min, the surrounding is different from previous studies (Gong and Agrawal 2016) be-
stress still raised and reached up to 284 MPa, whereas the stress cause the global interaction was considered in this study.

© ASCE 04021074-11 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Mises stress of floor in exposed deck segments at different times.

Fig. 15. Mises stress of roof in exposed deck segments at different times.

Fig. 16(c) shows that a remarkable increment in bending moments reduced up to 15% of the tension force, and the force of further ca-
was caused to the decks near the fire source. Before the fire accident, bles varied no more than 1%.
the bridge decks had reasonable bending moments that were limited As for the pylons, Fig. 18 compares the longitudinal displace-
at most positions and comparatively larger near the piers and the mid- ment and internal force at four representative times. The results of
span of auxiliary spans. Exposed to the fire underneath, the deck mo- both south and north pylons are shown side by side referring to
ment above the fire centerline jumped from 6.5 MN · m before the the annotated midspan. It is shown that the two pylons had differ-
fire to 88.3 MN · m after only 10 min. The bending moment contin- ent deformed directions under the fire condition. The south pylon,
uously increased up to 139.5 and 170.1 MN · m at 20 and 30 min, which is on the side that fire occurs, deformed southward, and the
respectively. Also, the bending moments (upper edge under tension) north pylon was also pulled to the midspan with a magnitude
of the deck segments between the anchors of Cables S13 to S30 were smaller than the south pylon. At the pylon tops, the south and
significantly increased by the fire, but the decks at other locations north pylons deformed 20.5 and 15.5 cm after the 30-min fire ex-
were not affected obviously. posure, respectively. The pylons kept their axial force during the
Although the stayed cables were not directly exposed to fire, the fire because the loads applied to them only included their gravity
cable force was influenced notably. As shown in Fig. 17, the axial and the deck gravity transferred from cables, both of which did
force of the three cables neighboring the fire centerline was in- not change.
creased. The axial force of Cable S20 increased by 32% and 59% However, a great variation was introduced to longitudinal
after 10 and 30 min, respectively. This force increment is caused bending moments. The bottom bending moment of the south
by the reduction of the corresponding deck segments shown in pylon decreased, and the moments of the pylon sections anchored
Fig. 16(a). The neighboring cables, S22–S26 and S11–S16, by the cables close to fire significantly increased. Also, the bottom

© ASCE 04021074-12 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


(a)
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

(b)

(c)

Fig. 16. Structural behaviors of decks at different times: (a) deflection; (b) axial force; and (c) bending moment.

Fig. 17. Cable force variation at different times.

bending moment of the north pylon, which is far away from the roof and floor in the exposed deck segments, including the time his-
fire source, increased 1.06 times at 30 min, endangering its struc- tories of centroids and the distributions at 30 min. Basically, be-
tural safety and even of the global bridge. Because the two pylons cause of the strong connections from vertical diaphragms, the
are not constrained in the longitudinal direction, the significant roof and floor had a similar deformation profile that the central
variation of their bending moments is caused by the variation of part deflects most severely. Their difference is the central region
cable force. For the north pylon, the slight variation of the tension of the floor was corrugated by fire, which means the material had
force of the anchoring cables gives rise to a prominent change of deteriorated to its limit before failure. This is due to the stiffness
the bending moment at the pylon bottom because its mechanics loss caused by the high temperature in the central region, which
characteristic is close to a column, which is very sensitive to lat- is demonstrated in Figs. 7(d) and 11.
eral actions. Deformation evolutions for the centroids of the roof and floor ex-
Compared to existing studies, it is shown that more comprehen- perienced three stages. From the fire ignition to 7 min, the centroids
sive results were obtained considering the cooperative actions of all had a consistent deflection. The thermal expansion is mainly respon-
structural members. Thereby, a global view is required for deter- sible for the deformation because the thermal properties varied much
mining the performance of large-scale bridges in localized fires more significantly than the mechanical property when the surface tem-
by undertaking a multiscale simulation for the entire bridge. perature was below 300°C, as demonstrated in Figs. 7 and 12. During
the second phase from 7 to 16 min, the floor deformed more appar-
ently than the roof because of the temperature-induced material dete-
Mechanism of Behavior Evolution
rioration. However, the temperature propagated slower after 16 min
To analyze the mechanism of the structural behavior evolution when the thermal conduction exceeded radiation becoming the main
caused by fire, Fig. 19 shows the vertical displacement of the manner of heat transfer. The central floor region was softened

© ASCE 04021074-13 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

(a) (b) (c)

Fig. 18. Structural behaviors of the pylons at different times: (a) longitudinal deformation; (b) axial force (positive representing compression); and
(c) bending moment.

