Research Paper BLAST LOAD GENERATION METHODS ON BRIDGES · 2015-04-13 · 36 Int. J. Struct. &...
Transcript of Research Paper BLAST LOAD GENERATION METHODS ON BRIDGES · 2015-04-13 · 36 Int. J. Struct. &...
36
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
BLAST LOAD GENERATION
METHODS ON BRIDGES
Binol Varghese1* and Ajith M S1
Due to different accidental and intentional events like terrorist attacks related to importantstructures all over the world, explosive loads have received considerable attention in recentyears. The design and construction of bridge to provide life safety in the face is receiving renewedattention from structural engineers. Analysis of highway bridges under blast loads requiresaccurate generation and application of blast loads and good understanding of the behavior ofcomponents of bridge. The purpose of this paper is to introduce some ideas about blast loadgeneration method like pressure wave method, detonation simulation method, hybrid blast loadmethod and multi-Euler domain method. Also verification of blast load results using hybrid blastload method and multi-Euler domain method included in this paper.
1 Government College of Engineering, Kannur, India.
*Corresponding author:Binol Varghese � [email protected]
ISSN 2319 – 6009 www.ijscer.com
Vol. 3, No. 3, August 2014
© 2014 IJSCER. All Rights Reserved
Int. J. Struct. & Civil Engg. Res. 2014
Research Paper
Keywords: Detonation simulation method, Hybrid blast load method, Multi-Euler domainmethod, Pressure wave method
INTRODUCTION
The number and intensity of domestic andinternational terrorist activities, including theSeptember 11, 2001 attack on World TradeCenter towers in New York, have heightenedour concerns towards the safety of ourinfrastructure systems. Terrorists attack targetswhere human casualties and economicconsequences are likely to be substantial.Transportation infrastructures have beenconsidered attractive targets because of theiraccessibility and potential impacts on humanlives and economic activity.
Bridges are an integral part of national
highway system. Military assaults, terroristattacks and accidental explosions may causeserious damage to bridges. As a result,engineers and transportation office workers aremore active in the construction of strong bridgesto withstand potential blast attacks. Explosionaccident analyses, blast-resistant design andanti-terrorist and military weapon design havebecome more important areas. Damage effectanalyses and assessments of bridges underblast loading are very important in these areas.With the rapid development of computerhardware over the last decades, detailednumerical simulations of explosive events inpersonal computers have become possible.
37
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
Loads imposed on highway bridge
components during a blast loading event can
exceed those for which bridge components
are currently being designed. In some cases,
the loads can be in the opposite direction of
the conventional design loads. Consequently,
highway bridges designed using current
design codes may suffer severe damages
even from a relatively small sizes explosion.
Importance of Blast LoadGeneration Methods
The objectives of this paper are to extensivelyinvestigate finite element tools for simulationof blast load effects on bridge, investigateperformance of bridge under various blast loadgeneration methods like pressure wavemethod, detonation simulation method, hybridblast load method and multi-Euler domainmethod.
Although experimental verification usingscaled models is generally carried out indeveloping design guidelines for structuressubject to hazards such as earthquakes, wind,etc., this is not practical in case of blast loadsbecause of following three reasons.
1. It is very difficult to reproduce the same blastwave environment, even in the same testfield and using the same amount ofexplosive charge because of thetemperature, humidity and dust condition ofthe air. Consequently, it is very difficult tocarry out systematic experimental study ofdifferent parameters affecting behavior ofstructures subject to blast loads, particularlyof relatively complex structures.
2. It is difficult to ensure reliability of sensorsand data measurements, e.g., strain gauge,displacement sensors, etc., duringexplosion tests because of largedeformation and fragmentation of the teststructure. The sensors are likely to bedestroyed during first few milliseconds ofthese types of experiments. Most of theexperimental data / results are inferred fromvideos or pictures obtained during theexperiments.
