Engineering Structures 32 (2010) 21922209

Contents lists available at ScienceDirect

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

Seismic assessment of an existing non-seismically designed major bridge-abutmentfoundation system

Aman Mwafy a , Oh-Sung Kwon b, , Amr Elnashai c
a b c

Department of Civil and Environmental Engineering, United Arab Emirates University, P.O. Box. 17555, Al-Ain, United Arab Emirates Department of Civil, Architectural and Environmental Engineering, Missouri University of Science and Technology, Rolla, MO 65409, USA Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA

article

info

abstract
A comprehensive study carried out to assess the seismic response of a 59-span bridge using a refined inelastic modeling approach and considering SoilStructure Interaction (SSI) is summarized in this paper. The focus is on describing the methodology adopted to idealize the bridge and its foundation system, while only highlights from the extensive elastic and inelastic analyses are presented. The bridge represents a typical case of vulnerable complex bridges since it was built in the early seventies with minimal seismic design requirements at a distance of about 5 km from a major fault. The SSI analysis is significant in this study due to the length of the bridge, the massive and stiff foundation, and the relatively soft deep soil of the site. A series of three-dimensional dynamic response simulations of the entire bridge are conducted using several analysis tools to verify the developed analytical models. The performance-based assessment study employs 144 site-specific input ground motions representing three seismic scenarios, corresponding to 500, 1000 and 2500 years return periods, to identify areas of vulnerability in the 2164-meter bridge at various hazard levels. It is concluded that the seismic response of the bridge at the 500 years ground motions does not meet todays standards, while the demands under the effect of the 1000 years ground motions almost exceed the capacity of most bridge components. The demands significantly increase under the effect of the 2500 years earthquake scenario and considerably exceed the collapse limit states. The results clearly reflect the benefit of retrofitting different bridge components to mitigate the anticipated seismic risk. The presented assessment study contributes to improve public safety by exploiting the most recent research outcomes in predicting the seismic response of complex highway bridges, which are essential for developing reliable and cost-effective retrofit strategies. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 14 May 2009 Received in revised form 13 March 2010 Accepted 15 March 2010 Available online 18 April 2010 Keywords: Major highway bridges Seismic assessment Inelastic dynamic analysis Soilstructure interaction High seismicity regions Bridge deficiencies

1. Introduction Research carried out during the past two decades has led to significant changes in seismic design provisions of bridges. The introduction of the AASHTO Load and Resistance Factor Design (LRFD) Specifications [1] and the Guide Specifications for LRFD Seismic Bridge Design [2] is aimed at providing more uniform safety for different types of bridge systems. The release of the FHWA Retrofitting Manual for Highway Structures [3] also provides comprehensive procedures for assessment and retrofitting highway bridges based on recent experiences in the US, Japan, and other countries. This major revision in the FHWA Seismic Retrofitting Manual introduces a performance-based retrofit philosophy and defines several methods for the detailed evaluation of existing bridges. These revisions in design specifications

Corresponding author. Tel.: +1 573 341 4536; fax: +1 573 341 4729. E-mail address: kwono@mst.edu (O.-S. Kwon).

and retrofitting guidelines draw attention to the need for seismic assessment of complex highway bridges designed to preceding provisions to determine the level of risk associated with loss of serviceability or possible damage. This is particularly significant in the light of the continuous updates in seismic hazard maps for several regions (e.g. [4]). A large number of bridges were designed and constructed at a time when bridge codes were insufficient according to current standards. The deficiencies in highway bridges designed prior to the seventies result in excessive seismic displacements and large force demands that were substantially underestimated. The anticipated damage includes unseating and pounding of superstructure at abutments and expansion joints, shear and flexural failure of RC piers, beamcolumn joint failure, footing and abutment failure and amplification of response due to soft/liquefiable soil (e.g. [5]). The existing bridge inventory designed to previous provisions is thus likely to suffer damage when subjected to seismic scenarios comparable to those observed in severe earthquakes (e.g. San Fernando, USA, 1971; Loma Prieta, USA, 1989; Northridge, USA, 1994).

0141-0296/$ see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2010.03.022

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2193

In-depth performance-based seismic assessment studies of major highway bridges point the way towards improving the understanding of the complex seismic performance of similar structures. This performance-based evaluation approach requires bridges to satisfy different performance criteria for different levels of ground motion. For instance, the bridge may suffer minor damage but should be operational under frequent earthquakes with low intensity. Under infrequent earthquakes with large intensity, the bridge should provide an acceptable level of life-safety. Quantifying the level of risk associated with anticipated earthquake scenarios enables taking rational decisions to retrofit, replace or accept the risk. The preliminary screening and prioritization approach proposed by the FHWA Retrofitting Manual for Highway Structures [3] classifies bridges according to their Seismic Retrofit Categories (SRC) to identify minimum requirements for screening, evaluation and retrofitting. The FHWA Retrofitting Manual recommends performance criteria according to bridge importance and anticipated service life, with more rigorous performance required for important, relatively new bridges, than for standard bridges at the end of their service life. The manual also introduces a performance-based retrofit philosophy similar to that used for the performance-based design of new buildings and bridges. Following the preliminary screening process, the FHWA retrofit manual [3] recommends six evaluation procedures, with increasing order of rigor, for quantitatively assessing the seismic performance of existing highway bridges. Method E, which is based on the inelastic time history analysis, compares seismic demands with member capacities and determines the degree of damage to the structure at the desired earthquake levels. The latter method, which requires considerable computational effort and a significant level of skill to interpret the results, is recommended for irregular complex bridges with the potential for substantial inelastic behavior particularly when sitespecific ground motions are used. The major bridge assessed in the present study requires the most rigorous evaluation procedure recommended by the FHWA retrofitting manual. The primary objectives of this paper are therefore as follows:

main channel crossing is composed of two-span asymmetrical cantilever steel truss and ten-span steel girders, while approach spans are precast prestressed concrete girders. The substructure includes piers on deep caissons and bents on steel friction piles driven into the near surface silty sands and clayey materials. Bedrock is located more than 800 m below the sand, gravel, and hard clay strata. Fig. 1 shows a three-dimensional view of the I-155 bridge and pictorially summarizes the adopted three-dimensional analytical modeling approach, which is discussed in the following sections. This brief description of the Caruthersville Bridge highlights the pressing need to reliably assess its response under anticipated seismic scenarios. 3. Development of simulation models for inelastic analysis Detailed three-dimensional dynamic response simulations of the entire bridge including foundations and soil effects are carried out using a number of verified analysis platforms. The finite element analysis programs SAP2000 [8] and ZEUS-NL [9] are employed for elastic and inelastic analysis of the structure, respectively. The Pacific Earthquake Engineering Research (PEER) Center analysis platform OpenSees [10] is used for an inelastic simulation of the foundation and the underlying sub-strata. The SAP2000 analytical models are mainly employed for verifications of the ZEUS-NL fiber model before executing the extensive inelastic analysis. ZEUS-NL is mainly employed to estimate the capacities and demands from inelastic pushover and response history analyses. The latter finite element analysis platform was developed at Imperial College London and at University Illinois at UrbanaChampaign, and has been extensively employed in large projects (e.g. [11,12]). 3.1. Super- and sub-structure modeling Three different analytical models are developed for the bridge: (i) SAP2000 detailed model, (ii) SAP2000 simplified model and (iii) ZEUS-NL fiber model. The first model is developed to represent all sub- and super-structural components for elastic analysis. Steel and concrete cross-sections from the SAP2000 library are employed to realistically model different structural members for elastic analysis. This modeling approach, particularly for the superstructure, is computationally demanding for inelastic response history analysis. Also, the design philosophy of bridges relies on bridge piers to dissipate energy rather than the superstructure, which remains elastic. The detailed SAP2000 analytical model is therefore modified to reduce the number of elements and nodes to a manageable limit for inelastic analysis. With the exception of spans with truss, the superstructure is replaced with a number of cross sections with equivalent geometrical properties connected together using rigid arms. This simplification in the superstructure enabled reducing the number of elements and DOFs by about 50%. On the other hand, substructure members are refined by subdividing the columns to a number of elements to accurately monitor the inelastic response during timehistory analysis. Moreover, the SAP2000 joint constraints, which are not available in ZEUS-NL, are replaced with strong arms. The simplified SAP2000 model was transferred to ZEUS-NL for inelastic analysis. Due to the complex behavior of the truss, it was transferred to ZEUS-NL without simplifications. Based on available cross-sections in the library of SAP2000 and ZEUS-NL, equivalent cross-sections are adopted for modeling of the bridge members. In the detailed ZEUS-NL model, each structural member is assembled using a number of cubic elasto-plastic elements capable of representing the spread of inelasticity within the member cross-section and along the member length via the fiber analysis approach. Sections are discretized to steel, confined

Discuss the comprehensive methodology adopted to realistically assess the seismic response of a major bridge and its foundation system taking into account the most recent advances in the analytical modeling approaches of substructure, superstructure, abutments and foundations. Present sample results from the comprehensive vulnerability study to highlight the significance of advanced simulation approaches in identifying areas of vulnerability of complex bridges. Investigate the significance of the recent understanding about the seismicity of the site of the investigated bridge in providing insight into its seismic response at different hazard levels. 2. Description of the I-155 bridge at Caruthersville Bridge A-1700 at Caruthersville, which carries Route I-155 across the Mississippi River between Pemiscot County, Missouri, and Dyer County, Tennessee, is a major bridge that has a pressing need for detailed vulnerability assessment [6]. Although the 59-span 2164-meter bridge is about 5 km from a presumed major fault, it was constructed in the early seventies with minimal seismic design requirements by todays standards. According to the design information [7], Peak Ground Accelerations (PGA) of 0.10g and 0.06g were employed in the seismic design of the bridge in the transverse and longitudinal directions, respectively. Clearly this reflects the limited knowledge about the seismicity of the construction site at the time of designing the bridge; the whole field of study has advanced since the time of construction. The superstructure consists of eleven units separated by expansion joints and supported on a variety of elastomeric and steel bearings. The

2194

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 1. Three-dimensional analytical model of the I-155 bridge at Caruthersville, Missouri, including foundation and abutment effects: (a) three-dimensional simulation model of the entire bridge; (b) components of the main channel crossing; (c) OpenSees abutment model and sample pushover analysis results; (d) OpenSees model and sample pushover results of one of the bridge footings.

and unconfined concrete fibers. The stressstrain response at each fiber is monitored during the entire multi-step analysis. Therefore, the degradation of stiffness with increasing inelasticity is fully accounted for in the ZEUS-NL inelastic analysis. Equivalent gravity loads and mass are distributed in the ZEUSNL fiber model on the superstructure and along the piers height. The employed distributed mass elements utilize cubic shape function and account for both the translational and rotational inertia. The total weight of the bridge is 1562,549 kN (351,275 kip), which includes superstructure, substructure, non-structural members, pile caps and caissons. The corresponding mass is 159,336 kN s2 /m (10,918 kip s2 /ft). The superstructure weight is higher than the substructure in the approach spans, which is not the case in the steel girders and the truss spans. This is due to the lower weight of the steel members compared with the concrete counterparts and the lightweight deck of the steel and truss spans. Also, the massive weight of the caissons significantly increases the substructure weight of the steel and truss spans. As a result of the several deficiencies observed in structural members in the latest available inspection report of the bridge and the lack of reliable information confirming the actual material characteristics, nominal material properties are used in analysis. The cylinder concrete strength (fc ) is 21 MPa (3000 psi) for normal concrete and 28 MPa (4000 psi) for prestressed concrete. The

yield strength of reinforcing steel (fy ) is 40,000 psi. Structural carbon steel ASTM A36-69 (fy = 248 MPa or 36,000 psi) is used for truss members and steel beams. A bilinear model was used to idealize steel members and reinforcement. In this model, the loading and unloading in the elastic range follow a linear function throughout various loading stages with constant stiffness represented by the Youngs modulus of steel. In the post-elastic range, a kinematic hardening rule for the yield surface defined by a linear relationship is assumed [13]. A uniaxial constant active confinement concrete model was employed in the ZEUS-NL analytical model. This model has a good balance between simplicity and accuracy and includes enhancements in the cyclic degradation rules, inelastic strain, and shape of unloading branches [14]. The model, which incorporates the influence of confinement effects on the peak stress and strain as well as on the post-peak stressstrain relationship, can provide a good estimation of the cyclic response of RC members under cyclic and dynamic loading. As suggested from the bridge drawings, appropriate translational () and rotational (r) Degrees of Freedom (DOFs) are released at the wind transfer device of the truss, as shown from Figs. 1 and 2. The translational DOFs, X , and the rotational DOFs, rY and rZ , of the stringers and the deck are also released at certain expansion joints as indicated in the bridge drawings. Bridge bearings and expansion joints are realistically modeled using ZEUS-NL joint elements. The rotation about the transverse axis, rY , is only

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2195

(a) Released DOFs at the truss intermediate hinge.

(b) Gap elements controlling displacement.

Fig. 2. Modeling approach of the truss wind transfer device.

''

''

(a) ZEUS-NL model of the expansion shoe at Bents 15, 19 (left), 21 (right) and 25.

(b) ZEUS-NL model of the Bronze self-lubricating bearings at Bents 8, 14 (left), 26 (right), 31, 41 and 51.