(a) (b)

Fig. 19. Vertical deformation of the roof and floor in fire-exposed segments: (a) time histories; and (b) distributions.

seriously and was corrugated by stiffeners and diaphragms. Because respectively, and the bending moment of the deck above the fire in-
the floor temperature became more steady, as shown by the plateau creased 81.8, 51.2, and 30.6 MN · m, respectively. It is seen that
in Fig. 12, the behavior evolution of the exposed surface became the increments of all these quantities decreased as the fire
less apparent. developed.
In addition, the centroids deformed quickly first but then slowly, Because TAST reached steady very quickly after ignition, which
which is similar to the performance variation of the global bridge. is shown in Fig. 10(a), the temperature difference between the steel
The greatest behavior evolution of the global structure was ob- and thermal boundary was the greatest at the early stage of the fire.
served in the first 10-min combustion comparing the results of The temperature elevation during this stage is the fastest thereby, as
the structural components shown in Figs. 16–18. For example, in demonstrated in Fig. 12. Also, after the central region of the deck
the three 10-min time intervals of the 30-min combustion, the floor reaches a very high temperature, the thermal conduction
deck above the fire centerline deflected 21.0, 11.3, and 6.2 cm, through the plates accounts for the major contribution to the

© ASCE 04021074-14 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


temperature elevation and the structural performance evolution. Acknowledgments
Therefore, the deterioration of the structural behavior of the global
bridge is slowing down as the fire develops. This study was financially supported by the Natural Science Foun-
Figs. 14 and 15 demonstrate that the unbalanced stress development dation of Jiangsu Province (Grant No. BK20181278 and
in the girder resulted in a dramatic increment in the bending moments BK20201274). Thanks to Dr. Kevin McGrattan at NIST for his
above the fire: the stress in the exposed floor was elevated seriously due support in improving the FDS codes regarding the simulation of
to expansion, whereas the roof stress was decreased, composing a sealed obstructions subjected to ultra-strong fires. Valuable sugges-
much higher bending moment comparing with the state before the tions from Xiuqi Xi at University College London are also
fire. This behavior evolution in girders further gives rise to the variation acknowledged.
of the internal force in the entire bridge, including the force release of
cables and increment of bending moments in pylons.
During both thermal and structural simulations, the running pro-
cesses converged until the fire extinguishment at 30 min. However, Supplemental Materials
this does not indicate the bridge can survive the fire. For instance, the
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Łazienkowski Bridge in Warsaw resisted the strong fire exposure in Additional data is available online in the ASCE Library (www
2015. However, the bridge was rated as the high level damage by the .ascelibrary.org).
fire accident and it was rebuilt totally. Thereby the criteria for fire-
exposed bridges are complex requiring specific and comprehensive
insights. This paper does not identify failure criteria for the investi- References
gated cable-stayed bridge, and instead, the purpose of this paper is to
propose the simulation methodology, providing a numerical basis for Alos-Moya, J., I. Paya-Zaforteza, M. E. M. Garlock, E. Loma-Ossorio,
evaluating the bridge responses related to fires. D. Schiffner, and A. Hospitaler. 2014. “Analysis of a bridge failure due
to fire using computational fluid dynamics and finite element models.”
Eng. Struct. 68: 96–110. https://doi.org/10.1016/j.engstruct.2014.02.022.
Conclusions Alos-Moya, J., I. Paya-Zaforteza, A. Hospitaler, and E. Loma-Ossorio. 2019.
“Valencia bridge fire tests: Validation of simplified and advanced numer-
ical approaches to model bridge fire scenarios.” Adv. Eng. Software 128:
This paper proposes a numerical methodology for predicting the re-
55–68. https://doi.org/10.1016/j.advengsoft.2018.11.003.
sponse of fire-exposed bridges by coupling the CFD-based fire Alos-Moya, J., I. Paya-Zaforteza, A. Hospitaler, and P. Rinaudo. 2017.
