Research Progress and Accomplishments: 2000-2001

Accomplishments

A

Selection of

Papers

Chronicling

Technical

Achievements

of the

Multidisciplinary

Center for

Earthquake

Engineering

Research

2000 – 2001

Research Progress and

The Multidisciplinary Center for Earthquake Engineering Research (MCEER) is a national centerof excellence in advanced technology applications that is dedicated to the reduction of earthquakelosses nationwide. Headquartered at the University at Buffalo, State University of New York, the Centerwas originally established by the National Science Foundation (NSF) in 1986, as the National Centerfor Earthquake Engineering Research (NCEER).

Comprising a consortium of researchers from numerous disciplines and institutions throughoutthe United States, the Center’s mission is to reduce earthquake losses through research and the applica-tion of advanced technologies that improve engineering, pre-earthquake planning and post-earthquakerecovery strategies. Toward this end, the Center coordinates a nationwide program of multidisciplinaryteam research, education and outreach activities.

Funded principally by NSF, the State of New York and the Federal Highway Administration (FHWA),the Center derives additional support from the Federal Emergency Management Agency (FEMA), otherstate governments, academic institutions, foreign governments and private industry.

The Multidisciplinary Center for Earthquake Engineering Research

Research Progress and Accomplishments

2000 – 2001

Multidisciplinary Center for Earthquake Engineering ResearchUniversity at Buffalo, State University of New York

May 2001

MCEER-01-SP01Red Jacket Quadrangle, Buffalo, New York 14261

Tel: (716) 645-3391; Fax: (716) 645-3399; Email: [email protected] Wide Web: http://mceer.buffalo.edu

Foreword

by George C. Lee, Director,Multidisciplinary Center for Earthquake Engineering Research

iii

The vision of the Multidisciplinary Center for Earthquake Engineering Research (MCEER) is to helpfoster activities that will lead to more earthquake resilient communities. Its mission is to discover,nurture, develop, promote, help implement, and in some instances pilot test, innovative measures andadvanced and emerging technologies to reduce losses in future earthquakes in a cost-effective man-ner. Research findings show that relatively new buildings and infrastructure that are designed andconstructed to the current state-of-practice in earthquake engineering perform significantly betterthan older ones during earthquakes, and that the largest threat to society lies in the seismically vulner-able infrastructure designed and constructed at a time when earthquake-resistant design had not yetmatured.

With that knowledge in mind, it is MCEER’s view that the best way of achieving the stated vision ofearthquake resilient communities in the short term is to invest in two focused system-integrated en-deavors: the rehabilitation of critical infrastructure facilities that society will need and expect to beoperational following an earthquake, more specifically hospitals and lifelines; and the improvement ofemergency response and crisis management capabilities to ensure efficient response and appropriaterecovery strategies following earthquakes.

MCEER works with the entire earthquake loss reduction community, which consists of practicingengineers and other design professionals, policy makers, regulators and code officials, facility andbuilding owners, governmental entities, and other stakeholders who have responsibility for loss re-duction decision making, to ensure that research results are implemented to improve safety and ad-vance earthquake loss reduction for government, private industry, and the public-at-large.

The flowchart above schematically shows how the Center’s research interests contribute collec-tively to achieve the vision of more earthquake resilient communities. The papers in this volume

-

Earthquake Resilient Communities

Through Applications of Advanced Technologies

Infrastructures that Must be Available /Operational following an Earthquake

Intelligent Responseand Recovery

Hospitals

Water, GasPipelines

Electric PowerNetwork

Bridges andHighways

More

Earthquake

Resilient Urban

Infrastructure

System

Cost-

Effective

Retrofit

Strategies

iv

highlight efforts in intelligent response and recovery, hospitals, water and gas pipelines, electric powernetworks, and bridges and highways. These studies involve many different disciplines, whose workwill cohesively join together toward successful implementation of design and retrofit techniques toprotect urban infrastructures from earthquake damage.

This report is the third in our annual compilation of research progress and accomplishments. It isavailable in both printed and electronic form (on our web site in PDF format at http://mceer.buffalo.edu/publications/default.asp, under Special Publications).

If you would like more information on any of the studies presented herein, or on other MCEERresearch or educational activities, you are encouraged to contact us by telephone at (716) 645-3391,facsimile (716) 645-3399, or email at [email protected].

Earthquake Motion Input and Its Dissemination Via the Internet 1Apostolos S. Papageorgiou (Principal Author), Benedikt Halldorsson and Gang Dong

Overcoming Obstacles to Implementation: Addressing Political, Institutional and 9Behavioral Problems in Earthquake Hazard Mitigation Policies

Daniel J. Alesch and William J. Petak

Large Scale Experiments of Permanent Ground Deformation Effects on Steel Pipelines 21Koji Yoshizaki, Thomas D. O’Rourke, Timothy Bond, James Mason and Masanori Hamada

Experimental and Analytical Study of Base-Isolation for Electric Power Equipments 29M. Ala Saadeghvaziri and Maria Q. Feng

Recommended Changes to the AASHTO Specifications for the Seismic Design of 41Highway Bridges (NCHRP Project 12-49)

Ian M. Friedland, Ronald L. Mayes and Michel Bruneau

Literature Review of the Observed Performance of Seismically Isolated Bridges 51George C. Lee, Yasuo Kitane and Ian G. Buckle

A Risk-Based Methodology for Assessing the Seismic Performance of Highway Systems 63Stuart D. Werner

Analysis, Testing and Initial Recommendations on Collapse Limit States of Frames 73Darren Vian, Mettupalayam Sivaselvan, Michel Bruneau and Andrei M. Reinhorn

Centrifuge-Based Evaluation of Pile Foundation Response to Lateral Spreading 87and Mitigation Strategies

Ricardo Dobry, Tarek H. Abdoun and Thomas D. O’Rourke

Analysis and Design of Buildings with Added Energy Dissipation Systems 103Michael C. Constantinou, Gary F. Dargush, George C. Lee (Coordinating Author), Andrei M.Reinhorn and Andrew S. Whittaker

Using Cost-Benefit Analysis to Evaluate Mitigation for Lifeline Systems 125Howard Kunreuther (Coordinating Author), Chris Cyr, Patricia Grossi and Wendy Tao

Retrofit Strategies For Hospitals in the Eastern United States 141George C. Lee, Mai Tong and Yasuhide Okuyama

Passive Site Remediation for Mitigation of Liquefaction Risk 149Patricia M. Gallagher and James K. Mitchell

Contents

v

Advanced GIS for Loss Estimation and Rapid Post-Earthquake Assessment of 157Building Damage

Thomas D. O’Rourke, Sang-Soo Jeon, Ronald T. Eguchi and Charles K. Huyck

Seismic Evaluation and Retrofit of the Ataturk International Airport 165Terminal Building

Michael C. Constantinou, Andrew S. Whittaker and Emmanuel Velivasakis

Estimating Earthquake Losses for the Greater New York City Area 173Andrea S. Dargush (Coordinating Author), Michael Augustyniak, George Deodatis, Klaus H.Jacob, George Mylonakis, Laura McGinty, Guy J.P. Nordenson, Daniel O’Brien, Scott Stanford,Bruce Swiren, Michael W. Tantala and Sam Wear

Contributors

vii

Tarek H. Abdoun, Research Assistant Professor, Civil and Environmental EngineeringDepartment, Rensselaer Polytechnic Institute

Daniel J. Alesch, Professor, Department of Public and Environmental Affairs, University ofWisconsin – Green Bay

Michael Augustyniak, Supervising Planner, New Jersey State Police Office of EmergencyManagement

Timothy Bond, Laboratory Manager, Department of Civil and Environmental Engineering, CornellUniversity

Michel Bruneau, Deputy Director, Multidisciplinary Center for Earthquake Engineering Researchand Professor, Department of Civil, Structural and Environmental Engineering, University atBuffalo

Ian G. Buckle, Professor, Department of Civil Engineering, University of Nevada, Reno

Sungbin Cho, Research Associate, School of Policy, Planning and Development, University ofSouthern California

Michael C. Constantinou, Professor and Chair, Department of Civil, Structural andEnvironmental Engineering, University at Buffalo

Chris Cyr, Undergraduate Student, Wharton Risk Management and Decision Processes Center,The Wharton School, University of Pennsylvania

Andrea S. Dargush, Associate Director for Education and Research Administration,Multidisciplinary Center for Earthquake Engineering Research

Gary F. Dargush, Associate Professor, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

George Deodatis, Professor, Civil Engineering & Operations Research Department, PrincetonUniversity

Ricardo Dobry, Professor, Civil and Environmental Engineering Department, RensselaerPolytechnic Institute

Gang Dong, Post-Doctoral Research Associate, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Ronald T. Eguchi, President, ImageCat, Inc.

Selahattin Ersoy, Graduate Student, Department of Civil and Environmental Engineering, NewJersey Institute of Technology

Maria Q. Feng, Associate Professor, Civil and Environmental Engineering Department, Universityof California, Irvine

Ian. M. Friedland, Associate Director for Development, Applied Technology Council

Patricia M. Gallagher, Assistant Professor, Department of Civil and Architectural Engineering,Drexel University

Siang-Huat Goh, Research Assistant, Department of Civil and Environmental Engineering,Cornell University

viii

Juan Gomez, Graduate Student, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Patricia Grossi, Post-Doctoral Researcher, Wharton Risk Management and Decision ProcessesCenter, The Wharton School, University of Pennsylvania

Benedikt Halldorsson, Graduate Student, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Masanori Hamada, Professor, Department of Civil Engineering, Waseda University, Japan

Wilhelm Hammel, Senior Consultant, KPMG, Germany

Charles K. Huyck, Vice President, ImageCat, Inc.

Klaus H. Jacob, Senior Reserch Scientist, Seismology Department, Geology and Tectonics,Lamont-Doherty Earth Observatory

Sang-Soo Jeon, Graduate Research Assistant, Department of Civil and Environmental Engineering,Cornell University

Yasuo Kitane, Graduate Student, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Howard Kunreuther, Co-director, Wharton Risk Management and Decision Processes Center,The Wharton School, University of Pennsylvania

George C. Lee, Director, Multidisciplinary Center for Earthquake Engineering Research andSamuel P. Capen Professor of Engineering, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Li Lin, Associate Professor, Department of Industrial Engineering, University at Buffalo

Wei Liu, Ph.D. Candidate, Department of Civil, Structural and Environmental Engineering,University at Buffalo

James Mason, Graduate Research Assistant, Department of Civil and Environmental Engineering,Cornell University

Ronald L. Mayes, Independent Consultant, Moraga, California

Laura McGinty, GIS Specialist, Westchester County Geographic Information Systems, Departmentof Information Technology

James K. Mitchell, University Distinguished Professor Emeritus, Department of Civil Engineering,Virginia Polytechnic Institute and State University

James E. Moore, III, Associate Professor, Departments of Civil Engineering and Public Policyand Management, University of Southern California

Nobuo Murota, Bridgestone Company, Japan

George Mylonakis, Assistant Professor, Department of Civil Engineering, City University of NewYork

Guy J.P. Nordenson, Associate Professor, Civil Engineering & Operations Research Department,Princeton University

Daniel O’Brien, Earthquake Program Manager, New York State Emergency Management Office

Yasuhide Okuyama, Assistant Professor, Department of Architecture and Planning, University atBuffalo

Thomas D. O’Rourke, Thomas R. Briggs Professor of Engineering, Department of Civil andEnvironmental Engineering, Cornell University

Apostolos Papageorgiou, Professor, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

William J. Petak, Professor, School of Policy, Planning and Development, University of SouthernCalifornia

Rajesh Radhakrishnan, Graduate Student, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Oscar Ramirez, Graduate Student, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Ricardo Ramos, Graduate Student, Civil and Environmental Engineering Department, RensselaerPolytechnic Institute

Andrei Reinhorn, Professor, Department of Civil, Structural and Environmental Engineering,University at Buffalo

M. Ala Saadeghvaziri, Associate Professor, Department of Civil and Environmental Engineering,New Jersey Institute of Technology

Ramesh Sant, Ph.D. Candidate, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Masanobu Shinozuka, Fred Champion Chair in Civil Engineering, Department of CivilEngineering, University of Southern California

Ani N. Sigaher, Ph.D. Candidate, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Mettupalayam Sivaselvan, Ph.D. Candidate, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Scott Stanford, Geologist, New Jersey Geological Survey

Ernest Sternberg, Associate Professor, Department of Architecture and Planning, University atBuffalo

Bruce Swiren, Natural Hazards Program Specialist, Federal Emergency Management Agency,Region II

Michael W. Tantala, Ph.D. Candidate, Civil Engineering & Operations Research Department,Princeton University

Wendy Tao, Undergraduate Student, Wharton Risk Management and Decision Processes Center,The Wharton School, University of Pennsylvania

Craig E. Taylor, President, Natural Hazards Management, Inc.

Mai Tong, Senior Research Scientist, Multidisciplinary Center for Earthquake EngineeringResearch and the Department of Civil, Structural and Environmental Engineering, University atBuffalo

Panos Tsopelas, Assistant Professor, Catholic University of America

Emmanuel Velivasakis, Senior Vice President, LZA Technology

Darren Vian, Ph.D. Candidate, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Jon S. Walton, Vice President, MacFarland Walton Enterprises

Yinjuang Wang, Graduate Student, Civil and Environmental Engineering Department, RensselaerPolytechnic Institute

Sam Wear, GIS Manager, Westchester County Geographic Information Systems, Department ofInformation Technology

Stuart D. Werner, Principal, Seismic Systems and Engineering Consultants

ix

Andrew S. Whittaker, Associate Professor, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Katrin Winkelmann, Visiting Research Associate, Department of Civil, Structural andEnvironmental Engineering, University at Buffalo, from the University of Darmstadt

Eric Wolff, Graduate Research Assistant, Department of Civil, Structural and EnvironmentalEngineering, University at Buffalo

Yihui Wu, Research Scientist, Multidisciplinary Center for Earthquake Engineering Research andthe Department of Civil, Structural and Environmental Engineering, University at Buffalo

Tony Yang, Undergraduate Student, Department of Civil, Structural and Environmental Engineering,University at Buffalo

Koji Yoshizaki, Pipeline Engineer, Fundamental Technology Laboratory, Tokyo Gas Company, Ltd.,Japan

x

1

Research Objectives

The objectives of this task are to conduct research on seismic hazards,and provide relevant input on the expected levels of these hazards toother tasks. Other tasks requiring this input include those dealing withinventory, fragility curves, rehabilitation strategies and demonstrationprojects. The corresponding input is provided in various formats depend-ing on the intended use: as peak ground motion parameters and/or re-sponse spectral values for a given magnitude, epicentral distance and siteconditions; or as time histories for scenario earthquakes that are selectedbased on the disaggregated seismic hazard mapped by the U.S. GeologicalSurvey and used in the NEHRP Recommended Provisions for the Devel-opment of Seismic Regulations for New Buildings (BSSC, 1998).

Earthquake Motion Input and ItsDissemination Via the Internet

by Apostolos S. Papageorgiou(Principal Author), Benedikt Halldorsson and Gang Dong

National Science Foundation,Earthquake EngineeringResearch Centers Program

Prediction of the seismic hazard in tectonic regions of moderate-to-low seismicity is a difficult task because of the paucity of data that

are available regarding such regions. Specifically, the problem of earth-quake ground motion prediction in Eastern North America (ENA) is hin-dered by two factors: (1) the causative structures of seismicity in ENA arelargely unknown, and (2) the recorded strong motion database is verylimited (if at all existent), especially for moderate to large magnitude (M

W

≥ 6) events virtually for all epicentral distances. For these reasons, predic-tion of strong ground motion in ENA makes the use of well-founded physi-cal models imperative. These models should, among other things, providethe means to make extrapolations to large magnitudes and/or short dis-tances with confidence. (This is contrast to the western U.S. (WUS), wherethe abundance of recorded strong motion data makes prediction of groundmotion by empirical methods a viable procedure.)

Strong Motion Synthesis TechniquesOur task is the synthesis of strong ground motion input over the entire

frequency range of engineering interest. There are two approaches formodeling earthquake strong motion:

Apostolos S. Papageorgiou,Professor, Gang Dong,Post-Doctoral ResearchAssociate, and BenediktHalldorsson, GraduateStudent, Department ofCivil, Structural andEnvironmentalEngineering, University atBuffalo

2

(1) The Stochastic (Engineering)Approach, according to which,earthquake motion (acceleration) ismodeled as Gaussian noise with aspectrum that is either empirical,or a spectrum that is based on aphysical model (such as the SpecificBarrier Model) of the earthquakesource. This approach is expedientand therefore cost-effective, and hasbeen extensively used in the pastby engineers (using empirical spec-tra) and recently by seismologists(using spectra derived from physi-cal models of the source). The in-tent of this approach to strongmotion simulation is to capture theessential characteristics of high-fre-quency motion at an average sitefrom an average earthquake ofspecified size. Phrasing this differ-ently, the accelerograms artificiallygenerated using the EngineeringApproach do not represent any spe-cific earthquake but embody cer-tain average properties of pastearthquakes of a given magnitude.

(2) The Kinematic Modeling Ap-proach was developed by seismolo-gists. In this approach, the ruptureprocess is modeled by postulatinga slip function on a fault plane andthen using the Elastodynamic Rep-resentation Theorem to computethe motion (e.g., Aki and Richards,1980). There are several variant

forms of this approach dependingon whether the slip function (i.e.,the function that describes the evo-lution of slip on the fault plane)and/or the Green functions are syn-thetic or empirical. The KinematicModeling Approach involves theprediction of motions from a faultthat has specific dimensions andorientation in a specified geologicsetting. As such, this approach moreaccurately reflects the various wavepropagation phenomena and is use-ful for site-specific simulations.

Stochastic (Engineering)Approach

Recognizing that the StochasticApproach for synthesizing earth-quake strong motion time historiesis the most expedient method,ground motion synthesis effortswere initiated with this approach.The following computer codeshave been developed, and are avail-able to any interested user via theInternet:1. SGMP_ENA - Stochastic Ground

Motion Prediction for EasternNorth America: Random Vibra-tion Theory (RVT): is used to es-timate the mean/expectedvalues of peak ground motionparameters (i.e., peak accelera-tion, peak velocity and peak

The user community for this research is both academic re-searchers and practicing engineers who may use the seismicinput generated by the synthesis techniques that are devel-oped under this task for a variety of applications. These in-clude ground motions for scenario earthquakes, fordeveloping fragility curves and in specifying ground motioninput for critical facilities (such as hospitals) located in theeastern U.S.

Program 1: Seismic Evaluationand Retrofit of LifelineSystems

Program 2: Seismic Retrofit ofHospitals

3Earthquake Motion Input and Its Dissemination Via the Internet

displacement), and spectral re-sponse amplitudes (e.g., Rice,1944, 1945; Cartwright andLonguet-Higgins, 1956; Shino-zuka and Yang, 1971; Soong andGrigoriu, 1993).

2. SGMS_ENA - Stochastic GroundMotion Simulation for EasternNorth America: Syntheticground motions are generatedusing the Stochastic (Engineer-ing) Approach briefly describedabove (e.g., Shinozuka and Jan,1972; Shinozuka and Deodatis,1991; Boore, 1983; Grigoriu,1995).

3. RSCTHS – Response SpectrumCompatible Time Histories: Syn-thesis of ground motion time his-tories that are compatible withprescribed response spectra us-ing the Spectral RepresentationMethod (Deodatis, 1996).

Various earthquake source mod-els that have been proposed in thepublished literature [such as theSpecific Barrier Model (Papa-georgiou and Aki, 1983a,b; 1985;1988) and the ω2-model (Brune,1970; Frankel et al., 1996)], havebeen implemented (and are pro-vided as options) in computercodes SGMP_ENA and SGMS_ENA.[Both computer codes predict/syn-thesize ground motions that arecompatible with the site classifica-

tions of the 1997 NEHRP Provisions(BSSC, 1998).]

Of all the source models, the Spe-cific Barrier Model is favored be-cause it provides the mostcomplete, yet parsimonious, self-consistent description of the fault-ing processes that are responsiblefor the generation of the high fre-quencies, and at the same time pro-vides a clear and unambiguous wayof how to distribute the seismicmoment on the fault plane. The lat-ter requirement is necessary for theimplementation of the KinematicModeling Approach describedabove. We calibrated the SpecificBarrier Model using the availablestrong motion data base of ENAearthquakes, as well as inferredsource spectra for the 25 Novem-ber 1988 Saguenay earthquake (M

w

5.8) and 19 October 1990 MontLaurier earthquake (M

w 4.5)

Quebéc, Canada. The estimates ofthe global and local stress dropsthat resulted from the calibrationare believed to be representative ofearthquake sources in ENA. Thus, a"scaling law" of the source spectraof ENA earthquakes has been estab-lished. Such a “scaling law” allowsthe prediction/synthesis of groundmotion at magnitude-distanceranges that currently are not cov-

Engineering SeismologyLaboratory:

http://civil.eng.buffalo.edu/engseislab/

USGS National SeismicHazards Project of theEarthquake HazardsProgram:

http://geohazards.cr.usgs.gov/eq/

Period (s) 0.1 0.2 0.33 0.5 1.0 2.0Mw

R (km)(factor)

4.812

0.712

4.812

1.253

5.733

1.458

5.733

1.692

6.234

1.170

6.234

1.999

� Table 1. Scenario Earthquake Events for Six Oscillator Periods for New York, NY

Notes:Mw = Moment MagnitudeR (km) = Epicentral Distance(factor) = scaling factor used to adjust/scale the synthetic ground motions of the modal event so that

they are compatible with the Uniform Hazard Spectrum

4

ered by the existing strong motiondatabase of ENA.

In order to make it more conve-nient for the practicing engineer/designer, the ground motion syn-thesis capabilities described above

are linked with estimates of theseismic hazard produced by the USGeological Survey (USGS), the Na-tional Seismic Hazards Project ofthe Earthquake Hazards Program(http://geohazards.cr.usgs.gov/eq). Specifically, the disaggregatedseismic hazard (Frankel, 1995;Frankel et al., 1996; Harmsen et al.,1999; Harmsen and Frankel, 2001)for five cities in the eastern U.S.(Boston, MA; Buffalo, NY; Charles-ton, SC; Memphis, TN; and NewYork, NY) was obtained from theabove USGS website. For each city/site, the disaggregated seismic haz-ard is expressed in the form of a“modal event” for each one of sixselected oscillator periods (0.1, 0.2,0.33, 0.5, 1.0 and 2.0 seconds)[“modal event” is the seismic eventthat makes the most significantcontribution to the seismic hazardat a site for a specified oscillatorperiod; each “modal event” is char-acterized/specified by a momentmagnitude M

w and a distance R].

The spectral amplitude of the“modal event” at the correspond-ing oscillator period usually doesnot agree with the spectral ampli-tude of the Uniform Hazard Spec-tra that are produced by theProbabilistic Seismic HazardAnalysis (PSHA) and are incorpo-rated in building codes. [Usually, foreastern North America, the ampli-tudes of the modal events arehigher than the Uniform HazardSpectra, while in the western U.S.the reverse is true]. Current prac-tice recommends (Shome et al.,1998) to adjust/scale the entirespectrum of the “modal event” sothat its spectral amplitude matchesthat of the Uniform Hazard Spec-trum at the oscillator period towhich it corresponds.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

200

400

600

800

1000

Period (s)

Spe

ctra

l acc

eler

atio

n (c

m/s

2 )

Scaled Pseudo Spectral Accelerations for New York, NY (site B−C, SBM)

NEHRP (B−C) design spectraNEHRP (B−C) MCE spectra Uniform Hazard SA M

w=4.80, R=12.0 km

Mw=4.80, R=12.0 km

Mw=5.70, R=33.0 km

Mw=5.70, R=33.0 km

Mw=6.20, R=34.0 km

Mw=6.20, R=34.0 km

0 2 4 6 8 10 12−500

0

500

New York, NY. M*=6.2, R*=34.0 km, Ty*= 1.00 sec., Scale=1.170, Soil=BC (SBM)

Acc

. (cm

/s2 )

0 2 4 6 8 10 12−500

0

500

Acc

. (cm

/s2 )

0 2 4 6 8 10 12−500

0

500

Acc

. (cm

/s2 )

Time (s)

� Figure 1. This figure displays: (1) The Pseudo-Spectral Acceleration(PSA) spectra of the scenario events tabulated in Table 1; (2) theUniform Hazard Spectrum, which is obtained from the PSHA; (3) theMaximum Considered Earthquake (MCE) spectrum as specified bythe 1997 NEHRP Provisions on the basis of the "Uniform HazardSpectrum"; (4) the Design Basis Earthquake spectrum, which isobtained by multiplying the amplitudes of the MCE spectrum by thefactor (2/3).

� Figure 2. Scaled synthetic ground accelerations of the scenario eventscorresponding to oscillator periods 1.0 and 2.0 seconds. The soilconditions correspond to the boundary B-C of site classes of the 1997NEHRP Provisions (BSSC, 1998).


For each one of the selected fiveeastern North America cities, fol-lowing the procedure briefly de-scribed above, we generated andposted on the web (http://civil.eng.buffalo.edu/engseislab/)a suite of ground motion realiza-tions (scenario earthquake mo-tions). [The site class used for thesesimulations is the B-C NEHRP siteclass (BSSC, 1998), which is the ref-erence site condition for the USGSseismic hazard maps.] Further-more, the ground motion synthe-sis computer codes discussedabove are accessible and may beused to generate additional timehistories, possibly for different soilconditions, if the user deems it nec-essary.

As an example, consider NewYork, NY. Table 1 shows the six“scenario earthquake events” thathave been selected (based on thedisaggregated seismic hazard pro-vided by USGS) for New York, cor-responding to the six oscillatorperiods that were mentionedabove and essentially cover the fre-quency range of interest for mostpractical applications.

Figure 1 displays the following:(1) the Pseudo-Spectral Accelera-tion (PSA) spectra of theabovementioned scenario events(the Specific Barrier Model wasused as the source model); (2) theUniform Hazard Spectrum, whichis obtained from the PSHA; (3) theMaximum Considered Earth-quake (MCE) spectrum as speci-fied by the 1997 NEHRP Provisions(BSSC, 1998) on the basis of theUniform Hazard Spectrum; (4) theDesign Basis Earthquake spec-trum, which is obtained by multi-plying the amplitudes of the MCEspectrum by the factor (2/3). Notethat a scenario event for a specific

oscillator period has the same PSAvalue as the uniform hazard spec-trum at that period.

The scaled synthetic ground ac-celerations (two horizontal andone vertical components) of thescenario events corresponding tooscillator periods 1.0 and 2.0 sec-onds are displayed in Figure 2. Thesoil conditions correspond to theboundary B-C of site classes of the1997 NEHRP Provisions (BSSC,1998).

The consistency between thepredictions of RVT with the syn-thetic ground motions are demon-strated, as seen in Figure 3. Theagreement, as anticipated, is verysatisfactory.

Development of KinematicModeling Approach

Having completed the work re-lated to the implementation of theStochastic (Engineering) Ap-proach of strong motion synthesis,efforts are now focused on the de-velopment of the Kinematic Mod-eling Approach. As discussed in

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

500

1000

1500

2000

2500

3000

Period (s)

Spe

ctra

l Acc

eler

atio

n (c

m/s

2 )

New York, NY − RVT vs. realizations (SBM)

Uniform Hazard Spectrum NEHRP 1997 MCE spectrum M=6.2, R=34 km (realizations)M=6.2, R=34 km (RV theory)

� Figure 3. The PSA spectra of ten realizations of a scenario event arecompared with the predictions of Random Vibration Theory

6

previous research progress reports(Papageorgiou, 2000), two impor-tant elements are necessary for thismethod of ground motion synthe-sis: (1) Sub-event models which areused to synthesize composite earth-quake events; (2) Scattering effectsof the lithosphere which are veryimportant for stable tectonic envi-ronments such as that of EasternNorth America. We have made sub-stantial progress regarding the firstelement. Specifically, we have de-rived closed form solutions for theseismic radiation of a new class ofkinematic models (asymmetricalcircular and elliptical crack models)that will be used to simulate sub-events in the synthesis of strongground motion generated by largeearthquake events. These resultshave been recently presented in aforum consisting primarily of seis-mologists (“International Work-shop on the QuantitativePrediction of Strong-Motion andthe Physics of EarthquakeSources” held in Tsukuba, Japan,Oct. 23 - 25, 2000) and were verywell received. These results havebeen presented in papers recentlysubmitted to the Bulletin of theSeismological Society.

Conclusions – FutureResearch

The efforts regarding the devel-opment and implementation of theKinematic Modeling Approach tostrong motion synthesis focus onthe following aspects of themethod:• Scattering effects of the lithos-

phere: Scattering effects are avery important consideration inthe synthesis of ground motionin ENA. We have initiated thecollection of seismograms of

small events that have been re-corded in ENA (US and Canada)by various networks (includingbroadband, short period andstrong motion networks) withthe intention of performinganalysis that will provide valuesof important parameters (e.g.,quality factor, Q, and scatteringcoefficient g) of the scatteringcharacteristics of the lithospherein ENA. The objective is to syn-thesize results developed in thefield of Stochastic Seismologywith techniques for the genera-tion of evolutionary stochasticprocesses developed in the fieldof Probability Theory [such asthe Spectral Representation(Shinozuka and Deodatis, 1991,see also Applied Non-GaussianProcesses (Grigoriu, 1995)].

• Variable size sub-events: Wehave collected all the necessaryliterature related to Poisson Ran-dom Pulse Processes and arewell positioned to investigatethe consequences of using vari-able size sub-events on syntheticmotions.

• Sub-event Models: We have madesubstantial progress in develop-ing closed-form mathematicalexpressions of the far-field radia-tion of new kinematic models torepresent the sub-events. Wewant to extent this line of re-search to other models of sub-events such as asperity models.We thus aim at creating a “li-brary” of models of sub-eventsadequate to simulate the variousmodes of rupture of real faults.

• Validation: Any simulationmethod should be validated bycomparing the synthetic seismo-grams against the recorded onesfor as many earthquake eventsas possible. We have initiated

“The objective isto proposesimpleanalyticalfunctions todescribe nearsource pulses.Thesedescriptionswill be readilyuseful toengineers andmay beincluded infuture versionsof the buildingcode.”


such validation comparisons us-ing the 1988 Saguenay earth-quake event as a case study.

• We intend to address also theissue of near source ground mo-tions. In particular, there are twocompeting physical effects thatwould affect near-source groundmotions in ENA: Earthquakesources in ENA are characterizedby higher stress drops andshorter rise times as comparedto corresponding motions inCalifornia. Thus ENA near-source“killer pulses” are expected tobe stronger and of higher fre-quency (i.e., shorter duration).On the other hand, ENA earth-quake sources appear to occurat greater depths and thus thesource-to-station distance (andconsequently geometric attenu-ation) is greater. It is of greatpractical significance to investi-gate which one of the above twoeffects dominates. The ultimate

objective of such an investiga-tion is to propose simple analyti-cal functions to describe nearsource pulses. Such simple de-scriptions will be readily usefulto engineers and may be in-cluded in future versions of thebuilding code.

• Finally, in response to the needsof various MCEER investigators,we intend to provide groundmotion synthesis capabilities forsites in the western U.S. (e.g.,California). For this, we need tocalibrate the Specific BarrierModel using the extensivestrong motion database that ex-ists for the above tectonic re-gion. For this calibration, we planto exploit some recent develop-ments regarding the improvedmathematical description of thegeometric attenuation of groundmotion in the vicinity of an ex-tended earthquake source.

References

Aki, K. and Richards, P., (1980), Quantitative Seismology: Theory and Methods,Volume I, II, W. H. Freeman and Company, San Francisco, 948 pp.

Boore, D. M., (1983), “Stochastic Simulation of High-Frequency Ground Motions Basedon Seismological Models of the Radiated Spectra,” Bull. Seism. Soc. Am., Vol. 73, pp.1865-1894.

Brune, J. N., (1970), “Tectonic Stress and the Spectra of Seismic Shear Waves fromEarthquakes,” J. Geophys. Res., Vol. 75, pp. 4997–5009.

Building Seismic Safety Council (BSSC) (1998), 1997 NEHRP RecommendedProvisions for the Development of Seismic Regulation for New Buildings, FederalEmergency Management Agency, Washington, D.C., FEMA 302.

Cartwright, D. E. and Longuet-Higgins, M. S., (1956), “The Statistical Distribution of theMaxima of a Random Function,” Proceedings of the Royal Statistical Society, Vol.237, pp. 212–232.

Deodatis, G., (1996), “Non-stationary Stochastic Vector Processes: Seismic GroundMotion Applications,” Prob. Eng. Mech., Vol. 11, pp. 149–167.

Frankel, A., (1995), “Mapping Seismic Hazard in the Central and Eastern United States,”Seism. Res. Lett., Vol. 66, pp. 8–21.

8

References (Con’t)

Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyendecker, E.V., Dickman, N., Hanson,S. and Hopper, M., (1996), National Seismic-Hazard Maps: Documentation, U.S.Geological Survey, Open-File Report 96-532, 110 pp.

Grigoriu, M., (1995), Applied Non-Gaussian Processes, Prentice Hall, Englewood Cliffs,NJ, 442 pp.

Harmsen, S. and Frankel, A., (2001), “Geographic Deaggregation of Seismic Hazard inthe United States,” Bull. Seism. Soc. Am., Vol. 91, pp. 13–26.

Harmsen, S., Perkins, D. and Frankel, A., (1999), “Deaggregation of Probabilistic GroundMotions in the Central and Eastern United States,” Bull. Seism. Soc. Am., Vol. 89, pp.1–13.

Papageorgiou, A. S., (2000), “Ground Motion Prediction Methodologies for Eastern NorthAmerica,” Research Progress and Accomplishments: 1999-2000, MultidisciplinaryCenter for Earthquake Engineering Research, University at Buffalo, pp. 63-69.

Papageorgiou, A. S., (1988), “On Two Characteristic Frequencies of Acceleration Spectra:Patch Corner Frequency and fmax,” Bull. Seism. Soc. Am., Vol. 78, pp. 509–529.

Papageorgiou, A. S. and Aki, K., (1983a), “A Specific Barrier Model for the QuantitativeDescription of Inhomogeneous Faulting and the Prediction of Strong Ground Motion.I. Description of the Model,” Bull. Seism. Soc. Am., Vol. 73, pp. 693–722.

Papageorgiou, A. S. and Aki, K., (1983b), “A Specific Barrier Model for the QuantitativeDescription of Inhomogeneous Faulting and the Prediction of Strong Ground Motion.Part II. Applications of the Model,” Bull. Seism. Soc. Am., Vol. 73, pp. 953–978.

Papageorgiou, A. S. and Aki, K., (1985), “Scaling Law of Far-field Spectra Based onObserved Parameters of the Specific Barrier Model,” Pure Appl. Geophys., Vol. 123,pp. 354–374.

Rice, S. O., (1944), “Mathematical Analysis of Random Noise,” Bell Syst. Tech. J., Vol. 23,pp. 282–332.

Rice, S. O., (1945), “Mathematical Analysis of Random Noise (concluded),” Bell Syst. Tech.J., Vol. 24, pp. 46–156.

Shinozuka, M. and Deodatis, G., (1991), “Simulation of Stochastic Processes by SpectralRepresentation,” Appl. Mech. Rev., ASME, Vol. 44, pp. 191–204.

Shinozuka, M. and Jan, C., (1972), “Digital Simulation of Random Processes and ItsApplications,” J. Sound. Vib., Vol. 25, pp. 111–128.

Shinozuka, M. and Yang, J.-N., (1971), “Peak Structural Response to Non-stationaryRandom Excitations,” J. Sound Vib., Vol. 16, pp. 505–517.

Shome, N., Cornell, A., Bazzurro, P. and Carballo, J. E., (1998), “Earthquakes, Records andNonlinear Response,” Earthquake Spectra, Vol. 14, pp. 469–500.

Soong, T. T. and Grigoriu, M., (1993), Random Vibration of Mechanical and StructuralSystems, Prentice-Hall, Englewood Cliffs, NJ, 402 pp.

9

Research Objectives

This project is aimed at bridging the three planes, from basic research,through enabling processes, to engineered systems. At the basic re-search plane, we have been working to improve our collective under-standing about obstacles to implementing mitigation practices, ownerdecision processes (in connection with other MCEER projects), andpublic policy processes. At the level of enabling processes, we havebeen seeking to develop an understanding of how obstacles to greatermitigation can be overcome by improved policy design and processes.At the engineered systems plane, our work is intended to result inpractical guidelines for devising policies and programs with appropri-ate motivation and incentives for implementing policies and programsonce adopted.

This phase of the research has been aimed, first, at a thorough,multidisciplinary review of the literature concerning obstacles to imple-mentation. Second, the research has focused on advancing the state ofthe art by developing means for integrating the insights offered bydiverse perspectives on the implementation process from the severalsocial, behavioral, and decision sciences. The research establishes a basisfor testing our understanding of these processes in the case of hospi-tal retrofit decisions.

Overcoming Obstacles to Implementation:Addressing Political, Institutional and BehavioralProblems in Earthquake Hazard Mitigation Policies

by Daniel J. Alesch and William J. Petak


As development continues to concentrate in high risk earthquake areas,the probability increases that disastrous earthquakes will occur. Pub-

lic officials face the prospect of dealing with earthquake crises thatcould have been reduced significantly with the application of knowntechnologies. Although earthquakes are an uncontrollable force of na-ture, unnecessary losses in life and property and social disruptions aregenerally the result of not having implemented precautions that weknow could have mitigated the losses.

The primary research emphasis in earthquake hazard mitigation hasbeen on developing increased knowledge about the earthquake phe-nomena, increasing understanding of structural performance underearthquake conditions, developing advanced design methods and stan-dards, and improving building codes. Improved knowledge about the

William J. Petak, Professor,School of Policy, Planning,and Development,University of SouthernCalifornia

Daniel J. Alesch, Professor,Department of Public andEnvironmental Affairs,University of Wisconsin-Green Bay

10

physical and technical aspects ofthe problem has led to the adop-tion of public policies intended toreduce the probability of loss, butthey must be implemented appro-priately to actually reduce risk. To-day, there remains an inadequateunderstanding of the barriers, im-pediments, disincentives, and issuesassociated with implementing ap-propriate earthquake hazard miti-gation technologies and strategies,much less with overcoming thosebarriers. This research was under-taken because no complete con-ceptual model or empiricallyvalidated model exists to explainmitigation adoption processes forthe rehabilitation of existing struc-tures, either across or within spe-cific stakeholder groups.

BackgroundOur broadly-based assessment of

barriers to policy implementationresulted in a background workingpaper entitled Barriers to Success-ful Implementation of EarthquakeHazard Mitigation Policies. In it,we address three questions that arecentral to implementation:• First, what constitutes appropri-

ate, successful implementationof public policies concerning

earthquake hazard mitigation?That is, how can we determinewhether a policy has beenimplemented appropriately orsuccessfully?

• Second, what are the key vari-ables thought to affect the suc-cess of implementation? Whichof those variables can be con-trolled to help ensure success-ful implementation?