3. Experimental blast tests are also cost-prohibitive and can only be carried out atselect facilities.
BLAST LOAD
Blast Load Characteristics
A structure is likely to be subjected to varioustypes of hazards during its life time. Thesehazards can be subdivided into two generalcategories: man-made (blast) and natural(earthquakes, wind, etc.). For a successfulapproach to any system design, it is essentialto understand the nature of the hazard.Dynamic hazards can be described by theirrelative amplitudes and relative time(frequency) attributes. Figure 1 shows a
Figure 1: Amplitude-frequencyDistributions for Different Hazards
38
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
schematic representation of the amplitude-frequency relationships of several dynamichazards.
It is important to emphasize the principaldifferences between static, dynamic and short-duration dynamic loads. Typically, static loadsdo not produce inertia effects in the structuralresponse, are not time dependent, and areassumed to act on the structure for longperiods of time (e.g., gravity loads). Dynamicloads, such as induced by earthquake or windgusts, have strong time dependencies andtheir typical durations are measured in tenthsof seconds. Short-duration dynamic loads,such as those induced by explosions or debrisimpact, are non-oscillatory pulse loads, andtheir duration is about 1,000 times shorter thanthe duration of typical earthquakes. Structuralresponses under short-duration dynamiceffects could be significantly different thanthose during much slower loading cases. Amulti-hazard design of bridges must explicitlyaddress the effects conventional loads as wellas severe loading environments imposed bydifferent hazards.
An explosion is a very rapid release ofstored energy characterized by an audibleblast. Part of the energy is released as thermalradiation, and part is coupled into the air (air-blast) and soil (ground-shock) as radiallyexpanding shock waves. Air-blast is theprincipal damage mechanism. Air-blastphenomena occur within milliseconds and thelocal effects of the blast are often over beforethe structure (building or bridge) can globallyreact to the effects of the blast. Initial peakpressure intensity (referred to as overpressure)may be several orders of magnitude higher
For any blast loading, the total dynamicpressure (in psi) and the positive phaseduration (in milliseconds) are found in termsof equivalent weight, W, of the explosive in TNTand the distance from the blast centre, R.Generally, blast loads are defined in terms ofthe scaled distance parameter Z (ft per lb TNTequivalent) as
Z = R x W-0.33
With the scaled distance in the correct units,published curves can be used to find the totaldynamic pressure and the positive phaseduration. Similar blast waves are produced atidentical scaled distances when two explosivecharges of similar geometry and of the sameexplosive, but of different sizes, are detonatedin the same atmosphere.
Stand Off Distance
Stand-off distance refers to the direct,unobstructed distance between a generationpoint and its target. Figure 3 shows the stand-off distance in blast load analysis.
Figure 2: Simplified Modelof Unconfined Blast Load on Structures
than ambient atmospheric pressure in caseof air-blast. Figure 2 represents a simplifiedmodel of unconfined blast load on structures.
39
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
Height of Burst (HOB)
Height of burst refers to aerial attacks. It is thedirect distance between the exploding weaponin the air and the target. If the exterior buildingwalls are capable of resisting the blast load,the shock front penetrates through window anddoor openings, subjecting the floors, ceilings,walls, contents, and people to suddenpressures and fragments from shatteredwindows, doors, etc. Building components notcapable of resisting the blast wave will fractureand be further fragmented and moved by thedynamic pressure that immediately follows theshock front. Building contents and people willbe displaced and tumbled in the direction ofblast wave propagation. In this manner the blastwill propagate through the building.
Material Behaviors at High StrainRate
Blast loads typically produce very high strainrates in the range of 102 - 104 s-1. This highloading rate would alter the dynamicmechanical properties of target structures and,accordingly, the expected damagemechanisms for various structural elements.For reinforced concrete structures subjectedto blast effects the strength of concrete andsteel reinforcing bars can increasesignificantly due to strain rate effects. Figure
4 shows the approximate ranges of theexpected strain rates for different loadingconditions. It can be seen that ordinary staticstrain rate is located in the range: 10-6 – 10-5 s-
1, while blast pressures normally yield loadsassociated with strain rates in the range: 102 -104 s-1.