Fig. 3. ZEUS-NL models of bridge bearings.

allowed at the fixed shoes. Based on the drawings, these bearings are located at the substructuresuperstructure connection of Bents 27, 913, 14 (right), 1618, 20, 21 (left), 2224, 26 (left), 2730, 3240, 4250 and 5259. On the other hand, the sliding bearings are located at the two abutments and at the substructuresuperstructure connection of Bents 8, 14 (left), 15, 19, 21 (right), 25, 26 (right), 31, 41 and 51. Modeling of the steel expansion shoes at Bents 15, 19 (left), 21 (right) and 25 follows the analytical model suggested by Mander et al. [15]. This bearing type is highly susceptible to instability since the horizontal motion in the longitudinal direction is accommodated through a rocking motion. This motion causes friction resulting from a rolling resistance at the base and a Coulomb friction at the top hinge [16,17]. Mander et al. [15] observed that debris and uneven wear cause the experimental hysteresis loops to be irregular. The ZEUS-NL model of this bearing is shown in Fig. 3(a). The initial stiffness of this bearing, K1 , is 14 kN/mm with a post-yield stiffness, K2 , of 0.018k1 . The frictional coefficient is 0.04. The analytical model of the expansion bearing at Pier 19 (right) has also an initial stiffness, K1 , of 14 kN/mm with a perfectly plastic post-yield stiffness. However, the longitudinal displacement of this bearing is restricted to 6.0 in. The frictional coefficient is 0.20, as recommended for the low-type expansion bearings investigated by Mander et al. [15]. Movable bearings, which typically have small friction coefficient at the low velocity rates, may have higher friction under high

seismic deformation. For un-lubricated elastomeric bearings, this coefficient at high velocities ranges from 5% to 15%, or even higher at the low temperature (e.g. [18]). It was also concluded in previous experimental studies that the coefficient of friction slightly decreases again under the high velocities due to frictional heating. The Bronze Self-lubricating bearings at Bents 8, 14 (left), 26 (right), 31, 41 and 51 are idealized using the model shown in Fig. 3(b). The initial stiffness, K1 , is taken equal to 123 kN/mm based on the low type sliding bearings tested by Mander et al. [15]. The frictional coefficient of the Bronze Self-lubricating surface is taken equal to 0.10. The post yield stiffness is estimated from the geometry of the system. The horizontal movement of this bearing implies a vertical uplift of the superstructure. The post yield stiffness, K2 = W /r, where W is the weight and r is the radius of the bearing surface curvature. The top plates of these bearings are slotted to allow for 3.5 in. of movement. The displacement in the longitudinal direction is therefore restricted to this limit and the stiffness of the system significantly increases (K3 ) if the maximum displacement is reached, as shown from Fig. 3(a). The structural gaps at the abutments and the expansion joints are considered in the inelastic analysis performed using ZEUSNL. In the latter modeling approach, joint elements with tri-linear asymmetric elasto-plastic idealization capable of representing the slippage and collision are employed. Two modeling approaches were investigated to idealize the impact of stiffness at the expansion joints. In the first approach, the pounding is represented by a

2196

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 4. ZEUS-NL model of the expansion joint at Abutment 1, Bents 8, 14 (left), 26 (right), 31, 41, 51 and Abutment 60.

Fig. 5. Vertically exaggerated view of soil profile along the length of the bridge.

nonlinear spring (Hertz model). The impact stiffness increases in this approach from 2.92E6 MPa to 4.38E6 MPa at a penetration of 1.27 mm, and increases again at 2.54 mm to 8.76E6 MPa (e.g. [19, 20]). These values are controlled to ensure that the penetration of pounding is less than 2.54 mm (0.1 in.). Results obtained from this idealization were compared with a more simplified linear spring model employing the maximum impact stiffness of 8.76E6 MPa. It was concluded that the former model has a marginal effect on the response compared with the simplified linear model. It was therefore decided to use the simplified approach to model the impact stiffness. Fig. 4 shows the force versus relative displacement relationships of the joint elements representing the expansion joint. In this modeling, a positive relative displacement corresponds to an opening of the joint gap and a negative displacement corresponds to a closing of the gap. When the gap at the abutment and at the expansion joint undergoes a relative movement in the negative direction (joint close) exceeding the gap width, the joint element begins resisting further opening (collisions). The modes of vibration obtained from modal analysis reflect the high contribution of the steel truss to the seismic response of the bridge, as subsequently discussed. Accordingly, 4% damping ratio is considered in the SAP2000 model used for elastic analysis. The mass-proportional and stiffness-proportional parameters are calculated for Response History Analysis (RHA) based on the predominant periods of the structure in the two principle directions. The damping and material modeling are critical components in the prediction of inelastic seismic structural response [21]. Hence, they were thoroughly investigated before executing the ZEUS-NL inelastic analysis. Hysteretic damping, which is responsible for the dissipation of the majority of energy introduced by the earthquake action, is accounted for in the inelastic fiber formulation of the inelastic elements. A relatively small quantity of non-hysteretic type of damping should be also added to the inelastic model.

This is achieved through stiffness-proportional damping [22]. 1.0%, 1.5% and 2.0% stiffness-proportional damping ratios are investigated. It is noteworthy that these damping ratios were applied on substructure members, which are the main source of energy dissipation. A higher level of damping is applied on the superstructure, which is anticipated to remain in the elastic range. To compare the impact of damping on the response of the Caruthersville Bridge obtained from SAP2000 and ZEUS-NL inelastic analysis, an input ground motion developed to match the Uniform Hazard Spectrum (UHS) of the bridge for a 500 years return period is used. The record was scaled to a Peak Ground Acceleration (PGA) of 0.05g to avoid a significant inelasticity from the ZEUS-NL nonlinear analysis. The objective from the comparison is to tune the ZEUS-NL damping level and to verify its results with those obtained from the SAP2000 elastic analysis. The results indicated that the ZEUS-NL results are comparable to those obtained from SAP2000 when a 2% damping level is used. The displacement and the base shear demands from ZEUS-NL were significantly higher than those from SAP2000 when a lower level of damping is used. Based on these comparisons, it was decided to employ a 2% stiffness-proportional damping in inelastic analysis. This damping ratio is applied on substructure, which is the main source of energy dissipation. Higher level of damping is applied on superstructure, which is anticipated to remain in the elastic range. 3.2. Soil and foundation modeling Soil and foundation may have significant impact on the seismic response of bridge structures, particularly those with stiff foundation and relatively soft deep soil [23]. Refined inelastic simulations of the foundation and the underlying sub-strata are undertaken using the Pacific Earthquake Engineering Research Center analysis platform OpenSees [10]. The objective is to realistically estimate the soil properties required for SAP2000 and ZEUS-NL soil and

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2197

y x

y x

y x

Fig. 6. Different bridge footing profiles.