• Third, how can mitigation advo-cates help to ensure that publicpolicies adopted in an attemptto reduce the probability oflosses to life and property fromearthquakes are implementedsuccessfully? How can we helpensure that the resources de-voted to hazard reduction inpursuit of these policies are usedeffectively?

Our first paper is based on a re-view of the implementation litera-ture that has developed over thepast three decades. A rich body ofresearch on implementation existswithin the social and behavioralsciences, but very little of it ad-dresses natural hazards risk reduc-tion. Consequently, we’ve drawn ona broad base of literature to drawinferences about implementingearthquake hazard mitigation. The75 page draft report cuts across the

• The International Institutefor Applied SystemsAnalysis, Vienna, Austria,provided Professor Petakwith extended use of itsresearch links and materialsduring his recent sabbaticleave so that he was able toinitiate and develop theliterature review.

• Representatives from theSan Francisco Departmentof Building and Safety,structural engineers(Degenkolb Engineering,Comartin and Reis, DanielShapiro), the AppliedTechnology Council,California Seismic SafetyCommission, and urbanplanning consultantsreviewed the draftdocument.

• Social Science scholarsfrom MAE, PEER, andMCEER reviewed early draftdocuments.

The primary audience for our reports comprises twogroups. One group includes architects, engineers, develop-ers, builders, building owners, and organizations whoseactions or inactions are the targets of policies intended toget them to reduce risks to themselves and other membersof the community from natural hazard events. The secondgroup consists of policy and program designers and publicand private program managers with responsibilities inearthquake hazard risk reduction and mitigation.

11Overcoming Obstacles to Implementation

social and behavioral sciences,drawing on more than three dozensignificant studies, and resulting in37 basic propositions concerningimpediments to effective imple-mentation. The propositions areorganized around a model we de-veloped that attempts to identifykey inter-institutional nexuses inthe implementation process.

The ImplementationWeb

We view organizations as opensystems, comprising elements re-lated to one another in ways suchthat perturbations of one elementhave ramifications for the others.Organizations exist within environ-ments of varying complexity withwhich they inexorably interact.Organizational systems, we believe,are inherently unstable, requiringcontinual resources from the envi-ronment to survive and requiringcontinual adjustment simply tomaintain their relative position. Ourmodel enables us to examine mul-tiple organizations embracing allthe levels of government within thesystems framework.

In our work, we examined orga-nizational variables, includingpolicy makers, the program design-ers, the program implementingagency, staff assigned to implementthe program, and the target popu-lation. We were unable to find amodel in the literature that em-braces the contributions of the vari-ous social and behavioral scienceconcepts, constructs, and analysis.We chose, therefore, to create amodel based on general systemstheory to try to embrace thebreadth of implementation re-search. The original model was not

intended as a conceptual break-through; it is simply the most con-venient, useful way we know toorganize this diverse, complex bodyof research. Our model also con-tains non-organizational variablesthat give it substance: the problemgiving rise to the policy, the policyitself, the characteristics of the sys-tem, and characteristics of thesystem’s environment. Each ofthese is thought to affect implemen-tation. Finally, the model incorpo-rates dynamic elements. These arecharacteristics of the system thataffect in the model. Four pointsabout the model are particularlyimportant. First, the entire processis dynamic and, typically, iterative;policies are often revisited afterhaving been enacted. Second,policy gets defined and redefinedat each step in the implementationprocess as it is interpreted and re-ality-checked by the participants inthat organizational node. Third, ob-stacles to implementation can ariseat each link in the implementationprocess. They can also arise at thepoints at which organizations andprocesses are joined with one an-other. Fourth, the nature of the en-tire process itself may engenderobstacles to implementation, par-ticularly if the process is long andcomplex, involving lots of actorsand transactions.

To guide our work, we conceptu-alized a simple model to embracethe processes that extend frompolicy adoption through implemen-tation by operating agencies (seefigures 1 and 2). We describe thisas a complex web of expectationsand actions. The model extends toinclude organizations with imple-mentation roles as well as the char-acteristics of the implementation

• Our efforts to seek greaterunderstanding of how toensure increasedapplication of methods forearthquake hazard riskreduction by public andprivate organizations cutsacross the program focusesin lifelines, hospitals, andresponse and recovery,including communitysustainability andcommunity resilience.

• Hospital retrofit cases arebeing used as a test bed forfurthering our knowledgeand for developingpractical guides toimproving mitigationpractice, coordinating ourwork closely with that ofvon Winterfelt, Tierney,and others. Since theMCEER agenda is aimed atthe application of newtechnologies to thereduction of risk,implementationnecessarily cuts across allprograms and is relevantto each.

12

process itself. Our focus is onsets of organizations and thesystemic setting within whichthey play their implementationroles. We call these different setsof organizations primary targetorganizations, market interme-diaries, front line implementors,indirect implementors, nongov-ernmental policymaking partici-pants, and public policymakingorganizations.

Primary Actors in theImplementationProcess

We define Primary Target Organi-zations as architects, engineers, de-

velopers, builders, building owners,and organizations whose actions orinactions are the targets of policiesintended to get them to reducerisks to themselves and other mem-bers of the community from natu-ral hazard events. MarketIntermediary Organizations are de-fined as private organizations andpublic agencies that provide mort-gage monies, mortgage insurance,or insurance against losses fromnatural hazard events including, es-pecially, those whose policies andpractices affect the behavior of Pri-mary Target Organizations.

We have defined Front Line Imple-menting Organizations as agencies,typically public agencies, such asbuilding and safety and code en-forcement agencies, that arecharged with program implemen-tation. That is, they are organizationsthat are expected to allocate re-sources, including time and person-nel, to bring about the desiredeffects in the target organizations.We include the individuals withinthose organizations assigned to takeaction. Indirect Implementing Orga-nizations, on the other hand, arethose organizations charged withdesigning implementation pro-grams which mandate others, likelocal departments of building andsafety, to take action, inflict sanc-tions for not taking action, or pro-vide incentives to take action.Typically, these are viewed as fed-eral or state agencies responsiblefor ensuring that municipalities ad-minister mitigation programs.

Nongovernmental Policymaking Par-ticipants are an important elementin the implementation web. Theycomprise private organizations thatparticipate in policy development,such as professional associations(SEAOC, BSSC), and private interest

Target Population,Including

Intermediaries

Policy Making Body

Environmental

characteristics

Characteristics of thesystem as a whole Characteristics of

the program

Characteristicsof the Policy

Third PartyStakeholders

Program Design

Assigned Staff

Characteristics andperceptions of the

problem

Implementing agency

Target Population

Policy Making Body

Environmentalcharacteristics

Characteristics of thesystem as a whole

Characteristics ofthe program

Characteristicsof the Policy

Third Party

StakeholdersImplementing Agency

Program Design

Implementation Staff

Characteristics andperceptions of the

problem

IntermediaryAgency 1

Program Design

Staff

Program Design

Staff

IntermediaryAgency 2

� Figure 1. Implementation Within a Single Jurisdiction

� Figure 2. Implementation in a Multi-Level Governmental Setting


groups that seek to influence policy(ICBO, trade organizations). Someparticipants are fully engaged inpublic policy formation to the ex-tent that they are almost indistin-guishable from authorized publicpolicymakers in their sphere of in-fluence.

Policy Making Organizations aredefined for our purposes as publiclegislative, executive (or occasion-ally judicial) entities that adopt andauthoritatively state a policy in-tended to reduce risk to life andproperty from natural hazardevents

What ConstitutesSuccessful PolicyImplementation?

At the simplest level, "implemen-tation represents the faithful fulfill-ment of policy intentions by publicservants"(Calista, 1994, p. 117).Newcomers to business and gov-ernment often assume that a policy,once adopted, will be implementedin accord with the policy makers’intent. An increasingly rich bodyof research confirms what oldhands know – that is just not thecase. Practitioners and scholarshave come to understand thatpolicy adoption is simply one mile-stone in a continuing process ofaddressing an issue. Researchershave concluded, however, thatimplementation is a critical partof policy making process. Policyis adopted and adapted and drifts,morphs, and mutates through theimplementation process. The ex-tent of drift and mutation dependson a myriad of variables, only someof which can be controlled bypolicy makers. Calista’s assessmentof the field of study is that it has

evolved from one of viewing imple-mentation as simply the process ofcarrying out policy directives towhere implementation "is now in-tegral to the field of policy interven-tion, including recognizing itsinfluence on policy formulation"(Calista, 1994, p. 117). Evidencecontinues to mount demonstratingthat, often, policies are not imple-mented in accord with the policymakers’ intent. Indeed, it may bethat successful implementation isthe exception rather than the rule.Calista reports that the most preva-lent finding in implementation re-search is that outcomes are eitherdisappointing or unwitting (Calista,1994, citing Derthick, 1990). Oth-ers suggest that policy implemen-tation is "the continuation ofpolitics with other means" (Majoneand Wildavsky, 1978).

We have concluded that success-ful implementation is not the sameas solving the problem. It is entirelypossible that a program could beimplemented exactly as intendedand that the program is ineffectivebecause the problem transformedduring the implementation periodor because the program was flawedconceptually. If a local governmentprovides incentives for action byprivate actors, some will choose toparticipate and others will not. It ispossible that an overwhelmingmajority act in such a way as toconvince even the most jaded skep-tic that the policy has been imple-mented. Suppose, however, thatonly 10 percent of those targetedby the policy and eligible to partici-pate actually volunteer to partici-pate in the program or implementthe policy. Is policy implementationsuccessful? Presumably not, be-cause such a small proportion ofthe target was reached. Careful di-

14

agnosis of the implementation pro-cess might focus attention on a spe-cific aspect of the programdesigned to implement the policy,such as providing greater incentiveor engaging in more effective cam-paigns to make members of the tar-get audience more aware of theprogram. If each of the organiza-tions in the implementation netdoes precisely what is called for inan agency’s program plan, but, still,the private citizens or organizationstargeted for action fail to take thesteps that bring about clear intentof the public policy, is implementa-tion successful or has it failed?

We concluded, too, that success-ful implementation is a matter ofdegree. Consider the fanning out ofresponsibility for implementation.A federal agency looks to fifty states,each of which looks to a hundredor more municipalities, each ofwhich looks to several employees,each of whom tries to affect thebehavior of a dozen or more indi-viduals or firms. What proportionof the several hundred thousandpotential "implementations" in thisexample has to "take" for imple-mentation to be judged successful?Successful implementation isclearly, then, a relative concept. Wehave to think of it in terms of theextent to which it has occurredrather than whether it has occurred.Success, in the case of implemen-tation, is not a matter of absolutes.

Criteria for EvaluatingImplementation

We have identified six criteria forevaluating the extent to whichimplementation has been success-ful:

• Did the policy have the nomi-nally intended effect on the in-tended target population?

• To what extent were there un-intended side effects and werethose side effects adverse?

• To what extent did the variouselements of the implementationnetwork comply with policy di-rectives?

• What proportion of the targetwas reached?

• Did implementation take placewithin a reasonable time frame?That is, if one expected imple-mentation of a retrofit to be com-pleted within five years and ithas been fifteen and the job isnot yet completed, how effectivehas the implementation been?

• Were the costs of implementa-tion acceptable and reasonable?

Barriers toImplementation

We developed 37 propositionsconcerning obstacles to implemen-tation, including obstacles through-out the policy development andimplementation process. For Year 4,we focused our attention on exam-ining reasons private organizationsmight not implement risk reductionmeasures. We summarized thoseimpediments into four barriers.

Barrier 1. The organizationdoes not perceive itself at risk.

The first barrier to successfulimplementation of earthquake riskreduction is that the organizationfacing the risk does not perceiveitself at risk. Most practitioners inthe earthquake hazard field mayfind it difficult to comprehendthere are organizations that are not

“A rich body ofresearch onimplementationexists withinthe social andbehavioralsciences, butvery little of itaddressesnaturalhazards riskreduction.”


fully aware of their exposure andvulnerability, many are not, even inseismically active areas. In suchcases, appropriate means for over-coming the ignorance barrier byefforts to communicate the hazardalong with improving the targetorganization’s perceptions of expo-sure, vulnerability, and probable ef-fects.

Barrier 2. The organizationmay perceive itself to be atrisk, but does not perceive thatit can do something to reducethe risk.

Organizations may see them-selves as having both exposure andvulnerability, but not know what todo to reduce the probability of losswhen the event occurs. This condi-tion can exist when there is a smallinventory of acceptable risk reduc-ing actions or when no means haveyet been devised to mitigate thepotential loss. The organizationaldecision makers may have a fatalis-tic mind set that dictates againstattempts at risk reduction. Or, theproblem may be viewed as intrac-table by organizational decisionmakers.

Barrier 3. The organizationsees the risk and possibleactions, but doesn’t see takingrisk-reducing actions as inbeing in its best interest.

This barrier is not at all unlikely.Decision makers must weigh thesure costs of risk reduction againstthe possible benefits and, then,compare those costs and returnsagainst those of other possible in-vestments. Earthquake hazard miti-

gation does not always win thosecomparisons. Second, any risk re-duction measures must be congru-ent with organizational culture,goals, and priorities; if they are not,then the investment is directed to-ward other goals. Similarly, the pro-posed mitigations must becongruent with organizational mo-tivation.

Barrier 4. The individualorganization may be aware ofthe problem and riskreduction measures and beeager to reduce its risks, butfind that it is impossible totake action at this time.

Any of several reasons can keepan organization from acting whenit knows the risk, understands thereare ways to reduce risk, and is will-ing to move ahead. There may notbe space on the organizationalagenda right now. The organiza-tion may not have the capacity atpresent in terms of financialstrength or available skills. Or, itmay be that the organizationalenvironment presents sufficientambiguity to cause the organizationto defer action.

Overcoming Barriersto Implementation

In Year 4, we also worked to de-velop a document useful to the de-cision making community and tointegrate our work with otherMCEER tasks. Our second workingdraft, entitled Overcoming Barri-ers To Implementing EarthquakeHazard Mitigations: A PracticalGuide, is an attempt to developpractical means for overcoming

16

barriers that occur throughout theimplementation network. Theguidelines on which the draft isbased are summarized below.

Guideline 1: If a basic obstacle totaking precautions is an inaccurateassessment of risks by the target or-ganization, then the hazards profes-sional, if he or she expects to haveany impact, must provide the orga-nization with a clear, compellingstatement of the risks to the orga-nization.

Guideline 2: Risk reduction mea-sures are more likely when the Pri-mary Target Organizations see a clearlink between the potential hazard-ous event and adverse effects onthem and their businesses. Prudenthazard mitigators work to make thelinkage clear to those organizations.

Guideline 3: The probability of suc-cessful implementation of a policy(in either a multi-organizational set-ting or in a single organization) in-creases when both those involvedin policy making and implementa-tion have a similar perception of theproblem. It is sensible, then, to helpensure similar perceptions by ac-tive communication among the twoparties.

Guideline 4: Problems often shiftout from under solutions, renderingpolicies obsolete, ineffective, or dys-functional. Prudent implementorswork to ensure that policies aremodified as necessary to ensurethey are both effective and appro-priate, so that efforts at implement-ing the policy will not diminish.

Guideline 5: For an organization totake steps to reduce its exposureor vulnerability to earthquakes, keydecision makers in the organizationmust believe that practical stepsexist to reduce the risks associatedwith the event or condition andthat those steps are congruent with

the problem. To increase the prob-ability of implementation, designprofessionals and public officialsmust ensure that decision makerssee the link between the risk andthe solution.

Guideline 6: The prudentimplementor does not assume thatthe Primary Target Organization un-derstands the availability of a rangeof solutions and the possibility ofmodifying solutions to match thespecific needs of the individual or-ganization.

Guideline 7: The more a problemis viewed intractable, the less likelyit is that implementation will besuccessful. Hazard mitigators mustwork to reduce ignorance aboutthe phenomenon, reduce diversityin specific target populations, workwith reasonably sized target popu-lations, and become more skilled inunderstanding means for changingperceptions and behavior in the tar-get population.

Guideline 8: Successful implemen-tation of new policies and ap-proaches is more likely to occurpromptly in organizations that aretraditionally amendable to changeor have a culture that embraces in-novation.

Guideline 9: : The probability ofsuccessful implementation in-creases to the extent that actors inthe implementation process per-ceive congruence between meansand ends; that is, they will workharder to ensure implementation ifthey perceive that the policy andthe programs designed to imple-ment the policy are appropriate,given their perception of the prob-lem.

Guideline 10: The probability ofsuccessful implementation in eithera multi-organizational setting or ina single organization increases to

“We haveconcluded thatsuccessfulimplementationis not the sameas solving theproblem.”


the extent that various actors in theorganization have similar goals withrespect to risk reduction and buyinto the means selected for risk re-duction.

Guideline 11: Risk reduction mea-sures are more likely to be faithfullyimplemented when the organiza-tional leadership’s support is unam-biguous, the order is widelypublicized, the people chargedwith implementation have every-thing needed to implement themeasures, and those charged withimplementation have no doubt ofthe authority of the leadership toissue the decision.

Guideline 12: Unless the interestsof the various stakeholders, espe-cially those of the Primary Target Or-ganizations, are accommodated atsome minimally acceptable level, itis likely that mitigation policies andprograms will face guerilla action,be subject to subsequent wateringdown, and face court challenges.

Guideline 13: "Hazard mitigationis not a technical exercise; it is in-herently and often intensely politi-cal because mitigation usuallyinvolves placing cost burdens onsome stakeholders, and may involvea redistribution of resources. Haz-ard mitigators must, therefore, de-velop political as well as technicalsolutions" (Alesch and Petak, 1986).

Guideline 14: Policies are morelikely to be implemented success-fully when they are entrusted forimplementation to organizationsthat embrace the same goals andvalues as those implicit or explicitin the policy.

Guideline 15: Organizations willwork toward achieving successfulimplementation to the extent thatthey believe they can implementthe policy, that implementing thepolicy will achieve desired program

objectives, and that achieving theprogram objectives is consistentwith and supportive of theorganization’s primary objectives.

Guideline 16: Private organizationsare more likely to implement riskreduction practices when they seethat the risk poses a clear andpresent danger to their enterprise.Coupling risk reduction with rou-tine business concerns, such asproperty and casualty insuranceand related risk management con-cerns, helps bring it to the atten-tion of the organizational decisionmakers.

Guideline 17: Public policies in-tended to induce private parties toreduce natural hazard risks to theorganization and to the public atlarge are more likely to be imple-mented when the financial con-cerns of the private parties areacknowledged explicitly in thepolicy and provisions are made toalleviate financial burdens associ-ated with implementation.

Guideline 18: Other things beingequal, successful implementationdepends on entrusting implemen-tation to organizations with suffi-cient capacity to administer theprogram. If local government agen-cies are called upon to implementrisk reduction programs, theyshould be provided with the re-sources necessary to do the job.

Guideline 19: Implementation pro-ceeds more effectively when "theleaders of the implementing agen-cies possess substantial managerialand political skill and are commit-ted to statutory objectives (Sabatierand Mazmanian, 1979).

Guideline 20: Smaller organiza-tions may need technical assistance,in the form of consultants or self-help instructional materials, to aug-ment their staffs so they are capable

18

of making prudent choices con-cerning risk reduction for buildingsand structures.

Guideline 21: In complex organi-zational environments character-ized by instability and change, itmay be useful to test implementpublic risk reduction programsaimed at private organizations inpilot projects in a variety of settings.This will help avoid implementa-tion pitfalls that could come fromimmediate, widespread implemen-tation.

Guideline 22: If the purpose of apublic program is to induce privateorganizations to implement risk re-duction policies and practices, gov-ernmental organizations shouldwork to make it easy for the privateorganizations to understand the re-quirements and to facilitate imple-mentation by the privateorganizations.

Conclusions andFurther Research

In February 2001, we distributedboth drafts to a select group of four-teen practicing structural engi-neers, public officials, and membersof the earthquake hazard commu-nity who agreed to read both docu-ments and to participate in a oneday session in San Francisco to cri-tique our work and our products.The session, held in March, wasextremely successful. Because thereviewers were practitioners, thefocus was primarily on the sec-ond working draft. The partici-pants pointed out what wasuseful and what was not, helpedus clarify our target audience, andprovided guidance on how wecould make the draft more useful.At this writing, we are working to

integrate their critique into thedocument.

We have also distributed our firstdraft — the synthesis of the priorresearch on implementation — toa half dozen scholars drawn fromthe social and behavioral sciences.Each of these scholars is associatedwith one of the three engineeringresearch centers (MCEER, MAE, orPEER) and is actively engaged inearthquake hazard research. Thegroup of scholars collectively de-cided it would be appropriate tomeet in a central location to reviewtheir work. The group will have as-sembled by the time this report ispublished and will have provideda substantive critique of our draftreport. Along with the critiques bythe practitioners, this review willguide our development of finaldocuments.

Several tasks dominate our cur-rent research activities. The first iscompleting our review of the stateof the art in understanding ob-stacles to implementation andmeans for overcoming them. Wehave essentially completed our as-sessment of policy, intergovernmen-tal, interorganizational, political, andprocess variables. We still have workto complete on obstacles to imple-mentation associated with organi-zational behavior and decisionmaking. Second, we will revise ourdrafts based on critique from thescholars and practitioners whohave reviewed our work. Third, wewill begin case studies involvingseismic retrofits of hospitals as ameans for evaluating and extendingour set of propositions concerningbarriers to implementation andmeans for overcoming them.Fourth, we will continue the devel-opment of our conceptual model

“Case studiesinvolvingseismic retrofitof hospitals areplanned toevaluate andextend thepropositionsconcerningbarriers toimplementationand a means toovercome them.”


concerning the implementationprocess. Development of the con-ceptual models has been guided bythe existing research literature, butrefining and specifying the modelis dependent on the data elicitedin our stakeholder studies.

Beyond this year, we hope to ex-pand our hospital retrofit case stud-ies. Second, we will continue ourcurrent efforts to integrate our workwith other MCEER researchers,

References

Alesch, Daniel J., (1998), “Adopting And Implementing Performance-Based SeismicDesign Standards,” paper prepared for the Earthquake Engineering Research Institute,Oakland, CA.

Alesch, D.J., and Petak, W.J., (1986), The Politics and Economics of Earthquake HazardMitigation, Natural Hazard Research and Application Center, University of Colorado,Boulder, CO.

Bardach, Eugene, (1977), The Implementation Game: What Happens After a BillBecomes Law, Cambridge, Mass.: The MIT Press.

Calista, Donald, (1994), “Policy Implementation,” Encyclopedia of Policy Studies (ed.Stuart Nagel), Marcel Dekker, New York, NY., pp 117-155.

Calista, Donald, (1986), Linking Policy Intention and Policy Implementation: The Roleof the Organization in the Integration of Human Resources, Administration andSociety, Vol. 18, pp. 263-286.

Drabeck, Thomas E., Mushkatel, Alvin H. and Kilijanek, Thomas S., (1983), EarthquakeMitigation Policy: The Experience of Two States, University of Colorado, Boulder,Institute of Behavioral Science, Monograph 37.

Earthquake Engineering Research Institute, (1988), Incentives and Impediments toImproving the Seismic Performance of Buildings, Oakland, CA.

Godschalk, David R., Beatley, Timothy, Berke, Philip, Brower, David J., Kaiser, Edward J.,Bohl, Charles C., and Goebel, R. Matthew, (1999), Natural Hazard Mitigation:Recasting Disaster Policy and Planning, Island Press, Washington, D. C.

Lipsky, M., (1971), “Street Level Bureaucracy and the Analysis of Urban Reform,” UrbanAffairs Quarterly, Vol. 6, pp. 391-409.

Mazmanian, D. A. and Sabatier, P.A., (1989), Effective Policy Implementation, LexingtonBooks, Lexington, MA.

Majone, G. and Wildavsky, A., (1978), “Implementation as Evolution,” Policy StudiesAnnual Review, Vol. 2. H. E. Freeman (ed.), Sage, Beverly Hills. CA

May, Peter J. and Williams, Walter, (1986), Disaster Policy Implementation: ManagingPrograms Under Shared Governance, Plenum Press, New York.

Detlof von Winterfelt (UniversitySouthern California) and KathleenTierney (Disaster Research Center,University of Delaware). Third, wewill complete the development ofour conceptual model of the imple-mentation process. Finally, we expectto complete two monographs forpublication by MCEER focusing onmeans for overcoming barriers toimplementation both in the publicand private sectors.

20

Pressman and Wildavsky, (1984), 3rd edition, Implementation, University of CaliforniaPress, Berkeley, California

Sabatier, P. and Mazmanian, D., (1979), "The Conditions of Effective Implementation;a Guide to Accomplishing Policy Objectives,” Policy Analysis, No. 5, pp. 481-504.

Sabatier, P. and Mazmanian, D., (1981), “The Implementation of Public Policy: AFramework of Analysis,” in Effective Policy Implementation, (eds. D. A. Mazmanianand P. A. Sabatier), Heath, Lexington, MA.

Sabatier, P. A., (1986), “Top-Down and Bottom-Up Approaches to ImplementationResearch: A Critical Analysis and Suggested Synthesis,” Journal of Public Policy, Vol.6, pp. 21-48.

Taylor, Craig, Mittler, Elliot, and Lund, LeVal, (1998), Overcoming Barriers: LifelineSeismic Improvement Programs, Monograph No. 13, American Society of CivilEngineers, Reston, VA.

References (Con’t)

21

Research Objectives

The objectives of the research are to: 1) simulate in the laboratory full-scale permanent ground deformation (PGD) effects on steel pipelines withelbows, 2) develop an extensive and detailed experimental database onpipeline and soil reactions triggered by earthquake-induced PGD, and 3)refine and validate analytical models so that complex soil-pipeline interac-tions can be numerically simulated with the precision and reliability nec-essary for planning and design. To accomplish these objectives, aninternational partnership was organized, involving the Tokyo Gas Com-pany, Ltd., MCEER, and NSF through its program for US/Japan CooperativeResearch in Urban Earthquake Disaster Mitigation. The project combinesexperimental and analytical research performed at Tokyo Gas facilities withexperimental and analytical work undertaken at Cornell University. Theexperiments at Cornell represent the largest simulations of PGD effectson pipelines ever performed in the laboratory.

Large Scale Experiments ofPermanent Ground DeformationEffects on Steel Pipelines

by Koji Yoshizaki, Thomas D. O’Rourke, Timothy Bond, James Mason and Masanori Hamada


Tokyo Gas Company, Ltd.National Science Foundation,

U.S.- Japan CooperativeResearch for Mitigation ofUrban Earthquake Disasters

Koji Yoshizaki, PipelineEngineer, FundamentalTechnology Laboratory,Tokyo Gas Company, Ltd.

Thomas D. O’Rourke,Thomas R. Briggs Professorof Engineering, TimothyBond, Laboratory Manager,and James Mason,Graduate ResearchAssistant,Department of Civil andEnvironmentalEngineering, CornellUniversity

Masanori Hamada,Professor, Department ofCivil Engineering, WasedaUniversity

During earthquakes, permanent ground deformation (PGD) can dam-age buried pipelines. Earthquake-induced PGD can occur as sur-

face fault deformation, liquefaction-induced soil movements, and land-slides. There is substantial evidence from previous earthquakes, suchas the 1983 Nihonkai-chubu (Hamada and O’Rourke, 1992), the 1994Northridge (O’Rourke and Palmer, 1996), and the 1995 Hyogoken-Nanbu (Oka, 1996) earthquakes, of gas and water supply pipeline dam-age caused by earthquake-induced PGD. More recent earthquakes,including the 1999 Kocaeli and Duzce earthquakes in Turkey, and the1999 Chi-chi earthquake in Taiwan, have provided additional evidenceof the importance of liquefaction, fault rupture, and landslides throughtheir effects on a variety of highway, electrical, gas, and water supplylifelines.

Gas and other types of pipelines must often be constructed to changedirection rapidly. In such cases, the pipeline is installed with an elbowthat can be fabricated for a change in direction from 90 to a few de-grees. Because elbows are locations where flexural and axial pipeline

22

• Cambridge University

• Kyoto University

• Rensselaer PolytechnicInstitute

• University of SouthernCalifornia

• University of Tokyo

• Yamaguchi University

deformations are restrained, con-centrated strains can easily accu-mulate at elbows in response toPGD.

The response of a pipeline el-bow, deformed by adjacentground rupture and subject to theconstraining effects of surround-ing soil, is a complex interactionproblem. A comprehensive andreliable solution to this problemrequires laboratory experimentson elbows to characterize theirthree-dimensional response toaxial and flexural loading, an ana-lytical model that embodies soil-structure interaction combinedwith three-dimensional elbow re-sponse, and full-scale experimen-tal calibration and validation ofthe analytical model.

To resolve this problem, an in-ternational team was organized.The principal participants are To-kyo Gas Company, Cornell andWaseda Universities. The researchalso involves the University ofCambridge, UK, Rensselaer Poly-technic Institute, and the Univer-sity of Southern California.Waseda University leads a consor-tium of Japanese university par-ticipants that include Kyoto andYamaguchi Universities and theUniversity of Tokyo. The work was

performed as part of MCEER Pro-gram 1 on the Seismic Retrofitand Rehabilitation of Lifelines andthe NSF program for U.S./JapanCooperative Research in UrbanEarthquake Disaster Mitigation.

MCEER has a long history ofproductive collaboration with theJapanese earthquake engineeringresearch community. Seven U.S./Japan workshops on the earth-quake performance of lifeline fa-cilities and countermeasuresagainst liquefaction have been co-sponsored by MCEER, and theproceedings of these workshopshave been published and distrib-uted by MCEER. The latest in thisseries of workshops (O’Rourke, etal., 1999) was held in Seattle, WAin conjunction with the 5th U.S.Conference on Lifeline Earth-quake Engineering. The nextworkshop is planned during theforthcoming year in Tokyo, Japan.U.S. participants in the work-shops include MAE, MCEER, andPEER researchers.

Experimental andAnalytical Models

One of the deformation condi-tions of interest is illustrated inFigure 1a that shows a pipeline

The users of this research include: public and private util-ity companies, including gas distribution companies, suchas Tokyo Gas and Memphis Gas, Light and Water; and waterdistribution companies, such as the Los Angeles Departmentof Water and Power (LADWP), and the East Bay MunicipalUtility District (EBMUD). The research is also of interest toengineering design and consulting companies. The experi-mental data and analytical modeling procedures developedfor this project are of direct relevance for underground gas,water, petroleum and electrical conduits.

23Large Scale Experiments of Permanent Ground Deformation Effects on Steel Pipelines

Program 1: SeismicEvaluation and Retrofit ofLifeline Networks

Program 2: Seismic Retrofit ofHospitals, Task 2.7a, CostBenefit Studies ofRehabilitation UsingAdvanced Technologies

with an elbow subjected to PGDconsistent with lateral spreadand/or landslides. Although lateralspreads and landslides involvecomplex patterns of soil move-ment, the most severe deforma-tion associated with thesephenomena occurs at the elbowsand near the margins between thedisplaced soil mass and adjacent,more stable ground. The deforma-tion along this boundary can besimplified as abrupt, planar soildisplacement. Pipelines that canbe sited and designed for abruptlateral displacement will be ableto accommodate complex pat-terns of deformation that fre-quently involve a more gradualdistribution of movement acrossthe pipeline. Abrupt soil displace-ment also represents the principalmode of deformation atfault crossings.

Figure 1b illustratesthe concept of thelarge-scale experiments.A steel pipeline with anelbow is installed underthe actual soil, usingfabrication and compac-tion procedures encoun-tered in practice, andthen subjected toabrupt lateral soil dis-placement. The scale ofthe experimental facil-ity is chosen so thatlarge soil movementsare generated, inducingsoil-pipeline interactionunaf fected by theboundaries of the testfacility in which thepipeline is buried. Theground deformationsimulated by the experi-ment represents defor-mation conditions

associated with lateral spread,landslides, and fault crossings, andtherefore applies to many differ-ent geotechnical scenarios. In ad-dition, the experimental data andanalytical modeling products areof direct relevance for under-ground gas, water, petroleum, andelectrical conduits.

A modeling technique, namedHYBRID MODEL, was developedfor simulating large-scale pipelineand elbow response to PGD(Yoshizaki et al., 1999 and 2001).The model uses shell elements forthe elbow where large, localizedstrains occur. Shell elements arelocated over a distance of 20times the pipe diameter from thecenter point of the elbow. Theshell elements are linked to beamelements that extend beyond this

(a) PGD Effect of Buried Pipelines with Elbows

9m5m

1.3m

ElbowCompacted sand

Welded steelpipeline

Fixed box

Displacement ofMovable box

: 1.1 m

(b) Experimental Concept

� Figure 1. Experimental Concept for PGD Effects on Buried Pipelines with Elbows

24

distance. Soil-pipeline interactionunder PGD is characterized by p-y and t-z curves, linking soilstresses on the pipe to the rela-tive displacement between them.Lateral soil-pipeline interaction ischaracterized on the basis of labo-ratory experiments originally per-formed at Cornell (Trautmannand O’Rourke, 1985) and dupli-cated at Tokyo Gas experimentalfacilities.

In-plane bending experimentswere conducted by Tokyo Gas(Yoshizaki et al., 1999 and 2001)on full-scale specimens of steel el-bows in both the closing andopening modes. No leakage wasobserved in the closing mode,

even when the opposing ends ofthe elbow were deformed intocontact with each other andstrains as high as 70% were mea-sured. In contrast, leakage was ob-served in the opening mode forall cases except those in whichthe deformation was restrictedbecause of the experimental load-ing device. The HYBRID MODELwas used to simulate the deforma-tion behavior of the elbows usinglinear shell elements. Very goodagreement was achieved betweenthe analytical and experimentalresults for all levels of plastic de-formation.

Large ScaleExperiments

Figure 2 shows a plan view ofthe experimental setup that con-sisted of five main components,including a test compartment (Aand C in the figure), pulley load-ing system (F), sand storage bin(G), sand container hoisted fromstorage bin to test compartment(not shown), and data acquisitionsystem (H). The test compart-ment was composed of a movablebox (A) and fixed box (C) withinwhich the instrumented pipelinewas installed and backfilled. TheL-shaped movable box had insidedimensions of 4.2 m by 6m by 1.5m deep. It was constructed on abase of steel I-beams positionedover Teflon sheets that were fixedto the floor. The Teflon sheetsprovided a low-friction surface onwhich the movable box was dis-placed by a pulley loading system.The fixed box, which was an-chored to the floor, was designedto simulate stable ground adja-

Movable boxTeflon sheetFixed box100-mm-diameter pipe90-degree elbowPulley systemSand storage binData aquisition system

A

C

F

GH

H

H

D E

ABCDEFGH

.B

2m

� Figure 2. Plan View of Experimental Setup


cent to a zone of PGD similar tothat illustrated in Figure 1.

A 100-mm-diameter pipelinewith 4.1-mm wall thickness wasused in the tests. It was composedof two straight pipes welded to a90-degree elbow (E). The shortsection of straight pipe (D) was5.4 m long, whereas the longestsection was 9.3 m. Both ends ofthe pipeline were bolted to reac-tion walls. The elbows were com-posed of STPT 370 steel (JapaneseIndustrial Standard, JIS-G3456)with a specified minimum yieldstress of 215 MPa and a minimumultimate tensile strength of 370MPa. The straight pipe was com-posed of SGP steel (JIS-G3452)with a minimum ultimate tensilestrength of 294 MPa. About 150strain gauges were installed onthe pipe to measure strain duringthe tests. Extensometers, loadcells, and soil pressure meterswere also deployed throughoutthe test setup.

The pipeline was installed at a0.9-m depth to top of pipe in eachof four experiments. In each ex-periment, soil was placed at a dif-ferent water content and in situdensity. All experiments were con-ducted to induce opening-modedeformation of the elbow.

The experimental facility wasdesigned with the assistance ofthe HYBRID MODEL that wasused to simulate various testingconfigurations and compartmentdimensions. Significant character-istics of the experimental facilityare its size and volume. The stor-age bin for the sand was overthree stories tall, with a capacityfor 75 tons. Approximately 60tons of sand were moved from thestorage bin into the test compart-ment for each experiment with a

container that was hoisted withthe overhead conveyor. The sandwas placed and compacted in150-mm lifts with strict controlson water content and in situ den-sity. One of the most significantchallenges during the testing wasthe movement and controlledplacement of such large volumesof sand with water content thatwas intentionally changed foreach experiment.

The movable box was pulled byan overhead crane with an 8 to 1mechanical advantage obtainedthrough the pulley system shownin the figure. The maximum ca-pacity of the loading system was

� Figure 3. Overhead View of Test Compartment Before(top) andAfter (bottom) an Experiment

26

1 m of lateral displacement and784 kN. The rate of displacementof the movable box was approxi-mately 16mm/s.

Figure 3 shows the ground sur-face of the test compartment be-fore and after an experiment.Surficial cracks can be seen in thearea near the pipeline elbow andthe abrupt displacement planebetween the movable and fixedboxes after the test. In all cases,planes of soil slip and crackingreached the ground surface, butdid not intersect the walls of thetest compartment to any appre-ciable degree.

Figure 4 shows an overheadview of the test compartment af-ter soil excavation to the pipelinefollowing one of the experiments.Because each experiment was rununtil a total displacement of about1m, the analytical models can betested and calibrated through abroad range of deformation andstrain in the elbow. The deformedshape of the pipeline can be seenclearly in the figure. Its shape isremarkably consistent with theshape shown by the finite element

simulations discussed in the nextsection.

Analytical andExperimental Results

Finite element analyses wereconducted with the HYBRDMODEL to check the ability of theanalytical simulations to capturekey aspects of the pipeline andelbow response to abrupt lateraldisplacement. Figure 5a comparesthe deformed pipeline shape ofthe analytical model with mea-sured deformation of the experi-mental pipeline. There isexcellent agreement between thetwo, and there is obvious agree-ment between the analytical de-formation and the overhead viewof the deformed pipeline in Fig-ure 4. Figure 5b shows the mea-sured and predicted strains undermaximum ground deformation onboth the tensile (extrados) andcompressive (intrados) surfacesof flexure along the pipeline. Fig-ures 5c and d show the measuredand analytical strains around thepipe circumference in which theangular distance is measured fromextrados to intrados of pipe, cor-responding to 0 and 180o, respec-tively. In Figure 5d, the data pointwith an upward arrow indicatesthe maximum strain measuredwhen the gauge was discon-nected during the experiment. Be-cause the disconnection occurredbefore maximum deformation ofthe elbow, it is likely that the ac-tual strain was larger than thevalue plotted. Overall, there isgood agreement for both the mag-nitude and distribution of mea-sured and analytical strains.

� Figure 4. Overhead View of Deformed Experimental Pipeline


The large-scalee x p e r i m e n t shave provided arich and compre-hensive databasefor understand-ing pipeline re-sponse to largePGD, particularlyunder conditionswhere local con-straint at an el-bow leads tocomplex three-dimensional re-sponse and highstrain concentra-tions. Althoughthe interpreta-tion of the ex-p e r i m e n t a lresults is still inprogress, the datahave alreadyshown that, witha p p r o p r i a t emodi f icat ions ,the current gen-eration of analyti-cal models is ableto simulate realperformance in areliable way.

