You are on page 1of 13

Engineering Structures 32 (2010) 3180–3192

Contents lists available at ScienceDirect

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

Numerical simulation of a cable-stayed bridge response to blast loads, Part I:


Model development and response calculations
Edmond K.C. Tang, Hong Hao ∗
School of Civil and Resource Engineering, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia

article info abstract


Article history: Many researchers have conducted comprehensive experimental and numerical investigations to examine
Received 7 August 2009 civilian structures’ response to explosive loads. Most of the studies reported in the literature deal with
Received in revised form building structures and structure components. Studies of bridge structures subjected to blast loads are
8 June 2010
limited. This study performs numerical simulations of dynamic responses of a large cable-stayed bridge
Accepted 8 June 2010
Available online 10 July 2010
under explosive loadings. All numerical simulations are carried out using the LS-DYNA explicit finite
element code. This paper describes the bridge under consideration, blast load estimation, finite element
Keywords:
model, material model, and detailed numerical simulation results of the bridge to blast loads from a 1000
Cable-stayed bridge kg TNT equivalent explosion at 0.5 m from the bridge tower and pier, and 1.0 m above the deck. Damage
Blast loading mechanism and severity of the bridge tower, pier and deck are examined. The companion paper Hao and
Numerical simulations Tang (2010) [19] presents intensive numerical simulation results of the bridge components under blast
Damage mechanism loads of different scaled distances, progressive collapse analyses of the bridge after either one of the four
Bridge components main bridge components is damaged, and the safe scaled distance for bridge protection before initiating
catastrophic collapse. The effectiveness of FRP strengthening of bridge concrete back span for blast load
resistance is also investigated.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction easy accessibility and likelihood of causing the most undesirable


impact on human lives along with economical disaster [12]. This
Responses of concrete structural components under blast enlightens the importance for researchers and structural engineers
loadings are commonly evaluated using simplified approaches to gain a better understanding of bridges’ performance when they
such as the elastic plate theory [1,2], Timoshenko beam theory are exposed to high intensive blast loads.
[3,4] and analysis of equivalent single degree of freedom system Two of the simulations of bridge structure responses to blast
(SDOF) [5,6]. These approaches are very efficient and also yield loads were conducted by Winget et al. [13] and Islam and Yaz-
good predictions of structural responses. However, as stated in Shi dani [14] who had separately investigated the response of an
et al. [7], it is difficult to define an appropriate damage criterion AASHTO concrete girder bridge under blast impact using uncou-
for these approaches and they may not give reliable predictions pled SDOF system and commercial software STAAD.Pro, respec-
of localized damages. Consequently, with the advancement in tively. Both researchers considered various detonation positions in
computer technology and computational mechanics techniques, order to gain a better perspective of the bridge performance against
more and more investigations of blast effect on structures blast loadings. Results indicated that the studied bridge type is
now employ physics based finite element model in conducting highly vulnerable to failure under the impact of conventional
structure response simulations, and these numerical simulations truck bomb. In addition, Winget et al. [13] concluded that bridge
are proved yielding to reliable structural response predictions response under blast loads is highly dependent on the bridge ge-
[8–11]. ometry, such as clearance of the bridge deck from ground and
Although efforts have been spent on modelling blast effects the confinement effect resulted from the deep girder, as these
on civilian structures, mainly on building structures, relatively factors can significantly enhance the magnitude of blast pressure
less attention has been put towards bridge structures under acting on the structure. Therefore blast resistant bridge design
blast loads. It has become evident in the past years that bridge and implementation of retrofitting option needs to be developed
structures are possible targets of terrorist attacks owing to their and adopted in bridge construction. Throughout the years, design
guidelines have been proposed to improve the performance of
concrete girder bridges against extreme dynamic loadings [15,16].
∗ Corresponding author. Tel.: +61 8 6488 1825; fax: +61 8 6488 1018. However, these recommendations are made based primarily on
E-mail address: hao@civil.uwa.edu.au (H. Hao). findings from previous terrorist incidents on commercial buildings.
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2010.06.007
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3181