foundation modeling. Fig. 5 shows a vertically exaggerated view of soil profile along the length of the bridge. The site consists of 8 types of soil materials. Stratum 1 and 2 are cohesive material with unconfined compressive strength of 67 and 48 kPa, respectively. This range of strength is for medium to stiff clay. The upper layers of approach spans are covered by either stratum 1 or 2. These strata are modeled with pressure-independent soil material model. Strata 3 to 8 are cohesionless materials. The friction angle of the materials ranges from 30 to 42. The in-situ properties of the cohesionless material at deeper layer are expected to have higher stiffness due to the large confinement. Since the shear modulus of each stratum is given as a constant value regardless of the depth of the stratum, it is assumed that the shear modulus is calculated at the mid-depth of each stratum. The effective reference pressure at the mid-depth is approximately calculated and used as the input parameters for soil and foundation modeling. Strata 3 to 8 are thus modeled with pressure-dependent material, which can dilate or contract with shear deformation depending on the initial density ratio. Based on the density ratio of various soil profiles, strata 3 and 4 are assumed to be medium sand, strata 5 and 6 as mediumdense sand and strata 7 and 8 as dense sand. The dilatation and contraction parameters in the pressure-dependent material models are chosen from the suggested values by the developer of the material model [24]. The resistance of soil medium surrounding pile largely depends on the contact surface area normal to the direction of the pile movement. Due to the large number of piles and the amount of analyses required, it is computationally demanding to model each pile using several finite elements. A single brick element with an equivalent projected area is therefore used for idealizing the pile section and reducing the computational demands. The equivalent element used in analysis has the similar properties of the actual piles in terms of flexural and axial rigidity. Based on the soil profile, number of piles and batter angle, thirteen soilfoundation profiles were idealized using OpenSees. Unlike the pile caps at expansion joint bents, those at other locations have battered piles. The number of piles varies from 9 to 112, depending on the supporting loads. Bent 19, 20, and 21 are supported on massive

Fig. 7. Controlling different foundation systems using a single node for inelastic pushover analysis.

caisson. Fig. 6 describes different bridge foundation profiles. The foundation and soil medium are all modeled with 8 node brick elements. The side boundary of the soil medium is restrained in the horizontal translation as shown in Fig. 1(d). Vertical DOFs of side boundary are released to allow settlement due to gravity loads. All DOFs of the bottom nodes of the soil medium are restrained. To reduce the controlling nodes within the pile cap and to prevent local deformations, it is assumed that the pile caps and caissons behave as rigid bodies. All foundation profiles, except caissons, are thus controlled using a single node connected to eight boundary nodes, as shown in Fig. 7. Symmetry is utilized to reduce the FE mesh and computational demands for certain types of foundation profiles (e.g. Class 6). On the other hand, it was difficult to control the response of the caissons using a single node due to their massive stiffness. Hence, a number of nodes are used to control their response, as shown in Fig. 8. Displacement-controlled pushover analyses are carried out using the above-mentioned refined FE models to evaluate the loaddeformation relationship of

2198

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 8. Modeling of the bridge caissons for OpenSees pushover analysis: (a) translational force; (b) rotational moment.

Fig. 9. Idealizing bridge footings using tri-linear relationships: (a) response of footing Class 2 under cyclic and monotonic loading; (b) adopted ZEUS-NL tri-linear model.

each foundation class in Fig. 6. The loaddeformation relationships of various foundation profiles are derived from the reactions at the control nodes. 4. Foundation and abutment response under cyclic and monotonic loading 4.1. Bridge foundations Sample results from the extensive foundationsoil analyses are presented below. The loaddeformation relationship of foundation Class 2 under the effect of both cyclic and monotonic loading is shown in Fig. 9(a). The thin solid line is from cyclic loading, while the thick solid line is from monotonic loading. The results confirm that the backbone of the hysteretic curve follows the monotonic pushover curve. It was decided based on this comparison to analyze other foundations profiles under a monotonic loading to estimate their loaddeformation relationships. For linearization of the nonlinear stiffness of translational DOFs, the maximum force is assumed to be twice of the reaction from dead loads, which corresponds to inertial force from 2g of horizontal acceleration. Idealized tri-linear relationships are subsequently used as soil springs for inelastic analysis. Fig. 9(b) depicts the adopted ZEUS-NL tri-linear model used to idealize bridge foundations. Figs. 1012 show the iso-deformation contours and the corresponding loaddeformation curves obtained from OpenSees pushover analysis of foundation Classes 5, 6 and 9, respectively. The full results of these comprehensive analyses for all foundation classes are given elsewhere [25]. Tri-linear idealizations are adopted to simplify the monotonic pushover curves of different foundation classes. The yield displacement and post-yield stiffness are chosen so that the tri-linear model closely represents the loaddeformation curve obtained from OpenSees pushover analysis. The idealized loaddeformation relationships are therefore used to model the foundation system in SAP2000 and ZEUS-NL analytical models. A summary of the foundation stiffness properties obtained from the OpenSees analyses is given in Table 1.

4.2. Abutments The analytical model of the bridge abutments developed using OpenSees along with the soil profile is described in Figs. 1(c) and 13. Stratum 1 and 2 are cohesive material with undrained strength (Su) of 67 and 48 kPa (0.7 and 0.5 tsf), respectively. All other strata are cohesionless material with friction angle ranging from 30 to 42. The inelastic analysis carried out in the longitudinal and transverse directions of the bridge indicated that the response is almost linear. Results of pushover analyses showed a degrading stiffness but at very high level of force. Table 2 shows a summary of the abutment stiffness properties in the longitudinal and transverse direction of the bridge. The vertical translation and rotations of the two abutments are fully restrained. The abutment modeling approach obtained using OpenSees is compared in Table 3 with a more simplified approach suggested by Caltrans [26]. In the latter approach, it is assumed that wing walls do not significantly contribute to the transverse horizontal resistance. Therefore, the transverse horizontal stiffness of the abutment is estimated using the translational stiffness of the abutment pile group. In the longitudinal direction, the abutment embankment fill stiffness is estimated based on the finding of a large-scale abutment testing [27]. The initial passive stiffness from this testing (11.5 kN/mm/m) is adjusted relative to the back-wall height [26]. A maximum passive pressure was recommended in the latter study, allowing the back-wall to break off in order to prevent inelasticity in the foundation system. The stiffness in active action is assumed to be one fifth of the passive stiffness [19]. A bilinear elasto-plastic relationship is therefore adopted to model the longitudinal behavior of the abutment. The vertical translation and rotations of the abutment are fully restrained. It is clear from the comparison shown in Table 3 that the refined OpenSees model results in a significantly different response at the investigated low level of input ground motion (PGA = 0.05g). This confirms the significance of the refined modeling approach adopted in the present study for different components of the Caruthersville Bridge.

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2199

Fig. 10. Iso-deformation contours and the corresponding loaddeformation curves obtained from the inelastic pushover analysis of foundation Class 5.

5. Performance criteria and analysis procedures To assess the seismic performance from inelastic static and dynamic analysis results, reliable definition of response parameters is needed. Two particular limit states in the response

of the bridge are required to be defined; one at which significant yield occurs and one at which the first indication of failure is observed. Local yield is assumed when the strain in the main longitudinal tensile reinforcement exceeds the yield strain of steel. The response from the inelastic timehistory analysis is

2200

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 11. Iso-deformation contours and the corresponding loaddeformation curves obtained from the inelastic pushover analysis of foundation Class 6.

monitored and any occurrence of yielding in the main longitudinal reinforcement is reported. On the other hand, the definition of failure or collapse criteria should represent an accurate or a slightly conservative estimation of structural collapse due to its significance role. The adopted failure or collapse performance criteria are as follows:

(i) Drift: The drift limit is defined as the ratio of relative displacement between the top and the base of the bridge piers to their height. Several values for the drift collapse limit state have been suggested in the literature and in seismic codes. A drift limit in excess of 3.0% is considered in the present study as an indication of a global collapse.