One of the most important as-pects of the research has been toclarify the effects of moisturecontent on the pressures gener-ated by soil-pipeline interactionduring PGD. Current analyticalmodels use p-y curves derivedfrom laboratory test results withdry sand. Virtually all sand in thefield is placed with measurablewater contents that affect its insitu density and shear deforma-tion characteristics. The large-scale experiments have shownthat the failure surfaces and de-

formation patterns in soil adja-cent to the pipeline during PGDare significantly different for dryand partially saturated sand.

SummaryLarge-scale experiments spon-

sored by Tokyo Gas were success-fully completed to evaluate theeffects of earthquake-inducedground rupture on welded steelpipelines with elbows. The ex-perimental set-up involved thelargest full-scale replication of

-6

-4

-2

0

2-10 -8 -6 -4 -2 0 2

Distance from the corner to south (m)

Dis

tanc

e fr

om th

e co

rner

to e

ast (

m)

Experiment

FEA

40

20

0

-200 5 10 15

Distance from the east edge, Lp (m)

Experiment (extrados)Experiment (intrados)FEA (extrados)FEA (intrados)Elbow

Lp

20

10

0

-100 45 90 135 180

Angle, (degree)

Experiment (Longitudinal)Experiment (Circumferential)FEA (Longitudinal)FEA (Circumferential)

Experiment (Longitudinal)Experiment (Circumferential)FEA (Longitudinal)FEA (Circumferential)

40

30

20

10

0

-10

-200 45 90 135 180

Angle, (degree)

B

B

Section B-B

(a) Deformation of the pipeline after the test (b) Distribution of axial strain in the longitudinaldirection

(c) Strain distribution at the cross section of theconnecting part between the elbow and shortleg of the straight pipe (Section A-A)

(d) Strain distribution at the cross section of theelbow (Section B-B)

Str

ain,

%

Str

ain,

%S

trai

n,%

A A

Section A-A

� Figure 5. Comparison Between Analytical and Experimental Results

28

References

Hamada, M. and O’Rourke, T.D., Eds., (1992), Case Studies of Liquefaction and LifelinePerformance During Past Earthquakes, Vol. 1, NCEER-92-0001, National Center forEarthquake Engineering Research, University at Buffalo, April.

Oka, S. (1996), “Damage of Gas Facilities by Great Hanshin Earthquake and RestorationProcess,” Proceedings, 6th Japan-U.S. Workshop on Earthquake Resistant Design ofLifeline Facilities and Countermeasures Against Soil Liquefaction, NCEER-96-0012,MCEER, University at Buffalo, pp.111-124.

O’Rourke, T. D. and Palmer, M. C. (1996), “Earthquake Performance of Gas TransmissionPipelines,” Earthquake Spectra, Vol. 20, No. 3, pp.493-527.

O’Rourke, T.D., Bardet, J.P., and Hamada, M., (1999), Proceedings, 7th US-Japan Workshopon Earthquake Resistant Design of Lifeline Facilities and Countermeasures AgainstSoil Liquefaction, MCEER-99-0019, MCEER, University at Buffalo, Nov.

Trautmann, C.H. and O’Rourke, T.D, (1985), “Lateral Force-Displacement Response ofBuried Pipe,” Journal of Geotechnical Engineering, ASCE, Vol.111, No.9, pp.1077-1092.

Yoshizaki, K., Hosokawa, N., Ando, H., Oguchi, N., Sogabe, K. and Hamada, M., (1999),“Deformation Behavior of Buried Pipelines with Elbows Subjected to Large GroundDisplacement,” Journal of Structural Mechanics and Earthquake Engineering,No.626/I-48, pp.173-184 (in Japanese).

Yoshizaki, K., O’Rourke, T. D. and Hamada, M., (2001), “Large Deformation Behavior ofBuried Pipelines with Low-angle Elbows Subjected to Permanent GroundDeformation,” Journal of Structural Mechanics and Earthquake Engineering, I-2130,pp. 41-52.

ground deformation effects onpipelines ever simulated in thelab. The tests allow for calibrationof a sophisticated soil-pipeline in-teraction analytical program de-veloped in conjunction with theexperimental work. The experi-mental data and analytical mod-eling products are of directrelevance for underground gas,water, petroleum, and electricalconduits.

Additional work at Cornellsponsored by Tokyo Gas is

planned in the forthcoming yearto investigate further the p-y char-acterization for partially saturatedsand as a function a water con-tent and compaction effort. Workalso is planned with collaboratinguniversities for developing thenext generation of analyticalmodel that will represent the soilas a continuum with specific con-stitutive relationships capable ofsimulating large ground deforma-tion and its interaction with bur-ied pipelines.

29

Research Objectives

The goals of this project are to develop rehabilitation strategies and en-hance seismic design guidelines for key equipments in power substations,which are one of the most critical facilities in a power system. Further-more, it will provide the knowledge base to be integrated into the overallloss estimation model for the entire power network. To achieve these goals,key substation equipments are identified and effectiveness of base-isola-tion to increase their seismic resilience are assessed using a comprehen-sive experimental and analytical study.

Experimental and Analytical Studyof Base-Isolation for Electric PowerEquipments

by M. Ala Saadeghvaziri and Maria Q. Feng

Critical power system facilities, such as substations, sustained signifi-cant damage in California and Japan earthquakes and more recently

during the 1999 Chi-Chi, Taiwan, and the Izmit, Turkey earthquakes. Func-tionality of electric power systems, especially in the age of informationtechnology, is vital to maintaining the welfare of the general public, sus-taining economic activities and assisting recovery, restoration, and recon-struction of the seismically damaged built environment. Furthermore,enhanced seismic design and rehabilitation will ensure long-term reliabil-ity and longevity of critical equipments. In order to be able to assess thereliability of power systems and to develop seismic resistant mitigationstrategies, critical components must be identified and their seismic per-formance evaluated. Transformers and their bushings are among the mostcritical components in a complex power system and their seismic perfor-mance during past earthquakes has not been satisfactory. Figure 1 showsdamage to a transformer in Izmit-2 substation during the Izmit, Turkeyearthquake of August 17, 1999 (EERI Newsletter, 1999).

Generally, there are several modes of substation transformer failure dur-ing an earthquake, namely movement and turn over of unanchored trans-formers, anchorage failure that can cause ripping of the transformer caseand oil leakage, foundation failure causing rocking and tilting, and failureof the gasket and oil leakage due to the interaction between the trans-former and the bushing. Some transformers are supported on rails forease of installation and because such a set up allows for air circulation toprovide additional cooling, thus enhancing corrosion resistance. Another


M. Ala Saadeghvaziri,Associate Professor andSelahattin Ersoy,Graduate Student,Department of Civil andEnvironmentalEngineering, New JerseyInstitute of Technology

Maria Q. Feng, AssociateProfessor, Civil andEnvironmentalEngineering Department,University of California,Irvine

Nobuo Murota, BridgestoneCompany, Japan

30

form of damage in anunanchored transformeris large movement thatcan cause damage to othercomponents connected tothe unit, such as controlcables and bushings. Dam-age due to the latter fac-tor (i.e., interactionbetween transformer-bushing and other equip-ments) can be critical andcan cause significant dam-age and delays in service.

It is more common to anchortransformers where considerationsare given to seismic loads. This canbe done by either bolting the trans-former to its footing slab or bywelding it to steel embedded in theslab (ASCE, 1999). Designing theanchorage at the supports requiresconsideration to large forces notonly due to gravity and horizontalseismic forces but also from over-turning moments in both direc-tions. Furthermore, otherappendages such as radiators andreservoirs are attached to a typicaltransformer. These appendages cancause significant torsional forcesthat must also be considered in thedesign of the supports. In additionto strength, the anchorage systemmust have adequate stiffness andtight tolerances to prevent initia-tion of impact forces that can dam-age internal elements or excitehigher modes that can damagebrittle porcelain members.

There are many cases of bolt orweld failure during past earth-quakes (ASCE, 1999). However,with better attention to the detailsin the design of supports, their per-formance during an earthquake canbe improved. Implementing well-designed anchorage for retrofit ofexisting transformers, nevertheless,can be difficult and costly. Further-more, in many situations, for bothnew and existing transformers, awell-designed anchorage may onlychange the mode of failure to thefoundation (i.e., the next weak linkin the system). The failure shownin Figure 1 appears to be due tofoundation settlement.

Boundary gaps due to back andforth motion of transformers androcking of transformers and theirfootings due to soil-structure inter-action have been observed duringpast earthquakes (ASCE, 1999).Therefore, in many cases, the useof base-isolation for transformers

• National Center forResearch in EarthquakeEngineering (NCREE),Taipei, Taiwan

• Earthquake ProtectionSystem, Inc., Richmond,California

• Bridgestone Corporation,Kanagawa-ken, Japan

Primary users of this research include utility companiesand owners of electrical substation equipments, manufac-turers of electrical equipments, and manufacturers of base-isolation devices. The research results can further be usedby structural and electrical engineers for design and ret-rofit of electrical power equipments.

� Figure 1. Transformer Failure During the August 17,1999, Izmit, Turkey Earthquake (EERI, Oct. 99)

31Experimental and Analytical Study of Base-Isolation for Electric Power Equipments

may be the only suitable remedy toalleviate these problems, especiallyfor existing transformers in highseismic regions. Base-isolation willalso reduce the input accelerationinto the bushing and will lessen theinteraction between the trans-former and the bushing, which hasbeen the cause of many bushingdamages during past earthquakes.Furthermore, by reducing the iner-tia forces, base-isolation can alsoprevent the possibility of internaldamage.

The after effect of an earthquakeon reliability and longevity of atransformer is directly related tothe level of shaking of internal ele-ments. High levels of uncontrolledshaking may very well reduce thelife expectancy and reliability ofinternal elements. Internal damageis normally difficult to observe anddocument because of limitationswith post-earthquake inspectionsof the transformer internal system.However, there are reports of casesof internal damage to transformers(ASCE, 1999).

Satisfying the mobility require-ment for maintenance purposes isanother advantage of base-isolationover a tightly designed anchoragesystem. An issue with the use ofbase-isolation that demands carefulconsideration is the possible ad-verse effect of relatively large dis-placements on the response ofinter-connecting equipments, espe-cially bushings. Among possibleremedies to be considered are abalanced approach to the design ofthe isolation system (displacementvs. inertia reduction), appropriatedesign of conductor slacks, and useof flexible conductor connections.Therefore, a successful applicationof this technology requires in-depth understanding of the re-

sponses of the individual systemsinvolved as well as their interac-tions.

Within this study, two base-isola-tion systems are being investigatedunder a collaborative effort amongseveral institutions and industrialpartners. The following sectionsdiscuss the experimental programalong with some of the findings.The two systems considered are afriction pendulum system and ahybrid system consisting of slidingand rubber bearings. It representsthe first effort in testing base-iso-lated large-scale transformer-bush-ing systems using an earthquakesimulator.

Experimental StudyAn extensive series of tests

were conducted on a transformermodel supporting a bushing. Theprimary objective was to com-pare the response of a fixed basedtransformer-bushing system tothat of the system when isolated.The testing was conducted on theearthquake simulator at the Na-tional Center for Research onEarthquake Engineering (NCREE)in Taiwan in collaboration withthe manufacturers. Uniaxial, bi-axial, and triaxial excitations wereconducted employing severalearthquake records with PGAs inthe range of 0.125g to 0.5g. Thetesting schedule also includedwhite noise tests to identify dy-namic characteristics of the bush-ings and the transformer model,resulting in more than 200 tests.

Considering the payload capac-ity of the earthquake simulator, thetransformer model was designed toweigh 235.5 kN. The model is afour-layer steel frame structure

“This studyrepresents thefirst effort intesting base-isolated large-scale trans-former-bushingsystems usingan earthquakesimulator.”

32

with lead blocks loaded to repre-sent the internal core and coil of atransformer.

Consistent with the practice em-ployed by TaiPower, the bushing wasattached to the transformer modeltop plate at a right angle. The con-nection represents some of the im-portant structural aspects of thesupport of the bushing, which hasbeen shown to be very important in

quantifying the important interactionbetween the two components.

Figure 2 shows isometric and el-evation views of the entire model. Ascan be seen from this figure, the bush-ing projects inside the transformermodel as is the case in an actual situ-ation. Two types of bushings, 69 kVand 161 kV, were used in the experi-ment. Several earthquake recordswere used for the testing, amongthem were the 1940 El-Centro, 1994Sylmar (Northridge), and 1995Takatori (Kobe) records. Transformerfrequency was measured to be 12.5Hz in both x and y directions. Thefrequencies for the 69 kV and 161kV bushings were 27.0 Hz and 12.5Hz in both x and y directions, respec-tively. The equivalent damping ratiofor each component was around 2%.

Friction PendulumSystem Results

One of the most recent base iso-lation systems to improve the earth-quake resistance of structures isFriction Pendulum System (FPS),shown in Figure 3. Detailed descrip-tion of the system along with indepth discussion of experimentaland analytical results can be foundelsewhere (Ersoy, et. al 2001;Saadeghvaziri and Ersoy, 2001). Forstructures isolated by this system,

the period of vibration (similar toa pendulum) only depends on ge-ometry (in this case radius of cur-vature) and the gravitationalconstant, and it does not dependon the mass. Therefore, FPS bear-ings are one of the most suitableisolation devices for a systemwith relatively small weight com-pared to a high-rise building(such as substation equipments).Furthermore, through friction,

SECTION VIEW

Spherical Concave

Surface

Articulated Friction Slider

Bearing MaterialBearing Material

Articulated Friction Slider

Spherical ConcaveSurface

SECTION VIEW

Articulated FrictionSliderBearing Material

Spherical ConcaveSurface

� Figure 2. Two Views ofTransformer-Bushing Model

� Figure 3. Photograph and Cross-Section View of an FPS Isolator


the system can provide a high levelof damping. Four 18.64" radius FPSbearings at the four corners wereused to support the model for theisolated case.

The test results for Sylmar recordare tabulated in Table 1 for isolatedand non-isolated cases. In thesetables, A

2 and A

3 show the accel-

eration values (in g) at the bottomand the top of the transformermodel, respectively. A

B1, A

B2, A

B3, A

B4

represent the accelerations at dif-ferent locations along the bushing.A

B1 is the bottom of the bushing

and AB4

is the top. Note that thebushing projects inside the trans-former model. That is, it is con-nected to the transformer modelsomewhere between points A

B2,

and AB3

. Thus, the bottom of thebushing (i.e., point A

B1) can have a

response as large as the top of thebushing (A

B4). D

2, and D

3 (in mm)

are the relative displacement valuesof the transformer model at thebottom and top, respectively.

In comparing these numbers apoint should be noted with regardto the base-isolated results for Case

PGATarget 0.1250 - - 0.1250 - - PGATarget 0.3750 0.2500 - 0.3750 0.2500 -PGAReal 0.1468 - - 0.1376 - - PGAReal 0.3699 0.2094 - 0.3809 0.2454 -A2 0.1602 - - 0.0778 - - A2 0.4213 0.2167 - 0.3642 0.5833 -A3 0.2457 - - 0.1173 - - A3 0.6257 0.4479 - 0.7031 1.0592 -AB1 0.2715 - - 0.1664 - - AB1 0.9549 1.1157 - 1.4774 4.6122 -AB2 0.2507 - - 0.1280 - - AB2 0.6118 0.5967 - 0.9029 2.7761 -AB3 0.5127 - - 0.1418 - - AB3 1.3414 1.1542 - 0.7015 1.2588 -AB4 0.9188 - - 0.2323 - - AB4 2.6399 2.5825 - 1.4507 3.0900 -D2 0.8638 - - 7.7837 - - D2 2.6765 3.3448 - 65.0059 79.1818 -

1

D3 1.2757 - - 7.8814 - -

5*

D3 4.8021 5.5605 - 71.0776 80.1011 -PGATarget 0.2500 - - 0.2500 - - PGATarget 0.2500 0.1250 0.1250 0.2500 0.1250 0.1250PGAReal 0.2377 - - 0.2817 - - PGAReal 0.2362 0.1160 0.1193 0.2676 0.1526 0.1239A2 0.2586 - - 0.1186 - - A2 0.2570 0.1280 - 0.1367 0.1273 -A3 0.4555 - - 0.2044 - - A3 0.4323 0.2539 0.1537 0.2017 0.2005 0.1728AB1 0.4888 - - 0.3748 - - AB1 0.5214 0.4160 0.1606 0.4075 0.4234 0.1745AB2 0.4374 - - 0.2824 - - AB2 0.4129 0.2671 0.1784 0.3052 0.3440 0.1727AB3 1.0256 - - 0.2111 - - AB3 1.0605 0.5432 0.1880 0.2478 0.2714 0.1633AB4 1.8672 - - - - - AB4 2.0097 1.1053 0.1611 0.4534 0.6593 0.1762D2 1.6892 - - 26.2215 - - D2 1.4588 1.8311 - 20.8579 22.5745 -

2

D3 2.4049 - - 26.6045 - -

6

D3 2.9084 3.3387 - 22.2175 22.3609 -PGATarget 0.3750 - - 0.3750 - - PGATarget 0.3750 0.2500 0.2500 0.3750 0.2500 0.2500PGAReal 0.3781 - - 0.3793 - - PGAReal 0.3833 0.2161 0.2362 0.4031 0.2414 0.2316A2 0.4191 - - 0.1661 - - A2 0.4150 0.2123 - 0.2205 0.2330 -A3 0.6664 - - 0.3323 - - A3 0.6180 0.4404 0.2844 0.4769 0.2743 0.3494AB1 0.8257 - - 0.5718 - - AB1 1.1053 1.0184 0.4418 1.0202 0.7539 0.3314AB2 0.6020 - - 0.4200 - - AB2 0.6412 0.5460 0.3679 0.7059 0.4971 0.3358AB3 1.4243 - - 0.2564 - - AB3 1.2734 1.0664 0.3858 0.5316 0.3782 0.3358AB4 2.7489 - - 0.6234 - - AB4 2.6668 2.3398 0.3550 1.0828 0.7979 0.3340D2 2.2553 - - 46.3317 - - D2 2.5987 3.7217 - 60.4434 75.6554 -

3

D3 3.0519 - - 46.7391 - -

7

D3 4.8967 5.9023 - 96.8357 65.5843 -PGATarget 0.2500 0.1250 - 0.2500 0.1250 -PGAReal 0.2313 0.1093 - 0.2637 0.1462 -A2 0.2513 0.1082 - 0.1227 0.1158 -A3 0.4282 0.1155 - 0.1883 0.2008 -AB1 0.4819 0.4807 - 0.3981 0.4484 -AB2 0.4267 0.2835 - 0.2722 0.3728 -AB3 0.9657 0.6177 - 0.2255 0.2739 -AB4 1.7932 1.3097 - 0.4328 0.6308 -D2 1.5519 1.6678 - 21.1447 21.5628 -

4

D3 2.8855 3.3082 - 22.5074 21.7414 -

Fixed Base Base Isolated Fixed Base Base IsolatedCaseNo.

Input andResponse x y z x y z

CaseNo.

Input andResponse x y z x y z

� Table 1. Fixed and FPS Isolated Results for Several Cases Using the Sylmar (Northridge) Record

Note:* For the base isolated case, the displacement limit of the FPS bearing is reached, causing impact.

34

5. Inspection of the displacementresults for this case indicates thatthe displacement capacity of thebearing has been reached causingimpact. Thus, the results are not use-ful for comparison to the fixed basesituation. However, they do high-light the fact that vertical motionhas a noticeable effect on the re-sponse of FPS isolated structures.In this case, vertical motion hascaused reduction in the horizontaldisplacements. Thus, for the 3-Dcase (Case 7 in Table 6), unlike the2-D case, the displacements arewithin the bearing capacity.

The response acceleration mapsfor one case is shown in Figure 4.In this figure, x-axis shows the ac-celeration values normalized withrespect to PGA. The y-axis showsdifferent locations along the heightof the test specimen ranging fromthe top of the shake table to thetop of the bushing. As one can seefrom this figure, acceleration re-sponse of the transformer model isreduced significantly at differentlevels throughout its height. The

level of acceleration reduction de-pends on the type of earthquakerecord used. That is, in addition toacceleration level and nature of theinput (e.g., 1-D vs. 2-D), the groundmotion characteristics affect thelevel of acceleration reductions.

The experimental results can besummarized as follow:• Inertia reductions depend on

peak ground acceleration (PGA)and bearing radius. The FPS sys-tem is more effective in reduc-ing inertia forces for higherPGAs. Furthermore, both inertiareductions and maximum dis-placements are affected by theearthquake record used.Records with dominant periodin the vicinity of the isolatorperiod reduce the isolator effec-tiveness.

• FPS bearings can provide, on theaverage, 60% acceleration reduc-tions within their displacementlimits. This number is with re-spect to the isolation level. Fora flexible system (as seen fromthe acceleration maps) accelera-tions are different at various lev-els, and the effectiveness of thebase-isolation is more apparentwhen one considers the entirepicture. For example, there is asignificantly greater reductionin the bushing acceleration thanthat of the transformer.

• Coupling of responses in twohorizontal directions does exist,which is due to dependency offrictional characteristics on totalvelocity. However, the effect tendsto diminish for higher PGAs, sinceat higher velocities, frictional con-stants are less sensitive to themagnitude of velocity.

• The vertical component ofground motion has an effect onthe response of the FPS bear-

Sylmar 3D-x

0 1 2 3 4 5 6 7 8 9

A/PGA

fps0250gx0125gy0125gz

fps0375gx0250gy0250gz

non0250gx0125gy0125gz

non0375gx0250gy0250gz

Shake Table

Bottom of Transformer

Top of Transformer

Middle of Bushing

Top of Bushing

FIXED BASE

FPS

ISOLATEDBushing topBushing mid.Trans. top

Trans. Bottom

Shake Table

fps0250gx0125gy0125gzfps0375gx0250gy0250gznon0250gx0125gy0125gznon0375gx0250gy0250gz

x

.

� Figure 4. Acceleration Maps: Triaxial Simulation, FPS Bearings, Sylmar(Northridge) Record


ings. This effect is expected tobe more pronounced for nearfield earthquakes (higher PGAs)and for sites where filtering ofthe motion due to local soil con-ditions is possible.

Hybrid Base-IsolationResults

It is difficult to use normal rub-ber bearings for seismic isolationof lightweight structures, such astransformers, due to their limitedability to elongate the natural pe-riod of the entire isolated systemwithout buckling. This difficultywas alleviated in this study by com-bining sliding bearings (to supportthe entire weight) with rubberbearings (to provide restoringforces). The hybrid system wastested under a transformer-bushingmodel, as discussed, and the resultswere compared to the fixed basesituation to investigate its effective-ness.

Four sliding bearings were in-stalled at the corners of the trans-former model and two rubberbearings were placed at the middleon two opposite sides, as shown inFigures 5. The sliding bearingscarry the entire weight of the trans-

former model and the bushing,while the rubber bearings providea horizontal restoring force with-out sustaining any vertical load.Detailed information on the designof the hybrid-isolation system andexperimental results can be foundin (Murota and Feng, 2001).

Peak response accelerationsalong with acceleration maps overthe height of the transformermodel and bushing, with and with-out the base isolation, are shownas a function of PGA in Figures 6and 7 for the El Centro groundmotion. Under other ground mo-tion records, the peak responsesshow a similar trend. Like the FPSbearings, the hybrid system is veryeffective in reducing the accelera-tions, especially in terms of the re-sponse of the bushing top. Withoutbase isolation, the peak accelera-tion at the top of the bushingreached 3.66 g, resulting in an am-plification factor of 10.80.

On the other hand, for the base-isolated case, the peak response ac-celeration at the top of bushing was0.354(g) with an amplification fac-tor equal to 1.05. Note that in thesediscussions, including those on theFPS bearings, the PGA referred toand shown on the correspondingfigures is the target PGA. The actual

S-W N-W

RB RB

SLB SLB

S-E N-E2200

y

xSLB: Sliding BearingRB: Rubber Bearing

1900

SLB SLB

“Researchefforts over thepast severalyears haverevealed thatunderstandingthe seismicinteractionamong keyequipments of asubstation iscritical toproperlyassessing theirseismicperformance.”

� Figure 5. Layout and Experimental Set-up of the Hybrid Sliding-Rubber Bearings

36

or real input acceleration may havebeen different due to difficulty inexactly matching the intendedPGA. Of course, response param-eters (such inertia reduction) arecalculated with respect to actualacceleration, not the target accel-eration.

The hybrid base isolation be-comes more effective as the PGAbecomes larger (Figure 7), whichis typical of a sliding isolation sys-

tem. It is observed that for the baseisolated system, the transformerresponse is not sensitive to theground motion. For the fixed-basedcase, the interaction between thetransformer (12.5 Hz) and the bush-ing was observed. As a result, theresponse of the 161kV bushing,with 12.5 Hz frequency, becamelarger than that of the 69kV bush-ing, which has a dynamic frequencyof 27 Hz.

Under triaxial shaking, the trans-former-bushing system showed sig-nificant difference in responsefrom those under uniaxial and bi-axial shaking. The response accel-erations at the tops of both 161kVand 69kV bushings for the isolatedcases were amplified, and in somecases (especially for the 69kV bush-ing), the response exceeded that ofthe fixed-based system. Figure 8shows the acceleration responsemaps (amplification factors) for theworst case.

Detailed inspection of the resultsindicates that this is due to the ef-fect of vertical motion on the fric-tion force acting on the slidingbearings. Vertical records are gen-erally rich in frequency content,resulting in high frequency fluctua-tions in the frictional force (20-30Hz), which in turn causes excita-tion of high frequency modes. Thus,the reason for the 69 kV bushing,which has a fundamental fre-quency around 27 Hz, to be moreaffected by triaxial excitation. Thisis valuable information that has notbe observed and/or investigatedbefore. It also has significant impli-cations for other structures (e.g.,response of secondary systems ina building within the framework ofperformance-based design) and itwill be investigated further.

0 2 4 6 8 10 120

1

2

3

4

5

6

7

Acc. / PGA

Is: x0.25gIs: x0.375gIs: x0.5gFb: x0.25gFb: x0.375g

Bushing top

Bushing mid

Trans. top

Trans. bottom

Platform

El Centro 1940 NS

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.1 0.2 0.3 0.4 0.5 0.6

PGA (g)

Maxim

um

Resp

on

se

Acc.(

g)

IsolatedTransformerBottom

Isolated:TransformerTop

Fixed Base:TransformerTop

Fixed Base:Bushing Top

� Figure 6. Acceleration Maps: Uniaxial Simulation, Hybrid Bearings, El-Centro Record.

� Figure 7. Peak Response Accelerations Under Uniaxial Shaking


In summary, the proposed hybridsliding-rubber bearing isolation sys-tem is quite effective in reducingthe response acceleration of atransformer-bushing system underuniaxial and biaxial earthquakesimulator tests. It should be notedthat the seismic performance of anactual transformer-bushing systemequipped with the proposed isola-tion system will be even better,because the isolation period of anactual transformer will be muchlonger than that of the transformer-bushing model used in the test dueto a much heavier weight of theactual transformer. Among ongoingobjectives of the project are nu-merical studies and modificationsto the design of the hybrid systemto improve its effectiveness undertriaxial motion.

Analytical Study

SDOF Model

Over the past decade or so, therehave been several analytical andexperimental studies on the seis-mic performance of FPS isolators.These works proved the effective-ness of FPS in reducing inertiaforces within the displacementlimit of the device. Practical appli-cations of the system in the designof new structures and the rehabili-tation of existing ones, includingthe historic Ninth Circuit U.S. Courtof Appeals in San Francisco (Mokha,et al., 1996), have taken place. Amore recent work by Almazan, etal., (1998) addresses several otherimportant aspects of modeling andresponse such as constitutive rela-tionships (small vs. large displace-ment), and refinement of thestructural model.

During the initial phase of thisstudy and parallel with the shaketable tests, an extensive analyticalwork was conducted. Using SDOFidealization and equilibrium of theforces involved, the differentialequations of motion were estab-lished and solved using IMSL rou-tine IVPAG (Ersoy et al., 2001).Although the focus of this study ison application of FPS devices totransformers, it also addresses ad-ditional parameters and aspects ofresponse that either have not beeninvestigated before or have onlybeen considered on a limited ba-sis. Therefore, some of the findingsare general and could have appli-cations in the design of FPS bear-ings for other structures as well.Among the parameters consideredare ground motion characteristics,bi-directional motions, the effect ofvertical motion, and isolation ra-dius.

Inertia reduction and the maxi-mum displacement of the systemwere the criteria used in evaluat-ing the seismic response and theeffectiveness of FPS bearings. Basedon the results of the parametric

0 2 4 6 80

1

2

3

4

5

6

7x-dir.

Acc. / PGA

Is:x0.25gy0.125gz0.125g

Is:x0.375gy0.25gz0.25gFb:x0.25gy0.125gz0.125gFb:x0.375gy0.25gz0.25g

Bushing top

Bushing mid

Trans. top

Trans. bottom

Platform

� Figure 8. Acceleration Maps: Triaxial Simulation, Hybrid Bearings, El-Centro Record

38

study, charts for inertia reductionand the maximum displacementwere developed. Shown in Figure9 is a chart corresponding toground motion with peak accelera-tion of 1.0g. The challenge in de-sign will be the selection of radiusof curvature of the bearing suchthat there is a balance between thedesired inertia reduction and thedisplacement limits vis-à-vis bush-ing interaction with interconnect-ing equipments.

As can be seen from figure 9, thebearings can provide large inertiareductions. For example, for a sys-tem with radius of curvature equalto 1.5 m (60" inches) the inertiareduction is about 60%. That is, thetransformer acceleration will be0.4g when the peak ground accel-eration is 1.g.

For this radius, the system periodis 2.5 sec (i.e., frequency of 0.4 Hz)regardless of the weight of thetransformer. However, the associ-ated large displacement needs to beaccommodated. To this end, thereis a need for a simplified 3-D modelto investigate the interaction be-tween the transformer-bushing and

interconnecting elements. To de-velop the knowledge base toachieve this goal, finite elementmodels of typical transformers andbushings were developed, as dis-cussed below, for time historyanalysis.

Finite Element Model

A power transformer is com-posed of six parts: transformer tank,radiators, reservoir, core and coil,oil, and bushings. The transformertank is the main structural compo-nent of power transformers. It hascore and coil centrally placedwithin it and the tank is completelyfilled by mineral oil. Radiators andreservoir are appendages and theyare externally attached to the trans-former tank.

A finite element model of apower transformer (55 MVA, 230/135 kV) from a substation in NewJersey is shown in Figure 10. Thetransformer weighs 1,335 kN (300kips), and the radiators (on theside) and reservoir (on the top)weigh 120 kN and 40 kN, respec-tively. Thus, one can see the possi-bility of torsional response andhigher demands on the supports inlight of the relative weights of theappendages compared to theweight of the transformer itself.

The finite element packageANSYS was used to develop the fi-nite element models. The trans-former tank was modeled by shellelements. Braces around the trans-former were modeled by offsetbeam elements. Currently, the coreand coil inside the transformerwere modeled as mass elements.Radiators and reservoir were mod-eled by 3-D solid elements. The con-tained oil inside the transformer

300

325

350

375

400

425

450

475

500

525

550

575

600

0.5 0.8 1.0 1.3 1.5 1.8 2.0 2.3 2.5 2.8 3.0 3.3 3.5 3.8 4.0

Radius of Curvature of Bearing (m)

Dis

pla

cem

ent(m

m)

35

40

45

50

55

60

65

70

75

80

Inert

iaR

eduction

(%)

displ.

inert. red.

PGA = 1.00g

f=0.10

Radius of Curvature of Bearing (m)

Dis

plac

emen

t (m

m)

Iner

tia R

educ

tion

(%)

disp.inert. red.

� Figure 9. Analytical Displacement-Inertia Reduction Chart for FPS System(input PGA = 1.0 g)


was modeled as solid with modu-lus of elasticity equal to the bulkmodulus of the fluid.

The seismic response of bushingswas dominated by the behavior ofthe gaskets between the porcelainunits. The common failure mode in-volved movement of the upper por-celain unit relative to its supportflange, causing oil leakage Therefore,the analytical model for the bushingsuses simple beam elements withequivalent density and stiffness torepresent porcelain units, dome, andaluminum support.

Gaskets between these elementswere modeled using nonlinearaxial and shear springs. For a fixedtransformer, the translational de-grees of freedom were removed atthe location of the supports. Thesoil-structure interaction may be in-vestigated using Winkler founda-tion elements to evaluate the levelof stresses in the subgrade. Cur-rently, time history analyses usingvarious earthquake records are be-ing conducted.

ConclusionsResearch efforts over the past

several years have revealed that un-derstanding the seismic interac-tions among key equipments of asubstation (transformers, bushings,foundation, and interconnecting el-ements) is critical to conducting aproper assessment of their seismicperformance. Thus, the thrust of fu-ture efforts will be to evaluate seis-mic response, and propose designand rehabilitation guidelines, basedon system performance. This willbe achieved through further ex-perimental tests and by developinga simplified model that can accu-rately represent critical elements ofa substation in order to investigatetheir interactions in detail througha parametric study. The model willbe developed based on the resultsof ongoing 3-D finite element analy-ses and the experimental resultsconducted over the past severalyears.

� Figure 10. Finite Element Model of a Transformer-Bushing System

40

Almazan, J. L., De La Llera, J. C., and Inaudi, J., (1998), "Modeling Aspects of StructuresIsolated with the Frictional Pendulum System," Earthquake Engineering andStructural Dynamics, 27, 845-867.

ASCE Manuals and Reports on Engineering Practice No. 96, (1999), Guide to ImprovedEarthquake Performance of Electric Power Systems, Edited by A. J. Schiff, ASCE,Virginia.

EERI Newsletter, (1999), "The Izmit (Kocaeli), Turkey Earthquake of August 17, 1999,"Vol. 33, Number 10, October.

Ersoy, S., Saadeghvaziri,M. Ala, Liu, G. Y., and Mau, S.T., (2001), "Analytical andExperimental Evaluation of Friction Pendulum System for Seismic Isolation of Trans-formers," Earthquake Spectra, Journal of Earthquake Engineering Research Institute,(tentatively accepted).

IEEE, (1998), IEEE Std. 693-1997, Recommended Practices for Seismic Design ofSubstations, Piscataway, NJ, IEEE Standards Department.

Mokha, A., Amin, N., Constantinou, M., and Zayas, V., (1996), "Seismic Isolation Retrofitof Large Historic Buildings," J. Structural Engineering, ASCE, Vol. 122, pp. 298-308.

Murota, N., and Feng, M.Q., (2001), "Hybrid Base-Isolation of Bushing-TransformerSystems," Proceedings, 2001 Structures Congress, ASCE, Washington, DC, May 21-23.

Saadeghvaziri, M. Ala, and Ersoy, S., (2001), "Evaluation of Seismic Response ofTransformers and Effectiveness of FPS Bearings for Base-Isolation," Proceedings, 2001Structures Congress, ASCE, Washington, DC, May 21-23.

References

41

Research Objectives

The ATC/MCEER Joint Venture is developing new specifications for theseismic design of highway bridges that can be recommended for incorpora-tion into the AASHTO LRFD Bridge Design Specifications. The recommendedspecifications will be performance-based and address state-of-the-art aspectsof highway bridge seismic design, including the latest approaches for repre-sentation of seismic hazard, design and performance criteria, improved analy-sis methods, steel and concrete superstructure and substructure design anddetailing, and foundation design.

Recommended Changes to the AASHTOSpecifications for the Seismic Design of HighwayBridges (NCHRP Project 12-49)

by Ian M. Friedland, Ronald L. Mayes and Michel Bruneau

In August 1998, a joint venture of the Applied Technology Council (ATC)and MCEER initiated work on a project to develop the next generation

of seismic design specifications for highway bridges in the United States.The project is sponsored by the American Association of State Highwayand Transportation Officials (AASHTO) and is being conducted by the Na-tional Cooperative Highway Research Program (NCHRP) of the Transpor-tation Research Board. NCHRP Project 12-49, "ComprehensiveSpecifications for the Seismic Design of Bridges," will result in the devel-opment of specifications and commentary which are expected to be in-corporated into the AASHTO LRFD Bridge Design Specifications. Thesewill be supplemented by a series of design examples demonstrating theapplication of key features of the new specifications.

The recommended specifications and commentary are currently beingassessed by the AASHTO Highway Subcommittee on Bridges and Struc-tures. The AASHTO Subcommittee is expected to make a decision regard-ing their adoption in May 2001. During the course of the project, threedrafts of specifications and commentary will have been produced and re-viewed, prior to completion and submission of the final draft to AASHTO.

The new specifications are expected to incorporate all current and state-of-the-art practices in highway bridge seismic design, and will be perfor-mance-based. They will address the latest approaches for representationof seismic hazard, design and performance criteria, analysis methods, steeland concrete superstructure and substructure design and detailing, foun-dation design, and soil behavior and properties. The specifications arealso intended to address the differences in seismic hazard, soils, and bridgeconstruction types found throughout the United States, and therefore are

National Cooperative HighwayResearch Program of theNational ResearchCouncil’s TransportationResearch Board

Applied Technology Council(ATC)/MCEER Joint Venture

Ian M. Friedland, AppliedTechnology Council (P.I)

Christopher Rojah, AppliedTechnology Council(Administrative Officer)

Ronald Mayes, consultant(Technical Director)

Donald Anderson, CH2MHill, Inc.

Michel Bruneau, Universityat Buffalo

Gregory Fenves, Universityof California at Berkeley

John Kulicki, Modjeski andMasters, Inc.

John Mander, University ofCanterbury

Lee Marsh, BERGER/ABAMConsultants

Geoffrey Martin, Universityof Southern California

Andrzej Nowak, Universityof Michigan

Richard V. Nutt, consultantMaurice Power, Geomatrix

ConsultantsAndrei Reinhorn, University

at Buffalo

42

If the seismic design provisions being developed underthe NCHRP 12-49 project are adopted by AASHTO in wholeor in part, they will become the national standard underwhich all highway bridges throughout the United States aredesigned. In addition, AASHTO specifications also often be-come the basis for the specifications adopted and used inmany foreign countries. As a result, the users of this re-search will be the practicing bridge engineering commu-nity nationwide.

to be national in scope. These newspecifications will be a marked de-parture from the design philosophy,approach, and requirements cur-rently in use in the United States.

Technical SummaryA number of important changes

in philosophy and approach will beincluded in these new specifica-tions, when compared to the cur-rent AASHTO highway bridgeseismic design provisions. Includedin these are the following:

Design Criteria

A performance-based design cri-teria has been included in the newspecifications and commentary(see Figure 1). Specifically, theproject recommends adoption of adual-level performance criteria, andtwo levels of seismic performanceobjectives, as shown in Table 1.

The recommended seismic per-formance objectives are perfor-mance-based and each Statedetermines the desired perfor-mance objective for any particularbridge. This is a change from cur-rent AASHTO definitions of "other,"

"critical," and "essential" bridges fordefining performance levels.

The upper level earthquake con-sidered in these provisions is des-ignated the Maximum ConsideredEarthquake (MCE). In general, theMCE ground motions have a prob-ability of exceedance of 3% in 75years, which is an approximate re-turn period of 2,500 years. How-ever, MCE ground motions arebounded deterministically to valueslower than 2,500-year ground mo-tions adjacent to highly active faults.It should be noted that the 75 yearbasis is equivalent to the theoreti-cal design life of a highway bridge,as specified in the AASHTO provi-sions.