Hence, there is a possibility that the recommendations may not be


applicable to bridges as their structural design and behaviour are
vastly different from each other and from building structures.
Unlike ordinary concrete girder bridges, above deck explosion
on cable supported bridges may not only give rise to a localized
crater on bridge deck but also detachment of cables resulted from
concrete weakening at the anchor locations. Thus, this action
can lead to instability of the overall structure when the main
supporting cables are damaged. Knowing this fact, Son et al. [17]
performed nonlinear finite element analysis to investigate the blast
resistibility of steel orthotropic box deck of different steel materials
subjected to a typical vehicle bomb. Numerical calculations
showed that the orthotropic box decks constructed with high
ductility steel material gave better performance in comparison
to high strength steel. This is attributed to the greater brittle
behaviour of high strength steel under dynamic loadings which
leads to a poorer performance in blast events.
Suthar [18] studied the effect of combination of dead, live
and blast load on Chesapeake Bay Suspension Bridge using the
commercial software SAP2000. Blast loads of 500 pounds TNT
were placed at three different positions along the deck. The
author also carried out progressive collapse analysis by using
nonlinear hinges to redistribute the load to other members when
deformation capacities were reached. From the bending moments
and structural member deformations, conclusion was drawn that
although the suspension bridge experienced severe localized
damage resulted from the blast load, collapse of the suspension
bridge was unlikely for all the blast events considered.
The previous studies of cable supported bridge responses to
blast loads have primarily focused on the above deck explosion
events and ignored other possible blasting scenarios that could
result in bridge collapse. This study performs intensive numerical
simulations of dynamic response and damage of a large cable-
stayed bridge to blast loads. Owing to the page limitation, the
results are presented in two parts. This paper, the first part,
presents the detailed bridge information, blast loading simulation,
finite element model of the bridge developed using the commercial
software LS-DYNA, material models, and detailed numerical
simulations to predict damage of the four main components Fig. 1. Cable-stayed bridge model and cross section of individual component (not
to scale).
of the bridge, namely the pier, tower, concrete back span and
steel–concrete composite main span, subjected to typical blast
loads from a truck explosion (1000 kg TNT at 0.5 m distance of the bridge, such as the reinforcement ratio, reasonable assump-
from the tower and pier, and 1.0 m above the deck). The tions are made in the study. The model considered is therefore
simulated results are used to assess the performance of the not exactly the bridge under consideration, but a representative of
bridge components under these typical truck bombing loads. The large cable-stayed bridges. However, any assumption made in the
damage processes to these bridge components are observed and study is clearly stated. Other data given in the paper reflects the
true information of the bridge.
damage mechanisms are discussed. The accompany paper [19]
The 298 m tall towers are shaped such that their base is in a
of this study presents comprehensive numerical simulations of
circular shape with a diameter of 18 m and elongated by 6 m long
bridge responses to blast loads of different scaled distances,
straight sides. Both the straight and radius of the curved sections
determines the safe scaled distances to bridge components for
reduce with the tower height. The tower is circular with a diameter
bridge protection, performs progressive collapse analysis of the
of 14 m at the deck level (74 m) and the diameter is further
bridge under dead and live loads (traffic loads) after damage to
reduced to 11 m at 175 m above the ground level. This diameter
either one of the bridge components has occurred, and investigates
is maintained a constant for the full height at 298 m above the
the effectiveness of FRP strengthening for improving the bridge
ground. The wall thickness of the tower is a constant of 2 m up
blast loading resistance capacities. The results obtained in this
to the deck level and then reduces to 1.4 m at the 175 m level
study give information for owners of similar type of bridges on
and remains a constant hereafter. The piers are designed with twin
the probable responses when subjected to intensive blast loads and
hollow box sections. Their dimensions are 10 m in width and 4 m
can assist engineers to choose the most effective retrofit measures
depth with walls designed to be 0.5 and 1 m thick, as shown in
to bring in better resistance against blast loads.
Fig. 1. Both the back spans and main span of the twin deck system
are constructed to have a total width of 53.5 m, with a 21 m wide
2. Cable-stayed bridge model deck for traffic flows on each direction and joined together by
concrete or steel girders of 5.75 m on each side between them. The
2.1. Bridge details back spans of this bridge are concrete hollow box section girders
with both the flange and the web thickness assumed to be 0.5 m.
The dimensions of the cable-stayed bridge under consideration A similar shape is adopted for the main span, with 0.25 m thick
are given in Fig. 1. Owing to the limitations of certain information concrete pavements placed above the 20 mm thick steel plates.
3182 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