model and FEM-based thermomechanical model. First, the fire sce- “Valencia bridge fire tests: Experimental study of a composite bridge
nario is simulated in FDS within the fire-affected spatial domain. under fire.” J. Constr. Steel Res. 138: 538–554. https://doi.org/10
Then, the thermal exposure is extracted from FDS using a validated .1016/j.jcsr.2017.08.008.
interface. At last, the time-variant thermal and structural behavior is AISC (American Institute of Steel Construction). 2005. Specification for
captured utilizing the sequentially coupled thermomechanical FEM structural steel buildings. Chicago: AISC.
simulation. Aziz, E., and V. Kodur. 2013. “An approach for evaluating the residual
Unlike existing studies, this approach does not simplify the fire strength of fire exposed bridge girders.” J. Constr. Steel Res. 88: 34–42.
condition as spatially uniform temperature curves, and instead, re- https://doi.org/10.1016/j.jcsr.2013.04.007.
produces an inhomogeneous surrounding temperature field by con- Aziz, E. M., V. K. Kodur, J. D. Glassman, and M. E. Maria Garlock. 2015.
“Behavior of steel bridge girders under fire conditions.” J. Constr. Steel
sidering the fire-driven heat flow, obstructions, and environment.
Res. 106 (2015): 11–22. https://doi.org/10.1016/j.jcsr.2014.12.001.
This method is validated by predicting a close thermomechanical Beneberu, E., and N. Yazdani. 2018. “Performance of CFRP-strengthened
response with the measured values in a fire-column test, whose concrete bridge girders under combined live load and hydrocarbon fire.”
specimen had a similar loading characteristic with the stiffening J. Bridge Eng. 23: 04018042. https://doi.org/10.1061/(ASCE)BE.1943
girder in cable-stayed bridges. -5592.0001244.
To demonstrate the application of the proposed methodology, Buchanan, A. H., and A. K. Abu. 2017. Structural design for fire safety.
this paper performs a complex case study on a prototype long-span Hoboken, NJ: Wiley.
steel cable-stayed bridge with an orthotropic deck system, which CEN (European Committee for Standardisation). 2002. Eurocode 1:
was assumed to be exposed to an under-deck tanker fire. The gra- Actions on structures—Part 1–2: General actions—Actions on struc-
dient thermal response of fire-affected segments is simulated, and tures exposed to fire. Brussels, Belgium: CEN.
the variation of the global structural performance is captured. Re- CEN (European Committee for Standardisation). 2005a. Eurocode 3:
Design of steel structures—Part 1–2: General rules structural fire de-
sults show that a localized fire can introduce a significant structural
sign. Brussels, Belgium: CEN.
behavior evolution to the entire bridge, endangering the fire- CEN (European Committee for Standardisation). 2005b. Eurocode 4:
exposed deck segments and pylons by increasing their bending mo- Design of composite steel and concrete structures—Part 1–2:
ments dramatically. General rules—Structural fire design. Brussels, Belgium: CEN.
Undertaking a multiscale FEM simulation is necessary for large- Choi, J. 2008. “Concurrent fire dynamics models and thermomechanical
span bridges in localized fires. The interactions between fire- analysis of steel and concrete structures.” Ph.D. thesis, School of
exposed members and other components can thus be considered. Civil and Environmental Engineering, Georgia Institute of Technology.
With this methodology that can programmatically tackle complex Choi, J., R. Haj-Ali, and H. S. Kim. 2012. “Integrated fire dynamic and
geometries with an acceptable time cost, various bridge structures thermomechanical modeling of a bridge under fire.” Struct. Eng.
subjected to fires can be numerically investigated. Mech. 42 (6): 815–829. https://doi.org/10.12989/sem.2012.42.6.815.
Cui, C., A. Chen, and R. Ma. 2020. “Stability assessment of a suspension
bridge considering the tanker fire nearby steel–pylon.” J. Constr. Steel
Res. 172: 1–9.
Data Availability Statement NIST. 2020. “Fire dynamics simulator.” Gaithersburg, MD: NIST.
Garlock, M., I. Paya-Zaforteza, V. Kodur, and L. Gu. 2012. “Fire hazard in
All data, models, or code that support the findings of this study are bridges: Review, assessment and repair strategies.” Eng. Struct.
available from the corresponding author upon reasonable request. 35 (2012): 89–98. https://doi.org/10.1016/j.engstruct.2011.11.002.