The definitions of service levelsas shown in Table 1 are as follows:

Immediate–Full access to nor-mal traffic shall be available follow-ing an inspection of the bridge.

Significant Disruption – Limitedaccess (reduced lanes, light emer-gency traffic) may be possible aftershoring; however the bridge mayneed to be replaced.

The damage level definitionsshown in Table 1 are as follows:

None – Evidence of movementmay be present but there is no no-table damage.

Project Engineering PanelIan G. Buckle, University of

Nevada at Reno (chair)Serafim Arzoumanidis,

Steinman BoyntonGronquist & Birdsall

Mark Capron, Sverdrup Civil,Inc.

I. Po. Lam, Earth Mechanics,Inc.

Paul Liles, GeorgiaDepartment ofTransportation

Brian Maroney, CaliforniaDepartment ofTransportation

Joseph Nicoletti, URSGreiner Woodward Clyde

Charles Roeder, University ofWashington

Frieder Seible, University ofCalifornia at San Diego

Theodore Zoli, HNTBCorporation

43Recommended Changes to the AASHTO Specifications for the Seismic Design of Highway Bridges

Minimal – Some visible signs ofdamage. Minor inelastic responsemay occur, but post-earthquakedamage is limited to narrow flex-ural cracking in concrete and theonset of yielding in steel. Perma-nent deformations are not apparent,and any repairs could be made un-der non-emergency conditionswith the exception of superstruc-ture joints.

Significant – Although there is nocollapse, permanent offsets mayoccur and damage consisting ofcracking, reinforcement yield, andmajor spalling of concrete and steelyielding and local buckling of steelcolumns, global and local bucklingof steel braces, and cracking in thebridge deck slab at shear studs onthe seismic load path is possible.These conditions may require clo-sure to repair the damage. Partialor complete replacement may berequired in some cases.

Geometric and StructuralConstraints

In the initial phases of thisproject, an attempt was made to

develop a set of geotechnical per-formance objectives that would besimilar to those being developed forconcrete columns. A two-day work-shop was held to review initial draftproposals and to refine recom-mended criteria and approaches.The consensus of the workshopwas that the amount of acceptablefoundation and abutment move-ment should be related to geomet-ric and structural constraints bybridge type, rather than explicitvalues on foundation movements(see Figure 2). As a result, the rec-ommended specifications proposeconstraints that would implicitlyprovide foundation design limits forseismic loads to meet the variousperformance objectives. Since thisis the first time an attempt has beenmade at developing these con-straints, the specified values may

Performance ObjectiveProbability ofExceedance

for Design EarthquakeGround Motions

Life Safety Operational

Service Significant Disruption ImmediateRare Earthquake3% in 75 years Damage Significant Minimal

Service Immediate ImmediateExpected Earthquake50% in 75 Years Damage Minimal Minimal to None

DESIGNAPPROACHES

MinimalDamage

ModerateDamage

SignificantDamage

ConventionalDuctile Design

With LowR-factor(R<1.5)

SeismicIsolation(R<1.5)

ConventionalDuctile

Design withMediumR-Factor1.5<R<3

Control andRepairability

EnergyDissipation

ConventionalDuctile

Design withHigh

R-FactorR>3

SP

EC

IFIC

ATIO

NS

CO

MM

EN

TAR

Y

� Figure 1. Design Approaches

� Table 1. Performance Objectives

44

require additional work in the fu-ture to refine them.

Geometric constraints generallyrelate to the usability of the bridgeby traffic passing on or under it.Therefore, such constraints will usu-ally apply to permanent displace-ments that occur as a result of theearthquake. The ability to repair (orthe desire not to be required to re-pair) such displacements should beconsidered when establishing dis-placement capacities. When imme-diate service is desired, permanentdisplacements should be small ornon-existent, and should be at lev-els that are within accepted toler-ances for normal highwayoperations. When limited serviceis acceptable, the geometric con-straints may be relaxed. These maybe governed by the geometry ortypes of vehicles that will be usingthe bridge after an earthquake, andby the ability of these vehicles topass through the geometric ob-struction. Alternately, a jurisdictionmay simply wish to limit displace-ments to a multiple of those al-

lowed for immediate service. In thecase of a significant disruption,post-earthquake use of the bridgeis not guaranteed and therefore nogeometric constraints would berequired to achieve these goals.However, because life safety is at theheart of this performance level, ju-risdictions applying the provisionsshould consider establishing somegeometric displacement limits.

Structural constraints on displace-ments can be based on the require-ments of a number of structuralelements and can result from eithertransient displacements due toground shaking or permanent dis-placements resulting from groundmovement due to faulting,seismically induced settlements, lat-eral spreading, and so forth. Struc-tural damage to foundationelements is limited primarily topiles since footings and pile capsare usually capacity protected. Al-though pile damage can often beavoided at a reasonable cost whenpiles are capacity protected, requir-ing this for all piles could lead to

• The specifications beingdeveloped under thisproject draw heavily on theresults of many researchstudies conducted withinthe MCEER HighwayProject, and on relatedNSF-sponsored research,especially in the areas ofseismic hazardrepresentation andgeotechnical/soilsperformance and response.

Permanent DisplacementType

Possible Causes Mitigation Measures Immediate SignificantDisruption

Vertical Offset

Vertical Grade Break (2)

Horizontal Alignment Offset

Approach fillsettlementBearing failure

Interior supportsettlementBearing failureApproach slabsettlement

Bearing failureShear key failureAbutment foundationfailure

Approach slabsApproach fillstabilizationBearing type selection

Strengthen foundationBearing type selectionLonger approach slab

Bearing type selectionStrengthen shear keyStrengthen foundation

0.083 feet(0.03 meters)

Use AASHTO“Green Book”requirementsto estimateallowable

grade break

0.33 feet(0.1 meters)

Joint seal mayfail

0.83 feet(0.2 meters)To avoidvehicle impact

None

ShoulderWidth

(To avoidvehicle impact)

G G

G

1 2

� Figure 2. Geometric Constraints on Service Level (partial listing, taken from Table C3.10.1.2,Revised LRFD Design Specifications, Third Draft of Specifications and Commentary)


overly conservative, and thus ex-pensive, foundation designs. There-fore, it may be desirable to establishsome lateral displacement limits forpiles based on a limited amount ofstructural damage that is unlikelyto compromise the structural integ-rity of the bridge. These limits willbe based on the type and size ofthe piling, and the nature of the soilnear the head of the pile. Caltranshas attempted to establish such lim-its for its standard piling based onphysical testing. Other jurisdictionsmay attempt to do the same, or theymay perform more complex analyti-cal studies to establish similar dis-placement, capacity, and stiffnesslimits.

Earthquake Return Periods

Current AASHTO specificationsconsider a single-level earthquakehazard design event, based on a500-year return period. At the timethis event was incorporated intothe AASHTO specifications, it wasthe only event readily available as adesign value for seismic hazard rep-resentation in the United States.

The new provisions recommenda 2,500-year return period, whichprovides an equivalent 3% probabil-ity of exceedance in 75 years, beused as the upper level designevent. A lower level event with anapproximate 100-year return pe-riod, which would be based on a50% probability of exceedence in75 years, has also been recom-mended. In brief, the reasons forthese recommended return periodsinclude:• For a frequent or expected

earthquake (50% probability ofexceedance in 75 years), this isthe design level under which thebridge should remain essentially

elastic (R = 1.0). This will resultin a performance that is equiva-lent to an elastic (no damage)design for a 100 year flood, how-ever, is less conservative butsimilar to an elastic design for a100 mph base wind design.

• For a rare earthquake (3% prob-ability of exceedance in 75years), this design level is recom-mended in order to assure thata "no collapse" performance cri-teria for the MCE is satisfied.Since seismic loads increasemuch more significantly com-pared to wind and flood loadsas the return period increases,earthquake design needs to con-sider longer return period de-sign events.

A critical earthquake design issuein preventing collapse of a bridgeis to ensure that deck displace-ments, and hence seat widths forthe girders, can accommodate dis-placements from events that haveoccurred and can be realisticallyexpected to reoccur. A return pe-riod of 500 years does not comeclose to capturing the force and dis-placement levels that may have oc-curred during earthquakes in theNew Madrid region (1811-1812)and Charleston, South Carolina(1886). A return period of the or-der of 2,500 years is therefore re-quired to obtain realisticdisplacement levels that may haveoccurred during these earthquakes.

It is noted that the MCE groundmotion maps incorporate determin-istic bounds on the ground motionsadjacent to highly active faults.These bounds are currently appli-cable in California, along a portionof the California-Nevada border, incoastal Oregon and Washington,and in parts of Alaska and Hawaii.These deterministic bounds are

46

applied so that the ground motionsdo not become unreasonably largein comparison to the ground mo-tions that could be produced bymaximum magnitude earthquakeon the faults. Deterministic boundsare defined as 150% of the medianestimates of ground motions calcu-lated using appropriate attenuationrelationships, and assuming the oc-currence of a maximum magnitudeearthquake on the fault. However,they are limited to not less than 1.5 gfor the short-period spectral accelera-tion plateau and 0.6 g for 1.0-secondspectral accelerations.

Design Spectra Shape

The long period portion of the cur-rent AASHTO acceleration responsespectra is governed by the spectrumshape and the soil factor, and it de-cays as 1/T2/3. There was consider-able "massaging" of the factors thataffect the long period portion of thecurrent AASHTO spectra in order toproduce a level of approximately 50%

conservatism in the design spectrawhen compared to the ground mo-tion response spectra beyond 1 sec-ond period.

Analysis of ground motion data in-dicates that the acceleration spectragenerally decreases with period inthe long period range as 1/T or morerapidly. The shape of the long pe-riod portion of the recommended re-sponse spectra is less conservativethan the current spectral shape andit is recommended that the shape ofthe spectra decay as 1/T.

For periods longer than about 3seconds, depending on the seismicenvironment, use of the 1/T relation-ship for spectral acceleration may beconservative because the groundmotions may be approaching theconstant spectral displacement re-gion for which spectral accelerationsdecay as 1/T2. Either the 1/T rela-tionship may be conservatively usedor a site-specific study conducted todetermine an appropriate long-pe-riod spectral decay.

Construction of the design re-sponse spectra requires the spectralacceleration value at 0.2 seconds and1.0 second. The base curve con-structed with these values is thenmodified according to the 1994NEHRP short- (F

a) and long-period

(Fv) soil site factors, as shown in Fig-

ure 3.

Design Procedures andResponse Modification Factors

The recommended specifica-tions provide for three levels of de-sign and analysis procedures in themoderate-to-high seismic zones.This is in addition to the current"no seismic" or low-level minimumrequirements, which are defined asLevel 0:

Sar(1.0)T

Spec

tral

Acc

eler

atio

n, S

a

Period, T (sec)

Fa x Sar(0.2)

To = 0.2Ts

0 10.2 Ts 2 3 4

Sar(0.2)

Fv x Sar(1.0)T

Sar(1.0)

To

Sa=0.4Fa x Sar(0.2)

Note: Sar is spectral acceleration on rock Type B

Rock (Type B)response spectrum

Fv x Sar(1.0)Fa x Sar(0.2)Ts =

Soil responsespectrum

� Figure 3. Response Spectrum Construction


Level 0 – These are the cur-rent AASHTO Zone 1 provi-sions, which requireminimum seat widths andspecified design forces forfixed bearings of 10% deadload for Zone 1A and 25%dead load for Zone 1B.

Level 1 – This requires noformal seismic analysis butrequires the use of capacitydesign principles and mini-mum design details.

Level 2A – This is a one-stepdesign procedure based on ananalysis method referred to asthe "capacity spectrummethod" and is applicable tovery "regular" bridges. Thismethod has been incorpo-rated in some of the retrofitguidelines for buildings and pro-vides the designer with the abilityto assess whether or not a bridgedesigned for all non-seismic loadshas sufficient strength and displace-ment capacity to resist the seismicloads.

Level 2B – This is a one-step de-sign procedure based on an elastic(cracked section properties) analy-sis using either the Uniform Loador Multimode method of analysis.The analysis is performed for thegoverning design spectra (either50%/75-year or 3%/75- year event)and the use of a conservative R-fac-tor. The analysis will determine themoment demand at all plastic hingelocations in the column. Capacitydesign principles govern founda-tion and column shear design. Ifsacrificial elements are part of thedesign (i.e., shear keys) they mustbe sized to resist the 50%/75-yearforces, and the bridge must be ca-pable of resisting the 3%/75-yearforces without the sacrificial ele-ments (i.e., two analyses are re-

quired if sacrificial elements existin a bridge).

Level 3 – This is a two-step designprocedure using an elastic (crackedsection properties) analysis withthe multimode method of analysisfor the governing design spectra toperform preliminary sizing of themoment capacity of the columns.Capacity design principles governfoundation and column shear de-sign. A pushover analysis is thenperformed. The designer canchange the design forces in the col-umns provided they are not low-ered below those required by the50%/75-year event and that theysatisfy the displacement demand.

These provisions attempt to makethe Level 1 approach applicable toas wide a range of bridges as pos-sible. Level 2A will be applicableto very "regular" bridges that canbe idealized as Single-Degree-of-Freedom systems, until more devel-opment work is done to extend themethod.

Substructure Element Performance ObjectiveLife Safety Operational

SDAPD

SDAPE

SDAPD

SDAPE

Wall Piers-larger dimension 2 3 1 1.5Columns – Single and Multiple 4 6 1.5 2.5Pile Bents and Drilled Shafts – VerticalPiles – above ground 4 6 1.5 1.5

Pile Bents and Drilled Shafts – VerticalPiles – 2 diameters below ground 1 1.5 1 1

Pile Bents and Drilled Shafts – VerticalPiles – in ground n/a 2.5 n/a 1.5

Pile Bents with Batter Piles n/a 2 n/a 1.5Seismically Isolated Structures 1.5 1.5 1 1.5Steel Braced Frame – DuctileComponents 3 4.5 1 1.5

Steel Brace Frame – Nominally DuctileComponents 1.5 2 1 1

All Elements for Expected Earthquake 1.3 1.3 0.9 0.9Connections (superstructure toabutment; joints within superstructure;column to cap beam; column tofoundation)

0.8 0.8 0.8 0.8

� Table 2. Typical Response Modification Factors

48

Base response modification (R-factors) are significantly more lib-eral and comprehensive in the newprovisions than in current AASHTO.These values, as shown in Table 2,reflect the higher levels of seismicdemand and analyses required bythe recommended provisions. Theyalso require equivalently elasticanalysis for the 50%/75-year event.R-factors for all connections requireessentially elastic behavior, butshould not be used where capac-ity design principles are used to de-sign the connections.

Level 1, No Seismic DemandAnalysis

The "no seismic demand analy-sis" design procedures are an ex-tremely important part of therecommendations because they areexpected to apply as widely as pos-sible in Zone 2. The purpose of theprovisions is to provide the bridgedesigner the ability to design struc-tures without the need to under-take dynamic analyses. The bridgeelements are designed in accor-dance with the usual provisions,but the primary seismic resisting el-ements are specifically detailed forseismic resistance.

Capacity design requirementsexist for shear reinforcement andconfining reinforcement at plastichinge locations in columns. There

are also capacity design require-ments for detailing the connectionto pile and column caps. There areno additional design requirementsfor abutments or foundations withone exception – integral abutmentsmust be designed for passive pres-sure. The premise for this is basedon a parameter study that was con-ducted under the project whichdemonstrated that a foundation de-signed using a factor of safety of 3will perform well in these lowerseismic zones, assuming that thecolumn connection to the pile capis based on capacity design prin-ciples. These procedures will bepermitted for use in sites with liq-uefaction potential, but will requiresome additional considerations ifthe predominant moment magni-tude, MB, for the site exceeds 6.

The limitations on the applicabil-ity of the "no seismic demand analy-sis" provisions started with thedefinition of a "regular bridge" ascontained in the current AASHTOseismic design provisions. Addi-tional limits were added based onengineering judgment, and theseshould be reevaluated in the future.The limits as they currently existconsider (a) regularity, (b) axial loadlimits, (c) load path and force shar-ing, and (d) continuity. Table 3 pro-vides the definition of a "regular"bridge for these provisions.

Span-to-span continuity must beprovided by one or more of the fol-lowing means:• A superstructure that is mono-

lithic with the substructure• A superstructure seated on bear-

ings that has transverse restraint• Simply supported girders seated

on bearings must be fixed at onesupport end, and transverselyrestrained at the other

“If the seismicdesignprovisionsbeing developedunder thisproject areadopted byAASHTO, inwhole or inpart, they willbecome thenationalstandard underwhich allhighwaybridges in theU.S. aredesigned.”

Parameter ValueNumber of Spans 2 3 4 5 6

Maximum subtended angle forcurved bridge 20 20 30 30 30

Maximum span length ratio, fromspan to span 3 2 2 1.5 1.5

Maximum bent/pier stiffness ratiofrom span to span, excluding

abutments– 4 4 3 2

� Table 3. Parameter Definitions for “Regular” Bridges


• A superstructure supported onisolation bearings that act in alldirections that accommodatedisplacements

Foundations

Provisions for spread footings,driven piles, drilled shafts (or cast-in-drilled-hole piles), and abutmentshave been significantly improvedover similar provisions in the cur-rent AASHTO specifications. Theseare, in large part, based on a num-ber of advances made through re-search programs sponsoredrecently by the Federal Highway Ad-ministration, Caltrans, and others.For abutments, explicit recognitionof sacrificial elements has been pro-vided, including knock-off backwalls and other similar fuse ele-ments.

Steel Bridges

The current AASHTO Specifi-cations do not have seismic re-quirements for steel bridges,except for the provision of a con-tinuous load path to be identi-fied and designed (for strength)by the engineer. Consequently,within the scope of this projectsubtask, a comprehensive set ofspecial detailing requirementsfor steel components expectedto yield and dissipate energy ina stable and ductile manner dur-ing earthquakes were developed(including provisions for ductilemoment-resisting frame sub-structures, concentrically-bracedframe substructures, and end-dia-phragms for steel girder andtruss superstructures).

In July 2000, MCEER hosted aworkshop to review a draft of the

proposed steel seismic provisionsdeveloped as part of this project.The workshop's participants con-sisted of representatives fromacademia, professional practice,State DOT’s, FHWA, and the steelindustry, with experience in seismicdesign and steel bridges. The objec-tive of the workshop was to iden-tify whether the proposedprovisions were sufficiently com-prehensive and answered needsidentified in past projects withworkable solutions.

The proposed seismic provisionsfor steel bridges were generally wellreceived. It was also the consensusopinion that seismic provisions forsteel structures were desirable, andthat there may even be instanceswhere steel substructures may bethe preferred choice for seismicenergy dissipation (particularlywhen combined with other advan-tages such as rapid construction).The final draft of the proposed

R/C Slab

R/C Slab

R/C Slab

TADAS

a. Eccentrically Braced Frame(EBF) Ductile Diaphragms

b. Shear Panel System (SPS)Ductile Diaphragms

c. Steel Triangular-Plate Added Damping andStiffness (TADAS) Device Ductile Diagrams

� Figure 4. Structural Model - Steel Structures

50

Specifications reflects much of theexpert opinions expressed duringthis workshop. It is also structuredto allow the use of innovative steel-based energy dissipation strategies(as shown in figure 4), providedseismic performance can be experi-mentally substantiated.

Special ConsiderationsRegarding Seismic andGeotechnical Hazards

A number of special provisionshave been included to address im-portant considerations in seismichazard. Included in these are speci-fications related to vertical accelera-tions and their impacts on bridgeperformance, and near fault (and so-called "fling") effects. A significantamount of work is also being done

to address the issues of soil lique-faction and lateral spreading, espe-cially as this relates to theinfrequent but potentially largeupper-level 2,500-year event.

SummaryThe new specifications for the

seismic design of highway bridgesthat have been developed underNCHRP Project 12-49 are a signifi-cant departure from the proceduresand requirements found in currentAASHTO specifications. Work onNCHRP Project 12-49 was com-pleted in the spring of 2001, andthe AASHTO Highway Subcommit-tee on Bridges and Structures willreview the provisions resultingfrom the project, with possibleadoption and publication byAASHTO in late 2001.

51

Research Objectives

Despite the use of seismic isolation for the protection of bridges formore than 20 years, the performance of these bridges during strong groundshaking remains to be verified in the field. Numerous analytical and ex-perimental studies have been completed in many countries to date dem-onstrating the validity of the concept, but full-scale field verification understrong shaking has yet to be obtained. This is partly due to the fact that theworld population of these bridges is relatively small, but in addition, onlya small fraction of this population is instrumented with strong motioninstruments. However, recent earthquakes (even though small-to-moder-ate in size) have given some insight into the actual performance ofseismically isolated bridges, and the experience-database for these bridgesis growing. This research focuses on assembling currently available infor-mation on observed performance of isolated bridges and identifying ar-eas where further research may be necessary to improve the state-of-the-art.This paper is a summary of a comprehensive literature review recentlyundertaken on observed performance in recent earthquakes. Based oncollected information, possible future research needs are indicated.

Literature Review of the ObservedPerformance of Seismically IsolatedBridges

by George C. Lee, Yasuo Kitane and Ian G. Buckle

The fundamental concept of seismic isolation is straightforward: to re-duce earthquake-induced forces in a structure by lengthening its natu-

ral period, or by adding damping to the structure, or both. Most isolationsystems fall into this last category, i.e., they both lengthen the period andadd damping.

Although the concept of protecting structures from an earthquake, byintroducing seismic isolators, was first proposed almost 100 years ago, itis only recently that seismic isolation has become a practical strategy forearthquake-resistant design (Chopra,1995). In general, seismic isolatorshave the following four major functions (Skinner, 1993; Kelly, 1997):• To transmit vertical load (e.g., dead load) from one part of a bridge to

another, usually from the superstructure to the substructure, while al-lowing thermal and other movement effects to occur (i.e., to performthe same role as a conventional bearing).

Federal HighwayAdministration

Ian G. Buckle, Professor,Department of CivilEngineering, University ofNevada, Reno

Yasuo Kitane, GraduateStudent, Department ofCivil, Structural andEnvironmentalEngineering, University atBuffalo

George C. Lee, Director,Multidisciplinary Centerfor EarthquakeEngineering Research, andSamuel P. Capen Professorof Engineering, Universityat Buffalo

52

• To isolate the part of the bridgeabove the isolators by introduc-ing flexibility in the horizontalplane, or by limiting the horizon-tal shear force that may be trans-mitted to the isolated part.

• To provide sufficient rigidityunder low level loads such aswind and traffic loads, and mi-nor earthquakes.

• To introduce additional damp-ing into the bridge system, sothat relative displacementsacross the isolators can be con-trolled. In some cases, dampingis provided directly by the isola-tors; in others, additional devicesare installed alongside the isola-tors to provide this additionaldamping, such as viscous fluiddampers.

The most common isolators inuse today in the United States fallinto one of two categories: elasto-meric-based (e.g., lead-rubber bear-ings and high-damping rubberbearings), and friction-based (e.g.,Eradiquake bearings and friction-pendulum bearings).

In the 1970s, seismic isolationsystems began to be implementedin bridges as aseismic devices. Thetechnology spread quickly for thenext 30 years, and many applica-tions can now be found around theworld and particular in Canada,

Italy, Iceland, Japan, New Zealand,Taiwan, and the U.S. (See, for ex-ample, list of Seismically IsolatedBridges in the United States at http://www.eerc.berkeley.edu/prosys/usbridges.html.) Extensive researchand development efforts have madethis rapid technology transfer pos-sible. Numerous experiments in thelaboratory and computer analyseshave been performed to investigatethe effectiveness of the isolation tech-nique, and have demonstrated (intheory) the advantages of isolationdesign over conventional earthquakeresistant design.

However, the number ofseismically isolated bridges thathave experienced actual earth-quakes is very low, and the numberof instrumented isolated bridges,that have seen earthquakes, is evensmaller. As a consequence, there isvery little recorded data on the re-sponse of these bridges and the ef-fectiveness of isolation remains tobe conclusively proven in the field,particularly for large earthquakes.

Nevertheless, collecting and syn-thesizing the performance data thatis available is a worthy exercisebecause it gives insight on fieldexperience as well as indicatingwhere future research efforts mightbe directed. This paper is the resultof such an exercise. A comprehen-

List of Seismically IsolatedBridges in the United States:http://www.eerc.berkeley.edu/prosys/usbridges.html

Engineers who design bridges in seismically active areasthroughout the world will find information on the responseof isolated bridges during earthquakes very valuable. Bridgesthat have both experienced actual earthquakes and have beeninstrumented are rare, however, this type of information isnecessary to demonstrate the effectiveness of the isolationsystem. This survey is the first step in developing a databaseof performance characteristics of a variety of isolation de-vices, which will eventually affect design and manufacturingcriteria for these devices, and impact guidelines for their use.

53Literature Review of the Observed Performance of Seismically Isolated Bridges

sive literature review has been con-ducted on the observed perfor-mance of seismically isolatedbridges in recent earthquakes, andthe results are presented below.Observations are also made andsummarized.

Observed SeismicPerformance

A database, summarizing the actualperformance of seismically isolatedbridges in real earthquakes, is grow-ing gradually. This section summa-rizes some of this material for bridgeslocated in Japan, New Zealand, Tai-wan, Turkey and the U.S. Informationis presented country-by-country, in al-phabetical order.

Japan

The Miyagawa bridge, as shown inFigure 1, was built in 1991 and is a3-span continuous steel plate girderbridge with a total length of 108.5m and effective width of 10.5 m.This bridge was the first in Japanto be seismically-isolated. It useslead-rubber bearings (LRBs) for theisolators, which are designed toyield at a lateral force of 12% of theweight of superstructure. In Japan,it is common to restrain the trans-verse displacement of bridge isola-tors, and therefore many bridges inJapan are isolated in the longitudi-nal direction only. This ‘partial’ iso-lation is called ‘menshin’ design, andthe Miyagawa bridge is an exampleof this technique. In 1991, thebridge was subjected to groundmotions from an earthquake witha magnitude of 4.9 and an epicen-ter 30 km northeast of the bridgesite. The seismic response of thebridge was recorded by three ac-

celerometers in the bridge super-structure. As shown in Table 1, thepeak ground acceleration was small,the LRBs did not reach yield and thenonlinear behavior of LRBs was notsignificant. The maximum accelera-tion in the deck was about 50% ofthat recorded 10 m below the groundsurface, which illustrates the effec-tiveness of the isolation. Spectralanalysis showed that vibrations withfrequencies in the 3.0 to 5.0 Hz range,were successfully filtered out by theisolators. However, it was also foundthat vibrations with frequencies of1.2 Hz were amplified by a factor of2.0 (by the isolators) when com-pared to the ground motion. Nonlin-ear behavior of the isolators maysuppress this amplification in largerearthquakes but the importance ofconsidering the predominant groundmotion period, when selecting theisolated period, is illustrated(Kawashima et al., 1992).

The Yama-age bridge was built in1993 and is a 6-span continuouspre-stressed concrete box girderbridge with a total length of 246.3m and an effective width of 10.5 mto 13.5 m. The bridge is seismicallyisolated with high damping rubberbearings, and is again isolated in thelongitudinal direction only(menshin design). Seismic re-sponses of this bridge were re-corded during the 1994Hokkaido-toho-oki earthquake

� Figure 1. Miyagawa Bridge (Kawashima et al., 1992)

54

with magnitude 8.1, whose epicen-ter was about 1,000 km away fromthe bridge. Recorded peak accel-erations are shown in Table 1. Judg-ing from obtained accelerationresponse records, none of the seis-mic isolators were exercised intheir nonlinear range. Peak deck ac-celeration decreased to one-third ofthat at the pier top, in the longitu-dinal direction, while it increasedby 30% in the transverse direction(restrained direction). Spectralanalysis of the recorded responsesshowed that the vibration compo-nent, with a frequency equal to thefundamental frequency of the pier,was not transmitted to the deckthrough the seismic isolators(PWRI, 1995; Unjoh, 1996).

The Maruki-bashi bridge was builtin 1992 as a seismically isolatedbridge with lead-rubber bearings ofa circular shape. It is isolated in thelongitudinal and transverse direc-

tions. It is a 3-span continuous pre-stressed concrete box girder bridgewith a total length of 92.8 m andan effective width of 11.0 m. Thebridge was excited by ground mo-tions and its seismic response wasrecorded during the 1994 Sanriku-haruka-oki earthquake with magni-tude of 7.5. The epicenter wasabout 190 km from the bridge site.Recorded peak accelerations areshown in Table 1. The peak deckacceleration in the longitudinal di-rection decreased by 11% from thatat the pier top, due to the seismicisolators, and by 5% in the trans-verse direction. The peak deck ac-celerations were 1.7 times and 2.2times of the peak ground accelera-tions in the longitudinal and trans-verse directions, respectively.Judging from the recorded accel-eration response, it can be con-cluded that seismic isolatorsremained elastic and did not yield.Therefore, the effectiveness of seis-

Longitudinal Transverse VerticalDeck 0.0051 g 0.0043 g 0.0027 g

Pier Crest 0.011 g 0.0067 g 0.0025 gPeak Acc.Underground2 0.012 g 0.0091 g 0.0030 g

Deck/Underground 0.431 0.483 0.897Pier Crest/Underground 0.897 0.742 0.862

MiyagawaBridge1

AmplificationFactor

Deck/Pier Crest 0.481 0.652 1.04Deck 0.012 g 0.019 g ---

Pier Crest 0.038 g 0.014 g ---Peak Acc.Underground3 0.0093 g 0.0085 g 0.047 g

Deck/Underground 1.30 2.28 ---Pier Crest/Underground 4.05 1.71 ---

Yama-ageBridge1

AmplificationFactor

Deck/Pier Crest 0.320 0.133 ---Deck 0.098 g 0.12 g 0.058 g

Pier Crest 0.11 g 0.13 g 0.038 gPeak Acc.Underground4 0.065 g 0.059 g 0.038 g

Deck/Underground 1.51 2.07 1.52Pier Crest/Underground 1.69 2.19 0.997

Maruki-bashiBridge

AmplificationFactor

Deck/Pier Crest 0.890 0.947 1.52Deck 0.19 g --- ---

Pier Crest 0.20 g 0.36 g 0.077 gFooting 0.11 g 0.13 g 0.070 gPeak Acc.

Underground5 0.15 g 0.14 g 0.12 gDeck/Underground 1.30 --- ---

Pier Crest/Underground 1.39 2.64 0.655Footing/Underground 0.717 0.933 0.595

MatsunohamaBridge1

AmplificationFactor

Deck/Pier Crest 0.940 --- ---

NOTES 1. Transverse movement is restrained in these bridges2. 10 m below the ground surface3. 5 m below the ground surface4. 6 m below the ground surface5. 1 m below the ground surface

� Table 1. Peak Accelerations and Amplification Factors for Bridges in Japan


mic isolation is not very apparentfrom the recorded data. Spectralanalysis of the recorded data showsamplification factors (ground todeck) with two peaks at 2.5 Hz and16.7 Hz in the longitudinal direc-tion (2.5 Hz is the fundamental fre-quency of the entire bridge, and16.7 Hz is the frequency of thebridge pier itself). Moreover, it wasalso observed that the vibrationcomponent with the frequencyequal to the fundamental fre-quency of the pier, was filtered outbetween the pier top and the deckby the seismic isolators, in the lon-gitudinal direction (PWRI, 1995).

The Matsunohama bridge was sub-ject to the 1995 Hyogoken-nanbu(Kobe) earthquake of magnitude7.2 and epicentral distance 35 kmfrom the bridge site. The bridgewas built in 1991, and is a curved4-span continuous steel box girderbridge with a total length of 211.5m, a maximum width of 21.94 m,and a radius of curvature of 560 m.It is a seismically isolated bridge butonly in the longitudinal direction(menshin design). Both ends of thebridge are supported by pivotroller bearings, and two lead-rub-ber bearings are installed on eachof the three middle piers. Re-corded peak accelerations areshown in Table 1. The peak accel-eration of the deck was a littlesmaller than that of the pier top inthe longitudinal direction, in whichseismic isolation was effective.Judging from the displacement re-sponse obtained by integrating theacceleration records, the isolatorsreached yield. The maximum dis-placement was about twice theyield displacement, and about 8%of the design displacement for theisolators. Spectral analysis showedthat the vibration component, cor-

responding to the frequency of thepier (1.0 Hz), was not transmittedfrom the pier to the deck by theisolators. These observations inferthat the isolation system func-tioned as expected (PWRI, 1995;Naganuma et al., 1996; Fujino et al.,1997; Kouno et al., 1997).

New Zealand

The Te Teko (Rangitaiki River)bridge was built in 1983 and is a 5-span pre-stressed concrete U-beambridge with a total length of 103m. The bridge is seismically isolatedby lead-rubber bearings at each pierand plain rubber bearings at eachabutment. It was subject to the1987 Edgecumbe (Bay of Plenty)earthquake with magnitude of 6.3at a distance of 15 km from thebridge. The bridge was not instru-mented but the peak ground accel-eration, recorded about 11 kmsouth of the bridge, was 0.33g. Oneof two bearings on the abutmentwas dislocated due to a construc-tion deficiency (inadequate fasten-ing detail for the isolator) and as aconsequence, the bridge sustainedminor damage. The superstructuresuffered a small permanent dis-placement, a plastic hinge began toform in one pier that spalled a smallamount of cover concrete, and asacrificial knock-off device waspushed about 80 to 100 mm fromits original position. Based on thefact that the bridge suffered onlyminor damage despite the severeground shaking, it can be said thatthe bridge performed well, and thatit may not have been damaged atall if the construction deficiencieshad been avoided by better detail-ing (Mckay, 1990; Robinson, 1992;DIS, 1996).

“Testing ofisolated bridgemodels at fullscale is nowpossible due tomultiple shaketable facilitieswith biaxialmotions underthe NSF-NEESinitiative.”

56

Taiwan

The Bai-Ho bridge, as shown infigure 2, is a 3-span continuous non-prismatic prestressed concretegirder bridge with a total length of145 m and a maximum width of16.1 m. The bridge is seismicallyisolated with two lead-rubber bear-ings installed on each pier (see fig-ure 3) and two PTFE coated rubberbearings at each abutment. Shearkeys and specially designed steelrods are installed on both abut-ments to restrict the transversemovement of the superstructure.The bridge is therefore partially iso-

lated in the transverse direction.Construction of the bridge wasnearly completed but it was not inservice at the time of the 1021(1999) Gia-I earthquake of magni-tude of 6.0. Eleven accelerometershad been mounted in the bridgeprior to the earthquake and re-corded peak accelerations and am-plification factors are shown inTable 2. No damage was suffered.The peak acceleration in the deckin the longitudinal direction wasslightly higher than that in thefoundation, while the peak accel-eration in the transverse directionwas 2.5 times that of the founda-tion. Significant elongation of thestructure period was observed inthe longitudinal direction due tothe nonlinear behavior of the LRBs.The bridge performed very well inthis direction. On the other hand,two dominant vibration modeswere observed in the transverse di-rection and the peak transverse re-sponse was governed by its second

Long. Trans. Vert.Deck 0.18 g 0.26 g 0.11 g

Pier Top 0.15 g 0.12 g ---Foundation 0.15 g 0.10 g 0.036 gFree Field 0.17 g 0.18 g 0.053 g

Deck/Foundation 1.20 2.49 3.20Pier Top/Foundation 0.991 1.10 ---

Deck/Pier Top 1.19 2.26 ---

� Figures 2 and 3. Bai-Ho Bridge (top) and Isolation System of the Bai-Ho Bridge (bottom)

� Table 2. Peak Accelerations and Amplification Factors in the Bai-Ho Bridge


mode. It is clear that the transverserestraint provided at the abutmentsadversely affected the response inthis direction (compared to that inthe longitudinal direction). Thisbridge also experienced the Chi-Chi earthquake on September 21,1999, but no records were obtainedat the time (Chang et al., 1999).

Turkey

The Bolu viaduct on the TransEuropean Motorway (see figure 4)in the Kaynasli valley was still un-der construction when the Duzceearthquake (M=7.2) occurred inNovember 1999. Many of the struc-tural elements were complete atthe time, but guardrails and pavingremained to be installed. The bridgeis composed of two parallel viaductstructures with pre-stressed con-crete hollow T-girders and 58monolithic column footings. It hasa total length of 2,313 m and awidth of 17.5 m. A typical moduleof the bridge (as shown in figure5) is a 10-span viaduct with a maxi-mum span length of 39.6 m and amaximum pier height of 49 m. ThePTFE bearing and the crescentmoon-type steel energy dissipationdevice are shown in Figures 6 and7, respectively. The bridge isequipped with flat stainless steel-PTFE sliding bearings, a multidirec-tional crescent moon-type steelenergy dissipation device at thecentral support of the 10-span unit,viscous connecting devices atother supports, steel strands re-strainers at the expansion joints be-tween the 10-span units, andbi-directional steel energy dissipa-tion devices at the two abutments.The bridge was not instrumentedat the time of the earthquake, butpeak ground accelerations at the

site have been estimated as over 1g(the strong motion station inDuzce, near the epicenter andabout 5 km to the west of the via-duct, recorded accelerations in ex-cess 1.0g.)

The following observations weremade by Xiao and Yaprak (1999).Ground rupture was observed nearthe south side of the bridge, andthe fault crossed under the bridgenear Pier 46. Soil around the foot-ing of the pier was disturbed sig-nificantly and the pier appeared tohave rotated about 12 degreesabout its vertical axis. Damage to

� Figure 4. Bolu Viaduct on the Trans European Motorway (Xiao et al., 1999)

� Figure 5. Typical Module of the Bolu Viaduct (Xiao et al., 1999)

58

the columns of the viaduct was lim-ited to fine horizontal f lexuralcracks in the lower portion of sev-eral piers. Most of the structuraldamage was sustained by the super-structure and the abutments. Bothsuperstructures (of the twobridges) had a westward perma-nent displacement relative to thepiers with offsets at all girder ends.Some of the outer-most pre-stressed concrete girders slid off

their concrete pedestals. Cracksoccurred at the ends of many gird-ers. Severe concrete spalling andcrushing was observed for at leastone pre-stressed concrete girder.The south side wing walls of theeast abutments of the two bridgessustained severe damage due topounding by the girder ends. A per-manent southward transverse off-set of about 250 mm was observedat the ends of both the south andnorth bridges. Concrete crushingand inclination of the abutmentswas observed at the west end ofthe viaduct, caused by poundingbetween the girders and abutments(Xiao et al, 1999).