In addition, the stiffeners made of the 12 mm thick steel plates 2.3. Material model and strain rate effects
are placed at an equal space in the hollow steel deck as shown
in Fig. 1 to increase the stiffness of the main span deck. A total of Both CONCRETE DAMAGE REL3 (MAT_72REL3) and PSEUDO
224 parallel strand wire stay cables, with a diameter of 0.17 m, are TENSOR (MAT_16) material models are utilized to model the
arranged in a fan layout to maintain the bridge stability as shown concrete in the current study. The CONCRETE DAMAGE REL3
in Fig. 1. model is employed at places that are directly subjected to blast
pressures and regions in which they are taken to be affected by the
2.2. Element properties, boundary conditions and contact interface blast stress waves. The PSEUDO TENSOR model is used to model
the smeared material at places considered to be undisturbed by
Constant stress cube elements are employed for all concrete direct blast loads and hence detail analysis is not needed. The
members in this study. By conducting a numerical convergence test advantage of these two material models is that they can model
on various mesh sizes (1.56 mm, 3.13 mm, 6.25 mm, 12.50 mm the complex behaviour of concrete by specifying the unconfined
and 25.0 mm), it is found that the 6.25 mm mesh yields similar compressive strength only when no detailed concrete material
results with the smaller meshes tested under close proximity experiment data is available. Previous studies have proved that
explosion load (1000 kg TNT at 0.5 m) but with less simulation they are capable of reproducing experimental results and give
time. Due to the size of the bridge under consideration and the robust representation of structure response under blast loads
fact that the structural response under the effect of blast pressure [22,23]. In order to avoid computer overflow during calculation,
is highly localized, the 6.25 mm mesh is used only at positions the function MAT_ADD_EROSION is used to eliminate elements
close to the blast center. Mesh size then increases gradually as that do not further contribute to resisting the blast loads during
the distance from the blast center increases. Similar mesh sizes the analysis procedure. In the present study, the concrete mesh
are used for the Belytschko beam and shell elements in modelling will be deleted either when the tensile stress reaches 5 MPa and/or
the reinforcement bars and composite steel girder, respectively. the principal strain reaches 0.10. It should be noted that erosion
Since only limited information is available regarding the bridge,
must be used with caution since removing the concrete materials
it is assumed herein that all concrete components have a cover
violates the mass conservation of the structure.
depth of 50 mm over the reinforcement bars. In addition, all steel
PIECEWISE LINEAR PLASTICITY (MAT_24) and CABLE DISCRETE
reinforcements, with diameter ranging from 20 to 30 mm, are
BEAM (MAT_71) material model are employed for all the
spaced at 100 mm and 200 mm for the primary reinforcement
steel components and the elastic stayed cables on the bridge,
and stirrup, respectively. Perfect bond assumption is incorporated
respectively. The advantage of these material models is that
in the current study to model the connections between the steel
they allow users to define arbitrary stress strain curves and in
reinforcement and concrete. Hence, the bond slip between the
particular the arbitrary strain rate dependency for PIECEWISE
reinforcement and concrete is taken as a property which is fully
dependent on failure of concrete [20]. For places sufficiently away LINEAR PLASTICITY material model. The engineering stress strain
from the explosion center, smeared models (i.e. reinforcement curves employed are in accordance to the database in [24,25]. The
assumed to be uniformly distributed over concrete element) are material properties of the structural materials are tabulated in
used to minimize the computational effort. Table 1.
Foundation components (i.e. piles) are not considered in the The strengths of the structural materials are strain rate depen-
current study as no information is available for the bridge under dent as their dynamic mechanical properties can be enhanced sig-
consideration. Under this limitation, the nodes at the base of the nificantly when subjected to high strain rate impact such as blast
piers and towers are constrained against horizontal and vertical loads, which can generate a strain rate with magnitude greater
motions to simulate the fixed support conditions. than 1000 s−1 . Current study employs the dynamic increase factor
The penalty method approach is adopted to model the contact (DIF), a ratio of the dynamic-to-static strength against strain rate,
interfaces between meshes because of its effectiveness and to account for the material strength enhancement with strain rate
simplicity for explicit analysis. With this method slave nodes effect.
penetration is resisted via the imaginary normal interface springs The bilinear relationship developed by the CEB code [26] and
between the shooting nodes and contact surface [21]. Two contact Malvar and Ross [27] are applied in this study for the concrete
algorithms, CONTACT AUTOMATIC SURFACE TO SURFACE and strength enhancement. In tension, the DIF of the tensile strength
CONTACT AUTOMATIC NODES TO SURFACE, are employed for all is given by;
finite element simulations to prohibit mesh penetration. The first
ε̇ δ
 
contact algorithm is employed to avoid penetration at the interface ft
TDIF = = for ε̇ ≤ 1 s−1 (1)
of meshes with different material properties (e.g. at the interface fts ε̇ts
between pier and girder); while the second contact algorithm is  1/3
invoked to disallow slave node penetrations at locations where ft ε̇
TDIF = =β for ε̇ > 1 s−1 (2)
different mesh sizes are used. fts ε̇ts
Fig. 2 shows the finite element meshes of the four bridge
components. To reduce the required computer memory and where ft is the dynamic tensile strength at strain rate ε̇ in the range
computational time, detailed modelling with fine mesh is only of 10−6 – 160 s−1 , fts is the static tensile strength at ε̇ts , log β =
applied to the area near the explosion center. In particular, detail 6δ − 2 and δ = 1/(1 + 8fc0 /fco0 ), fc0 is the static uniaxial compressive
modelling is only applied to a height of 11.6 m for pier and 12.0 m strength of concrete (in MPa) and fco0 is taken as 10 MPa.
for tower, with the largest mesh size 100 mm as shown in Fig. 2. In compression, the equations are given as follows;
Above this region, smeared concrete material model is used with 1.026α
ε̇

the mesh size gradually increases from 200 to up to 5000 mm. fc
CDIF = = for ε̇ ≤ 30 s−1 (3)
Moreover, beyond the detail modelling region, the tower and pier fcs ε̇cs
are modelled as linear elastic. Therefore no damage is modelled fc
in blast response calculations. Similarly, detail modelling is only CDIF = = γ (ε̇)1/3 for ε̇ > 30 s−1 (4)
applied to a length of 30 m of the back span concrete deck and fcs
21 m of the main span steel–concrete composite deck. Beyond this where fc is the dynamic compressive strength at strain rate ε̇ , fcs
region, the mesh size increases gradually from 200 to 10000 mm is the static compressive strength at ε̇cs , log γ = 6.156α − 0.49,
(in longitudinal direction) with linear elastic material assumption. α = 1/(5 + 3fcu /4) and fcs is the static cube strength (in MPa).
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3183

Fig. 2. Finite element mesh of the bridge model.