Figure 3: Blast Loads on a Building
Figure 4: Strain Rates Associatedwith Different Types of Loading
External Blast Load on Structures
The blast loading on a structure caused by ahigh explosive detonation is dependent uponseveral factors:
1. The magnitude of the explosion.
2. The location of the explosion relative to thestructure (confined or unconfined).
3. The geometrical configuration of thestructure.
4. The structure orientation with respect to theexplosion and ground surface (above, flushwith, or below the ground).
ConWep
Accurately modeling blast dynamics is criticalin the assessment of vehicles and structuressubjected to blast loads. The current industrystandard for modeling blast effects inLangrangian Finite Element simulations isConWep, tabulated pressure data takendirectly from blast events. Conwep is limited,however, and may not always be physicallyrepresentative of the blast/structural interactionthat occurs in the field. ConWep can capture
40
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
shock front interaction and model blast surface.
BLAST LOAD GENERATION
METHODS
Blast wave pressure decreases rapidly withthe standoff distance. On a SDOF structure ablast load can be applied as a point load. Here,failure mechanisms of structure members areassumed to simplify a structure as a SDOFmodel. In this paper failure mechanisms ofbridge under blast load are investigated.
Various methods used for evaluatingstructural behavior under impact loads are:single or multi degree of freedom, pressure-impulse diagrams, and response surfacesdeveloped from finite element analysis. Butthese methods have relatively low accuracy inthe prediction of either load or structuralperformance. These methods have their ownadvantages and disadvantages.
Pressure Load Method
In this method, ConWep (a collection ofconventional weapon effect calculations fromthe equations and curves) has been mergedinto LS-DYNA to apply a pressure load onstructures. Advantages of this method are: itaccurately controls load magnitude and doesnot consume extra calculation time. However,analysis before failure of a structural face onlycan be done by this method. When elementsfail in the FEM simulation, eroding techniqueis used to avoid element distortion, i.e.,“damaged” elements or nodes are removedfrom the structure.
Blast load generated by ConWep actsdirectly on exposed structural surface. Henceit will be lost once the load contact surface (i.e.,
exposed structural surface) is eroded. Figure5 shows ConWep pressure directly acting onthe surface of a column. Blast pressure is lostwith the erosion of FEM elements in concretecover. Blast load continues to act on unerodedelements.
Figure 5: Application of ConWep BlastLoad Directly on Concrete Structure
Since the load is not transmitted to concretecore after the spalling of the concrete surface,the simulation in Figure 6 underestimates thedamage to the structure.
Figure 6: Representation of Blast LoadGeneration Using Pressure Load Method
Detonation Simulation Method
Detonation Simulation approach generatesblast loads through detonation of highexplosives using Arbitrary-Lagrangian-Eulerian (ALE) mesh. The simulation ofexplosion events requires large deformationof element mesh, and this exceeds the abilityof Lagrangian mesh (nodes moving along with
41
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
material transferring). Eulerian mesh (nodesfix while the material transfers among differentelements) is used to solve the largedeformation problem. Calculation of the actionbetween the structure with Lagrangian meshand the Eulerian mesh creates dynamicboundary problem which is difficult to handle(Gong, 2006). Therefore, ALE mesh waschosen to simulate the explosion environmentin the bridge blast research. An ALEformulation consists of a Lagrangian time stepfollowed by a “remap” or “advection” step. Theadvection step performs an incremental rezone(the positions of the nodes are moved only asmall fraction of the characteristic lengths ofthe surrounding elements). The topology of themesh is fixed in an ALE calculation unlike amanual rezone. An ALE calculation can beinterrupted like an ordinary Lagrangiancalculation and a manual rezone can beperformed if an entirely new mesh is necessaryto continue the calculation (Hallquist, 1998).
This method simulates the process ofdetonation and gives accurate evaluation ofincident blast wave pressure through theexplosive material. It is perfect for thesimulation of interaction action betweenstructure and close blast such as landmineexplosion (Wang, 2001). But in simulation ofcivil structures, standoff distance has to beconsidered. Explosion generates blast wavein the air and the air blast wave blows thestructure. As shown in Figure 7, blast pressurewaves are carried to column surface using airas a medium because of standoff distancebetween charge and structure.
This approach has several advantages overdirect application ConWep pressure:
• The blast wave load continues to act on thestructure after the eroding of structuralsurface elements.
• It can predict the reflection and diffractionof the blast wave.