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209 Table 1 Stiffness properties of selected foundation classes. Footing DOF X Y Z rx ry X Y Z rx ry X Y Z rx ry X Y Z rx ry X Y Z rx ry X Y Z rx ry X Y Z rx ry X Y Z rx ry k0 1.0190E+06 8.8900E+05 1.5880E+06 3.8420E+06 5.6010E+06 1.2460E+06 1.1720E+06 2.3180E+06 5.8250E+06 1.2630E+07 1.4600E+06 1.3990E+06 2.3980E+06 5.5420E+06 1.3330E+07 4.5970E+06 4.0710E+06 1.0070E+07 1.8470E+08 1.6350E+08 1.3870E+07 1.2750E+07 1.4590E+07 1.0370E+09 9.9430E+08 2.3610E+07 2.2670E+07 3.3840E+07 2.6840E+09 3.3160E+09 1.5830E+07 1.6250E+07 2.3900E+07 1.9730E+09 1.5450E+09 1.7000E+06 1.5170E+06 3.7330E+06 3.2170E+07 8.3320E+07 d0 9.0620E04 1.3240E03 3.4810E04 1.0000E03 1.0000E03 1.4530E03 1.7910E03 4.3630E04 1.0000E03 1.0000E03 9.6050E04 1.0670E03 3.6600E04 1.0000E03 3.5200E03 5.0000E03 5.0000E03 2.7800E03 2.0000E03 2.0000E03 5.0000E03 5.0000E03 2.9200E03 8.8140E04 8.9480E04 8.0000E03 8.0000E03 2.5200E03 5.0000E04 1.0000E03 8.0500E03 5.9000E03 4.8650E04 4.0530E04 7.1090E04 1.4190E03 1.7630E03 4.8170E04 1.0000E03 9.0000E04 k1 6.6800E+05 4.7530E+05 1.5310E+06 1.7650E+06 3.0330E+06 7.3990E+05 6.0320E+05 2.1950E+06 2.7070E+06 5.6610E+06 1.0180E+06 9.3220E+05 2.3120E+06 3.0980E+06 7.0020E+06 2.3350E+06 1.6960E+06 8.4600E+06 8.0370E+07 6.7970E+07 6.3950E+06 5.7650E+06 8.1440E+06 5.3190E+08 2.8610E+08 9.1250E+06 8.4070E+06 2.3350E+07 1.7570E+09 8.6050E+08 5.0920E+06 6.4540E+06 2.1390E+07 1.0300E+09 4.7120E+08 1.3630E+06 1.1250E+06 3.6460E+06 2.1060E+07 5.6540E+07 d1 4.5310E03 6.6210E03 1.7400E03 6.0000E03 6.0000E03 7.2640E03 8.9570E03 2.1820E03 6.0000E03 6.0000E03 4.8030E03 5.3330E03 1.8300E03 6.0000E03 8.8000E03 2.5000E02 2.5000E02 1.3900E02 8.0000E03 1.0000E02 2.5000E02 2.0000E02 1.4600E02 5.2890E03 4.4740E03 4.0000E02 4.0000E02 1.2600E02 3.0000E03 5.0000E03 4.8300E02 2.3600E02 2.4320E03 2.4320E03 3.5540E03 7.0970E03 8.8160E03 2.4090E03 6.0000E03 5.4000E03 k2

2201

Class 02

3.9710E+05 2.1890E+05 1.4130E+06 1.0750E+06 1.3120E+06 3.7110E+05 2.6620E+05 1.8170E+06 1.2540E+06 1.5270E+06 6.0370E+05 5.1470E+05 2.1600E+06 1.6860E+06 3.0680E+06 1.5070E+06 1.0480E+06 2.4140E+06 3.9390E+07 4.8650E+07 3.5870E+06 2.4440E+06 1.5810E+06 4.2390E+08 1.6510E+08 3.8890E+06 2.6990E+06 1.7960E+06 1.3810E+09 3.7480E+08 1.6560E+06 2.5670E+06 1.8410E+07 8.2870E+08 2.4890E+08 9.5200E+05 7.1490E+05 3.3580E+06 1.3060E+07 3.0010E+07

Class 04

Class 05

Class 06

Class 07

Class 08

Class 09

Class 10

Unit: force (kN), moment (kN m), length (m), rotation (rad). K0 : Initial stiffness. d1 : First branch displacement limit. K1 : Second branch stiffness. d2 : Second branch displacement limit. K2 : Third branch stiffness. Also refer to Fig. 9(b) for the definition of stiffness properties. Table 2 Summary of abutment stiffness properties. DOF X Y k0 1.3250E+06 9.2060E+05 d0 2.2180E04 3.2070E04 k1 1.2560E+06 8.5390E+05 d1 1.1090E03 1.6040E03 k2 1.0500E+06 7.3660E+05

Unit: force (kN), length (m). Table 3 Response of the bridge in the transverse direction using different abutment idealization. Bent Top displacement (mm) Caltrans 2 15 20 21 25 59 5.2 11.7 38.1 23.2 8.1 5.9 OpenSees 3.0 9.5 22.7 10.4 6.7 2.0 Diff. (%) 74 23 68 123 22 195 Base shear (kN) Caltrans 422 5 072 42 638 30 666 988 671 OpenSees 272 4 869 31 153 18 530 857 231 Diff. (%) 55 4 37 65 15 191

Response history analysis results from Rec1-500-SE scaled to a PGA of 0.05g.

(ii) Degradation of lateral resistance: Collapse is considered to have occurred when lateral resistance of a structure drops by more than 10% below its peak value. (iii) Formation of hinging mechanism: Such a mechanism involves plastic hinges at the base of all bents. Although this

mechanism is difficult to occur, particularly for the long bridge under consideration, the spread of yielding in substructural members along the length of the bridge segments, particularly for the main approach, namely Piers 15 to 21, is considered here as an indication of structural collapse.

2202

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 12. Iso-deformation contours and the corresponding loaddeformation curves obtained from the inelastic pushover analysis of foundation Class 9.

(iv) Capacity versus demand ratios: The capacity of different bridge components are estimated from extensive pushover analyses and hand calculations. This is carried out for:

Bearings. Expansion joints and truss intermediate hinge.

The capacity curves of the foundation and bents are compared with the demands from inelastic Response History analysis (RHA). Similarly, the capacities of all bridge bearings, including pinned

Bridge foundations. Bents.