Table 1
Material properties.
Concrete Steel Elastic cables
ρ (kg/m3 ) fc0 (MPa) ρ (kg/m3 ) E (GPa) Poisson’s Ratio fy (MPa) fult (MPa) ε ρ (kg/m3 ) E (GPa)

2400 60 7850 200 0.3 500 660 0.13 7850 200

ρ = mass density; fc0 = unconfined compressive strength; E = Young’s Modulus; fy = yield stress; fult = ultimate stress; ε= failure strain.
3184 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

are performed with these two traffic loading cases and only the
one that gives the severer responses will be presented.
Two methods are used to estimate the reflected peak blast pres-
sures. Due to the absent of blast pressure data for scaled distance
smaller than 0.067 m/kg1/3 in TM5-1300 [30], a commercial soft-
ware, AUTODYN [31], is employed to determine the air blast pres-
sure when the explosive charge has a 0.05 m/kg1/3 scaled distance.
Fig. 3. Traffic loading cases. The explosives are modelled with Jones–Wilkins–Lee (JWL) equa-
tion of state with an expression
Two approaches are adopted to account for the strain rate
ϑ ϑ ϑe
   
−r1 χ
effect for steel materials. For steel composite girder, the Cow- P = A1 1− e + A2 1− e−r2 χ + MPa (8)
per–Symonds relation is used to determine the DIF as given below; r1 χ r2 χ χ
where P is the hydrostatic pressure, χ is a specific volume, e is the
DIF = 1 + [ε̇/D] 1/g specific internal energy and A1 , r1 , A2 , r2 and ϑ are constants. Air

(5)
is modelled with the ideal gas equation of state in which pressure
whilst, the K&C model [28] is adopted to determine the DIF for is related to energy with the empirical formula
reinforcement which is given as
P = (γ − 1)ρ e kPa (9)

ε̇

DIF = (6) where ρ is the density of air, γ is a constant and e is the
10−4 specific internal energy. From the numerical simulation the
α = 0.074 − 0.040fy /414 (7) peak pressure and impulse of the explosion scenario under
consideration (explosive weight 1000 kg TNT, distance 0.5 m
where fy is the static yield strength (MPa) of the rebar. This and scaled distance 0.05 m/kg1/3 ) are approximately 1210.8 MPa
formulation is valid for reinforcement with yield stress ranging and 0.157 MPa-s, respectively. Verification of the simulated blast
from 290 to 710 MPa under the strain rate between 10−4 and history against TM5-1300 is carried out only for positions further
225 s−1 . away from the explosive center when the scaled distance is
larger than 0.067 m/kg1/3 due to the limitation of experimental
3. Traffic load and blast load estimation database. Fig. 4 shows the comparison of the numerically predicted
blast time history and those by TM5-1300. As shown the peak
Before applying blast load, traffic loads are applied to the bridge blast pressures and positive impulses are about 31% and 22%
in this study in accordance with the M1600 moving traffic load less than those given in TM5-1300. This is primarily attributed
as specified in Australian Standard AS 5100.2 [29], in which it to the insufficiently small meshes employed in the simulation.
specifies that the total weight of a truck is 720 kN, distributed Although discrepancies are identified between TM5-1300 and the
over a 15 m long and 5 m wide lane section. The traffic load simulated results, numerically predicted blast loads corresponding
consists of uniformly distributed load and truck axial load. To to the scaled distances less than 0.067 m/kg1/3 are still used in
minimize the modelling effort, these loads are all converted studying the structure response because the blast pressures cannot
to equivalent pressure in each segment, resulting in a 1.2 kPa be estimated by the TM5-1300 manual.
uniformly distributed pressure and a 9.6 kPa pressure from moving When the scaled distance is larger than 0.067 m/kg1/3 , the
truck load. It should be noted that in the previous studies of girder commercial software ATBlast is used to evaluate blast loads acting
bridge responses to blast load, the traffic load is not included in on the bridge. It calculates the blast loads developed in an open
the analysis. The reason to neglect traffic load is because including air explosion based on the empirical relations given in TM5-1300.
traffic load will increase the axial compression in bridge piers and For the above deck explosion cases considered in the study, it is
therefore increase the flexural resistance of the pier. Therefore assumed that the explosive is located at a 1 m height above the
including traffic load is considered unconservative [13,14]. For deck and the blast loading time histories are treated as a triangular
cable-stayed bridge considered in this study, however, this is not loading function. Under these circumstances (explosive weight
the case. For example, application of traffic loads on the main span 1000 kg TNT, distance 1 m and scaled distance 0.1 m/kg1/3 ) the
will reduce the axial compression in the piers. Moreover, existence peak reflected pressure and impulse at the point directly below the
of traffic loads will also affect the progressive collapse of the bridge. explosion center are 464.4 MPa and 0.383 MPa-s, respectively.
Therefore, in this study, two traffic loading cases are considered For an assumed explosion standoff distance of 0.5 m to the
as shown in Fig. 3. In Case 1, the traffic loads are assumed to be tower and pier, and 1.0 m above the deck, it is unreasonable
distributed on the entire bridge deck, wile in Case 2, traffic loads to assume uniform blast loads on the bridge structure because
are assumed only on the main span deck. Numerical simulations the distances to various points on the structure and the different

Fig. 4. Comparison of simulated pressure time histories with those estimated from TM5-1300.
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3185

(a) On ground components. (b) Bridge deck.