Stainless steel plates, bearing ma-sonry and anchor plates, and PTFEbearings fell from their supports atalmost all piers. Judging from thescratch signs on the stainless steelplates, these bearings were prob-ably dislodged at a very early stagein the earthquake (see figure 8).This is consistent with the expec-

� Figure 7. Energy Dissipation Device(Xiao et al., 1999)

� Figure 8. Bearing Failure of the Bolu Viaduct (Xiao et al., 1999)

� Figure 6. PTFE Bearing (Xiao et al., 1999)


tation that the bridge encountereda near-fault or on-fault pulse-typemotion during the earthquake, andthat it experienced excessive dis-placement. Energy dissipation de-vices at the east abutmentsustained distortion of the steelyielding elements and rupture ofthe anchor bolts of the devices. Nospan collapsed despite experienc-ing a level of ground shaking be-yond the original design levelduring this earthquake. In addition,this viaduct survived the previousearthquake (August 17) withoutdamage. The magnitude of the Au-gust earthquake was 7.4, and thepeak ground accelerations wereestimated as 0.39 g (longitudinal),0.31 g (transverse), and 0.50 g (ver-tical) (Xiao et al, 1999).

United States

The Sierra Point overpass, U.S.101, is a 10-span simply supportedcomposite girder bridge with askew angle of 59 degrees at thenorth abutment and 72 degrees atthe south abutment. Its total lengthand width are 184.8 m and 35.1 m,respectively. The bridge was con-structed in 1956, and retrofittedwith lead-rubber bearings in 1985.It was the first bridge to be retro-fitted in the U.S. using isolation andavoided the need to jacket the non-ductile columns and strengthen thefootings. However, the abutmentswere not modified for the largerclearances required for an isolationretrofit. They are therefore ex-pected to engage during low-to-moderate shaking, and to bedamaged in strong shaking.

The seismic response of thisbridge was recorded by five accel-erometers during the 1989 LomaPrieta earthquake. The peak

ground acceleration at the base ofthe bridge columns was 0.090 g inthe longitudinal direction and0.050 g in the transverse direction.Significant amplification occurredat these low level accelerations inthe steel superstructure of thebridge, due to the lock-up actionof the abutments, which preventedthe isolators from acting. No dam-age, however, was sustained (DIS,1996).

The Eel River bridge was retrofit-ted with lead-rubber bearings in1987. Two truss spans of the bridgewere seismically isolated, and eachspan has a span length of 90.0 m(Buckle et al., 1987). The bridgewas subjected to earthquakeground motions during the 1992Cape Mendocino (Petrolia) earth-quake, whose magnitude was 7.0and epicenter was located 22 kmfrom the bridge. The bridge wasnot instrumented but peak groundaccelerations obtained at thePainter Street Overpass, a similardistance from the epicenter, were0.55 g in the longitudinal direction,and 0.39 g in the transverse direc-tion. As a consequence, it is be-lieved that this is the strongestground motion to have been felt bya seismically isolated structure inthe United States. Judging from thedisplacement of the guardrail at-tached to the truss span, the maxi-mum relative displacementbetween the isolated trusses andthe abutment appear to have beenof the order of 200 mm (longitudi-nally) and 100 mm (transversely).The bridge came to rest in its origi-nal position and structural damagewas light and limited to concretespalling at joints (DIS, 1992; DIS,1996).

60

Observations• Seismically isolated bridges have

been researched and investi-gated by academia and engi-neers for many years. Due to thisextensive effort, seismic isola-tion design has become apractical option for earthquake-resistant design.

• Results from numerous com-puter simulations and shaketable experiments have shownthe advantages of seismically iso-lated bridges compared to non-isolated bridges. However, fewisolated bridges have experi-enced actual earthquakes andbeen able to demonstrate theireffectiveness, especially understrong ground shaking. Further-more, the majority of bridgesthat have been subject to earth-quakes, have not been instru-mented and performance mustbe inferred or extrapolated fromnearby sites. Also, in the major-ity of cases, the earthquakes thathave been captured are rela-

tively small and the isolatorshave either not engaged or haveremained in their elastic (andstiff) range. Recorded perfor-mance of an isolated bridgeduring strong ground motion isnot yet available and thus theeffectiveness of seismic isola-tion, especially during a largeearthquake, remains to beproven in the field.

• An exception to the previousitem might be the performanceof several isolated bridges dur-ing the June 17 and 21, 2000,earthquakes in Iceland, whichmeasured 6.6 and 6.5 on theRichter scale, respectively. Thesebridges include the Thjórsábridge (see Figures 9 and 10), theSog River bridge, the Stóra LaxáRiver bridge, and the ÓseyrarRiver bridge (Higgins, 2000).Lead-rubber bearings are in-stalled in all four bridges, whichwere subjected to large groundmotions during these earth-quakes. The peak ground accel-eration was 0.84g at one of the

� Figures 9 and 10. Thjorsa Bridge in theSouth Iceland Lowlands(top left) andclose-up of the isolation bearings–bothends of the truss tipped away from theriver by 2 cm (bottom right)


bridge sites, but no significantdamage was found. Recordedresponses are currently beinganalyzed at the Earthquake En-gineering Research Center, Uni-versity of Iceland (EERCUI).Results are expected later thisyear (2001).

• Our experience-database of iso-lated bridges is very sparse.Many more seismically isolatedbridges need to be instrumentedwith strong motion arrays, inorder to address this urgentneed and increase the likelihoodthat field verification will beavailable in the near future.

• In the absence of field data, test-ing of isolated bridge models atfull scale, in structural testinglaboratories, could be under-taken. Such work has not beenpreviously possible on a mean-ingful scale, but the current de-velopment of multiple shaketable facilities with biaxial mo-tions, under the NSF-NEES initia-tive, makes this option apractical reality.

• The performance of isolatedbridges in the near field of ac-tive faults has been a concernsince the Northridge earthquakein 1994. The damage sustainedby the Bolu Viaduct in Turkey, de-scribed above, gives substance

to this concern. An urgent needis a rigorous investigation of theBolu Viaduct to determine thecause for the damage, and iden-tify changes, if any, that shouldbe made to U.S. design andmanufacturing criteria for isola-tors.

• It is clear from several of theabove case studies that smallearthquakes rarely yield the iso-lators, which then perform intheir elastic and stiff range. Ac-cordingly, the observed amplifi-cation factors are higher thanwould be expected for an iso-lated bridge and this highlightsthe need for a class of isolatorsthat have the ‘intelligence’ tooptimize their properties ac-cording to demand.

• Various types of seismic isolatorshave been proposed, developed,and tested to date. However,only a few of these have beenimplemented in the field due toconcerns about performanceunder extreme events and theirdurability and maintenance.Comparative performance ofnew devices, using either elas-tomeric isolators or friction pen-dulum devices as benchmarks,should be undertaken to con-tinue to advance the state-of-the-art.

References

Buckle, I.G. and Mayes, R.L., (1987), “Seismic Isolation of Bridge Structures in theUnited States of America,” Proceedings of the Third U.S. Japan Workshop on BridgeEngineering, Public Works Research Institute, Ministry of Construction, Tsukuba,Japan, pp. 379-401.

Chang, K. C., Wei, S. S., Chen, H. W., and Tsai, M. H., (1999), Field Testing andAnalysis of a Seismically Isolated Bridge, Report No. CEER R88-07, Center forEarthquake Engineering Research, National Taiwan University, Taipei, ROC (inChinese).

62

References (Con’t)Chopra, Anil K., (1995), Dynamics of Structures: Theory and Applications to

Earthquake Engineering, Prentice Hall, NJ.

Dynamic Isolation Systems, (1992), Performance of the Eel River Bridge in thePetrolia Earthquakes of April 25-26, 1992, Dynamic Isolation Systems, Inc., CA.

Dynamic Isolation Systems, (1996), Earthquake Performance of Structures on Lead-Rubber Isolation Systems, Dynamic Isolation Systems, Inc., CA.

Fujino, Y., Yoshida, J., and Abe, M., (1997), "Investigation on Seismic Response ofMenshin Bridge Based on Measured Records," Proceedings of the 24th Symposiumon Earthquake Engineering Research, pp. 325-328 (in Japanese).

Kawashima, K., Nagashima, H., Masumoto, S., and Hara, K., (1992), "Response AnalysisBuckle, I. G. and Mayes, R. L., (1987), "Seismic Isolation of Bridge Structures in theUnited States of America,” Proceedings of the Third U.S.–Japan Workshop on BridgeEngineering, Public Works Research Institute, Ministry of Construction, Tsukuba,Japan, pp. 379-401.

Kelly, James M., (1997), Earthquake-Resistant Design With Rubber, Second Edition,Springer-Verlag, UK.

Kouno, T. and Kawashima, K., (1997), "Vibration Characteristics of Menshin BridgeBased on Measured Records," Proceedings of the 24th Symposium on EarthquakeEngineering Research, pp. 329-332 (in Japanese).

Mckay, G.R., Chapman, H.E., and Kirkcaldie, D.K., (1990), "Seismic Isolation: NewZealand Applications," Earthquake Spectra, Vol. 6, No. 2, pp. 203-222.

Naganuma, T., Horie, Y., Kobayashi, H., and Sasaki, N., (1996), "The Analysis of SeismicResponse of the Bridge with Menshin Bearings During the Hyogo-ken NanbuEarthquake," Proceedings of the Fourth U.S.-Japan Workshop on Earthquake ProtectiveSystems for Bridges, Japan, December 9 and 10, 1996, Technical Memorandum of PWRI,No. 3480, Public Works Research Institute, Japan, pp. 259-265.

Public Works Research Institute (1995), Study on Seismic Response Characteristicsof Menshin Bridges Based on Measured Records, Technical Memorandum of PWRI,No. 3383, Public Works Research Institute, Japan.

Robinson, W.H., (1992), "Seismic Isolation of Bridges in New Zealand," Proceedingsof the Second U.S.-Japan Workshop on Earthquake Protective Systems for Bridges,Japan, December 7 and 8, 1992, Technical Memorandum of PWRI, No. 3196, PublicWorks Research Institute, Japan.

Skinner, Robert Ivan, Robinson, William H., and McVerry, Graeme H. (1993), AnIntroduction to Seismic Isolation, John Wiley & Sons Ltd, England.

Unjoh, S., (1996), "Analysis of the Seismic Response of Yama-age Bridge," Proceedingsof the Fourth U.S.-Japan Workshop on Earthquake Protective Systems for Bridges,Japan, December 9 and 10, 1996, Technical Memorandum of PWRI, No. 3480, PublicWorks Research Institute, Japan, pp. 149-163.

Xiao, Yan and Yaprak, Tolga Taskin, (1999), Performance of a 2.3 km Long ModernViaduct Bridge During the November 12, 1999 Duzce, Turkey, Earthquake, U.S.CStructural Engineering Research Report, Report No. U.S.C-SERP 99/01, Departmentof Civil Engineering, University of Southern California, Los Angeles, California.

63

A Risk-Based Methodology forAssessing the Seismic Performanceof Highway Systems

by Stuart D. Werner


Craig E. Taylor, President,Natural HazardsManagement Inc.

James E. Moore III,Associate Professor,Departments of CivilEngineering and PublicPolicy and Management,University of SouthernCalifornia

Sungbin Cho, ResearchAssociate, School of Policy,Planning, andDevelopment, University ofSouthern California

Jon S. Walton, VicePresident, MacFarlandWalton Enterprises

Stuart D. Werner, Principal,Seismic Systems andEngineering Consultants

Past experience has shown that effects of earthquake damage to high-way components (e.g., bridges, roadways, tunnels, etc.) may not only

include life safety risks and post-earthquake costs for repair of the com-ponents. Rather, such damage can also disrupt traffic flows and this, inturn, can impact the economic recovery of the region as well as post-earthquake emergency response and reconstruction operations. Further-more, the extent of these impacts will depend not only on the seismicresponse characteristics of the individual components, but also on thecharacteristics of the highway system that contains these components.System characteristics that will affect post-earthquake traffic flows include:(a) the highway network configuration; (b) locations, redundancies, andtraffic capacities and volumes of the system's links between key originsand destinations; and (c) component locations within these links.

From this, it is evident that earthquake damage to certain components(e.g., those along important and non-redundant links within the system)will have a greater impact on the system performance (e.g., post-earth-quake traffic flows) than will other components. Unfortunately, such sys-tem issues are typically ignored when specifying seismic performancerequirements and design/strengthening criteria for new and existing com-ponents; i.e., each component is usually treated as an individual entityonly, without regard to how its damage may impact highway system per-

Research Objectives

The objective of this research is to develop, apply, program, and dissemi-nate a practical and technically sound methodology for seismic risk analy-sis (SRA) of highway-roadway systems. The methodology’s risk-basedframework uses models for seismology and geology, engineering (struc-tural, geotechnical, and transportation), repair and reconstruction, systemanalysis, and economics to estimate system-wide direct losses and indirectlosses due to reduced traffic flows and increased travel times caused byearthquake damage to the highway system. Results from this methodol-ogy also show how this damage can affect access to facilities critical toemergency response and recovery.

64

• Ian G. Buckle, Universityof Nevada, Reno

• Ian Friedland, AppliedTechnology Council

• John Mander, Universityof Canterbury, NewZealand (formerly of theUniversity at Buffalo)

• Masanobu Shinozuka,University of SouthernCalifornia

Other Contributors• Howard Hwang, Center

for Earthquake Researchand Information,University of Memphis

• John Jernigan, Ellers,Oakley, Chester and Rike,Inc., Memphis, Tennessee

• Arch Johnson, Universityof Memphis

• W.D. Liu, ParsonsBrinckerhoff, Inc.,formerly with Imbsen &Associates, Inc.,Sacramento, California

• David Perkins, ArthurFrankel and Eugene(Buddy) Schweig, U.S.Geological Survey

• Maurice Power,Geomatrix Consultants,Inc., San Francisco

• Sarah H. Sun, ShelbyCounty Office of Planningand Development

• Edward Wasserman,Tennessee Department ofTransportation

• T. Leslie Youd, BrighamYoung University

formance. Furthermore, currentcriteria for prioritizing bridges forseismic retrofit represent the im-portance of the bridge as a trafficcarrying entity only by using aver-age daily traffic count, detourlength, and route type as param-eters in the prioritization process.No attempt is made to account forthe systemic effects associated withthe loss of a given bridge, or for thecombination of effects associatedwith the loss of other bridges in thehighway system.

To address these issues, currentand recent highway researchprojects conducted by MCEER andfunded by the Federal HighwayAdministration (FHWA) have in-cluded tasks to develop a SRAmethodology for highway systems.This paper describes this method-ology, presents results from a dem-onstration application of themethodology to the highway sys-tem in Shelby County, Tennessee,and summarizes plans for the fur-ther development, application, pro-gramming, and dissemination of themethodology. Further details onthis methodology and its applica-

tion are contained in the report byWerner et al. (2000).

MethodologyDescription

The SRA methodology (Figure 1)can be carried out for any numberof scenario earthquakes and simu-lations, in which a "simulation" isdefined as a complete set of systemSRA results for one particular setof input parameters and model un-certainty parameters. The modeland input parameters for one simu-lation may differ from those forother simulations because of ran-dom and systematic uncertainties.

For each earthquake and simula-tion, this multidisciplinary processuses geoseismic, engineering(geotechnical, structural, and trans-portation), network, and economicmodels to estimate: (a) earthquakeeffects on system-wide traffic flowsand corresponding travel times,paths, and distances; (b) economicimpacts of highway system damage(e.g., repair costs and costs of traveltime delays); and (c) post-earth-

This SRA methodology will provide cost and risk in-formation that will facilitate more rational evaluationof alternative seismic risk reduction strategies by deci-sion makers from government and transportation agen-cies involved with improvement and upgrade of thehighway-roadway infrastructure, emergency responseplanning, and transportation planning. Such strategiescan include prioritization and seismic strengtheningmeasures for existing bridges and other components,establishment of design criteria for new bridges andother components, construction of additional roadwaysto expand system redundancy, and post-earthquake traf-fic management planning.

65Assessing the Seismic Performance of Highway Systems

quake traffic flows to and from im-portant locations in the region. Keyto this process is a GIS data basethat contains four modules withdata and models that characterizethe system, seismic hazards, com-ponent vulnerabilities, and eco-nomic impacts of highway systemdamage (Figure 1b). These are in-corporated into the four-step SRAprocedure shown in Figure 1a.

This SRA methodology has sev-eral desirable features. First, it hasa GIS database, to enhance datamanagement, analysis efficiency,and display of analysis results.Second, the GIS database ismodular, to facilitate the incorpo-ration of improved data and mod-els from future research efforts.Third, the procedure can developaggregate SRA results that are ei-ther deterministic (consisting ofa single simulation for one or afew scenario earthquakes) orprobabilistic (consisting of manysimulations and scenario earth-quakes). This range of results fa-cilitates the usefulness of SRA fora variety of applications (e.g.,seismic retrofit prioritization andcriteria, emergency responseplanning, planning of system ex-pansions or enhancements, etc.).Finally, the procedure uses rapidengineering and network analy-sis procedures, to enhance itspossible future use as a real-timepredictor of component damagestates, system states and trafficimpacts after an actual earth-quake.

Recent DevelopmentsImproved procedures for char-

acterizing scenario earthquakes,

seismic hazards, bridge vulnerabili-ties, and transportation networkanalysis have been developed.These developments, which aresummarized below, are now incor-porated into a new pre-beta soft-ware package named REDARS 1.0(Risks due to Earthquake DAmageto Roadway Systems).

Initialization of Analysis

System Analysis for

Scenario Earthquake Em

and Simulation n(m)

Incrementation of

Simulations and

Scenario Earthquakes

Aggregate System

Analysis Results

Next Simulation

Next Earthquake

GIS Database

Initialization of Analysis

System Analysis for

Scenario Earthquake Em

and Simulation n(m)

Incrementation of

Simulations and


Aggregate System

Analysis Results

Next Simulation

Next Earthquake

GIS Database

System Module

Network InventoryTraffic DataO-D ZonesTrip TablesNetwork Analysis Models

Transportation Cost ModuleModels

Effects of Travel TimeIncreases

DataVehicle Types/OccupancyUnit Costs

Hazards ModuleSeismo-TectonicsTopographySoil ConditionsGround Motion AttenuationGeologic Hazard ModelsModel Uncertainties

Steps 1-4 ofSRA Procedure

Component ModuleData

StructuralRepair CostsTraffic States

ModelsLossFunctionalityUncertainties

System Module

Network InventoryTraffic DataO-D ZonesTrip TablesNetwork Analysis Models

Transportation Cost ModuleModels

Effects of Travel TimeIncreases

DataVehicle Types/OccupancyUnit Costs

Hazards ModuleSeismo-TectonicsTopographySoil ConditionsGround Motion AttenuationGeologic Hazard ModelsModel Uncertainties

Steps 1-4 ofSRA Procedure

Component ModuleData

StructuralRepair CostsTraffic States

ModelsLossFunctionalityUncertainties

(a) Overall Four-Step Procedure

(b) GIS Database

� Figure 1. Risk-Based Methodology for Assessing Seismic Performance ofHighway Systems

66


SRA of a highway system withspatially dispersed components re-quires use of scenario earthquakesto evaluate the simultaneous effectsof individual earthquakes on com-ponents at diverse locations (in-cluding systemic consequences ofdamages). Earthquake models nowbeing incorporated into REDARSare adaptations of work by Frankelet al. (1996) which was developedunder the United States GeologicalSurvey (USGS) National HazardMapping Program. Frankel et al.models for the Central (CUS) aresummarized later in this paper.Adaptation of models for California(which also builds on work by theCalifornia Division of Mines & Ge-ology) is now underway. All adap-tations feature a "walk-through"analysis, which is a natural way toassess system loss distributions andtheir variability over time.

Seismic Hazards

The ground motion models forthe SRA procedure include rockmotion attenuation characteristicsrepresentative of the region wherethe system is located, as well as am-plification of rock motions due tolocal soil conditions. For the Cen-tral United States, the Hwang andLin (1997) rock motion attenuationrelationships and soil amplificationfactors for NEHRP site classifica-tions meet these requirements.Models appropriate to other re-gions of the country are now be-ing incorporated. Liquefactionhazard models are based on workby Youd (1998), and include: (a)geologic screening to eliminatesites with a low potential for lique-faction; (b) use of modified Seed-

Idriss type methods to assess liq-uefaction potential during eachearthquake and (c) for those siteswith a potential for liquefactionduring the given earthquake, esti-mation of lateral spread displace-ment and vertical settlement usingmethods by Youd and by Tokimatsuand Seed (1987) respectively.

Component Models

Component models for highwaysystem SRA develop traffic state fra-gility curves, which estimate theprobability of a given traffic state(i.e., open lanes at various timesafter the earthquake) as a functionof the level of ground shaking orpermanent ground displacement atthe component site. Thus far, thisresearch has focused on develop-ing such models for bridges only.These models estimate the bridge’sdamage state (damage types, loca-tions, and extents) under a givenlevel of ground shaking or displace-ment, and then obtain correspond-ing traffic states by usingexpert-opinion damage-repair mod-els. The SRA methodology now in-cludes three options for modelingdamage states of bridges due toground shaking: (a) an elastic ca-pacity-demand approach byJernigan (1998); (b) a simplifiedbut rational mechanics-basedmethod by Dutta and Mander(1998) that develops rapid esti-mates of damage states based onbridge-specific input parametersinferred from the FHWA NationalBridge Inventory database; and (c)user-specified fragility curves,which can be developed for anybridge in the system, but are mostappropriate for complex or un-usual bridges. In addition, the SRAmethodology includes a first-order

• Seismic Vulnerability ofExisting HighwayConstruction (Project 106)

• Seismic Vulnerability of theNational Highway System(TEA-21 Project)


model for estimating bridge dam-age states due to permanentground displacement.

Transportation NetworkAnalysis

The SRA procedure contains twotransportation network analysismethods. For deterministic SRA fora limited number of scenario earth-quakes and simulations, a UserEquilibrium (UE) method is used.This is an exact mathematical solu-tion to an idealized model of userbehavior, which assumes that all us-ers follow routes that minimizetheir travel times. For probabilis-tic SRA involving many earth-quakes and simulations, a newAssociative Memory (AM) transpor-tation network analysis procedureis used. This method provides rapidestimation of network flows, rep-resents the latest well-developedtechnology for estimating trafficflows, is GIS compatible, and usestransportation system input datathat are typically available fromMetropolitan Planning Organiza-tions. It is derived from the artifi-cial intelligence field, and providesrapid and dependable estimates offlows in congested networks forgiven changes in link configurationdue to earthquake damage (Mooreet al., 1997).

DemonstrationApplication

System Description

The SRA methodology was usedin a demonstration application tothe highway system in ShelbyCounty, Tennessee. Shelby County

is located in the southwestern cor-ner of Tennessee, just east of theMississippi River. Its highway-road-way system contains a beltway ofhighways that surrounds the city ofMemphis, two major crossings ofthe Mississippi River, and majorroadways that extend from the cen-ter of Memphis to the north, south,and east (Figure 2). Traffic demandson the system are modeled by triptables that define the number oftrips between all of the origin-des-

Barlett

Government Center

Medical Center

President’s Island

Memphis Airport

University of Memphis

Shelby Farms

Federal Express

Germantown

Hickory Hill

� Figure 3. Origin-Destination Zones in Shelby County,Tennessee

� Figure 2. Shelby County Tennessee Highway System Model(7,807 Links and 15,604 Nodes)

68

tination (O-D) zones in the county.Figure 3 shows these O-D zones, aswell as those zones for which post-earthquake travel times were moni-tored in this SRA.

Input Data

The input data for this SRA are asfollows: (a) system input data de-scribing the roadway network ge-ometry, traffic capacities, O-Dzones, and traffic demands were de-veloped from a working file for the

County’s projected network for theyear 2020 that was provided by theShelby County Office of Planningand Development; (b) soils inputdata, in terms of NEHRP soil classi-fications and initial screening forliquefaction potential, were basedon local geology mapping carriedout by the Center of EarthquakeResearch and Development at theUniversity of Memphis; and (c) at-tribute input data for each of the384 bridges in the network werebased on data compilation effortsby Jernigan (1998).


This SRA was conducted as awalk-through analysis with a dura-tion of 50,000 years. Earthquakesoccurring during each year of thisduration were estimated by adapt-ing the Frankel et al. (1996) mod-els of the region. This generated2,321 earthquakes with momentmagnitudes ranging from 5.0 to 8.0.Each earthquake was located intoone of the 1,763 microzones (withlengths and widths of about 11.1

� Figure 4. Ground Motion Hazard. Spectral Acceleration atPeriod of 1.0 Sec.

� Figure 5. Liquefaction Hazard. Permanent Ground Displace-ment.

� Figure 6. Bridge Damage States


km) that encompassed the sur-rounding area.

Typical Results for OneScenario Earthquake andSimulation

To illustrate the form of the re-sults for one earthquake and simu-lation, we consider a scenarioearthquake with moment magni-tude of 6.9 centered about 65 kmnorthwest of downtown Memphis.For this event, ground shaking haz-ards and liquefaction hazards wereestimated by the previously notedmethods by Hwang and Lin (1997)and Youd (1998) respectively. (Fig-ures 4 and 5). Next, the Dutta-Mander (1998) approach was usedto estimate bridge damage states(Figure 6), and associated systemstates at various times after theearthquake were developed by ap-plying the first-order repair modelgiven in Werner et al. (2000) tothese damage states (e.g., Figure 7).Network analysis procedures sum-marized earlier in this paper werethen applied to each system state,to obtain corresponding system-wide traffic flows and travel timedelays. Finally, simplified economicanalysis methods adapted fromCalifornia Department of Transpor-tation models and summarized inWerner et al. (2000) were used toestimate corresponding economiclosses (due to commute time in-creases only).

Economic Losses

After results of the type shownabove are developed for each sce-nario earthquake and simulationduring each year of thewalkthrough, they can be aggre-

Exposure Times

100 years

50 years

10 years

1 year

Loss in Dollars (M)

Pro

babi

lity

of E

xced

ance

Pro

babi

lity

of E

xcee

danc

e

� Figure 7. System States at 7 Days (top) and 60 days(bottom) afterEarthquake

� Figure 8. Economic Losses due to Earthquake-Induced Increases inCommute Times in Shelby County, Tennessee

70

gated to obtain probablistic esti-mates of economic losses. Figure8 shows results of this type for ex-posure times of 1, 10, 50, and 100years. Deterministic estimates ofeconomic losses can also be ob-tained for selected individual earth-quake events.

Travel Times for SelectedLocations

For emergency planning pur-poses, it may be of interest to esti-mate how travel times to and/orfrom selected key locations in theregion may be affected by earth-quake damage to the highway sys-tem. Such results can be developedas aggregated probabilistic curves(similar in form to Figure 8) or asdeterministic estimates for selectedearthquake events (see Table 1).

Conclusions/FutureDirections

The risk-based methodology de-scribed in this paper estimates how

earthquake damage to highway sys-tems can affect post-earthquaketraffic flows and travel times. It isa technically sound and practicalapproach that will enable decisionmakers to consider system-widetraffic effects when evaluating vari-ous seismic risk reduction optionsfor highway components and sys-tems.

Although the basic SRA method-ology is in place, further work re-mains before it can be disseminatedand applied to highway systems na-tionwide. For example, theREDARS 1.0 pre-beta softwarepackage that is based on this meth-odology is now being indepen-dently validated and applied. Thiswill help to identify future direc-tions for further development ofthis software. Additional improve-ments now being planned include:(a) incorporation of models for es-timating scenario earthquakes andground motion hazards nationwide;(b) development of models for es-timating system-wide landslide andsurface fault rupture hazards; (c) de-

Origin-Destination Zone (see Fig. 3) Post-Earthquake Access Time7 Days

after EQ60 Daysafter EQ

150 Daysafter EQ

9 (Government Center in downtown Memphis) 43.8% 5.8% 2.0%28 (Major Hospital Center, just east of downtownMemphis)

44.6% 6.7% 2.0%

205 (Memphis Airport and Federal Expresstransportation center, south of beltway)

53.7% 4.0% 1.6%

73 (University of Memphis campus in centralMemphis)

21.6% 4.3% 1.5%

310 (Germantown, residential area east of beltway) 2.9% 0.9% 0.4%160 (President’s Island, Port of Memphis at MississippiRiver)

34.9% 6.1% 1.6%

246 (Hickory Hill, commercial area southeast ofbeltway)

3.9% 1.9% 1.1%

335 (Shelby Farms residential area northeast ofbeltway)

28.4% 4.8% 1.6%

412 (Bartlett, residential area north of beltway) 13.2% 3.0% 1.3%

� Table 1. Increase in Access Times to Locations in Shelby County due to Damage to Highway System fromEarthquake with Magnitude 6.8 centered 66 km Northwest of Downtown Memphis


velopment of improved compo-nent repair models; (d) develop-ment of vulnerability/fragilitymodels for retrofitted bridges aswell as other highway componentssuch as tunnels, slopes, pavements,

walls, and culverts; and (e) devel-opment of the system module toaccommodate post-earthquake traf-fic demands that differ from pre-earthquake demands.

References

Dutta, A. and Mander, J.B., (1998), “Seismic Fragility Analysis of Highway Bridges,” Proc.of INCEDE-MCEER Ctr-Ctr Workshop on Earthq. Engrg. Frontiers in Transp. Sys-tems, Tokyo Japan, June 22-23.

Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyendecker, E.V., Dickman, N., Hanson,S., and Hopper, M., (1996), National Seismic Hazard Maps, June 1996 Documen-tation, Preliminary Report prepared at USGS, Denver CO, July 19.

Hwang, H.H.M. and Lin, H., (1997), GIS-Based Evaluation of Seismic Performanceof Water Delivery Systems, University of Memphis, Memphis TN, Feb 10.

Jernigan, J.B., (1998), Evaluation of Seismic Damage to Bridges and Highway Sys-tems in Shelby County, Tennessee, Ph.D. Dissertation, Univ. of Memphis, MemphisTN, Dec.

Moore, J.E. II, Kim, G., Xu, R., Cho, S., Hu, H-H, and Xu, R., (1997), Evaluating Sys-tem ATMIS Technologies via Rapid Estimation of Network Flows: Final Report,California PATH Report UCB-ITS-PRR-97-54, Dec.

Tokimatsu, K. and Seed, H.B., (1987), “Evaluation of Settlements in Sand due to Earth-quake Shaking,” J. of Geotech. Engrg., ASCE, Vol. 113, No. 8, pp. 861-878.

Werner, S.D., Taylor, C.E., Moore, J.E. II, and Walton, J.S., (2000), A Risk-Based Meth-odology for Assessing the Seismic Performance of Highway Systems, ReportMCEER-00-0014, Multidisciplinary Center for Earthquake Engineering Research, Uni-versity at Buffalo, (Dec. 31).

Youd, T.L., (1998), Screening Guide for Rapid Assessment of Liquefaction Hazardat Highway Bridge Sites, MCEER-98-0005, Multidisciplinary Center for EarthquakeEngineering Research, University at Buffalo, May.

73

Research Objectives

The objective of this study is to develop a set of guidelines and analyticaltools for use by practicing engineers to determine the collapse limit state ofstructures. A program of shake table testing of simple frames through col-lapse was carried out, and analytical strategies to capture this behavior havebeen developed and made available on the web via MCEER’s users network.The research performed here helps to develop a better understanding of thecollapse mechanism and provides tools for further investigation.

Analysis, Testing and InitialRecommendations on CollapseLimit States of Frames

by Darren Vian, Mettupalayam Sivaselvan, Michel Bruneau and Andrei Reinhorn

As inelastic behavior is more extensively relied upon in the dissipationof seismic input energy, the destabilizing effect of gravity becomes

more significant in the structural evaluation of existing structures. How-ever, practicing engineers have limited confidence in the adequacy of cur-rently available analytical tools to accurately predict when collapse willoccur (i.e., the collapse limit state). As a result, there is a need to investigatethe seismic behavior of structures to enhance our understanding of thecondition ultimately leading to their collapse, and to ensure public safetyduring extreme events. While many experimental studies and theoreticaldamage models support these calculated values, it remains that few experi-mental studies have pushed the shake table tests up to collapse.

This paper presents the results of research to provide some of that datathrough a program of shake-table testing of simple frames through col-lapse, developments of analytical strategies to capture this behavior, andrecommendations for design. Note that every effort was made to ensurethat the experimental data is fully documented (geometry, material prop-erties, initial imperfections, detailed test results, etc.); it will be made broadlyavailable on MCEER’s Users Network such that these tests can be used ata later time by other researchers as a benchmark to which analytical mod-els can be compared. An example of such use is the development madeby the authors, described herein, using the benchmark test results.

Experimental ProgramFifteen specimens, each consisting of four columns, were tested to

failure in the course of this research. These were subdivided into threegroups of five with slenderness ratios of 100, 150, and 200. The dimen-


Darren Vian, GraduateStudent, MettupalayamSivaselvan, GraduateStudent, Michel Bruneau,Professor, and AndreiReinhorn, Professor,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

74

sions and mass used were variedwithin each group. Nominal speci-men column widths ranged from2.8 mm (1/8 in) to 9.8 mm (3/8 in).Column heights ranged from91.7 mm (5.41 in) to 549.9 mm(21.65 in). Individual columns were

cut from hot-rolled steel plate andthen milled to size. Mass appliedto each specimen column variedfrom approximately 150 kg to385 kg. Predicted fundamental pe-riod of vibration for the specimens,using nominal dimensions, variedfrom 0.191 sec up to 1.098 sec con-sidering the P-∆ effect. Note thatthe specimens were not intendedto be scaled models of actual struc-tures; therefore unscaled groundmotions were used.

A sample column layout is shownin Figure 1. A range of values foraxial capacity versus demand, P

u/

Pn, was used for each slenderness

ratio, where Pu is the weight of the

mass plates used in the test, and Pn

is the axial capacity of all columnsin the specimen, calculated usingthe AISC-LRFD specifications (AISC1993). This range of values isshown in Figure 2. A number ofinitial imperfections were carefullymeasured and documented, andmovement in the transverse direc-tion was prevented by f lexiblebraces verified to have no impacton behavior in the principal direc-tion, as reported in Vian andBruneau (2001).

A schematic of the test setup andinstrumentation is shown in Figure3. Some special attachments andmodifications to standard displace-ment transducers were developed(Vian and Bruneau, 2001) to ensurereliable measurements up to and

The experimental data and analytical models from thisresearch can be used by other researchers as a benchmarkfor comparison. Eventually, the research will enable struc-tural engineers to adequately predict when collapse of agiven structure will occur, thus ensuring public safety dur-ing strong earthquakes.

MCEER Users Network:http://civil.eng.buffalo.edu/users_ntwk

50.4 mmnominal

l1

12.7 mm

5 mmDIA

w1

w2

w3

l2

35.56 mm

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

75 100 125 150 175 200 225

Slenderness (kL/r)

Dem

and/C

apacity

,P

u/P

n

1

4

2

3

5

6

8

7

9

10

11

13

12

14

15

Specimen numbers next to value

� Figure 2. Specimen Axial Load-to-Strength Ratio versus Slenderness

� Figure 1. Typical Specimen

75Analysis, Testing and Initial Recommendations on Collapse Limit States of Frames

through collapse, but these detailsare beyond the scope of this paper.

A free vibration test was first per-formed on each specimen. Subse-quently applied were a number ofground motions progressively in-creasing in magnitude from approxi-mately two-thirds of the estimatedpeak elastic response to the esti-

mated peak inelastic response.Among all the data recorded andstored, it is noteworthy that the ac-tual total horizontal displacement ofa specimen was calculated using thecorrection shown in Figure 4.

SDOF SHAKING TABLE

UG

HORWEST

HOREAST

VE

RT

MASS PLATES

COLUMN HEIGHT = L

L/3

ACCWEST

ACCEAST

ACCTBL

INITIALPOSITION

DEFORMED

POSITION

STRAIN

DIRECTION OF SHAKING

DISPLACEMENT TRANSDUCER CHANNEL

ACCELEROMETER CHANNEL

STRAIN GAGE CHANNEL

LEGEND:+

+

Schem tic of Te t Set p nd In tr ment tion (Looking We

L

L

v

u

um

Magnet, PVC Collar,& Ball Joint Assembly

TransducerTube

TemposonicHousingSpecimen

InitialPosition

SpecimenDeformedPosition

ROD

� Figure 3. Schematic of Test Setup and Instrumentation (Looking West)

� Figure 4. Definition of Variables used in Horizontal Displacement

76

Summary of Test Results

The fundamental period of vibra-tion of each specimen is obtainedexperimentally via Fourier Spec-trum Analysis of the time historydata of free-vibration tests. Interest-ingly, the damping ratio was ob-served to vary as a function of theamplitude of linear elastic response.Results from the tests to collapseinclude:• Seismic response of shake table

tests including time history plotsof target table acceleration, mea-sured and filtered table accelera-tion, total mass acceleration,relative mass displacement, aswell as a plot of estimated baseshear including P-∆, V

p*, versus

relative displacement.• Time history plots of relative dis-

placement for each test in theschedule for comparison of dis-placements throughout therange of progressive collapse.

• Plots of estimated base shear, Vp*,

normalized by the plastic base

shear, Vyo

, versus displacementductility, µ, as well as similar plotsof base shear versus percentdrift, γ (=u

rel-max/L

avg).

Some typical results are shown inFigure 5.

Behavioral Trends

The value of the stability factor, in-dicated in the preceding discussion,has a significant effect on the re-sponse of the structure. In practicalbridge and building structures, θ isunlikely to be greater than 0.10, andis generally less than 0.060 (MacRaeet. al., 1993). Specimen 1 is the onlyone here that has a θ value near thatsuggested practical range for the sta-bility factor, with a value of 0.065.Specimens 2, 6, and 11 have stabilityfactors slightly larger than the likelyupper limit, at 0.123, 0.101, and 0.138,respectively. All other specimenshave a value of θ ≥ 0.155.

Three dimensionless accelerationparameters were compared with

0 10 20 30 40 50 60 70 80 90 100-100

-50

0

50

100

|urel-max

|= 82.4mm |ur|= 63.2mm

0 10 20 30 40 50 60 70 80 90 100-100

-50

0

50

100

|urel-max

|= 82.4mm |ur|= 63.2mm

-5 -2.5 0 2.5 5-2

-1

0

1

2

|Vp*max

/Vyo

| = 1.1

|mmax|= 3.77

-25 -20 -15 -10 -5 0 5-2

-1

0

1

2

|Vp*max

/Vyo

| = 1.17

|mmax|= 20.2

-30 -25 -20 -15 -10-2

-1

0

1

2

Trial 6 Trial 7

Trial 9 Trial 10 Trial 11

Time (s) Time (s)

Ductility, Ductility, Ductility,

V /

Vp

yo

V /

Vp

yo

V /

Vp

yo

Dis

plac

emen

t (m

m)

Dis

plac

emen

t (m

m)

� Figure 5. Typical Results from Tests to Collapse


five dimensionless displacementparameters. The following generalobservations can be made:• The elastic spectral acceleration,

Sa, ductility, µ, and percent drift,

γ, were observed to have inverserelationships with θ. In supportof this observation, these vari-ables are plotted in Figures 6and 7 versus the stability factorfor the next to last test (givensubscript "final"). This suggeststhat the structures may be lessable to undergo large inelasticexcursions before imminent in-stability as the stability factor in-creases.

• Specimen 1 was the only speci-men that underwent both a duc-tility greater than five (20.35),and a drift larger than 20% of thespecimen height (64%), prior tocollapse, as shown in Figure 7.Recall that this is the only speci-men that has a value of θ lessthan 0.1.