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2203

b a

Fig. 13. OpenSees abutment model and soil profile: (a) three-dimensional simulation abutment model; (b) pushover analysis results in the longitudinal direction; (c) abutment soil profile.

and sliding bearings, are compared with the maximum demands from inelastic RHA. Finally, the opening and closing of the joint gaps is monitored to allow for predicting the actual behavior of the structure and a possible bounding between different structural segments and any slippage at expansion bearings. An indication of structural damage is considered if the superstructure displacement in the longitudinal direction frequently exceeded the gap provided at the expansion joints, causing pounding between bridge segments. Total collapse is assumed if the displacement exceeded the seat width. As mentioned above, two analysis platforms are employed in the present study. Additionally, two analytical models were developed using SAP2000; a detailed and a simplified model. The latter idealization simplifies modeling of the superstructure, while refines the substructural members idealization, which are responsible for dissipating energy. This allows reducing the size of the simplified model, which is transferred to ZEUSNL for inelastic analysis. Bearings friction and expansion joints are considered only in the ZEUS-NL refined fiber model, which accounts for the material inelasticity and geometric nonlinearity. The kinematic soilstructure interaction is accounted for in all modeling approaches by restraining the pile caps and the caissons at their center of gravity with grounded springs representing the foundation stiffnesses estimated from the extensive OpenSees inelastic analyses. Three different analytical models are therefore employed in the present study. The following analyses are carried out using the three structural models:

Inelastic static pushover analyses in the longitudinal and

transverse directions using the ZEUS-NL model.

Elastic dynamic analyses using the SAP2000 and the ZEUS-NL

models.

Inelastic dynamic analyses in the longitudinal and transverse

directions using the ZEUS-NL models and 144 site-specific synthetic records representing three seismic scenarios. Eigenvalue analyses are first conducted to determine the noncracked horizontal and vertical periods of vibration and mode shapes. This simple analysis is used as an initial validation tool of the analytical models. Inelastic static pushover analyses are performed in both the longitudinal and transverse directions of the bents and the bridge foundations to evaluate their lateral capacities. Elastic and inelastic timehistory analyses are initially performed to verify the analytical models and tune the non-hysteretic damping level. Extensive inelastic RHA are then executed to examine the response of the structures under the effect of site-specific input ground motions with increasing severity. Sample results of these comprehensive analyses are discussed below. 6. Dynamic characteristics and verifications of the analytical models Sample results from the SAP2000 and ZEUS-NL free vibration analyses are presented in Figs. 14 and 15. The mode shapes and the corresponding periods of vibration obtained from the SAP2000 and ZEUS-NL models are presented. It is important to note that the

Free vibration analyses using the detailed and a simplified

SAP2000 models as well as the ZEUS-NL model.

2204

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

(a) Detailed model.

(b) Simplified model.

Fig. 14. Comparison of the bridge dynamic characteristics obtained from the detailed versus the simplified SAP2000 models.

(a) ZEUS-NL model (without reinforcement).

(b) SAP2000 simplified model.

Fig. 15. Comparison of the bridge dynamic characteristics obtained from the ZEUS-NL and SAP2000 models.

bridge bearings frictional resistance is disregard in these analyses. The significance of bearing friction on the dynamic characteristics of highway bridges was confirmed from several previous studies (e.g. [28]). Multi-linear elasto-plastic joint elements are therefore employed to idealize the movable bearings in the refined inelastic analyses carried out using ZEUS-NL, as explained above. The free vibration analyses confirmed that about 350 modes of vibration are required to reach a 90% mass participation in the two principle directions of the bridge. Higher modes of vibrations notably contribute to seismic response due to the length of the 2164-

meter (7100-foot) bridge and the non-uniform distribution of stiffness and mass of this complex bridge. High longitudinal vibrations are observed in the truss portal frames from the simple free vibration analyses, which reflect the vulnerability of these structural members. To verify the simplifications made in the superstructure modeling, the detailed and the simplified SAP2000 models are compared in Fig. 14. It is clear that the first mode shapes and their periods of vibration are comparable from the two models, which lends weight to the simplifications adopted in the superstructure idealization.

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209 Table 4 Surface input ground motions used for assessment of the Caruthersville Bridge. Scenario 1 Ground motion seta Rec1-500-SE Rec2-500-SE Rec3-500-SE Rec1-500-AS Rec2-500-AS Rec3-500-AS Rec1-1000-AS Rec2-1000-AS Rec3-1000-AS Rec1-2500-AS Rec2-2500-AS Rec3-2500-AS Return period (year) 500 Site effectsb
           

2205

Asynchronous ground motionsc

Number of records 12 12 12 12 12 12 12 12 12 12 12 12

500

1000

2500

a Each input ground motion set has twelve different ground motion records, corresponding to the six soil profiles at Bents 1, 11, 15, 19, 22 and 27 in the longitudinal and transverse directions of the bridge. b Only site effects were accounted for. c Incoherency, wave passage and site effects were accounted for.

Moreover, the SAP2000 simplified model is compared with ZEUSNL fiber model in Fig. 15. The effect of steel reinforcement, which increases stiffness and shortens periods of vibrations, was omitted from the latter model to obtain rational results from the two models. The results reflect the effectiveness of transferring the SAP2000 model to ZEUS-NL and confirm the rationality of the latter model for predicting the inelastic response of the bridge. Adding the steel reinforcement to the model increases stiffness and shortening the periods of the bridge by up to 8%. 7. Comparison of seismic demand versus capacity Inelastic static pushover analyses are performed in both the longitudinal and transverse directions of bridge to evaluate the lateral capacity of the bridge bents. This analysis is also conducted for all foundation classes to estimate their capacities, as explained above. Extensive inelastic RHAs are finally executed using sitespecific input ground motions to examine the response of the bridge under various seismic scenarios with increasing severity. Capacities of the foundations, bents, bearings and expansion joints are compared with seismic demands at various hazard levels. Sample results from these comprehensive analyses are discussed below to highlight the significance of advanced simulation approaches in identifying areas of vulnerability of complex bridges. Detailed results from the comprehensive seismic assessment study carried out for the A-1700 Bridge under the effect of different seismic scenarios are presented elsewhere [25,29,30]. 7.1. Input ground motions Probabilistic Seismic Hazard Analyses (PSHA) for hard rock site conditions were performed for the bridge site [31]. Three hazard levels corresponding to return periods of 500, 1000 and 2500 years were considered. Fig. 16(a) shows the location of the Caruthersville Bridge with respect to the New Madrid seismic faults, while the developed Uniform Hazard Spectra (UHS) are shown in Fig. 16(b). Ten records of spectrum compatible ground motions for site class A were generated for each of the three hazard levels of 10%, 5% and 2% probability of exceedance in 50 years. Each of the records included two components aligned in the longitudinal and transverse directions with respect to the Caruthersville Bridge. For more details about seismic hazard analysis of the bridge site, reference is made to Ferenndez and Rix [31]. Three input ground motions were selected from the above-mentioned ten records for propagation through the thick embayment deposits [32]. Six