Fig. 5. Segmentation of the bridge components for blast analysis (Not to scale).

incident angles of the shock wave to these points on the structure Numerical results indicate that the blast pressure generated
differ, which makes the reflected pressures at different points on from this close range explosion causes a highly localized damage
the structure very different. For example, the AUTODYN simulation to the pier wall facing the explosion with a completely crushed
gives the peak reflected pressure on bridge pier and tower at area measuring about 10 m wide and 5.6 m high, and a damage
the point directly facing the 1000 kg TNT explosive (0.1 m above zone of the concrete extends to a larger area of about 10.2 m high.
the ground) with a 0.5 m standoff distance 1210.8 MPa. The The front wall of the pier is predominantly damaged by highly
corresponding peak reflected pressure reduces quickly to 0.95 MPa localized crushing and spalling of concrete elements resulted from
at the height of 10 m above the ground because of the increased the stress wave created by the blast loads. The reason for this
standoff distance and very large incident angle. In this study, the is because the blast pressure at such a small scaled distance is
bridge structure is divided into many segments. Blast loads on characterised by an extremely high magnitude of reflected peak
the structure in each segment are assumed to be the same and blast pressure (e.g. 1210.8 MPa at 0.1 m height) and very short
are estimated using the center point of the respective segment as positive phase duration (i.e. 0.259 ms). As a result, the stress
the reference point. The segments’ dimensions for the on ground waves cause only direct crushing and spalling of the concrete,
components (tower and pier) and bridge decks are displayed in instead of flexural or shear failure which is often associated with
Fig. 5. blast loadings with a relatively lower pressure and longer positive
A total of 105, 95 and 85 blast load time histories are used for phase duration to allow for global structure deformations. More
the pier, tower and bridge deck, respectively. The reason for more discussions regarding these failure modes are given in [19].
load curves defined for the pier is because it is fully engulfed by As shown in the damage states in Fig. 6, the pressure
the air blast pressure when explosive charge is placed at a 0.5 m engulfment around the pier also pushes the side walls of the
standoff distance. Hence, blast loads are also applied to the side hollow section to deform inward, and causes some damage on
and rear surfaces of the pier. All analyses are carried out in a PC the side walls. The side walls are not as significantly damaged
with a Pentium 4 CPU processor with main frequency 3.40 GHz and because the pressures acting on them are substantially smaller.
2.00 GB RAM memory. The simulation time for each case scenario However, the web of the pier is heavily damaged as shown in Fig. 7.
ranges from 240 to 340 h. Moreover, owing to the stress wave transmitted from the web, the
rear wall of the pier also suffers intensive damage although the
blast pressures acting on the rear wall are smaller. The extended
4. Dynamic response calculations
concrete damage area on rear wall is in fact larger than that in the
front wall, as shown in Fig. 8. This is because of the lower tensile
4.1. Pier response calculation strength of concrete material than its compressive strength. The
rear wall experiences primarily spalling damage due to concrete
Fig. 6 illustrates the snapshots of concrete damage contours tensile failure. Although the rear wall experiences more concrete
and the corresponding damage states of the pier subjected to damage, the estimated stresses in the reinforcement bars in the
the blast pressure generated at a scaled distance of 0.05 m/kg1/3 front wall are larger than those in the rear wall as shown in Fig. 9.
(explosive placed on the ground facing the center of the pier Fig. 9 shows the axial stress time histories in the reinforcement
with a standoff distance of 0.5 m and an explosive weight of bars at different locations in the front face and rear face of the pier
1000 kg TNT equivalent). It should be noted that the contour value front and rear walls. As shown the reinforcements in the front wall
between 0 and 1 indicates the concrete element stress states in at ground level and 2 m above the ground level are damaged almost
the range of elastic and elastic–plastic, and that between 1 and immediately with the peak axial stress reaches about 787 MPa,
2 represents the material softening [21]. It should also be noted and then drops rapidly to zero, indicating the breakage of the
that explosion on ground surface will damage the ground, produce reinforcements. The reinforcement bar at 5 m level also reaches a
a crater and generate a strong ground shock. All these will affect peak axial stress of about 787 MPa and breaks, but at a delayed time
the bridge response [32]. This study, however, concentrates on because the explosive is placed at the ground level. The axial stress
the bridge structure responses to direct air blast pressures only. in the reinforcements at 7 and 9 m level in the front wall reaches
Further studies of the combined effect of air blast pressure, ground 685 MPa and 501 MPa, respectively. No breakage of reinforcements
shock and ground crater on bridge structure damage are deemed at these levels is observed, but they all display plastic deformations.
necessary. It should be noted that because of the strain rate effect, the ultimate
3186 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

(a) Damage contour. (b) Damage state.

Fig. 6. Snapshots of pier damage (front view) to blast load of scaled distance 0.05 m/kg1/3 .

stresses of the steel reinforcements are approximately 1.2 times on the overall bridge stability. Progressive collapse analyses of the
greater than its static ultimate stress (i.e. 660 MPa). Therefore an bridge after the pier damage will be carried out in [19].
ultimate tensile stress of 787 MPa is recorded before the breakage
of the reinforcement. 4.2. Tower response calculation
The axial stresses in the reinforcement bars in the rear wall
are similar to those in the front wall, but they exhibit a lower Fig. 10 shows the snapshots of concrete damage contour and
ultimate stress and a delayed response as compared to those in the damage states of the bridge tower when subjected to an explosion
front wall. This is because the stress wave propagation through the at a scaled distance of 0.05 m/kg1/3 . As shown, similar to the
web causes a delay in rear wall response and reduces the loading pier, localized damage due to crushing and spalling of concrete
rate, and hence the strain rate effect on material properties. The is observed on the front wall of the tower facing the explosion
ultimate stress of the reinforcement bars in rear wall is 772 MPa center. No obvious shear and flexural damage is observed. The
before they break. completely crushed area is about 10.5 m wide and 5.8 m high,
Because of the complete damage of the front and rear walls, the which is smaller in comparison to that of the pier. This is because
entire section of the pier is lost. This will have a significant effect the tower has a thicker wall of 2.0 m and more reinforcement bars
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3187

Fig. 7. Damage of the pier web.

