4th International Conference on Earthquake EngineeringTaipei, TaiwanOctober 12-13, 2006Paper No. 142

ANALYTICAL SEISMIC ASSESSMENT OF HIGHWAY BRIDGES

WITH SOIL-STRUCTURE INTERACTIONOh-Sung Kwon1 and Amr S. Elnashai2

the Meloland Road Over-crossing (MRO)bridge.

ABSTRACTThe paper describes a new approach and its practical implementation for the analysis of the inelasticdynamic response of bridges with soil-structure interaction using multiple analysis platforms. Thedevelopment is presented through an application test bed; the Meloland Road Over-crossing (MRO)bridge. The structure was heavily instrumented and has been shaken several times by significantearthquakes, rendering it an ideal benchmark to test the accuracy of the new development. Theapproach embankments, abutments, and supporting pile groups are modeled as three-dimensionalfinite element idealizations in OpenSees with realistic soil material models. The bridge structure ismodeled in Zeus-NL, the Mid-America Earthquake Center analysis platform. The mode shape andfundamental properties of the soil-structure system are compared with those determined from systemidentification of the bridge based on measured ground motion. Response history analysis isconducted by distributed computation of geotechnical and the structural model on four separatecomputers, representing four geographically distributed simulation sites. In addition to the earthquakeassessment observations given in the paper, the results and comparisons indicate that the analysis oflarge interacting structure-foundation-soil systems using distributed computation is an extremelypowerful tool that provides accurate results at considerable savings in computing effort. Thepresented framework is also suitable for hybrid-distributed NEES-like simulations.Keywords: Soil-Structure Interaction, Distributed computing, Bridge seismic response.

INTRODUCTIONSoil deposits supporting structure have mainly two effects on the behavior of systems subjected toearthquake loading: (i) local site amplification and (ii) soil-structure interaction (SSI) effects. Toaccount for both effects most accurately, the structure as well as a large soil domain should bemodeled as a whole. This type of full SSI analysis requires a tremendous amount of computational

effort. At the current state of the development of computational environments, there have been only afew attempts to analyze the full soil-structure system, such as the Humboldt Bay Bridge Project(Zhang et al., 2004) used as a test bed in the Pacific Earthquake Engineering Research Center. Thechallenges reported indicate that running many such analyses for the purposes of probabilisticassessment for example is still in the realm of the future. Many practical approaches account for theSSI effect by modeling the stiffness of supporting soil and applying surface ground motions. Thisapproach is approximate in that the inertial interaction between soil mass and structural mass is notaccounted for and also the stiffness of the soil foundation is mostly based on empirical relationships.

DESCRIPTION OF THE MRO BRIDGEThe MRO Bridge was instrumented with 26 accelerometers in 1978 and 6 more instruments in 1992.Twelve instruments placed as a downhole array were also installed to monitor propagation of wavesfrom deep soil layers to surface. The 1979 Imperial Valley Earthquake (ML = 6.6) was the largestrecorded event at the site with a peak ground acceleration of 0.3 g. The recorded ground motion fromthis earthquake has been extensively studied in the 1980s and 1990s (Norris 1986; Werner 1987;Vrontinos et. al., 1993; Zhang and Makris, 2001 among many others). These investigations as well asrecorded ground motions from five earthquake events are used to validate the analytical approachpresented in this paper. The MRO Bridge is located over Interstate 8 approximately 0.5 km from thefault rupture of the 1979 Imperial Valley earthquake. The bridge consists of two spans of pre-stressedbox-girder decks monolithically connected to the center pier. The abutments are placed on fill. Sevenpiles support each abutment. Each side of abutment has 5.9 m of wing-wall. The pier at the center ofthe bridge has a diameter of 1.5 m and is 7.9 m is high from the top of piles. A total of 18 longitudinalreinforcement bars are used in the pier, the foundations of which are supported on 25 timber pilesspaced at 0.91 m. Figure 1 shows the configurations of MRO Bridge with the location ofaccelerometers in the transverse direction. Reference is made to Zhang and Makris (2001) for moreinformation about the MRO Bridge.