• This method can account for the mutualinteraction between structures and blastwave. This interaction cannot be ignoredwhen the structural material yields underblast wave load with the elastic modulusapproaching zero.
Hybrid Blast Load Method
It is a new approach that can simulate loadson structural elements similar to ConWeppresented in order to overcome limitations ofexisting approaches. In this approach,ConWep pressure generated for a specificcharge weight is transferred to an air layer nearstructural element. The blast wave frontpropagates through the surrounding air layerand the air mesh interacts with Lagrangianstructure element to apply the load on structuralelements (Figure 8).
This approach has advantages of bothpressure load method and detonationsimulation. It produces correct pressure fieldwith the same arriving time of blast wave asConWep, and can simulate wave reflectionand diffraction. When simulation involves
Figure 7: Representationof Blast Load Generation UsingDetonation Simulation Method
42
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
complex geometry or non-air blast waves suchas inter-explosion in box girder and airpressure near deck, other experiment data orprogram such as BlastX are necessary tocalibrate the load effect parameters. Figure 9represents of blast load generation usinghybrid blast load method.
pressure and incident blast pressure, thesmaller for each load segment, the moreaccurate the simulation results. In this paper,one load segment is set for each structuremember.
• The air density of the air layer sustainingthe segment load. When the load is appliedto air layer before the structure, the air layerwill be compressed first and form a blastwave front. Then the blast wave front moveson and transfers the blast load. Usually, thewave front air density is 2-6 times of airdensity of approximately 1.29 kg/m3. It isfound that the smaller the air density (1/10to 1/100 of the normal air density), thebigger the pressure transferred to structuralelements. If the air density of the load airlayer is set to smaller than 1.29×10-2 kg/m3,the transferred load drops down. It has beenobserved from simulation results that theblast pressure is transmitted accurately tostructural elements when air density ofsurrounding air layers is 1/10 of normal airdensity. In this case, the density ofcompressed air layer is approximately twiceof the normal air density value of 1.29 kg/m3. This value is close to shock densityparameter obtained from ConWep.
Multi – Euler Domain Method
The FEM analysis of explosions is time-consuming as well as requires too many inputparameters. Therefore, only small-scalestructural components could be simulated byFEM., As the Euler domain should be largeenough to cover the entire target in the airrange, the analysis for much larger structureshad to rely on supercomputers. However, it isdifficult to conduct analysis of full-scalestructures such as long-span bridges or tall
Figure 8: Simulation Under Blast WaveUsing Hybrid Blast Load Method
Figure 9: Representationof Blast Load Generation
Using Hybrid Blast Load Method
In this approach, when pressure load isapplied on air layer segment, the pressure ofnearest air element should have the samepeak as the applied load.
The two key factors in this approach are:
• The arrangement of load segments. SinceConWep reflected air blast pressure with ascale factor is used to simulate the load onstructures including internal reflected blast
43
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
buildings subjected to blast loading as thelimitation of element numbers and the poorhandling of the geometry of the Euler domain(the Euler domain could be modelled only ascuboids). To solve this problem, a multi-Eulerdomain method is proposed.
In this method, the initial Euler domain isbuilt as the carrier of the blast remapping inputfile that was mentioned earlier, and contains apart of the structure. With the propagation ofthe pressure along the structural longitudinaldirection, a new Euler domain is inserted andconnected with the previous one. Thesubsequent Euler domains are introducedgradually, while the previous domains areremoved when the boundary pressuredecreases. Using this method, the peakpressure can be controlled to a certain airrange with fewer Euler elements, thus avoidingany waste of computational resources.
In the multi-Euler domain analysis, to ensurecontinuous transmittance of the blast wave, asecond Euler domain was built and connectedwith the initial Euler domain as the pressurereached the boundary surface.
COMPARISON ON BLAST
LOAD GENERATION
METHODS
A comparative study of various blast loadgeneration methods was done in the previouschapter. The analysis results of multi-Eulerdomain method and hybrid blast load methodare discussed below.
Multi-Euler Domain Method
Bridge Model
In this paper, a 24 m single-span RCcomposite steel slab-on-girder bridge system
was adopted. The bridge deck has two layersof steel reinforcements, and the five steelgirders are equally spaced at 2 m apart.