Fig. 8. Damage of the rear wall at 0.5 sec.

a: Front wall reinforcement. b: Rear wall reinforcement.

Fig. 9. Axial stress time histories of reinforcement bars in bridge pier front and rear wall.

as compared to the pier, thus leads to higher resistance capacity the center area at the ground level and 2 m above the ground break
to blast load. Another factor that may have attributed to relatively immediately after the application of the blast load. The stresses of
smaller damage is the shape of the tower. Since the tower section these two bars reach 787 MPa and drop to zero almost instantly
tapers linearly from the ground level to the 175 m height, both because of the very large pressures applied to the tower front wall.
the pressure incident angle and the scaled distance at each level The axial stresses in the reinforcement bars at 5 m and above are
is altered more than those of the pier. Hence it results in a greater less than 503 MPa, indicating the reinforcement bars at these levels
reduction in magnitude of the reflected blast pressures acting on are not yielded because the stress is only slightly larger than the
the tower wall. Unlike the pier, there is no web inside the tower to static yield strength of the reinforcement at 500 MPa. The axial
transmit stress wave and to push the rear wall. As such no damage stresses in the reinforcement bars 1 m from the damaged zone
is generated on the rear wall of the tower under this explosion at different heights are all less than 501 MPa, indicating again
scenario, as shown in Fig. 11. that these reinforcement bars are most likely not yielded. These
Fig. 12 shows the axial stress time histories in reinforcement observations indicate that, unlike the pier in which the damage is
bars at different levels in the center area of the front wall, and at observed in the entire cross section, the damage from the direct
1 m from the damaged zone. As shown the reinforcement bars in blast loading is limited to an area of the tower wall facing the
3188 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

Fig. 10. Snapshots of tower damage to blast load of scaled distance 0.05 m/kg1/3 .

Da ma getotofront
Damage fr on t result in the tower structure to lose stability and induce bridge
surface
su rfac eonly
on ly
collapse. This will be studied and discussed in the accompany
paper [19].

4.3. Back response calculation

Response of the back span concrete deck to blast pressures


generated by 1000 kg TNT explosion at 1 m above the deck (scaled
distance 0.1 m/kg1/3 ) are calculated. Fig. 13 shows the snapshots
of the damage contours and damage states of the deck at different
time instants. As shown in Fig. 14, the damaged area measured on
the deck surface from this blast scenario is approximately 30 m by
21 m, or the entire area with detail modelling. Therefore, the exact
Fig. 11. Top view of the damaged bridge tower. damage area is not known in this simulation because the smeared
area is modelled with a linear elastic assumption. It is likely that
explosion center. No damage in the curved sides and rear wall of a larger damage area extended to more than 30 m is generated.
the tower section is generated. However, this damage may still However, owing to the limitation in computer memory, detailed
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3189

a Fig. 15 shows the reinforcement axial stress time histories.


As shown the reinforcements directly below and 5 m from the
detonation point experience an abrupt increase in axial stress
to magnitudes 787 MPa and 730 MPa, respectively before they
break. At other points further away from the explosion center, the
reinforcements exhibit permanent plastic deformation, breakage
is not observed. Failure of steel reinforcement in the bottom flange
of the hollow deck is only identified at positions near the edge of
the deck. The ultimate stress recorded at 3 m away from the edge
of the deck is approximately 721 MPa before failure. The ultimate
stresses of reinforcement bars at different locations are different.
The ultimate stress in reinforcement bars close to the explosion
center is higher than that in those away from the explosion center
because of different strain rate effect.