distinct soil profiles along the length of the bridge were also selected for use in one-dimensional seismic site response analyses. Therefore, each of the three input ground motions used in the site response analyses produced twelve different ground motion records, corresponding to the soil profiles at Bents 1, 11, 15, 19, 22 and 27 (refer to Fig. 5) in the longitudinal and transverse directions of the bridge. The soil profile, shear wave velocity, and modulus reduction and damping curves that were used for site amplification are presented elsewhere [30]. Fig. 16(c) depicts the response spectrum of an input ground motions generated for bedrock in the longitudinal direction of the bridge and those propagated to the surface for the 2500 years seismic scenario. It is clear that the surface motions have lower amplification in the short period range compared with bedrock motions. The former ground motions significantly amplify in the mid period range, which is in the range of 0.11.0 s for the 500 years records and above 1.0 s for the 1000 and 2500 years return period records [32]. Given the length of the 2164-meter (7100-foot) bridge, the ground motion spatial variability were also included in the propagated records. In addition to local site effects, two additional effects were considered [33,3,34]: (i) geometric incoherence effects due to the scattering in the heterogeneous ground, and (ii) wave-passage effects, where non-vertical waves reach different points on the ground surface at different times, producing a time shift between the motions at those points. The coherency function by Abrahamson et al. [33], which provides quantitative measure of similarities in two ground motions recorded at different seismic stations, was applied to compute incoherency at Bents 1, 11, 15, 19, 22 and 27 [34]. The input ground motions were propagated to the ground surface through the corresponding six soil profiles after including the geometric incoherence and wavepassage effects. Four different scenarios of ground motions were generated following the above-mentioned procedures. Table 4 shows the surface input ground motions used for assessment of the Caruthersville Bridge. In total 144 input ground motions are used in the present assessment study. Ground motions due to site effect only are used since they represent the typical site-specific ground motions used in inelastic dynamic analysis. Spatially varied motions due to site effect, wave passage effect and ground motion incoherency are also employed since they represent the real earthquake effect. The development of spatially variable ground motions is discussed in more detail by Tsai and Hashash [34] and Mwafy et al. (2010). The latter study also investigated the impact of wave passage and ground motion incoherency, and concluded that asynchronous ground motions have significant implications on the seismic behavior of extended bridges.

2206

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 16. Site-specific input ground motions: (a) location of the Caruthersville Bridge; (b) bedrock Uniform Hazard Spectra (UHS) for return periods of 500, 1000 and 2500 years; (c) input ground motions generated for bedrock and those propagated to the surface for the 2500 years seismic scenario.

(a) Capacity of expansion joints vs displacement demand (most critical ground motion set in longitudinal direction).

(b) Capacity vs demand at the top columns of Bents 1921 (ground motions are in transverse direction).

(c) Bearing force capacity versus demand (most critical ground motion set in transverse direction). Fig. 17. Capacity versus demand of different bridge components under the 500 years earthquake scenario.

7.2. Capacity versus demand under the 500 years earthquake scenario Due to the length of the bridge and the non-uniform distribution of stiffness and mass, higher modes of vibrations notably contribute to the seismic response. Inelastic response history anal-

yses carried out in the transverse directions of the bridge indicate that the drift demands are acceptable (less than 0.63%), with the exception of the high displacement at the truss intermediate hinge. Ground motion set 2 (Rec2-500-AS) significantly amplifies the truss deformation demands compared with other records. The

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

2207

(a) Mapping of plastic hinge formation (most critical ground motion set in transverse direction).

(b) Capacity versus bending moment demands of foundation class 3 (ground motions are in transverse direction).

(c) Bearing force capacity versus demand (most critical ground motion set in longitudinal direction). Fig. 18. Response of different bridge components under the 1000 years earthquake scenario.

relative displacements between piers along the length of the bridge are inconsistent as a result of the difference in stiffness, which causes high demands on the superstructure. Higher drift demands are observed in the longitudinal direction of the bridge (1.1%) compared with the transverse direction due to its lower stiffness. High deformations are observed in the truss portal frames, which are also observed from free vibration analysis. The high relative displacement demands in the longitudinal direction cause pounding at the two abutments and the expansion joint of Bent 8, as shown in Fig. 17(a). The response of the foundation system are acceptable in both directions, while high demands are observed at the top columns of Bents 1521, as shown in Fig. 17(b). The demands of other bents are moderate to high. Several plastic hinges are observed when applying the load in the transverse direction, particularly at the top columns of Bents 1521. The non-uniform distribution of stiffness along the height of these piers causes high

stress concentrations at the top columns. A number of bridge bearings are vulnerable under the effect of this earthquake scenario, particularly the bearings at the expansion joints, as shown in Fig. 17(c). 7.3. Capacity versus demand under the 1000 years earthquake scenario The displacement and force demands increase under this seismic scenario by up to 200% and 100%, respectively, compared with the 500 years seismic scenario. The maximum drift demands in the longitudinal (3.8%) and the transverse (2%) directions are unacceptable. Exceeding the 3% drift limit is considered in the present study as an indication of extensive structural damage. The high relative displacement demands in the longitudinal direction

2208

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209

Fig. 19. Bearing capacity versus demand under the 2500 years ground motions (most critical record in the longitudinal direction).

cause pounding at all expansion joints. Unacceptable displacement demands (934 mm) are observed in the transverse direction at the truss intermediate hinge, particularly under the effect of Ground motion set 2 (Rec2-1000-AS). Extensive damage and yielding are also detected in several bents. The high ductility demands imposed on the top columns of Bents 1521 cause severe damage. The spread of plastic hinges is extensive, as shown in Fig. 18(a), while a severe damage in several bearing is confirmed, as confirmed from Fig. 18(c). Unacceptable inelastic response is observed in some footings under the effect of the 1000 years seismic scenario, as shown in Fig. 18(b). 7.4. Capacity versus demand under the 2500 years earthquake scenario Displacement demands significantly increase in the longitudinal direction of the bridge by up to 90% compared with those from the 1000 years input ground motions. The results obtained from Ground motion set 2 (Rec2-2500-AS) confirm that the drift at the top of bents (7.25%) considerably exceeds the collapse limit state. The very high relative displacement demands cause collision between all bridge segments. Significantly higher force demands are also observed in the bents and the foundation system compared with those observed from the 1000 years earthquake scenario. The inelastic response of the foundation system is confirmed, while extensive damage and yielding at the base of major bents supporting the steel truss are detected. Most of the bridge bearings are vulnerable (capacity/demand ratio is less than unity) under the effect of the 2500 years earthquake scenario, as shown in Fig. 19. The results confirm the need to retrofit different bridge components to mitigate the potential seismic risk. 8. Conclusions This paper presents highlights of a project carried out to assess the seismic response of a 59-span bridge, considering SoilStructure Interaction (SSI). The focus is on describing the methodology adopted to idealize the bridge and its foundation system, while only sample results from the extensive elastic and inelastic analyses under the effect of 144 site-specific input ground motions are presented. The bridge was built in the vicinity of a major source of earthquakes; the New Madrid Seismic Zone, and includes typical deficiencies of bridges constructed without current seismic provisions. The refined three-dimensional simulations of the bridge and its foundation system carried out using verified analysis tools and a detailed site-specific seismic hazard study enabled the identification of areas of vulnerability of the investigated bridge and assessment of its response at three hazard levels, corresponding to 500, 1000 and 2500 years return periods. The study confirmed that simplifying modeling assumptions of different bridge components may have significant impact on the seismic response of complex bridges. Higher modes of vibration notably