© ASCE 04021074-15 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074


Gong, X., and A. K. Agrawal. 2015. “Numerical simulation of fire damage Silva, J. C. G., A. Landesmann, and F. L. B. Ribeiro. 2014.
to a long-span truss bridge.” J. Bridge Eng. 20 (10): 04014109. https:// “Performance-based analysis of cylindrical steel containment vessels
doi.org/10.1061/(ASCE)BE.1943-5592.0000707. exposed to fire.” Fire Saf. J. 69: 126–135. https://doi.org/10.1016/j
Gong, X., and A. K. Agrawal. 2016. “Safety of cable-supported bridges .firesaf.2014.08.013.
during fire hazards.” J. Bridge Eng. 21 (4): 04015082. https://doi.org Silva, J. C. G., A. Landesmann, and F. L. B. Ribeiro. 2016.
/10.1061/(ASCE)BE.1943-5592.0000870. “Fire-thermomechanical interface model for performance-based analy-
Harik, I. E., A. M. Shaaban, H. Gesund, G. Y. S. Valli, and S. T. Wang. sis of structures exposed to fire.” Fire Saf. J. 83: 66–78. https://doi.org
1990. “United States bridge failures, 1951–1988.” J. Perform. /10.1016/j.firesaf.2016.04.007.
Constr. Facil 4 (4): 272–277. https://doi.org/10.1061/(ASCE)0887 Wardhana, K., and F. C. Hadipriono. 2003. “Analysis of recent bridge
-3828(1990)4:4(272). failures in the United States.” J. Perform. Constr. Facil 17 (3):
Kamikawa, D., Y. Hasemi, K. Yamada, and M. Nakamura. 2006. 144–150. https://doi.org/10.1061/(ASCE)0887-3828(2003)17:3(144).
“Mechanical responses of a steel column exposed to a localized fire.” Wickström, U., D. Duthinh, and K. B. McGrattan. 2007. “Adiabatic surface
In Proc., 4th Int. Workshop on Structures in Fire, 225–234. Aveiro, temperature for calculating heat transfer to fire exposed structures.”
Portugal: Universidade De Aveiro Campus Universitário De Santiago. In Proc., 11th Int. Interflam Conf., 943–953. London: Interscience.
Karlsson, B., and J. G. Quintiere. 2000. Enclosure fire dynamics. Boca Wu, X.-q., T. Huang, F. T. K. Au, and J. Li. 2020. “A localized fire model
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Raton, FL: CRC Press. for predicting the surface temperature of box girder bridges subjected to
Kodur, V., E. Aziz, and M. Dwaikat. 2013. “Evaluating fire resistance of tanker truck fire.” Fire Technol. 56: 2059–2087.
steel girders in bridges.” J. Bridge Eng. 18 (7): 633–643. https://doi Zhang, C., G.-Q. Li, and R. Wang. 2013. “Using adiabatic surface temper-
.org/10.1061/(ASCE)BE.1943-5592.0000412. ature for thermal calculation of steel members exposed to localized
Peris-Sayol, G., I. Paya-Zaforteza, J. Alos-Moya, and A. Hospitaler. 2015. fires.” Int. J. Steel Struct. 13 (3): 547–556. https://doi.org/10.1007
“Analysis of the influence of geometric, modeling and environmental /s13296-013-3013-2.
parameters on the fire response of steel bridges subjected to realistic Zhang, C., J. G. Silva, C. Weinschenk, D. Kamikawa, and Y. Hasemi.
fire scenarios.” Comput. Struct. 158: 333–345. https://doi.org/10.1016 2016. “Simulation methodology for coupled fire–structure analysis:
/j.compstruc.2015.06.003. Modeling localized fire tests on a steel column.” Fire Technol.
Quiel, S. E., T. Yokoyama, L. S. Bregman, K. A. Mueller, and 52 (1): 239–262. https://doi.org/10.1007/s10694-015-0495-9.
S. M. Marjanishvili. 2015. “A streamlined framework for calculating Zhang, G., V. Kodur, C. Song, S. He, and Q. Huang. 2020. “A numerical
the response of steel-supported bridges to open-air tanker truck model for evaluating fire performance of composite box bridge girders.”
fires.” Fire Saf. J. 73: 63–75. https://doi.org/10.1016/j.firesaf.2015 J. Constr. Steel Res. 165: 105823. https://doi.org/10.1016/j.jcsr.2019
.03.004. .105823.

© ASCE 04021074-16 J. Bridge Eng.

J. Bridge Eng., 2021, 26(10): 04021074

You might also like