b 4.4. Main span response calculation

Fig. 16 shows the snapshots of damage of the steel–concrete


composite main span deck to blast loads. As shown, the concrete
pavement is severely damaged with an area of about 21 m by 21 m
as shown in Fig. 17. The cause of concrete failure may attribute to
both the direct blast pressure and the reflection of tensile stress
wave from the steel deck. The concrete pavement layer in the
damaged area is eroded away quickly after the explosion. Akin to
the back span, the exact damage area of the concrete pavement
layer is not known owing to the fact that detailed modelling is only
extended to 21 m of the main span deck. Failure of the steel deck
is observed within an area of approximately 11.5 m by 10 m, but
plastic deformation extends to the entire damaged area of 21 m by
21 m, as shown in Fig. 17. The reason for the less severe damage
Fig. 12. Axial stress time histories of reinforcement bars in tower front wall (a: on the steel component is because of its high material strength
along the center of the wall, b: 1 m from the edge of the damaged zone). and ductility which enabled the steel material undergo a greater
deformation before failure.
Fig. 18 shows the maximum effective stress time histories
modelling is only applied to the 30 m range. Nevertheless, as can be
recorded on the steel deck at various locations. As shown, at the
seen in the figure, the damage intensity reduces substantially with
point directly below the explosion center the maximum stress
the increase in distance from the point directly below the explosion
reaches 748 MPa, indicating failure of the steel element. Failure
center. The damage beyond the area with detailed modelling is
of steel element is also observed at 2 m away from the point
relatively insignificant, and is assumed not to greatly affect the
directly below the explosion center with an ultimate stress of
subsequent progressive collapse analyses that will be carried out
about 716 MPa. No steel failure is observed at 6 m distance from
in [19]. the explosion center, but steel elements display obvious plastic
A larger damaged area as compared to the pier and tower deformations. No steel failure in the bottom surface of the hollow
is generated on concrete back span deck although the scaled section is observed.
distance is larger. This is because the pressures generated are Only one cable, located close to the explosion center, is found
distributed over a larger structural area as compared to the pier not to provide any support for the bridge resulted from the steel
and tower case owing to the larger standoff distance. In the pier girder failure under the blast load. Adjacent cables, which are 18 m
and tower case, the standoff distance is 0.5 m, which makes the apart, maintain their load-carrying capacity. Fig. 19 shows the time
shock wave incident angle to the structure very large, thus reduces histories of the two cables after explosion. As shown, the cable near
the reflected pressure quickly on structure surface. As for the deck the explosion center loses load-carrying capacity not because of
case, the standoff distance is 1.0 m, the incident angles of blast the failure of the cable itself, but the anchorage owing to the failure
pressure are therefore relatively smaller, which makes the blast of the composite deck. The maximum axial force in the cable is only
pressure distributed to a larger area. 11.6 MN, substantially lower than the design axial force capacity of
Failure of concrete at the bottom of the hollow section deck 40.2 MN for each cable.
is also observed. It first occurs near the web section at the edge
of the bridge deck component. The failure area is measured to 5. Conclusions
be approximately 24 m by 18 m and concentrates at the region
directly beneath the explosion, as shown in Fig. 14. Detail modelling and damage analyses of a long span cable-
As shown in Fig. 14, three stayed cables are found to have stayed bridge under close proximity blast loads are carried out
failed in this blast event and are found to be detached at forces of in this study. All numerical calculations are performed using an
magnitudes below the breaking force of the cables (i.e. 40.2 MN). explicit finite element code LS-DYNA. This paper presents the finite
This suggests the failure of cables is caused by the damage of the element and material models of the bridge, and the simulation
bridge deck that results in the loss of anchorages to the cables. In results of responses of bridge structures (pier, tower, concrete
fact, simulations were conducted to study the cable responses to back span and steel–concrete composite main span) to blast
direct blast pressure (results are not shown here) and it was found loads generated from a 1000 kg TNT explosion on ground surface
that the direct blast load did not break the cable that is nearest 0.5 m from the pier and tower, and 1.0 m above the back span
to the explosion center. Therefore the detachments of cables are and main span deck, respectively. It is found that the failure is
caused by loss of anchorage owing to deck damage. caused predominantly by compressive crushing and spalling of
3190 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

Fig. 13. Snapshots of concrete back span damage to blast load of scaled distance 0.1 m/kg1/3 .

Fig. 14. Damaged area and detached cables of concrete back span.

concrete materials with localized steel reinforcements and steel However, some cable detachments are observed because of the
plate failure. Complete failure of the bridge pier section and loss of anchorage owing to the deck failure. Concrete back span is
concrete back span deck is observed, whereas severe damage is most vulnerable to blast load. It suffers the most significant damage
noted on the front wall of the tower structure facing the explosion among the four bridge components.
center. Complete concrete pavement failure in an area close to the Although the damaged areas of each bridge component are
explosion center is also observed in the steel–concrete composite significant, they are still localized in view of the large dimension of
main span deck, but only limited failures are observed in the steel the bridge. These localized damages may, however, cause bridge
girder. Direct blast pressures do not cause failure of the stay cables. to lose stability. In the accompany paper [19], more simulation
E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192 3191

(a) Top deck. (b) Bottom deck.

Fig. 15. Steel reinforcement axial stress time histories.

Fig. 16. Snapshots of the main span steel–concrete composite deck response to blast load of scaled distance 0.1 m/kg1/3 .

results of these bridge components to blast loads of different has occurred. The safe scaled distances for bridge protection are
scaled distances are presented. Progressive collapse analyses of determined. The effectiveness of FRP strengthening for bridge
the bridge are carried out after damage of each bridge component protection against blast loadings is also examined [19].
3192 E.K.C. Tang, H. Hao / Engineering Structures 32 (2010) 3180–3192