contributed to the seismic response of the I-155 bridge due to the length of the bridge and the non-uniform distribution of stiffness and mass. Under the 500 years ground motions, the response of the bridge was unacceptable due to the observed yielding and damage in a number of bents and bearings. The demands corresponding to the 1000 years ground motions almost exceeded the collapse limit state and the capacity of bridge components. Indications of yielding in foundations were also observed. The displacement and force demands under the effect of the 2500 years earthquake scenario significantly increased compared with the 1000 years input ground motions and exceeded by far the collapse limit states. The presented assessment study confirmed the need to retrofit different bridge components to mitigate potential seismic risk. The tools and procedures used for this assessment are applicable to similar situations of complex bridges constructed in a mix of material and structural systems. Acknowledgements This study was funded by the Missouri Department of Transportation (MoDOT) with Jacobs Civil Inc. as main contractor, and the Mid-America Earthquake Center (MAE) as sub-contractor. The MAE Center is an Engineering Research Center funded by the National Science Foundation under cooperative agreement reference EEC 97-01785. All findings and conclusions are those of the authors and do not represent Jacobs or MoDoT. References
[1] AASHTO. AASHTO LRFD bridge design specifications. 4th ed. Washington (DC): American Association of State Highway and Transportation Officials; 2007. [2] AASHTO. Guide specifications for LRFD seismic bridge design. 1st ed. Washington (DC): American Association of State Highway and Transportation Officials; 2009. [3] FHWA. Seismic retrofitting manual for highway structures: part 1-bridges. Publication No. FHWA-HRT-06-032. Washington (DC): Federal Highway Administration; 2006. [4] USGS. United States National Seismic Hazard Maps, US. Geological Survey Earthquake Hazards Program. http://earthquake.usgs.gov/ (Jan. 16, 2009). [5] Elnashai AS, Mwafy AM. Seismic response and design. In: Parke G, Hewson NR, editors. ICE manual of bridge engineering. 2nd ed. London: Thomas Telford Ltd.; 2008 [Chapter 5]. [6] Mwafy AM, Kwon O-S, Elnashai AS. Inelastic seismic response of a 59-span bridge with soil-structure interaction. In: Proceedings of the 14th world conference on earthquake engineering. 2008. [7] TDOT & MODOT. As-built drawings of a bridge over mississippi river near Caruthersville Missouri-bridge no. A-1700. Tennessee Department of Transportation & Missouri Department of Transportation; 1974. [8] CSI. SAP2000structural analysis program. Berkeley (California): Computers and Structures Inc; 2006. [9] Elnashai AS, Papanikolaou V, Lee D. ZEUS-NLa system for inelastic analysis of structures. Urbana (IL): Mid-America Earthquake Center, University of Illinois at Urbana-Champaign; 2008. [10] McKenna F, Fenves GL. et al. OpenSeesopen system for earthquake engineering simulation. Berkeley (CA): Pacific Earthquake Engineering Research Center; 2006. [11] Jeong S-H, Elnashai AS. Analytical assessment of an irregular RC frame for fullscale 3D pseudo-dynamic testing-part I: analytical model verification. J Earthq Eng 2005;9(1):95128.

A. Mwafy et al. / Engineering Structures 32 (2010) 21922209 [12] Mwafy AM, Elnashai AS. Importance of shear assessment of concrete structures detailed to different capacity design requirements. Eng Struct 2008; 30(6):1590604. [13] Elnashai AS, Izzuddin BA. Modelling of material nonlinearities in steel structures subjected to transient dynamic loading. Earthq Eng Struct Dyn 1993;22:50932. [14] Martinez-Rueda JE, Elnashai AS. Confined concrete model under cyclic load. Mater Struct 1997;30:13947. [15] Mander JB, Kim DK, Chen SS, Premus GJ. Response of steel bridge bearings to the reversed cyclic loading. Report No. NCEER 96-0014. National Center for Earthquake Engineering Research; 1996. [16] DesRoches R, Leon RL, Dyke S. Response modification of bridges. Mid-America Earthquake Center Research Report. Urbana (IL): University of Illinois at Urbana-Champaign; 2003. [17] Nielson BG. Analytical fragility curves for highway bridges in moderate seismic zones. Ph.D. thesis. Georgia: Georgia Institute of Technology; 2005. [18] Priestley MJN, Seible F, Calvi GM. Seismic design and retrofit of bridges. NY: Wiley; 1996. [19] Choi E, DesRoches R, Nielson B. Seismic fragility of typical bridges in moderate seismic zones. Eng Struct 2004;26:18799. [20] Muthukumar S, DesRoches R. A Hertz contact model with non-linear damping for pounding simulation. Earthq Eng Struct Dyn 2006;35(7):81128. [21] Mwafy AM, Elnashai AS, Sigbjrnsson R, Salama A. Significance of severe distant and moderate close earthquakes on design and behavior of tall buildings. Struct Des Tall Special Build 2006;15:391416. [22] Priestley MJN, Grant DN. Viscous damping in seismic design and analysis. J Earthq Eng 2005;9(1):22955 [special issue]. [23] Kwon O-S, Elnashai AS. Multi-platform simulation of highway over-crossing bridge with consideration of soil-structure-interaction. J Struct Eng, ASCE 2008;134(4):65160.

2209

[24] Yang Z, Lu J, Elgamal A. OpenSees geotechnical simulation capabilities and user manual. San Diego: University of California; 2005. [25] Elnashai AS, Mwafy AM, Kwon O-S. Seismic assessment of the I-155 bridge at Caruthersville with SSI. Urbana (IL): Mid-America Earthquake Center, University of Illinois at Urbana-Champaign; 2007. [26] Caltrans. Caltrans seismic design criteria. Sacramento (CA): California Department of Transportation; 2004. [27] Maroney BH. Large scale bridge abutment tests to determine stiffness and ultimate strength under seismic loading. Ph.D. thesis. Davis (CA): University of California Davis; 1995. [28] Mwafy AM, Elnashai AS, Yen W-H. Implications of design assumptions on capacity estimates and demand predictions of multi-span curved bridges. J Bridge Eng, ASCE 2007;12(6):71026. [29] Kwon O-S, Mwafy AM, Elnashai AS. Analytical assessment of seismic performance evaluation procedures for bridges. Performance-based seismic design of bridges. ACI Special Publication. 2010 [in press]. [30] Mwafy AM, Kwon O-S, Elnashai AS, Hashash YMA. Wave passage and ground motion incoherency effects on seismic response of an extended bridge. J Bridge Eng, ASCE (2010) [in press]. [31] Fernandez JA, Rix GJ. Seismic hazard analysispreliminary results. Seismic retrofit study of bridge A-1700 route I-155. Pemiscot County (Atlanta, Georgia): Georgia Institute of Technology; 2006. [32] Hashash Y. Seismic site response analysis. Seismic retrofit study of bridge A1700 route I-155. Pemiscot County (Urbana, IL): Univ. of Illinois at UrbanaChampaign; 2006. [33] Abrahamson NA, Schneider JF, Stepp JC. Empirical spatial coherency functions for application to soil structure interaction analyses. Earthq Spectra 1991;7: 127. [34] Tsai C-C, Hashash YMA. Evaluation of two approaches to simulate spatially variable ground motions. J Earthq Eng 2010;14(2):293308.