Central Pier FoundationIn this study, an inelastic three-dimensional finite elements (FE) model is assembled for the pile group ofthe central pier. Timber piles have diameters of 32 cm at the top and 20 cm at the bottom. The modulus ofelasticity of the piles is 1.24x107 kPa following Maragakis et al. (1994). In the FE model, the piles areassumed to be prismatic with rectangular cross section, to reduce the mesh size. Therefore, modulus ofelasticity is adjusted so that the flexural rigidities are matched. Strength parameters such as friction anglesand cohesion are important for the large strain response of soil while the elastic shear modulus and bulkmodulus have dominant roles in the small strain response. Norris (1986) reported friction angles andcohesion of the studied site. Recently, as part of the ROSRINE project (Anderson, 2003), field andlaboratory tests of the MRO Bridge site were conducted. Although dynamic properties of the soil arereported in the literature, not all the parameters required for the soil material model in OpenSees wereavailable. Hence in this study the unknown material properties are inferred from the cohesion of clay andthe friction angle of sand. In Maragakis et al. (1994), the clay layers are located between 0~2.7 m, 6m~10.7 m, and 15m and below grade. Each clay layer has cohesion of 35.9 kPa, 76.6 kPa, and 86.2 kPa,respectively. Based on these values, shear moduli are chosen as 60,000 kPa, 150,000 kPa, and 150,000kPa, for each clay layer, respectively, following Yang et al. (2005). Poisson’s ratios of the clay layer areassumed to be 0.4. The other intermediate layers are silty sand with friction angle of 32 to 33 degree andrelative density of 45 ~ 52%. All layers with cohesionless soil are in the range of medium sand (Yang etal., 2005).It is expected that the behavior of a pile group is more governed by global soil–pile group interactionrather than soil–pile interaction of individual pile, thus interface elements are not used for this analysis.Cylindrical soil medium with diameter of 48 m is modeled, which is 10.5 times larger than the pile capdimension. The depth of the soil medium is 17 m, Figure 2. All elements in the soil mesh are 8-node brickelements. The top of the pile cap is controlled by a control node, which connects the top nodes of pierwith rigid frame elements.

The material properties of the embankment fill are based on Zhang and Makris (2001) where density of

ρs = 1600 kg/m3 and shear wave velocity of Vs = 110 m/sec are used. These values correspond to shearmodulus of 19.4 MPa, which corresponds to soft clay in Yang et al. (2005). In this study, pressureindependent material (Von Mises type material) is used with the properties shown in Figure 3. All theelements are modeled with 8 node brick elements. The dimension of supporting ground is 124 m by 105m in plane with 18 m of depth. The overall dimension of abutment is similar to the dimensions describedin previous studies. Total 1675 elements are used to model supporting ground, embankment, abutment,and pile groups. A control node is placed at the top of abutment and tied with top nodes on abutment withrigid frame element.Eigen value analysis is conducted for the embankment model. The first transverse and longitudinal modesare shown in Figure 4. The 1st mode is in longitudinal direction with fundamental period of 0.319 sec. The2nd mode is in transverse direction with fundamental period of 0.314 sec. These modal properties will beused in the following section to define lumped mass of embankment for hybrid simulation.

Bridge Deck and PierThe bridge experienced the 1979 Imperial Valley earthquake without noticeable damage. Since thestructure experienced a few earthquakes, it is assumed that there are cracks in the concrete sectionreducing its stiffness. Young’s modulus of 22 GPa for concrete is assumed as in the report by Zhang andMakris (2001). The young’s modulus of cracked section was reported by Douglas and Reid (1982,Ec = 20 ~ 25 GPa) and Dendrou et al. (1985, Ec = 20 GPa), so the selected values is within the range ofprevious studies. The density of concrete is assumed to be 2400 kg/m3. For the deck, an equivalent sectionwith the same area (A) and moment of inertia (Ix, and Iy) as the actual structure is used. The torsionalstiffness is not considered when determining the equivalent section as structural displacement isdominated by flexural deformation of pier and deck.

Therefore an effective embankment mass is lumped at the abutment-bridgeconnection. The effective mass of the embankment is determined from its natural frequency and stiffness.From the initial stiffness evaluation, the transverse stiffness of embankment is 741 MN/m. Assuming thatthe mass is lumped at the bridge-embankment connection, the mass can be calculated from the transverseperiod of embankment as M = kT 2/(4π 2) = 1848 ton.

The damping ratio was identified from recorded ground motion, which has impact type acceleration at theearly stage of earthquake event using the simple approach of logarithmic decrement. The damping ratio atthe fundamental period of the structure is identified to be 4%.

ID Dateyr/mo/dy

ML Lat Long Depth(km)

Epic. Dist. (km)

PGA (g)

Availablerecord1

GM01 79/10/15

6.6 32.614

115.318

12.1 21.5 0.3 B

GM02 99/10/16

7.1 34.594

116.271

6.0 216.0

0.016

D

GM03 00/04/09

4.3 32.692

115.392

10.0 10.4 0.043

B, D

GM04 00/06/14

4.2 32.896

115.502

5.1 14.6 0.015

B, D

GM05 00/06/14

4.5 32.884

115.505

4.9 13.5 0.009

B, D

GM06 02/02/22

N/A N/A N/A N/A N/A 0.039

B, D