References

[1] Symonds PS. Elastic, finite deflection and strain rate effects in a mode
approximation technique for plastic deformation of pulse-loaded structures.
J Mech Eng Sci 1980;22(4):189–97.
[2] Yankelevsky DZ. Elasto-plastic blast response of rectangular plates. Internat J
Impact Eng 1985;3(2):107–19.
[3] Ross TJ. Direct shear failure in reinforced concrete beams under impulsive
loading. Ph.D. thesis. California (USA): Stanford University; 1983.
[4] Krauthammer T, Assadi-Lamouki A, Shanaa HM. Analysis of impulsively loaded
reinforced concrete structural elements. Comput Struct 1993;48(5):851–60.
[5] Low HY, Hao H. Reliability analysis of direct shear and flexural failure modes
of RC slabs under explosive loading. Eng Struct 2002;24:189–98.
[6] Shope RL. Response of wide flange steel columns subjected to constant axial
load and lateral blast load. Blacksburg, Ph.D. thesis. Virginia (USA): Virginia
Polytechnic Institute and State University; 2006.
[7] Shi Y, Hao H, Li Z-X. Numerical derivation of pressure-impulse diagrams for
prediction of RC column damage to blast loads. Internat J Impact Eng 2007;
Fig. 17. Damage of the main span steel–concrete composite deck under blast load
35(11):1213–27.
of scaled distance 0.1 m/kg1/3 . [8] Luccioni BM, Ambrosini RD, Danesi RF. Analysis of building collapse under
blast loads. Eng Struct 2004;26(1):63–71.
[9] Motley MR, Plaut RH. Application of synthetic fiber ropes to reduce blast
response of a portal frame. Internat J Struct Stability Dynam 2006;6(4):
513–26.
[10] Zhou XQ, Kuznetsov VA, Hao H, Waschl A. Numerical prediction of concrete
slab response to blast loading. Internat J Impact Eng 2008;35(10):1186–200.
[11] Riedel W, Mayrhofer C, Thoma K, Stolz A. Engineering and numerical tools for
explosion protection of reinforced concrete. Internat J Protective Struct 2010;
1(1):85–101.
[12] Mahoney EE. Analyzing the effects of blast loads on bridges using probability,
structural analysis and performance criteria. Master of Science Thesis,
Department of Civil and Environmental Engineering, The University of
Maryland, College Park, USA; 2007.
[13] Winget DG, Marchand KA, Williamson EB. Analysis and design of critical
bridges subjected to blast loads. J Struct Eng, ASCE 2005;131(8):1243–55.
[14] Islam AKMA, Yazdani N. Performance of AASHTO girder bridges under blast
loading. Eng Struct 2008;30:1922–37.
[15] AASHTO. American association of state highway and officials: load and
resistant factor design specifications, Washington, DC, USA; 2003.
[16] FHWA. Recommendations for bridge and tunnel security. The Blue Ribbon
Panel on Bridge and Tunnel Security, Federal Highway Administration,
Fig. 18. Maximum effective stress time histories recorded on steel elements of the Washington, DC, USA; 2003.
main span steel–concrete composite girder. [17] Son J, Astaneh-Asl A, Rutner M. Performance of bridge decks subjected to blast
load. In: Proceedings (CD-Rom), 6th Japanese-German Bridge Symposium;
2005.
[18] Suthar KN. The effect of dead, live and blast loads on a suspension bridge.
master of science thesis, Department of Civil and Environmental Engineering,
College Park (USA): University of Maryland; 2007.
[19] Hao H, Tang EKC. Numerical simulation of a cable-stayed bridge response to
blast loads, Part II: damage prediction and FRP strengthening. Eng Struct 2010;
32:3193–205.
[20] Lan S, Crawford JE, Morrill KB. Design of reinforced concrete columns to
resist the effects of suitcase bombs. In: Proceedings of the 6th Asia-Pacific
Conference on Shock & Impact Loads on Structures. 2005, pp. 325–332.
[21] LS-DYNA. LS-DYNA Version 970 Keyword User’s Manual. Livermore Software
Technology Corporation; 2006.
[22] Malvar LJ, Crawford JE, Wesevich JW, Simons DA. A plasticity concrete material
model For DYNA3D. J Impact Eng 1997;19:847–73.
[23] Malvar LJ, Crawford JE, Morrill KB. 1999. K&C Concrete material model,
release III: automated generation of material model input. Report TR-99-24,
Karagozian & Case Structural Engineers, Burbank, California, USA.
[24] ArcelorMittal. Mittal products — automotive application: high strength steels.
2006. Available [Online]: http://www.mittalsteel.com.
[25] Warner RF, Rangan BV, Hall AS, Faulkes KA. Concrete structures. Australia:
Addison Wesley Longman; 1998.
[26] Comite Euro-International du Beton. Concrete structures under impact
Fig. 19. Cable axial force time histories.
and impulsive loading. CEB Bulletin 187. Switzerland: Federal Institute of
Technology Lausanne; 1990.
Acknowledgements [27] Malvar LJ, Ross CA. Review of strain rate effects for concrete in tension. Amer
Concr Inst Mater J 1998;95(6):735–9.
[28] Malvar LJ. Review of static and dynamic properties of steel reinforcing bars.
The authors would like to thank Australian Research Council for Amer Concr Inst Mater J 1998;95(6):609–16.
financial support under grant No. DP0774061 for conducting this [29] Australian Standard. AS5100.2-2004, Bridge design — Part 2: design loads.
Sydney (Australia): Standards Australia; 2004.
research work. The first author would also like to acknowledge
[30] TM5-1300. Structures to resist the effects of accidental explosions, U.S.
the University of Western Australia for provision of the UPA Department of the Army, Navy and Air Force Technical Manual, Washington,
scholarship and computing assistance during the research study. DC, USA; 1990.
Support from the State Key Laboratory of Science and Technology [31] AUTODYN. Theory manual, Century Dynamics, 2005.
[32] Wu CQ, Hao H. Modelling of simultaneous ground shock and airblast pressure
of Beijing Institute of Technology with its collaborative research on nearby structures from surface explosion. Internat J Impact Eng 2005;31:
scheme under project number KFJJ08-3 is also acknowledged. 699–717.

You might also like