Six sub-Euler element domains are definedand built along the longitudinal direction for thecomplete release of the blast wave on thewhole bridge. Each Euler domain has a 2 mlength with 10 cm mesh size for consideringthe time-step of each cycle and thecomputational resources. The height and widthof the domain are adjusted by the bridgedimensions under different blast events. A listof virtual gauges is set for each air domain tocapture the pressure time histories. Twoblasting scenarios are analyzed with differentcharge weights and detonation locations, i.e.,above- deck and under- deck cases. A 100kg charge weight of TNT is used to model amedium-scale explosive device for bothcases.
Results
Above - Deck Detonation with Medium-Scale TNT
The deck girder system is originally designedfor two purposes. One is to carry the loads(superstructure dead load and traffic liveloads). Second is to stabilize the substructurefrom sudden collapse under extreme dynamicloads. First blast waves will damage the uppersurface of the deck. Then the downwarddisplacement and pressures will betransferred to the steel girder. The materialdamage propagates radially outward on thebridge deck. Since some restrains areprovided by the supporting steel girders, forfirst several milliseconds the damage on theconcrete is mainly concentrated along thelongitudinal direction above the supporting
44
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
girders. When the pressure reaches the freeedges of the midspan, further damage isidentified because of the rarefaction wavealong the front face. A low-pressure wave isgenerated at the outer corner, where vorticesare shed. As the blast shock wave passes thefront surface slight damage appears aroundthe fixed boundary of the bridge, which iscaused by the vibration of the bridge insteadof being caused by the direct shock of the blastwaves.
A maximum stress of 195 MPa appearsaround 1.2 ms near the surface between themiddle girder and the deck and much higherstress of 240 MPa is obtained at around 5.4ms in the reinforcements.
Under-Deck Detonation with Medium-Scale TNT: Because typical RC slab-on-girder bridges are designed to withstandupward bending moments and negative shearforces with rapid dynamic loads, explosionbelow the bridge deck and girders may beamplified in those confined regions. Furthergenerate large lateral forces that may causelarge deformations or flexural failure, especiallyif the standoff distance is small. In addition,separation of the girder system and the largebending in the steel girders will make thesituation much more severe.
Because the reinforcements are coincidentwith the concrete, the maximum upwarddisplacement of 216.8 mm represents thedeformation of the bridge deck. Comparedwith the concrete material, the steel girdersexperience a smaller horizontal deformationof 88 mm without obvious failure, becausesteel has good ductility and high strength.Figure 10 shows comparison of above deck
with medium scale TNT detonation and belowdeck with medium scale TNT detonation inmulti-Euler domain method
Hybrid Blast Load Method
Bridge Model
The bridge model adopted in this study was a2 lane, 3 span RC bridge with maximum span18.9 m. The bridge deck width is 12.2 m andthickness 0.33 m. It is supported over 6 (3 x 2groups) piers of height 4.9 m and cross section0.91 x 0.91 m. The bridge structure model isput into an air mesh so that the blast air waveload can be transferred to the structure. An airALE mesh shown in Figure 11.
Figure 10: Comparison of Resultsin Multi-Euler Domain Method
Figure 11: Blast Load Simulationon Highway Bridges
Results
Load due to blast can cause severe damageto bridge components.
45
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
It shows that the bearing compressive forcehas been increased 2 – 5 times of the serviceload (Table 1). During blast load event structuralcomponents undergo large displacement.Shear displacement of a pier could be severalinches at failure. Structural members may alsoundergo large strains; structure steel mayexhibit plastic strains as high as 20%. The coreconcrete material near the pier bottom iselastic under low and medium levels of blastloads. The highest concrete stress reachesapproximately 27.6 MPa while the concretecompressive strength is 55.2 MPa. Figure 12shows the peak pressure distribution of thedeck under the explosive in hybrid blast loadmethod.