Overall, these Figures show a highdependence of ultimate inelasticbehavior upon the stability factor fora P-∆ affected structure. For thespecimens tested in this research,those that had a value of θ equal toor greater than 0.1, tended to have arelatively low level of inelastic behav-ior before collapse of the structure.Structures with θ < 0.1 were able towithstand ground motions withhigher spectral accelerations, expe-rience larger values of ductility, andaccumulate larger drifts, than thosewith θ > 0.1. The more slender struc-tures, characterized by a larger θvalue, will undergo relatively smallinelastic excursions prior to collapse.

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Stability Coefficient,

y=0.0175x-1.8118

R =0.86042

S

(

g)a-

final

0

5

10

15

20

25

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

y=0.4503x-1.0465

R =0.7152


final

70%

60%

50%

40%

30%

20%

10%

0%

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7


y=0.0386x-0.6638

R =0.45682

(%)

final



� Figure 6. Spectral Acceleration versus Stability Coefficient

� Figure 7. Plots of Displacement Ductility and Drift versus Stability Coefficient

78

Analytical Modelingand Verification usingData Generated Fromthe Experiments

Data from the experiments per-formed in this research can be usedin the verification of time historyanalysis programs in the modelingof inelastic structural behavior upto collapse. In these studies, an at-tempt was made to formulate aflexibility-based planar beam-col-umn element, which can deforminelastically until it looses stabilityor deteriorates and cannot sustaingravity loads. The formulation hasno restrictions on the size of rota-tions, using one co-rotational framefor the element to represent rigid-body motion, and a set of co-rota-tional frames attached to theintegration points, as customarilyused to represent the constitutiveequations. The formulation shownin the next section provides an en-hancement to the existing modelsby adding the inelastic behavior inan element, which is stable understatic and reversible loads withlarge deformations using the flex-

ibility approach. The solution pro-cedure associated with the modelallows verifying their performanceup to complete collapse.

Element Equations

Figure 8 shows the deformedshape of an Euler-Bernoulli beamin co-rotational coordinates at-tached to the initially straightcenterline of the beam. The non-linear strain displacement relation-ships are given (Huddleston, 1979)by:

ddxθ ε φ= +( )1 (1)

ddx

ξ ε θ= +( )1 cos (2)

ddxη ε θ= +( )1 sin (3)

where (ξ,η) is the coordinate ofa point which was at (x,0) beforedeformation, θ is the angle madeby the tangent to the center-linewith the horizontal, ε is the axialstrain of the centerline and φ is thecurvature.

u2

u1

u3u3+u3

u1

θ

θ+θ

u2 P’(ξ,η)

P”(ξ+ξ,η+η)

P(x,0)

� Figure 8. Euler Bernoulli Beam Subjected to Large Deformation


Considering a small perturbationabout this deformed position, theincremental compatibility condi-tions are given by:

ddx

˙˙ ˙θ εφ ε φ= + +( )1 (4)

ddx

˙˙ cos sin ˙ξ ε θ ε θ θ= − +( )[ ]1 (5)

ddx

˙˙ sin cos ˙η ε θ ε θ θ= + +( )[ ]1 (6)

Integrating these equations overthe length of the element and per-forming a series of integrations byparts, the following variationalequation is obtained:

˙

˙

˙

˙

˙˜̇

˙u b b=

=

=∫ ∫u

u

u

dx dxL L1

2

30 0

T Tεφ

ε

(7)where u

1 = axial displacement,

u2 = rotation at left end and u

3 =

rotation at right end as shown inFigure 8, φ̃ = (1+ε) φ and is foundto be the work conjugate of the co-rotational moment. This agreeswith the result of Reissner (1972).b is a matrix given by:

b =− −

−

cossin( )

sin( )

( ) ( )

θ θξ

θξ

η ξξ

ξξ

L L

L L1

(8)

Constitutive Model

In this work, the inelastic behav-ior of the members is captured ina global sense. The relationships be-tween stress resultants (axial force,bending moment, etc.) and gener-alized strains (centerline strain, cur-vature, etc.) are used directly

instead of stress-strain relationships.Simeonov and Reinhorn (2001)derived a smooth three-dimen-sional plasticity model for arbi-trarily shaped yield functions andused it to represent the constitu-tive behavior of beam-columncross-sections. This model is basedon a parallel-spring plasticity model(Park and Reinhorn, 1986, andNelson and Dorfmann, 1995) withan extension of Bouc-Wen hyster-etic model (Wen, 1976, Sivaselvanand Reinhorn, 2000) and its gener-alization to multiple dimensions byCasciati (1989). The model deter-mines the generalized force, F, vec-tor (forces and moments) as:

F = F + Fe h(9)

Fe = aε (10)

˙ ˙F I a K I - B0h H H= −( ) [ ]1 2 ε (11)

where, Fe and F

h are the elastic

and hysteretic components of F, ais the diagonal matrix of post-yieldrigidity ratios, ε is the vector ofstrains and curvatures, K

0 is the ini-

tial tangent rigidity matrix, I is theidentity matrix and H

1 and H

2, the

step functions for yielding and forloading reversals, respectively, aredescribed below. B is the force-moment interaction matrix:

B =( )( ) ( )( ) ( ) ( )

∂∂

∂∂

∂∂

∂∂

Φ Φ

Φ Φ

F F 0

F 0 F

h h

h h

I - a K

I - a K

T

T (12)

where Φ(Fh) is the yield function.

H1(F

h) is the step function the de-

noted yielding given by:

HN

1 = ( )Φ Fh (13)

where N is a parameter that gov-erns the smoothness of the transi-tion from the elastic to the plasticstate. When N tends to infinity thenthe model collapses to a bilinear

80

model. H2(F

h, &ε ) is the step func-

tion denoting loading/unloading,given by:

H h2 1 2= ( ) +η ηsgn ˙F Tε (14)

where η1 and η

2 are parameters

that govern the shape of the un-loading curve. η

1 + η

2 = 1 to en-

sure compatibility with classicalplasticity theory. It can be noticedthat Eq. (7) is an implementationof the principle of virtual forces inrate form.

The compatibility equations (7)and the constitutive equations (9)to (14), are written as a set of dif-ferential-algebraic equations (DAE)along with the global equations ofmotion as proposed by Simeonovet al., (2000). The resulting systemof equations is solved by the Back-ward Difference Method using theroutine DASSL (Brenan et al., 1996).

Verification Study

The results of the above develop-ments are compared to those fromother computational solutions aswell as with the collapse experi-ments described above. The param-eters of the example for verificationwere chosen to correspond tospecimen 10a (Vian and Bruneau,2001). This specimen consists offour columns each 137 mm tall andhaving a square cross-section ofside 3.1 mm (see Figure 9). In thefirst step, the constitutive model ofthis cross-section is derived. Theplastic interaction function of thesquare cross-section is shown inFigure 10. The yield function forpositive and negative directions ofthe bending moment can be com-bined to obtain a single yield func-tion as follows:

Φ P M p m p,( ) = + − −2 12 2 4 (15)

where p = P/PP, m = M/M

p, P

P=

plastic axial force and Mp

= plasticmoment.

The formulation for this study isusing the yield function and asmoothness parameter N=2 in Eq.(13) to represent the transitionfrom initial yield to the plastic state.Equation (15) is substituted in equa-tions (9) to (14) to obtain the nec-essary constitutive model.

m

p

1.0

1.0

-1.0

-1.0

21p m+ =2

1p m− =

� Figure 9. Typical Specimen

� Figure 10. Yield Surface of a SquareCross-section


The results obtained from the cur-rent formulation and those from thefinite element analysis programABAQUS for monotonic loadings, areshown in Figure 11 with good agree-ment. Figure 12 shows monotonicdeflections of columns leading tocomplete loss of strength assumingactual elastic material behavior. Theslight difference between the resultsstems from the representation ofmoment-curvature-behavior in thecurrent approach vs. ABAQUS. Thepresent formulation uses the smooth-ness parameter N=2 to represent thepost-yield behavior. While this is suf-ficient for structural steel sections, amore accurate description is re-quired for a square section. Effortsare underway to incorporate the ex-act moment-curvature equation pre-sented by Stronge and Yu (1993).

In the dynamic analysis, the speci-men was subjected to historiesused in the experiment withoutconsidering the initial state of themodel (previously subjected to se-ries of base motions) and the re-sult is shown in Figures 13 and 14.

The time history results differfrom those obtained in the experi-ment (see Figure 14). However,when the initial deformation of themodel is considered, the analysismodel shows same pattern as theexperiment as shown in Figure 15.

Comparison withNCHRP 12-49Proposed P-∆ Limits

The National Cooperative High-way Research Program (NCHRP),Project 12-49, under the auspicesof the Transportation ResearchBoard, is investigating seismic de-sign of bridges from all relevant as-pects. At the conclusion of this

60.0

50.0

40.0

30.0

20.0

10.0

0.0

-10.0

0 2 4 6 8 10 12

Time (s)

Hor

izon

tal D

ispl

acem

ent (

mm

)

-20.0

-15.0

-10.0

-5.0

0.0

5.0

10.0

15.0

20.0

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40

Lateral Displacement

La

tera

lF

orc

e

NEW FORMULATION

ABAQUS

(P/Pcr=0.25)

New FormulationAbaqus

(P/P =0.25)cr

Lateral Displacement

Late

ral F

orce

Theoretical Response inDisplacement Control

New FormulationAbaqus

P/P =0.25

P/P =0.50

P/P =0.75

cr

cr

cr

600.0

500.0

400.0

300.0

200.0

100.0

0.00.0 0.5 1.0 1.5 2.0 2.5 3.0

Lateral Displacement (in)

Late

ral F

orce

(lb

)

� Figure 11. Analytical Results for Linear Elastic Material

� Figure 12. Static Analysis with Nonlinear Material

� Figure 13. Horizontal Displacement from Dynamic Analysis

82

project, proposed revisions to thecurrent LRFD specifications forhighway bridges will be presentedto the American Association of StateHighway Transportation Organiza-tions (AASHTO) for review andpossible implementation. Includedin the proposed revisions are howadditional demands from P-∆ affectstructural performance. The mostrecent proposed provision, as ofthis writing, states:

The displacement of a pier orbent in the longitudinal andtransverse direction must satisfyproposed AASHTO LRFD Equation3.10.3.9.4-1:

∆m CW

PH≤ ⋅ ⋅

⋅0 25. (16)

where:∆

m = R

d∆; R

d is the factor related

to response modification factor andfundamental period; ∆ is the dis-placement demand from the seis-mic analysis; C is the seismic baseshear coefficient based on lateralstrength; W is the weight of themass participating in the responseof the pier; P is the vertical load onthe pier from non-seismic loads;and H is the height of the pier.

For analysis of the specimens inthis research, the W/P ratio isequal to unity, and the measuredexperimental displacements, u

rel,

and estimated base shear coeffi-cient, C

s*, can substitute for ∆

m

and C, respectively.Figure 16 compares the proposed

limit with the peak experimentalresponses. The estimated baseshear coefficient, C

s*, is plotted as

a function of the maximum drift, γ.Results for specimens with θ < 0.25(1, 2, 4, 6, 7, 11, and 12) are shown.During the initial tests, when theproposed limit was satisfied, noneof these specimens failed. Due torepeated inelastic action, the cumu-lative drifts of the structure in-creased, eventually causingprogressive collapse and violatingthe proposed limit. Collapse alwaysoccurred only after the limit wasexceeded in a prior test, thus vali-dating the proposed criterion. Asshown in Figure 16, the remainingspecimens, for which θ ≥ 0.25,never satisfied the drift criteria,even for those tests that remained

Horizontal Displacement

Time (s)

60

50

40

30

20

10

0

-10

0 0.5 1.51 2 2.5 3 3.5

Dis

plac

emen

t (m

m)

Hor

izon

tal D

ispl

acem

ent

Time

Hor

izon

tal D

ispl

acem

ent

Time

a) Computational(with initial offset)

b) Experimentala) Computational(with initial offset)

b) Experimental

� Figure 15. Displacement Response at the Verge of Collapse

� Figure 14. Displacement History from Experiment


in the elastic range. The stabilityfactor for these specimens, how-ever, is well above the practicalrange discussed previously; there-fore, the limit violation is of no con-sequence.

ConclusionsThe experimental data generated

by this project provides a well-documented database of shaketable tests of a SDOF system sub-jected to earthquakes of progres-

(a) Cs* vs drift - θ < 0.25

0.00

0.25

0.50

0.75

1.00

0% 10% 20% 30% 40% 50% 60% 70%

drift, γ

Cs*

Specimen 1 Specimen 2



Specimen 12 NCHRP Limit

(b) Cs* vs drift - θ ≥ 0.25

0.00

0.25

0.50

0.75

1.00

0% 10% 20% 30% 40% 50% 60% 70%

drift, γ

Cs*

Specimen 5b Specimen 8

Specimen 9 Specimen 10b


NCHRP Limit

(a) C vs drift - < 0.25s*drift,

drift,

(b) C vs drift - > 0.25s*

Cs*

Cs*

Specimen 1Specimen 4Specimen 7Specimen 12

Specimen 2Specimen 6Specimen 11NCHRP Limit

Specimen 5bSpecimen 9Specimen 13NCHRP Limit

Specimen 8Specimen 10bSpecimen 14

Satisfies limit

Satisfies limit

Violates limit

Violates limit

NCHRP 12-49 proposed P- limit

NCHRP 12-49 proposed P- limit

� Figure 16. Test Results Comparison with NCHRP 12-49 Limits

84

sively increasing intensity up to col-lapse due to instability. This datawill be useful for, and shared with,other researchers who may wish tovalidate or develop algorithms ca-pable of modeling inelastic behav-ior of steel frame structures up toand including collapse. The datapresented here will also be locatedon the Internet (with all interme-diate data files) for immediate ac-cess by other researchers.

An attempt was made to analyti-cally model the collapse using anadvanced flexibility based formu-lation with large deformation in-elastic behavior. The verificationsand the refinement of the modelare done using the experimentaldata. Parametric studies can be per-formed with the model to supportfurther codes and standards devel-opment efforts.

The research presented heredemonstrated a number of impor-

tant points that must be consideredin the design of slender steel struc-tures. The stability coefficient, θ,has the most significant effect onthe behavior of the structure. As θincreases, the maximum attainableductility, sustainable drift, and spec-tral acceleration, which can be re-sisted before collapse, all decrease.When this factor is larger than 0.1,the ultimate values of the maxi-mum spectral acceleration, dis-placement ductility, and driftreached before collapse are allgrouped below values of 0.75 g, 5,and 20%, respectively. Stability co-efficient values less than 0.1 tendto increase each of those responsevalues significantly. The researchperformed here helped develop abetter understanding of the col-lapse mechanism and provides thetools for further work to modify orimprove current design standards.

AISC, (1993), Load and Resistance Factor Design Specification for Structural SteelBuildings, American Institute of Steel Construction, Chicago, IL, December1.

Brenan, K. E., Campbell, S.L., and Petzold, L. R., (1996), Numerical Solution of Initial-value Problems in Differential-Algebraic Equations, Philadelphia, Society forIndustrial and Applied Mathematics.

Casciati, F., (1989), "Stochastic Dynamics of Hysteretic Media," Structural Safety, Vol.6, No. 2-4, pp. 259-269.

Huddleston, J. V., (1979), "Nonlinear Analysis of Elastically Constrained Tie-Rods underTransverse Loads," International Journal of Mechanical Sciences, Vol. 21, No. 10,pp. 623-630.

MacRae, G. A., Priestley, M.J.N., Tao, J., (1993), P-∆ Effects in Seismic Design, ReportNo. SSRP-93/05, Department of Applied Mechanics and Engineering Sciences,University of California, San Diego.

Nelson, R. B. and A. Dorfmann, (1995), "Parallel Elastoplastic Models of Inelastic MaterialBehavior," Journal of Engineering Mechanics, Vol. 121, No.10, pp. 1089-1097.

Park, Y.J. and Reinhorn, A.M., (1986), "Earthquake Responses on Multistory BuildingsUnder Stochastic Biaxial Ground Motions," Proceedings of the Third U.S. Conferenceon Earthquake Engineering, Charleston, South Carolina, Vol. 2, pp. 991-1001, August1986.

References


Reissner, E., (1972). "On One-Dimensional Finite-Strain Beam Theory: the PlaneProblem," Journal of Applied Mathematics and Physics (ZAMP), Vol. 23, No. 5, pp.795-804.

Simeonov, V. K., (1999). Three-Dimensional Inelastic Dynamic Structural Analysis ofFrame Systems, Ph.D. Dissertation, Civil, Structural and Environmental Engineering,University at Buffalo.

Simeonov, V. K., Sivaselvan and A. M. Reinhorn, (2000), "Nonlinear Analysis of StructuralFrame Systems by the State-Space Approach," Computer-Aided Civil &Infrastructure Engineering, Vol. 15, No. 2, pp. 76-89.

Simeonov, V. K. and Reinhorn, A.M., (2001), Three-Dimensional Inelastic DynamicStructural Analysis of Frame Systems , Technical Report MCEER-01-xx,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo(in review).

Sivaselvan, M. V. and Reinhorn, A.M., (2000), "Hysteretic Models for DeterioratingInelastic Structures," Journal of Engineering Mechanics, Vol. 126, No. 6, pp. 633-640.

Stronge, W. J. and Yu, T.X., (1993), Dynamic Models for Structural Plasticity, London,New York, Springer-Verlag.

Vian, D. and Bruneau, M., (2001), Experimental Investigation Of P-Delta Effects ToCollapse During Earthquakes, Technical Report MCEER-01-0001, MultidisciplinaryCenter for Earthquake Engineering Research, University at Buffalo.

Wen, Y. K., (1976), "Methods of Random Vibration of Hysteretic Systems," Journal ofthe Engineering Mechanics Division, Vol. 102, No. 2, pp. 249-263.

References (Cont’d)

87

Research Objectives

A first objective of the research is to identify and quantify the mecha-nisms and parameters determining the hazard to deep foundations andsuperstructure caused by the lateral spreading. Two main situations havebeen identified, depending on pile foundation bending response controlledby the pressure of the liquefied soil or by that of a shallow nonliquefiedlayer. Field evidence, centrifuge results, and analyses have shown the shal-low nonliquefiable layer to be more critical and more amenable to retrofit-ting. A second objective of the research is to develop and evaluate strategiesto retrofit deep foundations by decreasing the pressure exerted by thisnonliquefied shallow layer, with emphasis in cost-effective and advancedmaterials. A third objective is to develop fragility curves for nonretrofittedand retrofitted foundations.

Centrifuge-Based Evaluation of PileFoundation Response to LateralSpreading and Mitigation Strategies

by Ricardo Dobry, Tarek H. Abdoun and Thomas D. O'Rourke


National Science Foundation:U.S.–Japan CooperativeResearch for Mitigation ofUrban EarthquakeDisasters


The effects of liquefaction on deep foundations are very damaging andcostly. Permanent lateral ground deformation or lateral spreading is a

main source of distress to piles, either alone or in combination with iner-tial superstructural forces and moments arising during shaking and actingon a soil already weakened by rising water pore pressures. Cracking andrupture of piles at shallow and deep elevations, rupture of pile connec-tions, and permanent lateral and vertical movements and rotations of pileheads and pile caps with corresponding effects on the superstructurehave been observed (Figure 1). This has affected buildings, bridges, portfacilities and other structures in Japan, the U.S. and other countries in-cluding the 1989 Loma Prieta, California and the 1995 Kobe, Japan, earth-quakes (Hamada and O'Rourke, 1992; O'Rourke and Hamada, 1992;Tokimatsu et al., 1996; Dobry and Abdoun, 2001).

Examination and analysis of case histories have revealed important as-pects of the phenomenon and highlighted its complexity. It is essentiallya kinematic soil-structure interaction process involving large ground andfoundation permanent deformations, with the deep foundation and su-perstructure responding pseudostatically to the lateral permanent displace-ment of the ground.

While in some cases the top of the foundation displaces laterally a dis-tance similar to that in the free field, like in Figure 1 where both the

Ricardo Dobry, Professor,Tarek H. Abdoun,Research AssistantProfessor, YinjuangWang, Graduate Student,Ricardo Ramos, GraduateStudent, Civil andEnvironmentalEngineering Department,Rensselaer PolytechnicInstitute

Thomas D. O’Rourke,Thomas R. Briggs Professorof Engineering and Siang-Huat Goh, ResearchAssistant, Department ofCivl and EnvironmentalEngineering, CornellUniversity

88

ure 1). More damage tends to oc-cur to piles when the lateral move-ment is forced by a strongnonliquefied shallow soil layer(end-bearing pile No. 2 in Figure 1),than when the foundation is freerto move laterally and the forcesacting on them are limited by thestrength of the liquefied soil (float-ing Pile No. 1 in Figure 1).

Lateral spreading has been iden-tified as a major hazard to pile foun-dations of hospital buildings, andcentrifuge modeling as a key toolto identify and quantify mecha-nisms, calibrate analyses and evalu-ate retrofitting strategies for pilefoundations. Figure 2 shows the100 g-ton RPI geotechnical centri-fuge used for this research, whichis located at the RPI campus in Troy,New York. This centrifuge, originallycommissioned in 1989 with sup-port from MCEER (then NCEER),has in-flight earthquake simulationcapability allowing base shaking tobe applied to the base of the model.It was recently selected by NSF to-gether with other earthquake en-

ground and foundation moved hori-zontally about 1 m, in others itmoves much less due to the con-straining effect of the superstruc-ture, or of the deep foundation'slateral stiffness including pilegroups and batter piles. The foun-dation may be exposed to large lat-eral soil pressures, includingespecially passive pressures fromthe nonliquefied shallow soil layerriding on top of the liquefied soil.In some cases, this soil has failedbefore the foundation with negli-gible bending distress and verysmall deformation of the founda-tion head and superstructure(Berrill et al., 1997); while in oth-ers the foundation has failed firstin bending (Figure 1) and/or hasexperienced excessive permanentdeformation and rotation at the pileheads. The observed damage andcracking to piles is often concen-trated at the upper and lowerboundaries of the liquefied soillayer where there is a suddenchange in soil properties, or at theconnection with the pile cap (Fig-

• Mourad Zeghal, AssistantProfessor and VivianKallou, Graduate Student,Civil and EnvironmentalEngineering Department,Rensselaer PolytechnicInstitute

• Masanori Hamada,Professor, WasedaUniversity, Japan

Engineering design firms, foundation and consulting en-gineers, hospital authorities, state transportation depart-ments, and port and harbor authorities will all be interestedin the results obtained through this research. The work ispart of a broader geotechnical task focusing on ways toremediate potentially dangerous sites and/or rehabilitatedeep foundations of hospitals and other structures vulner-able to earthquakes, by reducing or eliminating the effectsof soil liquefaction, including those due to permanent lat-eral and vertical ground deformations. Lateral spreadinghas been identified as a major hazard to deep foundations,and centrifuge modeling is a key tool to identify and quan-tify mechanisms, calibrate analyses and evaluate retrofit-ting strategies for pile foundations. This centrifuge-basedresearch provides the foundation component to the effortto mitigate the seismic hazard to hospitals with the help ofcost-effective and advanced materials and technologies.

89Pile Foundation Response to Lateral Spreading and Mitigation Strategies

gineering experimental sitesthroughout the U.S. to form theGeorge E. Brown, Jr. Network forEarthquake Engineering Simulation(NEES, www.eng.nsf. gov/nees). Ad-ditional information on the centri-fuge equipment used in thisresearch, results from otherprojects and the basic principles ofcentrifuge modeling, can be foundat the RPI web site (www.ce.rpi.edu/centrifuge), which also hasuseful links to other relevant websites; see also summary articles byDobry et al. (1995) and Dobry andAbdoun (1998, 2001). In additionto the centrifuge experimentsthemselves done at RPI, this cen-trifuge-based research has includedother analytical, laboratory, casehistory review and retrofitting strat-egy components, conducted eitherat Cornell University or in close co-operation between the RPI andCornell teams. The RPI-Cornelljoint centrifuge-based research onlateral spreading effects on piles

started in 1995 with support fromNCEER and NSF and has continuedsince then with current supportfrom both MCEER and NSF. Thetechnical discussion below is di-vided in three parts: case of pilebending response to lateral spread-ing controlled by the pressure ofthe liquefied soil, case of responsecontrolled by shallow nonliquefiedsoil layer, and pile retrofitting strat-egies and results.


� Figure 1. Damage to pile foundations due to lateral spreading under NFCH building, 1964Niigata earthquake, Japan (Hamada, 1992, 2001)

Pile 1 Pile 2N-ValuePile 1 Pile 2

NonliquefiableLayer

LiquefiableLayer

Firm Base

8.5

to

9.0m6.0m

to

6.5m

2.5m

to

29 kN 290 kN0.0

(m)

2.5

4.0

6.0

9.0

10.0

10 20

~0.7 m

~0.5 m

Pile Cap

1.6m

1.5m

( 1.5 m, say 1.5 m or

11.5 m below G.S. )

Pile 1 Pile 2N-ValuePile 1 Pile 2

NonliquefiableLayer

LiquefiableLayer

Firm Base

8.5

to

9.0m6.0m

to

6.5m

2.5m

to

3.0m

29 kN 290 kN0.0

(m)

2.5

4.0

6.0

9.0

10.0

10 20

~0.7 m

~0.5 m

Pile Cap

1.6m

1.5m

( 1.5 m, say 1.5 m or

11.5 m below G.S. )

>_

� Figure 2. 100 g-ton geotechnical centrifuge with in-flightshaking capability at RPI

90

Pile Bending ResponseControlled by theLiquefied Soil

Figure 3 shows centrifuge pileModel 3, simulating the bendingresponse of a pile foundation sub-jected to the lateral pressure of aliquefied soil due to lateral spread-ing. These and other experimentswere conducted using the rectan-gular, flexible-wall laminar box con-tainer sketched in Figure 3. Thislaminar box is comprised of a stackof up to 39 rectangular aluminumrings separated by linear rollerbearings, arranged to permit rela-tive movement between rings withminimal friction. In Model 3 as wellas in all other lateral spreadingspreading experiments, the laminarbox and the shaker under it are in-clined a few degrees to the proto-type horizontal direction tosimulate an infinite mild slope andprovide the shear stress biasneeded for a lateral spread. The flex-ibility of this box container is dem-onstrated by the large permanent

deformations and strains attainedin the experiments (Figure 5).

In the test of Figure 3, the soilprofile consists of two layers of fineNevada sand saturated with water:a top liquefiable layer of relativedensity, D

r = 40% and 6 m protoype

thickness, and a bottom slightlycemented nonliquefiable sand layerhaving a thickness of 2 m. The pro-totype single pile is 0.6 m in diam-eter, 8 m in length, has a bendingstiffness, EI = 8000 kN-m2, and isfree at the top. The pile model isinstrumented with strain gages tomeasure bending moments alongits length, and a lateral LVDT at thetop to measure the pile head dis-placement. The soil is instrumentedwith pore pressure transducers (pi-ezometers) and accelerometers, aswell as with lateral LVDTs mountedon the rings of the flexible wall tomeasure soil deformations in thefree field. A prototype inputaccelerogram consisting of 40 sinu-soidal cycles of a peak accelerationof 0.3 g was applied to the base,which liquefied the whole top layerin a couple of cycles and induceda permanent lateral ground surfacedisplacement in the free field ofabout 0.8 m.

Results of this experiment areshown in Figures 4 and 5. As soonas the top sand layer liquefied atthe beginning of shaking, it startedmoving laterally downslopethroughout the shaking, with themaximum displacement at all timesmeasured at the ground surface,and with this surface ground dis-placement increasing monotoni-cally with time to its final value D

H

= 0.8 m at the end of shaking. Themaximum bending moment alongthe pile at any given time occurredat the interface between the twosoil layers, that is at a depth of about

RPI 100 g-ton geotechnicalcentrifuge facility:

http://www.ce.rpi.edu/centrifugeThe complete visualization or

"movie" produced from thedata recorded in thecentrifuge test illustratedby the single frame ofFigure 5 can be viewed inthis web site. To see thismovie, after entering theweb site, go to Researchand then to Visualization.

NEES Initiative: http://www.eng.nsf.gov/nees

50 g

Input Motion

Z = 6

Z = 8

AccelerometerLVDTPore Pressure

TransducerStrain Gage

Nevada sand (40%)

α = 2

Z = 0 m

Slightly

Cemented sand

� Figure 3. Lateral spreading pile centrifuge model in two-layer soil profile(Abdoun, 1997)


6 m. Figure 4 shows the time his-tory of this prototype bendingmoment for Model 3, measured atz = 5.75 m; the plot reveals that themoment increased to a maximumM

max = 110 kN-m at a time,

tª ≈ 17 sec, with the moment de-creasing afterwards despite thecontinuation of shaking and thecontinuous increase of the soil de-formation in the free field. The pilehead displacement in the same fig-ure also reached a maximum atabout 17 sec and decreased after-wards. Clearly at this time the liq-uefied soil reached its maximumstrength and applied a maximumlateral pressure to the pile, with thesoil flowing around the pile, exhib-iting a smaller strength and apply-ing a smaller pressure afterwards;as a result, the model pile bouncedback and the bending momentsdecreased. The two photos in Fig-ure 6 - taken after the centrifugetests - illustrate this flow of lique-fied soil around the pile in othertwo models where colored sandhad been placed around the pile.

Figure 5 summarizes the state ofthe system during a repeat ofModel 3, at the time when the pilehead displacement and the bend-ing moment at a depth of about 6m attained their maximum values.This is a frame taken from the visu-alization of the experiment pro-duced from the measurements (thewhole visualization may be viewedat the RPI centrifuge web site). Thedisplaced shape of the box con-tainer indicates the lateral spread-ing in progress, with concentrationof permanent shear straining in thelower part of the liquefied soil; thisbox shape was obtained from thelateral LVDTs placed on the sidewalls. This distorted shape is alsocopied as a white mesh to the right

side of the pile for direct compari-son between ground and pile dis-placements as well as to visualizethe larger movement of the liquefied

Mom

ent

(kN

-m)

0

15

30

45

0

20

40

60

80

0 10 20 30 40

0

40

80

120

Mom

ent

(kN

-m)

0

15

30

45

0

20

40

60

80

Time (sec)

0 10 20 30 40

0

40

80

120

Ground surface

Pile head

Z = 5.75 m

Dis

p.(c

m)

Dis

p.(c

m)

� Figure 4. Prototype lateral displacement of soil and pile and ground surface,and pile bending moment at a depth of 5.75 m in model of Figure 3 (Abdoun,1997)

� Figure 5. Frame taken out of visualization of two-layer centrifuge model ofFigure 3, produced from the recorded data (Kallou et al., 2001; to seewhole visualization, visit http://www.ce.rpi.edu/centrifuge)

92

soil flowing around the pile, com-pared with the displacement of thepile itself. The blue color in the up-per part of the loose sand layer indi-cates complete liquefaction asmeasured by the piezometers, whilethe green color in the lower part ofthe layer indicates lower excess porepressure due to dilative cyclic stress-strain response of the liquefied

sand in that part of the shaking cycle.At other times corresponding to dif-ferent parts of the shaking cycle, thewhole layer is blue and hence com-pletely liquefied.

In addition to Model 3 summarizedin Figures 3 to 5, similar centrifugetests of a single pile with a pile cap,with densification around the pile tosimulate pile driving, and with 2x2pile groups indicated that, while M

max

still occurs at a depth of about 6 msometime during the shaking, thevalue of M

max increases with the area

of pile foundation exposed to the soillateral pressure and decreases in thepile groups due to the contributionto moment of the axial forces in thepiles (frame effect). Simple limit equi-librium calculations with a constantassumed maximum pressure of theliquefied soil along the pile, p

l , indi-

cate that values of pl of the order of

10 kPa explain well all measuredtrends and values of M

max in this se-

ries of centrifuge tests.

Direction of the free fieldlateral displacement

Model 5a Model 5b

� Figure 6. Photos showing flow of liquefied sand around the pile in the downslope direction in two-layer centrifuge models (Abdoun, 1997). The photos were taken after the test in models wherecolored sand had been placed in a circular ring around the pile

� Figure 7. Concept used to develop undrained triaxial extension model forthe lateral loading of liquefied soil on the pile (Goh and O’Rourke, 1999;Goh, 2001)


The physical origin and basicmechanisms determining the behav-ior of the liquefied soil, including thelateral pressure on pile foundationsand values such as p

l and M

max mea-

sured in these centrifuge tests, are notyet well understood and are the sub-ject of intense research. The Cornellteam has proposed the explanationsketched in Figure 7, with p

l and M

max

controlled by the peakundrained shear strength ofthe saturated sand loaded inthe extension mode (Goh andO'Rourke 1999; Goh, 2001).Based on p-y curves generatedanalytically from triaxial exten-sion tests conducted at Cornellusing the same Nevada sandand relative density of the cen-trifuge tests, nonlinear Beam-on-Winkler -Foundation (BWF)analyses of centrifuge Model 3were able to predict closelythe measured bending re-sponse (Figure 8).

Pile Bending ResponseControlled by ShallowNonliquefied Layer

Figure 9 shows centrifuge Model2, where a strong shallownonliquefied soil layer increases sig-nificantly the bending response ofthe pile foundation to lateralspreading. The shallow top layer

Input Motionα = 2

LVDTPore Pressure

TransducerStrain Gage Accelerometer

50 g

Nevada sand

(Dr=40%)

Slightly

Cemented sand

Slightly

Cemented sand

Input Motionα = 2

LVDTPore Pressure


50 g

Nevada sand

(Dr=40%)

Slightly

Cemented sand

Slightly

Cemented sand DH =40cm

-200 0 200

0

2

4

6

8

10

So

il d

epth

(m

)

Moment (kN-m)

� Figure 8. Comparison between predicted and measured pile bending moment incentrifuge model of Fig. 3 at the lower boundary of liquefied soil using triaxialextension undrained loading approach (Goh and O’Rourke, 1999; Goh, 2001)

� Figure 9. Lateral spreading pile centrifuge model in three-layer soil profile (Abdoun, 1997)

94

consists of a 2-m thick (in proto-type units), free draining, slightlycemented sand. Model 2m, notshown here, is similar to Figure 9but with a mass added aboveground to evaluate the combinedeffects of lateral spreading and in-ertial loading.

Figures 10-11 summarize themain characteristics of the bendingresponse of Model 2 (only lateralspreading), which is also typical ofother pile models tested in this 3-layer soil profile. The same as inModel 3 discussed before, the 6-mthick noncemented sand layer liq-uefied early in the shaking afterwhich the lateral spreading in-creased monotonically, reaching avalue D

H = 0.7 m at the end of shak-

ing (Figure 11). The pile bendingmoments in the top 2 m first in-creased with time of shaking andthen decreased after passive failureof the top nonliquefiable layeragainst the pile (Figure 10); whilethe bending moments near thebottom increased monotonicallyand never decreased, as the bottom

nonliquefiable layer did not fail. Thevalues of maximum bending mo-ments at 2 m and 8 m are close to300 kN-m, much greater than thosemeasured in 2-layer tests such asshown in Figure 3, which did notexceed 170 kN-m even when a pilecap was added.

The shapes of the bending mo-ment profiles at various times pre-sented in Figure 10 indicate that thedeformed shape of the pile had adouble curvature caused by the topand bottom soil layers loading thepile in opposite directions. Thisdouble curvature was confirmed bythe fact that when the top soil layerfailed, the pile head and cap"snapped" in the downslope direc-tion (Figure 11), showing that atvery shallow depths, the pile waspushing the soil rather than theother way around. Both the passivefailure of the top layer and themoment concentrations at the topand bottom boundaries of the liq-uefied layer indicated by the figuresare consistent with the experiencefrom earthquake case histories.

-200 0 200

0

2

4

6

8

10-200 0 200

0

2

4

6

8

10-200 0 200

0

2

4

6

8

10-100 -50 0 50 100

0

2

4

6

8

10

DH =40cm

Moment ( kN -m)

Soi

ld

epth

(m)

Soi

ld

epth

(m)

Model 2

-200 0 200

0

2

4

6

8

10

Model 2m

DH =20cm DH =70cm

-200 0 200

0

2

4

6

8

10-200 0 200

0

2

4

6

8

10

DH =10cm

-100 -50 0 50 100

0

2

4

6

8

10

� Figure 10. Measured bending moment response along pile in lateral spreading centrifuge models without(Model 2) and with (Model 2m) inertial loading (Wang, 2001)


These moment concentrations arealso predicted by theory (e.g.,Meyersohn, 1994; Meyersohn et al.,1992; Debanik, 1997). Another in-teresting aspect of Figures 10 and11 is that the bending momentsvary linearly within the liquefiedlayer, suggesting that they are es-sentially controlled by the loadingof the top and bottom layers, withthe pressure of the liquefied soilbeing negligible. The values of M

max

at z = 2 m and z = 8 m are higherthan the corresponding values ofM

max at z = 6 m for the 2-layer soil

profiles, such as in Figure 4, whichwere controlled by the strength ofthe weaker liquefied soil. The au-thors have successfully calibrateda limit equilibrium method to pre-dict M

max in some of these 3-layer

pile centrifuge models, after incor-porating basic kinematic consider-ations to allow for the change inpile curvature (and hence of thesign of the passive soil pressure onthe pile) within the topnonliquefied soil layer.

The comparisons in Figures 10and 11 between Models 2 and 2mreveal interesting aspects of therole played by superstructural in-ertia in the lateral spreading pro-cess. For depths greater than 2 or3 m, the effect of lateral spreadingpredominates and the inertial load-ing due to the mass can be ignored.However, at shallow depths of lessthan 2 m, that is in the topnonliquefiable layer, the bendingmoments of the two centrifugemodels are very different, withthose of Model 2m changing rap-idly with time due to the combinedeffect of inertia and lateral spread-ing. However, even in Model 2m themaximum moments still tend toconcentrate at the upper and lowerboundaries of the liquefied layer.

Despite the rapid change in shal-low bending moments due to themass, when the top soil layer failedin passive in Model 2m, the pilehead and cap "snapped" in thedownslope direction, exactly thesame as in Model 2 (Figure 11),showing that the soil failure mecha-nism was still controlled by lateralspreading.

Another factor which has beenstudied in the centrifuge for the 3-layer soil model is the influence ofthe superstructural stiffness thatfield case histories has shown to beimportant. This has been done bythe addition of lateral and rotationalsprings above ground connected tothe pile head, such as spring k inFigure 12 (Ramos, 1999). As ex-pected, the analysis of these cen-trifuge results has requiredsignificant kinematic consider-ations and parameters, even whensimple limit equilibrium calcula-tions are conducted. On the other

0 20 40 60 80 100

0

2

4

6

8

100 20 40 60 80 100

0

2

4

6

8

10

. ..

cemented sand

(Dr=40%)Nevada sand

Slightly

Model 2

0 20 40 60 80 100

0

2

4

6

8

10

cemented sandSlightly


(Dr=40%)Nevada sand


Model 2m

0 20 40 60 80 100

0

2

4

6

8

10

Measured Soil Disp. Measured Pile Disp.Estimated Pile Disp.