Table 1: Bearing Forces Under Blast Load
Load Maximum Maximum
Tension (kN) Compression (kN)
Low 1480 -1078
Medium 1760 -2125
High 2648 -775
Serviceload -176 -440
Figure 12: The Peak Pressure DistributionOf The Deck Under The Explosive
The core concrete shows highly nonlinearbehavior for all levels of loading. After thecompressive blast load hits the pier surface,the compressive stress wave propagates inconcrete material. When the compressivewave propagates to the back surface of thepier, it reflects back as a tensile stress wave,causing tensile strain on the back surface.
The following observations on the behaviorof bridge components subject to blast loadscan be made from the preceding simulationcase:
• The damage mode for a bridge under blastloads is similar to that for seismic loads.During low and medium levels of blast load,it is in an elastic range and during high levelsof load it goes into an inelastic range.
• The damage mode is as a plastic hingeformed in the middle of column. It has biggerenergy absorption ability than the localdamage mode, such as compressivefailure of concrete directly facing the blastwave load.
CONCLUSION
In this paper, a multi-Euler domain method andhybrid blast load method has been developedand applied for blast effects simulation of aRC composite slab-on-girder bridge. By usingthe hybrid blast load method, the large andcomplex structural response of bridges underblast loading can be analyzed. Compared withthe multi-Euler domain method, which issuitable only for structural components andshort span bridges, the new approach showsoutstanding efficiency in reducing the numberof elements by up to a hundredfold for differenttypes of bridges, as well as shortening the
46
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
calculation time in each cycle by 40 to 60%,and it is much more convenient to use hybridblast load with varying mesh size.
Hybrid blast load method has proved to bebetter than multi-Euler domain methods, as faras the following aspects are considered
• Multi-Euler method involves complicatedcalculations. The hybrid blast load methodis easier as compared to multi-Eulerdomain method.
• Since multi-Euler domain method involvescomplex calculations, it is difficult to modelthe structure.
• For more complex geometry, hybrid blastload method is found to be easier andappropriate as compared to multi-Eulerdomain method
• Multi-Euler domain method is anapproximate method. It approximates thegeometry which leads to pressures andimpulses higher than that obtained bynumerical methods.
• Multi-Euler domain leads to a conservativesystem. The results obtained by multi-Eulerdomain method are found to beoverestimated as compared to hybrid blastload method. And hence the design usingthis result will be uneconomical.
SCOPE FOR FUTURE STUDY
The number and intensity of domestic andinternational terrorist activities have heightenedour concerns towards the safety of ourinfrastructure systems. Terrorists attack targetswhere human casualties and economicconsequences are likely to be substantial.Transportation infrastructures have been
considered attractive targets because of theiraccessibility and potential impacts on humanlives and economic activity. A study on blastload generation methods of a highway bridgehas been carried out in this report.
Some of the future needs in this area aredescribed in the following.
• Very few data on blast effects to calibratematerial models for concrete, steel andelastomer have been found. In order toimprove reliability of the finite elementsimulation, it is necessary to calibratematerial models using blast tests on typicalbridge components, e.g., concrete;reinforce concrete, stringers, elastomericbearings, etc.
• In order to fully understand the dynamics ofbridge components during blast loads, anddevelop guidelines, there is a need toinvestigate different variations of bridgecomponents, e.g., wall type piers, circularpiers, continuous bridges, detailing atfootings and in piers, etc.
• There is a need to develop detailedknowledge and guidelines on performancebased multi-hazard seismic-blast design ofbridge components.
REFERENCES
1. Pan Y, Chan B and Cheung M (2013),“Blast Loading Effects on an RC Slab-on-Girder Bridge Super structure Usingthe Multi-Euler Domain Method”, Journalof Bridge Engineering, Vol. 18, No. 11.
2. Yi Z, Agrawal A, Ettouney M and AlampalliS (2013), “Blast Load Effects on HighwayBridges I: Modeling and Blast Load
47
Int. J. Struct. & Civil Engg. Res. 2014 Binol Varghese and Ajith M S, 2014
Effects”, Journal of Bridge Engineering,December.
3. Yi Z, Agrawal A, Ettouney M and Alampalli
S (2014), “Blast Load Effects on HighwayBridges II: Failure Modes and Multihazard Correlations”, Journal of BridgeEngineering, January .