Soi

l dep

th (

m)

Soi

l dep

th (

m)

Displacement (cm) Displacement (cm)

� Figure 11. Snapping of pile in downslope direction in centrifuge modelswithout (Model 2) and with (Model 2m) inertial loading (Wang, 2001)

96

hand, some aspects of the analysisbecome simpler compared withthe case of k = 0 (Figure 3), in thatif the value of k is large enough,there is no double curvature of thepile at very shallow depths, and no"snapping" of the pile in thedownslope direction as in Figure11. That is, the constraining effectof spring k forces the lateral pres-sure of the nonliquefied layer onthe pile to act in the samedownslope direction at all depthsbetween 0 and 2 m.

Pile RetrofittingStrategies and Results

Both case histories and centri-fuge models have shown the greatimportance of the shallownonliquefiable soil in increasing thebending response of the pile foun-dation. Therefore, a promising reha-bilitation approach of existingfoundations is to replace the shal-low soil in a trench around pilesand pile cap by a frangible mate-

rial that will yield under constantlateral soil forces (Figure 13a). Thiswould decrease both bending mo-ments and foundation deforma-tions while allowing the groundlateral spreading to take place with-out interference from the founda-tion. As this retrofitting scheme alsodecreases the lateral resistance ofthe foundation to inertial loading,the desired frangible material se-lected, while yielding to static forceshould remain resilient under thetransient inertial loading. Alterna-tively, the trench surrounding thefoundation with frangible materialmay be located at some distancefrom the foundation so as to in-crease the resistance to inertialloading (Figure 13b).

A series of centrifuge models ofa single pile with pile cap in the 3-layer soil profile were conductedusing the retrofitting setups of Fig-ure 13, labeled respectively Strat-egy 1 and Strategy 2. Theseexperiments are listed in Table 1,which include also the benchmarknonretrofitted Models 2 and 2m,

� Input motion:

– 0.3 g

– 40 cycles

– 2 Hz

� Free rotation:

– Model 3free-24

– Model 3free- ∞– Model 3free- ∞

cap

� Fixed rotation:

– Model 3fixed -0

50 g

Input Motion

Z = 2

Z = 8

Z = 10

AccelerometerLVDTPore PressureTransducer

Strain Gage

Nevada sand (40%)

α = 2

Z = 0 mSlightly

Cemented sand

Slightly

Cemented sand

kkk

� Figure 12. Lateral spreading pile centrifuge model incorporating effect of superstructural stiffness(Ramos, 1999)


already discussed. Models 2r1,2mr1a and 2mr1b were done withStrategy 1, without and with a massabove ground, and Models 2r2 and2mr2 were conducted with Strat-egy 2. In both cases, a soft clay wasplaced in a trench either directlyaround or at some distance fromthe foundation. In future tests theuse of an artificial frangible mate-rial with higher resistance to tran-sient loading is planned (Wang,2001).

Figures 14 and 15 illustrate mea-surements and observations ob-tained from Models 2r1 and 2r2.The free field lateral ground dis-placements during shaking in cen-trifuge tests without and with pilefoundation retrofitting were essen-tially the same (Figure 11), consis-tent with the assumption that theyrepresent truly free field response.Figure 14 compares the bendingmoment response without andwith retrofitting. As expected, there

Input Motionα = 2

LVDTPore Pressure


50 g

Z = 8

Z = 10

Nevada sand

(Dr=40%)

Z = 0 m

Slightly

Cemented sand

Slightly

Cemented sandZ = 2

Input Motionα = 2

LVDTPore Pressure


50 g

Z = 8

Z = 10

Nevada sand

(Dr=40%)

Z = 0 m

Slightly

Cemented sand

Slightly

Cemented sandZ = 2

Input Motionα = 2

LVDTPore Pressure


50g

Z = 8

Z = 10

Nevada sand

(Dr=40%)

Z = 0 m

Slightly

Cemented sand

Slightly

Cemented sandZ = 2

Input Motionα = 2

LVDTPore Pressure


50g

Z = 8

Z = 10

Nevada sand

(Dr=40%)

Z = 0 m

Slightly

Cemented sand

Slightly

Cemented sandZ = 2

a) Model 2r1 (Strategy 1)

b) Model 2r2 (Strategy 2)

� Figure 13. Lateral spreading pile centrifuge models to evaluate retrofitting strategies(Wang, 2001)

98

is a dramatic reduction in the mo-ments in the top 2 m of pile in con-tact with the nonliquefiable soil.The maximum moment there wasclose to 300 kN-m in Model 2 andbecomes about 10 kN-m after ret-rofitting. A smaller reduction is alsoobserved for the maximum bend-

ing moment at the lower boundaryof the liquefied layer, at about 8 mdepth. Similarly, the pile head dis-placements at the end of the testswere reduced by a factor of two byretrofitting (from 85 to 40-50 cm,with D

H = 70 to 80 cm for the soil

in the free field). The photo ofModel 2r1 in Figure 15, taken afterthe test, illustrates the correspond-ing "crunching" of the soft clayagainst the pile cap in the upslopeside, and opening of a gapdownslope between soil and foun-dation. However, the counterpart tothis reduction in permanent bend-ing response to lateral spreading ofthe pile foundation was an increaseof transient pile accelerations anddisplacements, especially in thetests incorporating inertial loading(Models 2mr1a,b and 2mr2, notshown), due to the reduced lateralground support in the top 2 m ofthe foundation; future tests willaddress this problem.

DH =20cm DH =40cm

-200 0 200

0

2

4

6

8

10-200 0 200

0

2

4

6

8

10

DH =70cmDH =60cm

-200 0 200

0

2

4

6

8

10-200 0 200

0

2

4

6

8

10

Model 2 Model 2r1 Model 2r2

So

il d

epth

(m

)

Moment (kN-m)

Gap

Direction of the free field lateral displacement

� Figure 14. Measured bending moment response along pile in lateral spreading centrifuge models without (Model 2)and with (Models 2r1 and 2r2) foundation retrofitting (Wang, 2001)

� Figure 15. Plan view of retrofitted pile cap and ground after the test,Model 2r1 (Wang, 2001)


Conclusions andFuture Research

Case histories during earth-quakes have shown the signifi-cance of lateral spreading incausing damage to deep founda-tions and supported structures dur-ing earthquakes. The complexity ofthe problem requires use of cen-trifuge physical modeling to clarifymechanisms, quantify relations andcalibrate analysis and design pro-cedures. Centrifuge results so farhave clarified the deep foundationresponse, have shown significantagreement with field experience,and are being used to calibrate limitequilibrium and Beam-on-Winkler-Springs (p-y) analytical methods.Specifically, the importance of the

Test No. Cap Mass Retrofitting Comments2 Yes No No

2r1 Yes No Yes2r2 Yes No Yes2m Yes Yes No

2mr1a Yes Yes Yes2mr1b Yes Yes Yes Repeat of 2mr1a2mr2 Yes Yes Yes

shallow nonliquefiable soil layerriding on top of the liquefied soilin increasing foundation bendingresponse has been clarified. Retro-fitting strategies are being evalu-ated in the centrifuge, aimed atmitigating the effect of lateralspreading associated with the pres-sure of this shallow layer while pre-serving needed lateral resistance toinertial loading. Additional work isneeded to understand and quantifythe response of nonretrofitted andretrofitted pile foundations, withcentrifuge model experimentscombined with case studies andtheory, toward improving the state-of-practice of seismic design andretrofitting of deep foundationsagainst liquefaction.

ReferencesAbdoun, T. H., (1997), Modeling of Seismically Induced Lateral Spreading of Multi-

Layer Soil Deposit and Its Effect on Pile Foundations, Ph.D. Thesis, Dept. of CivilEngineering, Rensselaer Polytechnic Institute, Troy, NY.

Berrill, J. B., Christensen, S. A., Keenan, R. J., Okada, W. and Pettinga, J. K., (1997), Lateral-spreading Loads on a Piled Bridge Foundation, Seismic Behavior of Ground andGeotechnical Structures, (Seco E Pinto, ed.), pp. 173-183, Balkema, Rotterdam.

Debanik, C., (1997), Pile Response to Liquefaction-Induced Lateral Spread, Ph.D. Thesis,Cornell University, Ithaca, NY.

Dobry, R., Taboada, V. and Liu, L., (1995), “Centrifuge Modeling of Liquefaction EffectsDuring Earthquakes,” Keynote lecture, Proc. 1st Intl. Conf on EarthquakeGeotechnical Engineering (K. Ishihara, ed.), Tokyo, Japan, Nov. 14-16, Vol. 3, pp. 1291-1324.

� Table 1. Program of Centrifuge Tests to Evaluate Retrofit Strategies 1 and 2 (Wang, 2001)

100

References (Cont’d)Dobry, R. and Abdoun, T. H., (1998), “Post-Triggering Response of Liquefied Soil in The

Free Field and Near Foundations,” State-of-the-art paper, Proc. ASCE 1998 SpecialtyConference on Geotechnical Earthquake Engineering and Soil Dynamics (P.Dakoulas, M. Yegian and R. D. Holtz, eds.), University of Washington, Seattle,Washington, August 3-6, Vol. 1, pp. 270 – 300.

Dobry, R. and Abdoun, T. H., (2001), “Recent Studies on Seismic Centrifuge Modelingof Liquefaction and its Effect on Deep Foundations,” State-of-the-Art Paper, Proc. 4thIntl. Conf. on Recent Advances in Geotechnical Earthquake Engineering and SoilDynamics and Symposium to Honor Prof. W.D.L. Finn (S. Prakash, ed.), San Diego,CA, March 26-31, SOAP-3, Vol. 2.

Goh, S. H., (2001), Soil-pile Interaction During Liquefaction-Induced Lateral Spreading,PhD Thesis, Cornell University, Ithaca, NY.

Goh, S.-H. and O' Rourke, T. D., (1999), “Limit State Model for Soil-Pile Interaction DuringLateral Spread,” Proc. Seventh U.S.-Japan Workshop on Earthquake Resistant Designof Lifeline Facilities and Countermeasures Against Soil Liquefaction, Seattle, WA,August 15-17, Technical Report MCEER-99-0019, Multidisciplinary Center forEarthquake Engineering Research, University at Buffalo, (O'Rourke, Bardet andHamada, eds.), pp. 237-260.

Hamada, M., (1992), “Large Ground Deformations and their Effects on Lifelines: 1964Niigata Earthquake,” Ch. 3 of Case Studies of Liquefaction and Lifeline PerformanceDuring Past Earthquakes, Vol. 1: Japanese Case Studies, (Hamada and O’Rourke,eds.), 3-1 to 3-123.

Hamada, M., (2001), Personal Communication.

Hamada, M. and O’Rourke, T. D., (eds.) (1992), “Case Studies of Liquefaction and LifelinePerformance During Past Earthquakes,” Vol. 1: Japanese Case Studies, Technical ReportNCEER-92-0001, National Center for Earthquake Engineering Research, University atBuffalo, February.

Kallou, V., Abdoun, T. H., Zeghal, M., Oskay, C. and Dobry, R., (2001), Visualization ofCentrifuge Models of Lateral Spreading and Pile Bending Response (inpreparation).

Meyersohn, W. D., (1994), Pile Response to Liquefaction Induced Lateral Spread, Ph.D.Thesis, Dept. of Civil and Environmental Engineering, Cornell University, Ithaca, NewYork.

Meyersohn, W. D., O’Rourke, T. D. and Miura, F., (1992), “Lateral Spread Effects onReinforced Concrete Pile Foundations,” Proc. 5th US-Japan Workshop on EarthquakeDisaster Prevention for Lifeline Systems, Tsukuba, pp. 173-196.

O’Rourke, T. D. and Hamada, M., (eds.), (1992), “Case Studies of Liquefaction and LifelinePerformance During Past Earthquakes,” Vol. 2: United States Case Studies, TechnicalReport NCEER-92-0002, National Center for Earthquake Engineering Research,University at Buffalo, February.

Ramos, R., (1999), Centrifuge Study of Bending Response of Pile Foundation to aLateral Spread Including Restraining Effect of Superstructure, Ph.D. Thesis, Dept.of Civil Engineering, Rensselaer Polytechnic Institute, Troy, NY

Tokimatsu, K., Mizuno, H. and Kakurai, M., (1996), “Building Damage Associated withGeotechnical Problems,” Soils and Foundations, pp. 219-234, January.

Wang, Y., (2001), Evaluation of Pile Foundation Retrofitting Against Lateral Spreadingand Inertial Effects During Liquefaction Using Centrifuge Models, MS Thesis, Dept.of Civil Engineering, Rensselaer Polytechnic Institute, Troy, NY.


Other Publications

Abdoun, T.H., Dobry, R., O'Rourke, T.D. and Chaudhuri, D., (1996), “Centrifuge Modelingof Seismically-induced Lateral deformation During Liquefaction and its Effect on a PileFoundation,” Proc. Sixth Japan-U.S. Workshop on Earthquake Resistant Design ofLifeline Facilities and Countermeasures Against Soil Liquefaction, Technical ReportNCEER-96-0012, Multidisciplinary Center for Earthquake Engineering Research,University at Buffalo, pp. 525-539.

Abdoun, T.H. and Dobry, R., (1998), “Seismically Induced Lateral Spreading of Two-layer SandDeposit and its Effect on Pile Foundations,” Proc. Intl. Conf. Centrifuge'98 (T. Kimura,O. Kosakabe and J. Takemura, eds.), Tokyo, Japan, Sept. 23-25, Vol. 1, pp. 321-326.

Dobry, R., (1994), “Foundation Deformation Due to Earthquakes,” Proc. ASCE SpecialtyConference on Settlement ’94, College Station, TX, June 16-18, pp. 1846-1863.

Dobry, R., (1995), “Liquefaction and Deformation of Soils and Foundations under SeismicConditions,” State-of-the-art paper, Proc. Third Intl. Conf on Recent Advances inGeotechnical Earthquake Engineering and Soil Dynamics (S. Prakash, ed.), St. Louis,MO, April 2-7, Vol. III, pp. 1465-1490.

Dobry, R., Abdoun, T.H. and O'Rourke, T. D., (1996), “Evaluation of Pile Response Dueto Liquefaction-Induced Lateral Spreading of the Ground,” Proc. 4th Caltrans SeismicResearch Workshop, Sacramento, CA, July, 10 pages.

Dobry, R., Van Laak, P., Elgamal, A.-W., Zimmie, T. F. and Adalier, K, (1997), “The RPIGeotechnical Centrifuge Facility,” NCEER Bulletin, April, pp. 12-17.

Elgamal, A.-W., Dobry, R., Van Laak, P. and Nicolas-Font, J., (1991), “Design, Constructionand Operation of 100 g-ton Centrifuge at RPI,” Proc. Intl. Conf. Centrifuge'91 (H.-Y.Ko and F. G. McLean, eds.), pp. 27-34.

O' Rourke, T., Meyersohn, W. D., Shiba, Y. and Chaudhuri, D., (1994), Evaluation of PileResponse to Liquefaction-Induced Lateral Spread, Tech. Rept. NCEER-94-0026,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo, pp.445-455.

Ramos, R., Abdoun, T. H. and Dobry, R., (1999), “Centrifuge Modeling of Effect ofSuperstructure Stiffness on Pile Bending Moments Due to Lateral Spreading,” Proc.Seventh U. S.-Workshop Workshop on Earthquake Resistant Design of LifelineFacilities and Countermeasures Against Soil Liquefaction, Seattle, WA, August 15-17,10 pages.

Ramos, R., Abdoun, T. H. and Dobry, R., (2000), “Effect of Lateral Stiffness of Superstructureon Bending Moments of Pile Foundation Due to Liquefaction-induced LateralSpreading,” Proc. 12th World Conf. on Earthquake Engineering, Auckland, NewZealand, Jan. 30 - Feb. 4, 8 pages.

Taboada, V.M. and Dobry, R., (1998), “Centrifuge Modeling of Earthquake-Induced LateralSpreading in Sand,” J. Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 124,No.12, pp. 1195-1206.

Van Laak, P., Elgamal, A.-W. and Dobry, R., (1994a), “Design and Performance of anElectrohydraulic Shaker for the RPI Centrifuge,” Proc. Intl. Conference Centrifuge'94,Singapore, August 31-Sept. 2, (C.F. Leung, F.H. Lee and T.S. Tan, eds.), A.A. Balkema,Rotterdam, pp. 139-144.

Van Laak, P., Taboada, V., Dobry, R. and Elgamal, A.-W., (1994b), “Earthquake CentrifugeModeling Using a Laminar Box,” Dynamic Geotechnical Testing II (R. J. Ebelhar, V. P.Drnevich and B. L. Kutter, eds.), ASTM STP 1231, pp. 370-384.

Van Laak, P., Adalier, K., Dobry, R. and Elgamal, A.-W., (1998), “Design of RPI's LargeServohydraulic Shaker,” Proc. Intl. Conf. Centrifuge'98 (T. Kimura, O. Kosakabe and J.Takemura, eds.), Tokyo, Japan, Sept. 23-25, Vol. 1, pp. 105-110.

103

Research Objectives

This article is a summary of the progress made during 2000-2001 byMCEER researchers working in the subtask of facilitating technologies.These research progresses should be viewed from the context that thelong term (3-4 year) objective is to complete an MCEER monograph onAnalysis and Design of Buildings with Added Energy Dissipation Sys-tems. Although each individual researcher is advancing the state-of-the-art knowledge with his respective graduate students and researchcollaborators, it is the total systems integrated effort directed toward thepracticing professional that underpins the projects within this group. Thisundertaking is possible by using the “Center Approach” in earthquakeengineering research.

Analysis and Design of Buildings withAdded Energy Dissipation Systems

by Michael C. Constantinou, Gary F. Dargush, George C. Lee (Coordinating Author), Andrei M. Reinhornand Andrew S. Whittaker


MCEER’s research program 2 on the seismic retrofit of hospitals concen-trates on developing cost-effective retrofit strategies for critical fa-

cilities using new and emerging materials and enabling technologies.Current emphasis is given to hospitals that should remain functional dur-ing and immediately after damaging and/or destructive earthquakes.

The major disciplinary components of this program are geotechnical,structural, nonstructural, advanced technologies and high performancematerials, socio-economic issues and systems integration. One importantresearch task concerning the methods of analysis and design of buildingswith added emerging materials (e.g., composite infill walls) and/or en-abling technologies (e.g., damping devices) is carried out by a group ofresearch investigators under the title “Facilitating Technologies.” This taskis the heart of the systems integrated approach concerning the perfor-mance of a system (i.e., the performance of hospital buildings and con-tents with added earthquake protective systems so that the medicalfunctions can be carried out). To develop retrofit strategies for buildingsby using added materials and/or enabling devices, practicing engineersneed to know how to choose appropriate technologies to satisfy buildingperformance indices and/or objectives cost-effectively for given earthquakerisk. In view of FEMA 273/274 and NEHRP 2000, which encourage the

Michael C. Constantinou,Professor and Chair,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo, Panos Tsopelas,Assistant Professor,Catholic University ofAmerica, WilhelmHammel, SeniorConsultant, KPMG,Germany, and Ani N.Sigaher, Ph.D. candidate,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Gary F. Dargush, AssociateProfessor. Ramesh Sant,Ph.D. Candidate, andRajesh Radhakrishnan,Graduate Student,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Research Teamcontinued on page 2

104

use of emerging technologies toachieve building performance ob-jectives, there is a need to developprinciples and design guidelines forthese engineers. This is the funda-mental rationale of the researchtask on facilitating technologies.

MCEER investigators have mademany key contributions to the twomost advanced codes and guide-lines related to the implementationof passive energy dissipation sys-tems: FEMA 273/274 Guidelinesfor the Seismic Rehabilitation ofBuildings, published in 1998, andthe NEHRP 2000 Guidelines forSeismic Regulations for NewBuildings and Other Structures,that will be published in the nextfew months.

The FEMA 273 Guidelines and274 Commentary represented theculmination of more than a decadeof work funded by the FederalEmergency Management Agency,the National Science Foundationand other agencies. These docu-ments provide structural engineerswith new information on proce-dures for the analysis, evaluation,and design of existing and retrofitconstruction. Information is alsopresented on performance-basedearthquake engineering, modelingand analysis, steel, concrete, andtimber structures, foundations,

and nonstructural components.MCEER researchers have contrib-uted to FEMA 273/274 in two ar-eas: modeling and analysis, andseismic isolation and damping sys-tems (see Constantinou et al., 1998and Tsopelas et al., 1997).

The 2000 NEHRP RecommendedProvisions for Seismic Regulationsfor New Buildings and Other Struc-tures includes robust procedures forthe analysis and design of passive en-ergy dissipation systems (Appendixto Chapter 13) using force-basedmethods of analysis that are consis-tent with those methods used for theanalysis and design of conventionalconstruction. The development offorce-based analysis and component-checking methods for highly nonlin-ear or velocity-dependent energydissipation devices proved to be amost demanding task. MCEER re-searchers have developed the tech-nical underpinnings of the methodspresented in NEHRP 2000 (seeRamirez et al., 2000).

Since the publication of FEMA273/274 in 1998, many practicingengineers have been interested inusing these guidelines to retrofitexisting buildings with base isola-tion and/or energy dissipation de-vices. As a result, several majorretrofit projects have been com-pleted and their results publicized,

George C. Lee, Director,MCEER, and Samuel P.Capen Professor ofEngineering, Departmentof Civil, Structural andEnvironmentalEngineering, University atBuffalo, Mai Tong, SeniorResearch Scientist, MCEERYihui Wu, ResearchScientist, MCEER,and WeiLiu, Ph.D. candidate,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Andrei Reinhorn, Professor,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo, KatrinWinkelmann,VisitingResearch Associate,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo from the Universityof Darmstadt andMettupalayamSivaselvan, Ph.D.candidate, Department ofCivil, Structural andEnvironmentalEngineering, University atBuffalo

Andrew S. Whittaker,Associate Professor, OscarRamirez, GraduateStudent, and Juan Gomez,Graduate Student,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Structural engineers who develop cost-effective retrofit de-signs for existing buildings will find information regardinghow to select the best passive energy dissipation device sys-tem for a given type of building (e.g., high rise, low-mediumheight, steel, reinforced concrete, masonry, timber, rela-tively symmetrical, irregularly-shaped, etc.) and for givenearthquake risk very useful. Other procedures, such as howto optimally distribute damping devices throughout a struc-ture and how to consider multi-performance indices are alsoof critical importance.

105Analysis and Design of Buildings with Added Energy Dissipation Systems

particularly in the area of base iso-lation systems, in which MCEERresearchers (most notably, M.Constantinou and A. Whittaker)have acted as consultants. Becausebase isolation technology has alonger history of practical imple-mentation for buildings andbridges, more projects have beensuccessfully completed. However,only limited reports are available inthe open literature on passive en-ergy dissipation devices applied tobuildings. This is primarily due tothe relatively short history in ex-perience of practical applicationsand the fact that many availabledevices/approaches are reported inresearch publications, many devel-oped and/or improved by MCEERresearchers.

During the past two decades,many advances have been reportedon the performance and vibrationreduction properties of passiveenergy devices. As noted earlier,MCEER researchers have played asignificant role in the developmentof a variety of devices. This “device-based” line of pursuit will be con-tinued as new ideas, materials and/or devices become available.

A new systems-based approach isnow being undertaken to provideanswers to the many questions onchoosing appropriate devices. Re-garded as the “building-device sys-tem based” consideration, thesestudies emphasize the perfor-mance of the building with addedpassive energy dissipation devices.This new systems-based approachis the central theme of the researchtask on facilitating technologies,and this paper focuses on thesenew approaches.

To develop analytical models orto carry out experimental observa-tions for the systems-based study

is considerably more complicatedthan to model or to test the behav-ior of a single device.

First, existing buildings them-selves have different dynamic char-acteristics and are complicated tomodel. The performance of thedevices cannot be generalizedbased on one simple structure.Many multiple degree of freedom(MDOF) systems cannot be treatedas single DOF systems (decouplingassumption) with sufficient accu-racy. To develop simplified finiteelement (FE) analysis models forcomplicated MDOF systems is itselfa challenge.

When energy dissipation devicesare implemented in these MDOFstructures, the total building-devicesystems are generally nonlinear sys-tems. Much creativity and funda-mental research in structuraldynamics principles have to be pur-sued to develop a reasonablysimple and accurate analysis anddesign procedure. The current,MCEER studies may be groupedinto three separate categories.

Type I Projects: New Devices andSystems —continued developmentof new ideas in passive energy dis-sipation devices and semi-activesystems. The semi-active systemscan extend the range of traditionalpassive energy dissipation systemsand can be combined to providean added fail-safe feature to thesesystems.

Type II Projects: MDOF Modelingof Building-Device Systems—This cat-egory of studies is concerned withthe behavior and design of thebuilding-device system that mustbe studied by using multiple DOFmodels, including when the sys-tems are nonproportionallydamped. These studies providequantitative information on the er-

• Martin Johnson,GroupManager, EQEInternational

• Ali Karakaplan,President, LARSA, Inc.

• Charles Kircher,Principal, Charles Kircherand Associates

• Douglas Taylor,President, Taylor Devices,Inc.

• U.S. Department ofCommerce

106

rors when such buildings are ap-proximated by decoupled singleDOF systems.

Type III Projects: Analysis and De-sign Software—MCEER will con-tinue its development of analysisand design software for buildingswith added energy dissipation sys-tems. Some projects are fundamen-tal in nature to establish newapproaches while others empha-size the development of user-friendly simplified procedures fordesign professionals with accept-able accuracy.

The findings from research inthese three categories will culmi-nate in a monograph on Analysisand Design of Buildings withAdded Energy Dissipation Systemfor the design professional. Thefollowing sections provide brief de-scriptions of the progress made incurrent projects.

Scissor-Jack SeismicEnergy DissipationSystem(Type I Project)

Energy dissipation systems arebeing employed in the UnitedStates to provide enhanced protec-tion for new and retrofit buildingand bridge construction. The hard-ware utilized includes yielding steeldevices, friction devices, viscoelas-tic solid devices and mostly, so far,viscous fluid devices.

Engineers are familiar with andhave extensively used diagonal andchevron brace configurations for thedelivery of forces from energy dissi-pation devices to the structuralframe (Soong and Dargush, 1997;Constantinou et al., 1998). New con-figurations have been developed

which offer certain advantages, eitherin terms of cost of the energy dissi-pation devices, or in terms of archi-tectural considerations such as openspace requirements. Particularly, stiffstructural systems under seismicload or structural systems underwind load undergo small drift andthe required damping forces arelarge. This typically results in largerdamping devices and accordingly,greater cost. In other cases, energydissipation devices cannot be usedin certain areas due to open spacerequirements and the ineffective-ness of damping systems wheninstalled at near-vertical configura-tions.

Two recently developed configu-rations, the toggle-brace and thescissor-jack energy dissipation sys-tem configurations, offer advan-tages that overcome theselimitations. Both utilize innovativemechanisms to amplify displace-ment and accordingly lower forcedemand in the energy dissipationdevices. However, they are morecomplex in their application sincethey require more care in theiranalysis and detailing. The theoryand development of these systemshas been described inConstantinou et al. (2001) andConstantinou and Sigaher (2000).This section briefly presents thesenew configurations and comparesthem with the familiar chevronbrace and diagonal configurations.

The toggle-brace and scissor-jacksystems are configurations for mag-nifying the damper displacementso that sufficient energy is dissi-pated with a reduced requirementfor damper force. Conversely, theymay be viewed as systems for mag-nifying the damper force throughshallow truss configurations and


• Task 2.4a TechnologyPortfolio – Structural


then delivery of the magnifiedforce to the structural frame.

Figure 1 illustrates variousdamper configurations in a framingsystem. Let the interstory drift beu, the damper relative displace-ment be u

D , the force along the axis

of the damper be FD and the damp-

ing force exerted on the frame beF. It may be shown that

u uD = f (1)

F FD= f (2)

where ƒ = magnification factor.Expressions for the magnificationfactor of various configurations areshown in Figure 1. The significanceof the magnification factor may bebest demonstrated in the case oflinear viscous dampers, for which

F C uD o D= ˙ (3)

where ̇uD= relative velocity be-

tween the ends of the damperalong the axis of the damper. Thedamping ratio under elastic condi-tions for a single-story frame (asshown in Figure 1) with weight, W,and fundamental period, T, is:

βπ

= C gT

Wo f

2

4(4)

That is, the damping ratio is pro-portional to the square of the mag-nification factor. The toggle-braceand scissor-jack systems canachieve magnification factorslarger than unity. The systems canbe typically configured to have val-ues ƒ = 2 to 3 without any signifi-cant sensitivity to changes in thegeometry of the system. By con-trast, the familiar chevron-brace

and diagonal configurations have ƒless than or equal to unity.

For the purpose of comparison,consider the case of the use of alinear viscous damper with C

o =160

kN-s/m (= 0.9 kip-s/in) in the fram-ing systems of Figure 1 with weightW = 1370 kN (= 308 kip) and T =0.3 second. The resulting dampingratios are shown in Figure 1. Theeffectiveness of the toggle-braceand scissor-jack systems is clearly

MCEER Users Network:http://civil.eng.buffalo.edu/

users_ntwk

IDARC Users Group:http://civil.eng.buffalo.edu

Dia

go

na

lF

W

θCo

u

θcosf =

Ch

evro

n

Wu

F

Co001.f =

Lo

we

rT

og

gle

u

F

W

Coθ1

2θ

°90

)cos(

sin

21

2

θθθ+

=f

Up

pe

rT

og

gle

90

u

F

W

Co

θ2

θ1

°1

21

2 θθθ

θsin

)cos(

sinf +

+=

Re

ve

rse

To

gg

le

W

θ oC1

θ2

u

F

90

α

°

2

21

1cos

)cos(

cos θθθ

θα −+

=f

Scis

so

r-Ja

ck

3tan

cos

θψ=f

Ψ

θ3

Co

u

F

W

� Figure 1. Effectiveness of Damper Configurations in Framing Systems

108

demonstrated. It should be notedthat the configurations for thesetwo systems are identical to thosetested at the University at Buffalo.

It is clear in the results of Figure 1and in equations (1), (2) and (4) thatthe toggle-brace and scissor-jack con-figurations may provide substantialenergy dissipation capability with theuse of low output force devices. Thismay result in an important cost ad-vantage in systems that undergosmall drifts such as stiff structuralsystems under seismic load and moststructural systems under wind load.Such cases of small drift lead to a re-quirement for increased volume offluid viscous devices and accordinglyincreased cost. The use of the newconfigurations eliminates the neces-sity for large volume damping de-vices and may result in reduced cost.

Moreover, the scissor-jack systemmay be configured to allow foropen space, minimal obstruction ofview and slender configuration,which are often desired by archi-

tects. As an example, Figure 2 illus-trates the scissor-jack system testedat the University at Buffalo. Theopen bay configuration, the slen-derness of the system and the smallsize of the damper are apparent.

Damping Ratio as aSeismic Response Re-duction Measure ofNon-ProportionallyDamped Structures(Type II Project)

In FEMA 273/274 (1997), an im-portant design parameter for addeddevices is the effective dampingratio. As suggested in FEMA 273/274, displacements are reduced asthe effective damping ratio is in-creased. Some believe that placingmore dampers at the level of maxi-mum inter-story drift will achieveoptimized damping effects. This istrue for proportionally damped orsingle degree of freedom (SDOF)systems. It has been shown by Leeet al., (2001) that for multiple de-gree of freedom (MDOF) systems,higher damping ratios could, insome cases, increase the seismic re-sponse. For structures with addedpassive energy dissipation or seis-mic isolation devices, the dampingis no longer proportional nor neg-ligibly small. To study the dampingeffects on MDOF building struc-tures, the analysis method shouldconsider the effects of non-propor-tional damping.

For non-proportionally dampedMDOF systems, the damped modeshape is not orthogonal to thedamping matrix in the n-dimensionalphysical domain. The complex modeshapes are orthogonal in the state-space domain, where real valued

� Figure 2. Tested Scissor-Jack Damper Configura-tion


modal superposition methods can-not be directly applied. The “doublemodal superposition” approach de-veloped by Gupta and Law (1986) isused in this analysis, where the re-sponse is the superposition of “modaldisplacement” and “modal velocity.”These modes are not the un-dampedsystem modes, nor the complex sys-tem modes. The conventional modalanalysis routine of fast calculationcan be used in this approach and thedamping effects can be more readilyexplained by using structural dy-namic parameters.

Theoretical Background

In a proportionally damped sys-tem, the damping ratio is used todescribe damping effects. However,damping can affect the response ofan MDOF non-proportionallydamped system in many ways, in-cluding modal response, modalshape and natural frequencies, anddamping ratio.

For an MDOF system subjectedto earthquake excitation, the equa-tion of motion can be written as:

M U C U KU MU.. . ..

+ + = − b gu (5)

where M, C and K denote mass,damping and stiffness matrices, re-spectively; U is the relative displace-ment vector; U

b is a displacement

vector obtained by statistically dis-placing the support by unity in thedirection of the input motion; andu

g is the ground displacement.The double modal superposition

method obtains the response byequation (6):

, i U U U U U= = −∑ i

i

N

i id v

(6)

U Uid

id

i iv

iv

iX X= =ψ ψ , .

(7)

where U id and U i

v are real

value response associated with themodal shape, as defined by equa-

tion (8), and ψ id and ψ i

v are dis-

placement and velocity modeshapes, respectively. They are realvalue mode shapes and are calcu-lated by the state vector eigen-equa-tion. These mode shapes aredetermined not only by the systemmass and stiffness but also by theadded damping matrix. In somecases, the added damping willdominate the mode shapes (aswhen the fundamental modalshape collapse due to added damp-ing.). X

i is the modal response of

the following “modal” equation.In equation (8), ζ

i and ω

i are the

“modal damping ratio and circularfrequency of mode i”, and u

g is the

ground motion.•• •

˙̇X X X ui i gi i i + + = −2 2

iζ ω ω (8)

The damping ratio and naturalcircular frequency ζ

i, ω

i in the

above equation are determined bythe complex state vector eigen-value solution, like the mode shape.Because the state vector will varyfor different damping devicesadded, the modal shape, circularfrequency and damping ratio willnot be constant. Thus, the systemresponse defined by equation (6)cannot be always reduced by in-creasing the damping ratio.

Case Studies

To illustrate how system charac-teristics change and their impact onseismic response due to addeddamping, a four DOF frame struc-ture was analyzed. The first two vi-

“The resultsfrom modelingand analysis ofa Californiahospitalbuilding willprovide theengineeringcommunity witha three-dimensionalnonlinearanalysisplatform thatcurrently doesnot exist.”

110

bration modes of the frame areshown in Figure 3.

The added damping was limitedto a maximum of 30% of the effec-tive damping ratio (defined inFEMA 273 as a proportional ratio).Fifty-two linear viscous damperconfigurations were examined. The52 configurations were arranged byconsidering all possible damperlocations and their combinations,then changing the damping param-eters to make the first modal effec-tive damping ratio 5%, 10%, 20%and 30% (damping ratio by the

added device plus 2%). As listed inTable 1, dampers were installed onevery floor independently, everytwo floors, every three floors or onall four floors. There were 13 dif-ferent damper configurations, withfour damping ratios.

As noted above, the first mode ofthe complex damping ratio may bedifferent from the first mode of thecontrolling effective damping ratio.The first complex modal dampingratio for all 52 cases is listed in Table2. For small damping ratios (5%),the complex damping ratio is al-most the same, however, when thedamping ratio is increased to 10%,case 4 shows a dramatic change,while the other cases stay the same.Cases 3, 4, 7, 10, and 12 have verydifferent values when the control-ling ratio increases to 20%. Finally,the first complex damping ratio forcases 1, 3, 4, 7 and 12 becomes 99%when the ratio is increased to 30%.

The change in natural frequencyin the first mode is given in Table3. Both damping ratio and fre-quency changes are related to themodal change for different damperconfigurations.

To illustrate the non-proportionalmode shape change, the first dis-

Case 1 2 3 4 5 6 7 8 9 10 11 12 131st story X X X X X X2nd story X X X X X X X3rd story X X X X X X X4th story X X X X X X

Flo

or

Mode 1

Mode 2

4

3

2

1

0-1 -0.5 0 0.5 1

Flo

or

Case 1 2 3 4 5 6 7 8 9 10 11 12 135% 0.05 0.05 0.05 0.04 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.0510% 0.10 0.10 0.10 0.99 0.10 0.10 0.09 0.10 0.10 0.10 0.10 0.10 0.1020% 0.16 0.20 0.13 0.99 0.20 0.20 0.13 0.20 0.20 0.20 0.20 0.99 0.2030% 0.99 0.28 0.99 0.99 0.31 0.31 0.99 0.30. 0.30 0.30 0.30 0.99 0.30

� Table 2. Damping Ratio Comparison (First mode)

� Table 1. Damper Location

� Figure 3. First Two Modal Shapes


placement mode shape at the con-trolling effective damping ratios of5%, 10%, 20% and 30% are shownin Figure 4. The figure shows that,as the damping ratio becomeshigher, dramatic changes can resultin cases 1, 3, 4, 7 and 12 (the funda-mental mode collapsed). When thefundamental mode shape collapses,the second mode shape becomesdominant in the system response,as shown in Figure 5.

The relationship between themaximum displacement responseand the controlling effective damp-

ing is shown in Figure 6. Exceptfor case 4, the response decreasesas the damping ratio increases.When the controlling damping ra-tio is higher than 15%, differentdamper configurations will resultin different response reductionseven though they may achieve thesame effective damping ratio value.It may also be seen from the figurethat the maximum responses witha 30% effective controlling damp-ing ratio for some cases are higherthan those with 20%. Table 4 liststhe maximum displacement re-

Case 1 2 3 4 5 6 7 8 9 10 11 12 135% 2.03 2.02 2.03 2.05 2.02 2.02 2.03 2.02 2.02 2.02 2.02 2.03 2.0210% 2.06 2.04 2.07 1.17 2.03 2.07 2.02 2.03 2.03 2.03 2.03 2.05 2.0220% 2.22 2.12 2.24 0.50 2.06 2.07 2.23 2.03 2.07 2.08 2.09 2.14 2.0330% 1.96 2.29 1.50 0.32 2.12 2.15 1.41 2.04 2.14 2.17 2.20 1.30 2.04

Displacement Displacement

a) Damping = 5% b) Damping = 10%

c) Damping = 20% d) Damping = 30%

Mas

s no

.M

ass

no.

Mas

s no

.M

ass

no.

case 4

case 1case 4

case 12

case 3

case 7

case 12

case 4

� Table 3. Natural Frequency Comparison (First mode)(in Hz)

� Figure 4. Displacement Modal Shapes

112

sponses for these two damping ra-tio values, calculated for the ElCentro earthquake. As seen in thetable, the damping increases in twoways: the first is where the re-sponse increased after more damp-ers were added and theconfiguration was unchanged(cases 3 and 4). The second case iswhen more dampers are added indifferent locations and configura-tions, as shown in case 1 at 20%,case 3 at 30%, case 4 at 30% andthe rest of the cases are at 20%. Thisconclusion has also been obtainedusing other earthquake records,such as from the Northridge earth-quake.

Continuing Effort

It is important to understandthe limitations of using the damp-ing ratio as the key seismic re-sponse reduction measurement.For many non-proportionallydamped structures, the evaluationof performance should be basedon response history analysis. Inthis regard, a simple method tocope with the various non-pro-portional damping effects will bevaluable for engineering applica-tions. This team is currently at-tempting to develop such amethod.

Case 1 2 3 4 5 6 7 8 9 10 11 12 1320% 0.357 0.398 0.381 0.738 0.403 0.392 0.397 0.409 0.399 0.392 0.383 0.365 0.41430% 0.343 0.287 0.384 0.765 0.324 0.315 0.374 0.339 0.309 0.313 0.289 0.280 0.350

First 2 displacement modes for case 4 at 10% damping ratio

Mas

s no

.

Displacement

1st mode 1.17Hz, complex dampingratio=99%

2nd mode 2.09Hz, complexdamping ratio=4%


Mas

s no

.

Displacement

b) Case 7, 30% Damping Ratio

a) Case 4, 10% Damping Ratio

1st mode 1.41Hz, complexdamping=99%

2nd mode 2.35Hz, complexdamping=13%


Mas

s no

.

1st mode 1.30 Hz, complexdamping=99%

Displacement

2nd mode 2.37 Hz, complexdamping=21%

1st mode 1.30Hz, complexdamping=99%

2nd mode 2.37Hz, complexdamping=21%

c) Case 12, 30% Damping Ratio

� Figure 5. Mode Collapse

� Table 4. Maximum Displacement Response at the Top Level of the Frame Using Standard El Centro Earthquake Record


Optimal Design ofDamping Devices forMulti-story SteelFrames Based onMulti-performanceIndices(Type III Project)

FEMA 273, NEHRP 2000 and theBlue Book 1999 provide primarydesign guidelines for buildingswith supplemental damping de-vices. These procedures can ad-dress building design for differentperformance objectives using lin-ear or nonlinear methods after thedamping device configuration hasbeen determined. However, therule or procedure of how to opti-mally distribute these damping de-vices throughout the building andhow to consider multi-perfor-mance indexes is still not available.These procedures are very criticalfor practicing engineers. Since add-ing damping devices causes the

structure to be non-proportionallydamped, a “systems approach” con-sidering the change in characteris-tics is necessary.

The earthquake ground accelera-tion can be approximated as a fi-nite Fourier series expansion, andthe seismic response as the linearcombination of all responses to thesingle frequency excitation. Thus,to find the effective configuration,it is necessary to know the domi-nant frequency response (or domi-nant mode), the modalcomposition, and its mass partici-pating factors. The selection of theresponse will follow the perfor-mance index. With this knowledge,an optimization scheme for damp-ing devices can be developed.

Analysis and Design Procedure

The design procedure is outlinedas follows:

1. Calculate the characteristics ofeach potential configuration

Maximum displacement and damper configuration

Damping /%

Dis

plac

emen

t

case 12

case 3

case 4

� Figure 6. Maximum Displacement Response

114

2. Compare these with the origi-nal structural system’s character-istics, and identify for eachconfiguration:a. Natural frequencies (Some

modes may collapse into eachother due to added damping.This is particularly criticalwhen damping becomes non-proportional. It follows thatthe natural frequencies maychange significantly.)

b. Modal damping ratio (100%critical damping usually im-plies a mode collapsed, whichmay be a problem for modalresponse reduction)

c. Mode shapesd. Modal loading factors (varia-

tion in this characteristics maystrongly influence the abso-lute acceleration response)

e. Phase differences between themodes.

3. For a mass point of interest,check which natural frequen-cies will contribute most in theresponse, by using sinusoidal ex-citations of different amplitudes.Examine each correspondingsteady state response vector toidentify the largest contributionvector. A general excitation canbe simplified as a sum of this si-nusoidal excitation of variousamplitudes, multiplied by cer-tain envelopes. Thus, the re-sponse will be dominated by thecombination of n basic vectors.

4. The vectors computed above canbe further split into modal contri-butions. In this regard, compari-sons can be made among modalshapes, loading factors, natural fre-quencies (which provides infor-mation on relative phase lags) anddamping ratios (which provide in-formation on dynamic amplifica-tion factors). This information

illustrates the modes that are im-portant in the response of a givenmass point.

5. The damping distribution maybe very different if the optimi-zation target is selected as thestory drift or acceleration. Thedrift response is dominated bythe major modes. Thus, with ahigher modal participation fac-tor and a higher damping ratio,the drift responses will be fur-ther reduced. This is not the casefor acceleration response, whichincludes more higher mode con-tributions.

6. For acceleration response, if adamper is placed between theith and i+1th floors, then the i+1th

f loor acceleration is usuallylower than the ith floor. In addi-tion, i-1th floor acceleration willbe lower than that of ith floor.

Case Study

An 11-story steel frame buildingstructure was used as an optimaldesign example. The building’stypical plan and north-south eleva-tion are shown in Figures 7 and 8.This building had been designedwith supplemental damping de-vices. The typical device configu-ration in the frame is shown inFigure 9. A total of 120 damperswere added to the building, with24 dampers on the first story, andgradually decreasing as the storyheight increases.

If the performance index is se-lected as optimal story drift, thefirst story mass contributes less tothe dominant mode of the systemresponse. Better results may beobtained if the dampers in thisstory are redistributed. The opti-mized distribution removes most ofthe devices from the first story, and


adds them to the 6th to 11th stories.The same total number of damp-ers is used as in the original damperdistribution. With this optimizeddistribution, the total system damp-ing is increased by 4%. More im-portantly, the damping ratio of thedominant mode is increased by19%, and the largest story drift,which occurred in the 6th floor, isreduced by an average of 11% for aseries of spectra-compatible earth-quake. If the performance index isselected as acceleration, the opti-mal distribution will be different.

Current Efforts

The design and analysis proce-dure proposed here is based on fre-quency domain analysis and timedomain verification. Since responsespectrum or time history analysisare preferred in structural design,the method is currently being im-proved and optimized using re-sponse spectrum-based objectivefunctions.

ComputationalAseismic Design andRetrofit for PassivelyDamped Structures(Type III Project)

Over the past two decades, con-siderable effort has been directedtoward the development and en-hancement of protective systemsfor the control of structures underseismic excitation. In the area ofpassive energy dissipation systems,applications typically involve me-tallic yielding dampers, frictiondampers, viscous fluid dampers orviscoelastic dampers (e.g., Soongand Dargush, 1997; Constantinouet. al., 1998). Although the intro-duction of these new concepts andsystems presents the structural en-gineer with additional freedom inthe design process, many questionsalso naturally arise. In the case ofpassive energy dissipation systems,these questions range from perfor-mance and durability issues to con-cerns related to the sizing andplacement of damping elements.

One promising direction for fu-ture research involves the furtherdevelopment of the FEMA 273/274

� Figure 7. Floor Frame Plan � Figure 8. Elevation Frame Plan � Figure 9. Damper Distribution

116

and NEHRP 2000 design guidelinesbased upon additional numericalsimulations and practical experi-ence. Alternatively, one may envi-sion a dramatically different designprocess for passively damped struc-tures by adopting a computationalapproach. Such an approachshould incorporate the dynamics ofthe problem, the uncertainty of theseismic environment, the reliabilityof the passive elements and per-haps also some key socioeconomicfactors. With these requirementsin mind, one can conceptualizeaseismic design as a complex adap-tive system and begin to develop ageneral computational frameworkthat promotes the evolution of ro-bust, and possibly innovative, de-signs.

Complex Adaptive Systems

There is a broad class of systemsin nature and in human affairs thatinvolve the complicated interac-tion of many components or agents.These may be classified as complexsystems, particularly when the in-teractions are predominantly non-linear. Within this class are systemswhose agents tend to aggregate ina hierarchical manner in responseto an uncertain or changing envi-ronment. These systems have theability to evolve over time and toself-organize. In some cases, thesystem may acquire collectiveproperties through adaptation thatcannot be exhibited by individualagents acting alone. Key character-istics of these complex adaptivesystems are nonlinearity, aggrega-tion, flows and diversity (Holland,1995). Examples include the hu-man central nervous system, thelocal economy, a rain forest or amultidisciplinary research center.

Holland (1962, 1992) also devel-oped a unified theory of adaptationin both natural and artificial sys-tems. In particular, Holland broughtideas from biological evolution tobear on the problem. Besides pro-viding a general formalism forstudying adaptive systems, this ledto the development of genetic al-gorithms.

Computational Framework forAseismic Design

With these ideas, one can nowenvision a new aseismic designapproach based upon the creationof an artificial complex adaptivesystem. The primary research ob-jective is to develop an automatedsystem that can evolve robust de-signs under uncertain seismic en-vironments. With continueddevelopment, the system may alsobe able to provide some novel so-lutions to a range of complexaseismic design problems.

Figure 10 depicts the overall ap-proach for computational aseismicdesign and retrofit (CADR), borrow-ing terminology from biologicalevolution. Design involves a se-quence of generations within a se-quence of eras. In each generation,a population of individual struc-tures is defined and evaluated inresponse to ground motion realiza-tions. Cost and performance areused to evaluate the fitness, whichin turn determines the makeup ofthe next generation of structures.Performance is judged by perform-ing nonlinear transient dynamicanalysis. Presently, this analysis uti-lizes either ABAQUS (2000) or anexplicit state-space transient dy-namics code (tda). The implemen-tation of the genetic algorithm


controlling the design evolution isaccomplished within the public-domain code Sugal (Hunter, 1995).

Model Problem: Five-StorySteel Moment Frame

Consider an example of a five-story steel moment frame retrofitwith passive energy dissipators asshown in Figure 11. Three differ-ent types of dampers are available:metallic plate dampers, linear vis-cous dampers, and viscoelasticdampers. For each type, five dif-ferent sizes are possible. Conse-quently, a 20-bit genetic code isemployed to completely specifythe dampers used in each story ofany particular structure A∈α . Thus,for this problem, the set α contains220 (i.e., more than one million)possible structures. Figure 11 alsodefines a hierarchical approach inwhich different structural modelswith varying levels of complexityare utilized in each era. The idea isto first use simple models to widelyexplore the design space and thento employ more complicated andcomputationally expensive models

later in the design process. Cur-rently, a two-surface cyclic plastic-ity model is applied for the primarystructural system and metallic platedampers, while a coupledthermoviscoelastic model with in-elastic heat generation is used forthe viscoelastic dampers. Bothinterstory drift and story accelera-tion limits are set in order to estab-lish acceptable performance.

As a specific example, considerthe application of the CADR strat-egy to a typical five-story steel mo-ment frame based upon Era 1 (i.e.,

Computational Aseismic Design and Retrofit

Evaluate Era

Evaluate Generation

Evaluate Individual Structure

Define StructureRealize

Ground Motion(s)

Determine CostEvaluate Performance

(abaqus or tda)

Determine Fitness

SEQUENTIAL

PARALLEL

Computational Aseismic Design and Retrofit

Example: Five-story steel moment frame retrofit w/passive energy dissipators

Problem Definition:

2nd Era3rd Era

1st Era

…

Genetic Code 20-bit design descriptor

xxyy xxyy xxyy xxyy xxyy

Story: 1 2 3 4 5

xx =

00 None/Size A

01 Metallic

10 Viscous

11 VE

yy =

00 Size B

01 Size C

10 Size D

11 Size E

Example: Five-story steel moment frame retrofit w/passive energy dissipators

Problem Definition:

1st Era 2nd Era 3rd Era

Genetic Code 20-bit design Descriptor

Story: 1 2 3 4 5

00 None/Size A01 Metallic10 Viscous11 VE

00 Size B01 Size C10 Sise D11 Size E

xx = yy =xxyy xxyy xxyy xxyy xxyy

� Figure 10. Overall Framework for Computational Aseismic Design andRetrofit

� Figure 11. Problem Definition for Five-story Steel Moment Frame

118

lumped parameter) simulations.Let k

i and Wi represent the ith story

elastic stiffness and story weight,respectively. The baseline framemodel has uniform story weightsW

i =W =125 kips for i =1,2,...5 and

story stiffness k1=k

2=k

3 =193 kip/in,

k4 =147kip/in,k5 =87 kip/in. The

first two natural frequencies are1.07 Hz and 2.72 Hz. A two-surfacecyclic plasticity model is employedto represent the hysteretic behav-ior of the primary structure.

A retrofit strategy is now devel-oped to protect this structure situ-ated on firm soil in a simplifiedhypothetical seismic environmentthat can be represented by a uni-form distribution of earthquakeswith magnitude 7.2 ≤ M

w ≤ 7.8

and epicentral distance20 km ≤ r ≤ 30km. Each groundmotion realization is generated ac-cording to the model ofPapageorgiou (2000) for easternU.S. earthquakes. For the retrofit,it is assumed that linear viscous(visc) dampers, metallic yielding

(tpea) dampers and viscoelastic(ve) dampers are available and the20-bit genetic code defined in Fig-ure 11 is applied. Hypotheticaldevice cost data for various sizedampers were set as indicated inTable 5. Each increment in dampersize corresponds roughly to a dou-bling of the damping capacity.

For the automated design, a popu-lation of N

p = 40 individual struc-

tures was evolved for a total ofN

g= 40 generations. Within each

generation, each structure was sub-jected to a total of N

s= 10 seismic

events. Crossover and mutationoperators were used to evolve newstructures from an initially randompool. At the end of each genera-tion, one-half of the structures werereplaced with potentially new indi-viduals. As generations pass, gener-ally speaking, the average fitnessincreases, indicating that the popu-lation becomes enriched with morerobust structures. However, the evo-lution of average fitness is not mono-tonic, because the genetic algorithmcontinues to explore the designspace for better structures. Table 5also presents the five structures thathave appeared most frequently in thepopulation. These are high fitnessdesigns that have survived over manygenerations. The table data includesthe total number of earthquakes thateach of the five structures has expe-rienced and the success (or survival)rate. Notice, according to Table 5,that the high fitness designs mostoften utilize viscous dampers andthat the largest dampers are placedon the first story. In four of the highfitness designs, size C dampers ap-pear in the fourth story, suggestingperhaps that the second mode re-sponse also requires damping.

Allowable Drift =1.500 in.Allowable Acceleration =193.200 in/s2

Device Cost:A B C D E

visc 2.00 4.00 6.00 8.00 10.00tpea 2.00 4.00 6.00 8.00 10.00ve 2.00 4.00 6.00 8.00 10.00

High Fitness Designs:No.Trials:

2350 1900 570 410 320

DamperCost:

26.00 26.00 26.00 28.00 26.00

SuccessRate:

0.9655 0.9611 0.9614 0.9561 0.9625

Story 5 visc A visc A visc A visc A visc AStory 4 ve C ve C visc C ve C tpea BStory 3 visc B ve B ve B ve C tpea CStory 2 visc C visc C visc C visc C visc CStory 1 visc D visc D visc D visc D visc D

� Table 5. Five-Story Steel Moment Frame –Baseline (Case 1)


Concluding Remarks

In this research, a new computa-tional aseismic design and retrofit(CADR) approach is advocated.This approach centers on the de-velopment of an artificial complexadaptive system within which ro-bust aseismic designs may evolve.As a first phase of this research pro-gram, a genetic algorithm is appliedfor the discrete optimization of apassively damped structural system,subjected to an uncertain seismicenvironment. The results of pre-liminary applications, involving theseismic retrofit of multi-story steelmoment frames, suggest that con-tinued development of the ap-proach may prove beneficial to theengineering community. Currentefforts are underway to work withseveral MCEER Industry Partners toenhance the CADR software and todevelop applications associatedwith critical facilities.

Development ofAnalysis Tools forEngineeringCommunity(Type III Project)

Any implementation of protec-tive systems in design of new build-ings or bridges, or in their retrofit,requires modeling and analysis ofintegral systems including thestructures and the devices. Whenthese multiple-DOF systems areimplemented with energy dissipa-tion devices, the total building-de-vice system in general is nonlinear.Much creativity and fundamentalresearch in structural dynamicsprinciples have to be pursued inorder to develop a reasonably

simple and accurate analysis anddesign procedure for use by thepracticing professionals. Two on-going MCEER projects are de-scribed in the following sections.One is fundamental in nature toestablish new approaches whilethe other emphasizes the develop-ment of user-friendly simplifiedprocedures for the design profes-sionals.

Nonlinear Structural Analysisby the State Space Approach

The State Space Approach (SSA)is an alternative approach to theformulation and solution of initial-boundary-value problems involvingnonlinear distributed-parameterstructural systems. The response ofthe structure, which is spatiallydiscretized following a weak formu-lation, is completely characterizedby a set of state variables. These in-clude global quantities such asnodal displacements and velocitiesand element (or local) quantitiessuch as nodal forces and strains atthe integration points. The nonlin-ear evolution of the global statevariables during the response ofstructures is governed by physicalprinciples, such as momentum bal-ance, and the nonlinear inelasticevolution of the local variables isgoverned by constitutive behavior.The essence of the SSA is to solvethe two sets of evolution equationssimultaneously in time using directnumerical methods, in general as asystem of differential-algebraicequations. The proposed method-ology results in a more consistentformulation with a clear distinctionbetween spatial and temporaldiscretization.

“Muchcreativity andfundamentalresearch instructuraldynamicsprinciples haveto be pursued todevelop asimple andaccurateanalysis anddesignprocedure forpracticingprofessionals.”

120

Objectives and results A material nonlinear three-di-

mensional beam column elementand a fully geometric (equilibriumand nonlinear strain-deformationrelations) and material nonlineartwo-dimensional beam-column el-ement have been developed in thisframework based on a flexibilityformulation. A general three-dimen-sional interactive constitutivemacro-model has been developed.In this model, hysteretic degrada-tion can also be modeled using suit-able constitutive equations(Sivaselvan and Reinhorn, 2000).The resulting platform can studystructures near collapse. The basicapproach has been used to modela structure, which collapsed inshake table under severe lateralbuckling (see Vian et al., in this vol-ume).

The above models and solutionprocedure have been implementedin an object-oriented computerprogram that uses the graphicaluser interface (GUI) of the com-mercial structural analysis program,LARSA. Figure 12 shows the re-sponse of a three-dimensionalframe with hysteretic behavior tobi-axial non-proportional loading.Figure 13 shows results of analysisof extremely large deformations,

which allows an elastic beam to bebent into a circle. Figure 14 showsthe collapse pattern of a simplestructure while Figure 15 showsthat all the models are developedusing the macro model approachin which structures are representedby beam-column elements withhysteretic degradation.

Three Dimensional InelasticDynamic Analysis of Structureswith Protective Systems:IDARC3D Version 2.0

The nonlinear analysis of inelas-tic structures with energy dissipa-tion systems and base isolationswas the subject of research anddevelopment throughout the exist-ence of the Center’s activities. Theresearch work, both analytical andexperimental, resulted in a seriescomputer platforms, IDARC and3DBASIS, now available nationallyand internationally to the public atlarge through a dedicated UsersGroup (http://civil.eng.buffalo.edu).(See also Park et. al., 1987, Reinhornet. al., 1988, Kunnath et. al., 1989,Nagarajaiah et. al., 1989, Nagarajaiahet. al., 1991, Tsopelas et. al., 1991,Kunnath et. al., 1992, Nagarajaiahet. al., 1993, Tsopelas et. al., 1994,

216”

108”

108”

W8×31

typ.

W12×40

typ.

216”

108”

108”

W8×31

typ.

W12×40

typ.

216”

108”

108”

W8×31

typ.

W12×40

typ.

(a) D Frame

-4

-3

-2

-1

0

1

2

3

4

-4 -3 -2 -1 0 1 2 3 4

ux [in]

ux [in]

uz

[in

]

-4

-3

-2

-1

0

1

2

3

4

-4 -3 -2 -1 0 1 2 3 4

ux [in]

ux [in]

uz

[in

]

(b) Non-proportionalExcitation History

-30

-20

-10

0

10

20

30

-3 -2 -1 0 1 2 3

ux [in]

SSA

ANSYS

UNIAXIAL

Fx

[in

]

ux [in]

-30

-20

-10

0

10

20

30

-3 -2 -1 0 1 2 3

ux [in]

SSA

ANSYS

UNIAXIAL

Fx

[in

]

ux [in]

(c) Force-disp. in Long. Dir.

-15

-10

-5

0

5

10

15

-3 -2 -1 0 1 2 3

uz [in]

SSA

ANSYS

UNIAXIAL

uz [in]

Fz

[in

]

-15

-10

-5

0

5

10

15

-3 -2 -1 0 1 2 3

uz [in]

SSA

ANSYS

UNIAXIAL

uz [in]

Fz

[in

]

(d) Force-disp. in Trans. Dir.

� Figure 12. Inelastic Response of Frame Using 3D Interaction Elements


Reinhorn et. al., 1994, Valles et. al.,1996, and Reinhorn et. al., 1998.)

The authors undertook an expan-sion to the three-dimensional sys-tems of the code of IDARC in aredevelopment effort using an ob-ject-oriented approach. The result-ing software architecture willenable progressive growth by easyaddition of new models from otherresearch tasks and provides the out-come to the engineering commu-nity at large. The current focus ofwork is to provide the tools formodeling damping and other ad-vanced systems using a unified ap-proach to nonlinear systems. Thework done in cooperation withLARSA, Inc., a software developer,enables creation of a user friendlyand acceptable analysis platform.

Objectives and results The redevelopment includes

three main steps. The first step isto create a flexible and extendablesetup for the platform using anobject-oriented finite element pro-gramming approach. This modularframework clearly separates the dif-ferent elements of the program byencapsulating data in classes.Classes are black boxes, which pro-vide easy to use interfaces through-out the program. Thus if a new class

is added, the developer has to dealonly with the data and routines ofthe new class itself. Also, changesto one class will not affect the restof the program because of the en-capsulation. This simplifies the in-tegration of new parts.

In the second step, componentsare incorporated in the new plat-form IDARC3D Version 2.0: (i) en-ergy dissipation systems / dampersand (ii) a computing core capableof nonlinear analysis. The new pro-gram structure provides possibili-ties to easily add new elementssuch as base isolators, adapted fromthe platform 3D-BASIS also devel-oped by Reinhorn andConstantinou in multi-annualprojects. To ensure convenience,IDARC3D Version 2.0 operates ona PC and has a graphical user inter-face for input and output (I/O). TheI/O is decoupled from the core ofthe program so that it can bechanged without interfering withthe actual program.

The third step is to model andevaluate a structure with protectivesystems and to verify the results ofthe nonlinear analysis withIDARC3D 2.0 against results fromother standard analysis programs orexperiments.

P

(a) Undeformed

-2.0

0.0

2.0

4.0

6.0

8.0

10.0

12.0

-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

(b) Deformed

Axial Load

Ground Excitation

Axial Load

Ground Excitation

(a) SDOF system

-20.0

-15.0

-10.0

-5.0

0.0

5.0

10.0

15.0

20.0

-0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70

Displacement (in)

(b) Force-Disp.Response

),( yy uF

(a) Beam Model

Force-Displacement Reponse

-2000

-1500

-1000

-500

0

500

1000

1500

2000

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3

Displacement (m)

(b) Force-Disp.Response

� Figure 13. Eccentric AxiallyLoaded Elements

� Figure 14. Dynamic Collapse usingGeometric Nonlinear Beam

� Figure 15. Hysteretic Degradation

122

ABAQUS, (2000), Theory Manual, Version 6.1, Hibbitt, Karlsson and Sorensen,Pawtucket, RI.

Constantinou, M.C., Tsopelas, P., Hammel, W. and Sigaher, A.N., (2001), “Toggle-Brace-Damper Seismic Energy Dissipation Systems,” Journal of Structural Engineering,ASCE, Vol. 127, No. 2, pp. 105-112.

Constantinou, M.C., and Sigaher, A.N., (2000), “Energy Dissipation SystemConfigurations for Improved Performance,” Proceedings, 2000 Structures Congress,ASCE, Philadelphia, PA, May.

Constantinou, M.C., Soong, T.T., and Dargush, G.F., (1998), Passive Energy DissipationSystems for Structural Design and Retrofit, MCEER-98-MN01, MultidisciplinaryCenter for Earthquake Engineering Research, University at Buffalo.

Federal Emergency Management Agency, (1997), NEHRP Guidelines for the SeismicRehabilitation of Buildings, FEMA 273, Federal Emergency Management Agency,October.

Federal Emergency Management Agency, (1997), NEHRP Commentary on theGuidelines for the Seismic Rehabilitation of Buildings, FEMA 274, FederalEmergency Management Agency, October.

Gupta A. K. and Law J. W., (1986), “Seismic Response of Non-classically DampedSystems,” Nuclear Engineering and Design 91, pp. 153-159.

Holland, J.H., (1962), “Outline for a Logical Theory of Adaptive Systems,” J. Assoc.Comp. Mach., Vol. 3, pp. 297-314.

Holland, J.H., (1992), Adaptation in Natural and Artificial Systems, MIT Press,Cambridge, MA.

References

The implementation of the newdesign is realized using FORTRAN90 to be consistent with the previ-ous and existing IDARC programs.Much of the existing code is reusedin the new platform to minimizeprogramming new code. AlthoughFORTRAN 90 is not an object-ori-ented programming language, ob-ject-oriented features can besimulated with a reasonable effort.Also, source code written in FOR-TRAN 90 and C++ (as an examplefor an object-oriented program-ming language) can be used to formone platform together.

The documentation developedfor the new platform includesguidelines to further develop thesystem, a users manual and instal-

lation examples. The developer’smanual, which describes the modu-lar setup and provides the readerwith the necessary information tochange or extend the program, willbe published through the MCEERNetworking activities.

A California hospital building isbeing modeled for research byother center investigators, and abenchmark physical model testedon the shake table at University atBuffalo is being analyzed to providethe first example cases. The resultof this development will providethe engineering community with athree-dimensional nonlinear analy-sis platform that currently does notexist.


Holland, J.H., (1995), Hidden Order, Addison-Wesley, Reading, MA.

Hunter, A., (1995), Sugal Programming Manual, Version 2.1, University of Sunderland,England.

Kunnath, S.K., Reinhorn, A.M. and Lobo, R.F., (1992), IDARC Version 3.0: InelasticDamage Analysis of Reinforced Concrete Structures, Technical Report NCEER 92-0022, Multidisciplinary Center for Earthquake Engineering Research, University atBuffalo.

Kunnath, S.K. and Reinhorn, A.M., (1989), Inelastic Three-Dimensional ResponseAnalysis of Reinforced Concrete Building Structures (IDARC-3D): Part 1 –Modeling, Technical Report NCEER-89-0011, Multidisciplinary Center for EarthquakeEngineering Research, University at Buffalo.

Lee, G. C., Tong, M. and Wu, Y.H., (2001), “Some Design Issues for Building SeismicRetrofit Using Energy Dissipation Devices,” Int’l. Journal of Structural Engineeringand Mechanics, (in press).

Nagarajaiah, S., Li, C., Reinhorn, A.M. and Constantinou, M.C., (1993), 3D-BASIS-TABS:Computer Program for Nonlinear Dynamic Analysis of Three Dimensional BaseIsolated Structures, Technical Report NCEER-93-0011, Multidisciplinary Center forEarthquake Engineering Research, University at Buffalo.

Nagarajaiah, S., Reinhorn, A.M. and Constantinou, M.C., (1991), 3D-BASIS – NonlinearDynamic Analysis of Three-Dimensional Base Isolated Structures: Part II,Technical Report NCEER-91-0005, Multidisciplinary Center for EarthquakeEngineering Research, University at Buffalo.

Nagarajaiah, S., Reinhorn, A.M. and Constantinou, M.C., (1989), Nonlinear DynamicAnalysis of Three-Dimensional Base Isolated Structures (3D-BASIS), Technical ReportNCEER-89-0019, Multidisciplinary Center for Earthquake Engineering Research,University at Buffalo.

Papageorgiou, A.S., (2000), “Ground Motion Prediction Methodologies for Eastern NorthAmerica,” Research Progress and Accomplishments, 1999-2000, MultidisciplinaryCenter for Earthquake Engineering Research, University at Buffalo, pp. 63-67.

Park, Y.J., Reinhorn, A.M. and Kunnath, S.K., (1987), IDARC: Inelastic Damage Analysisof Reinforced Concrete Frame – Shear-Wall Structures, Technical Report NCEER-87-0008, Multidisciplinary Center for Earthquake Engineering Research, University atBuffalo.

Ramirez, O., Constantinou, M.C., Kircher, C.A., Whittaker, A.S., Johnson, M.W. and Gomez,J.D., (2000), Development and Evaluation of Simplified Procedures for Analysis andDesign of Buildings with Passive Energy Dissipation Devices, MCEER-00-0010,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo.

Reinhorn, A.M., Simeonov, V., Mylonakis, G. and Reichman, Y., (1998), IDARC Bridge: AComputational Platform for Seismic Damage Assessment of Bridge Structures,Technical Report MCEER-98-0011, Multidisciplinary Center for Earthquake EngineeringResearch, University at Buffalo.

Reinhorn, A.M., Nagarajaiah, S., Constantinou, M.C., Tsopelas, P. and Li, R., (1994), 3D-BASIS-TABS Version 2.0: Computer Program for Nonlinear Dynamic Analysis ofThree Dimensional Base Isolated Structures, Technical Report NCEER-94-0018,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo.

Reinhorn, A.M., Kunnath, S.K. and Panahshahi, N., (1988), Modeling of R/C BuildingStructures With Flexible Floor Diaphragms (IDARC2), Technical Report NCEER-88-0035, Multidisciplinary Center for Earthquake Engineering Research, University atBuffalo.

References, (Cont’d)

124

References, (Cont’d)

Simeonov, V.K., Sivaselvan M. and Reinhorn, A.M., (2000), “Nonlinear Analysis ofFrame Structures by the State-Space Approach,” Computer-Aided Civil andInfrastructure Engineering, (Special Volume in honor of Prof. Gear), Vol. 15, Jan2000, pp. 76-89.

Sivaselvan, M., and Reinhorn, A.M., (2000), “Hysteretic Models for DeterioratingInelastic Structures,” Journal of Engineering Mechanics, Vol. 126, No. 6, Jun. 2000,pp.633-640.

Soong, T.T. and Dargush, G.F., (1997), Passive Energy Dissipation Systems inStructural Engineering, J. Wiley, England.

Tsopelas, P., Constantinou, M.C., Kircher, C.A. and Whittaker, A.S., (1997), Evaluationof Simplified Methods of Analyses for Yielding Structures, NCEER-97-0012,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo.

Tsopelas, P.C., Constantinou, M.C., and Reinhorn, A.M., (1994), 3D-BASIS-ME:Computer Program for Nonlinear Dynamic Analysis of Seismically IsolatedSingle and Multiple Structures and Liquid Storage Tanks, Technical ReportNCEER-94-0010, Multidisciplinary Center for Earthquake Engineering Research,University at Buffalo.

Tsopelas, P., Nagarajaiah, S., Constantinou, M.C. and Reinhorn, A.M., (1991), 3D-BASIS-M: Nonlinear Dynamic Analysis of Multiple Building Base Isolated Structures,Multidisciplinary Center for Earthquake Engineering Research, University at Buffalo.

Valles, R.E., Reinhorn, A.M., Kunnath, S.K., Li, C. and Madan, A., (1996), IDARC2D,Version 4.0: A Computer Program for the Inelastic Damage Analysis ofBuildings, Technical Report NCEER-96-0010, Multidisciplinary Center forEarthquake Engineering Research, University at Buffalo.

Winklemann, K., (2001), “IDARC3D VERSION 2.0: Three Dimensional InelasticDynamic Analysis of Structures with Protective Systems,” MS Thesis, University ofDarmstadt (completed while Visiting Research Associate at the University atBuffalo).

125

Research Objectives

The purpose of this research is to examine how cost-benefit analysis (CBA)can be utilized to evaluate the attractiveness of mitigation for lifeline systemssubject to earthquake ground motion. We propose a framework for the CBAthat can be used in conjunction with work being completed by other re-searchers at MCEER (Shinozuka et al., 2000; Chang et al., 2000). With theirdevelopment of fragility curves and insight into specific utility lifelines sys-tems, our framework is useful for the next step in the analysis. In this paper,we use an example of a transportation system to show the CBA framework.Then, we consider two case studies to show the effects of the disruption ofutility lifeline service on two stakeholders in the analysis. First, the indirecteconomic loss to business owners is studied, and then, the cost to publicagencies to shelter displaced residents is considered.

The two lifelines serving as case studies for this work are the electricpower system in Shelby County, Tennessee [building on previous workdone at MCEER] and the water distribution system in Alameda and ContraCosta Counties, California [working with the East Bay Municipal UtilityDistrict (EBMUD)]. Our research provides a framework to link data fromthe physical and engineering sciences (i.e., seismology of the region andvulnerability of the lifeline) with the social sciences (i.e., costs of naturaldisasters and public policy implications of mitigation).

Using Cost-Benefit Analysis toEvaluate Mitigation for LifelineSystems

by Howard Kunreuther (Coordinating Author), Chris Cyr, Patricia Grossi and Wendy Tao

Cost-benefit analysis (CBA) is a systematic procedure for evaluating de-cisions that have an impact on society. There are different ways to

conduct a valid CBA, depending on the information one has and the na-ture of the problem at hand. We chose a simplified five-step procedure toillustrate this approach (Figure 1). A more comprehensive approach, whichincorporates several additional steps, is discussed in Boardman et al. (2001).The five-step procedure includes: defining the nature of the problem, in-cluding the alternative options and interested parties; determining thedirect cost of the mitigation alternatives; determining the benefits of miti-gation, via the difference between the loss to the system with and with-out mitigation; calculating the attractiveness of the mitigation alternatives;and, finally, choosing the best alternative.


Howard Kunreuther, Co-director, Chris Cyr,Undergraduate Student,Patricia Grossi, Post-Doctoral Researcher, andWendy Tao, UndergraduateStudent,Wharton RiskManagement and DecisionProcesses Center, TheWharton School, Universityof Pennsylvania

126

These steps were chosen keep-ing in mind the complex processof estimating losses to lifeline sys-tems and evaluating the benefits ofmitigation to the system. Previouswork performed at the WhartonSchool analyzed the cost-effective-ness of mitigation to residentialstructures (Kleindorfer andKunreuther, 1999). In this analysis,the reduction in damage to thestructure was accomplishedthrough a shift in the fragility (i.e.,vulnerability) curve in the analysisand a recalculation of the expectedloss. If the expected reduction inloss exceeds the cost of undertak-ing mitigation (e.g., step 4 in Fig-ure 1), then one can justifyinvesting in it.

The analysis of the benefits ofmitigating lifeline systems is a morecomplicated process than for a resi-dential structure. Lifeline systemshave unique characteristics, whichmake the calculation of damage tothe system difficult (Chung et al.,1995). First, the loss of function ofthe lifeline is dependent on manyparts of the system, often buried un-derground, spread across a largegeographic region rather than atone location (e.g., as in the case ofa residential structure). For ex-ample, in analyzing the functional

• David Lee and TimFuette, East Bay MunicipalUtility District

• Water distribution systemsin Alameda and ContraCosta Counties, California(working with East BayMunicipal Utility District)

The cost to mitigate is primarily undertaken by owners, buteveryone in a region benefits from uninterrupted or fasterrestoration of lifeline service after a disaster. Therefore, theprimary users of this research are the operators of lifelinesystems, which could be government agencies or private sec-tor organizations who fund the cost of implementing mitiga-tion measures. Cost benefit analyses can help determine ifthe expected reduction in loss exceeds the cost of undertak-ing mitigation. The thrust of the research is to encourage andjustify implementation of retrofit schemes for a variety oflifeline systems.

Step 5Choose Best Alternative

Step 4Calculate Attractivenessof Mitigation Alternatives

(NPV or B/C ratio)

Step 3Determine Loss to System

with and withoutMitigation Alternatives

Step 2Determine Direct Costs

of MitigationAlternatives

Step 1Specify Nature of Problem

- Alternative Options- Interested Parties

� Figure 1. Simplified Cost-Benefit Analysisfor Lifeline Systems

127Using Cost-Benefit Analysis to Evaluate Mitigation for Lifeline Systems

reliability of an electric power trans-mission network, one needs to con-sider the loss of connectivity ofdifferent substations, as well as thevulnerability of the various compo-nents of each substation, to the sys-tem as a whole. This process can beeven more complex when one mustconsider the collocation of differentlifeline systems. For example, if theelectric power system’s cables areadjacent to the water distributionsystem’s pipelines underground,damage to one can compound thedamage to the other.

Second, the damage to the sys-tem, measured in outage or servicedisruption, needs to be translatedinto dollar loss to the interestedparties in the cost-benefit analysis.The benefits of mitigation are thedifference between the loss with-out mitigation and the loss withmitigation. The costs of mitigatingthe system must also be estimated.While these mitigation costs aregenerally borne by the owner andoperator of the lifeline, everyonein the region benefits from the un-interrupted or faster restorationlifeline service after a disaster.1

Therefore, the attractiveness ofmitigation to a lifeline systemshould be viewed as beneficial tosociety as a whole, calculated via asocietal benefit-cost ratio. Thethrust of this research is to encour-age and justify funding for earth-quake hazard mitigation of lifelinesystems.

The Five-StepProcedure

Step 1: Specify Nature of the ProblemTo initiate a CBA, one needs to

specify the options that are being

considered and the interested par-ties in the process. Normally, onealternative is the status quo. In thecase of the current analyses, the sta-tus quo refers to the current vul-nerability of the lifeline systemwithout a mitigation measure inplace. The status quo is likely to bethe reference point for evaluatinghow well other alternatives per-form. In general, if there is sufficientpolitical dissatisfaction with theproposed mitigation options and/or the perceived expected benefits(i.e., reduction in lifeline disrup-tion) are considered to be less thanthe expected costs to mitigate thesystem, then the status quo will bemaintained.

For the utility lifeline problem weare studying in Shelby County, Ten-nessee, the status quo is the vulner-ability of the electric power systemcurrently in place. An alternativeoption is to retrofit or replace someor all of the substation equipmentcomponents in the electric powersystem so they are still functionalafter a severe earthquake. For ex-ample, the transformers in the sub-stations can be seismicallyretrofitted to withstand lateral load-ing. Alternatively, the high-voltagetransformer bushings can be re-placed before an earthquake occurs.

Each of the alternative optionswill impact a number of individu-als, groups and organizations in oursociety. It is important to indicatewho will benefit and who will paythe costs associated with differentalternative options when undertak-ing a CBA analysis. In the case of alifeline system, one needs to con-sider a broad set of interested par-ties. These include residents andbusiness owners affected by theearthquake, public sector agenciesthat must respond and fund the re-

Program 1: SeismicEvaluation and Retrofit ofLifeline Networks

• Task 1.3 Loss EstimationMethodologies

• Task 1.4 Memphis LifelineSystems Analysis

• Task 1.10 RehabilitationStrategies for Lifelines


128

covery process, as well as the gen-eral taxpayer that will bear someof the repair costs of the damagedlifeline system(s).

Step 2: Determine Direct Costs ofMitigation Alternatives

For each mitigation alternative(i.e., all alternatives except the sta-tus quo), one needs to specify thedirect cost to implement the miti-gation measure. For a lifeline sys-tem, the owner and operator of thesystem incurs the costs of mitiga-tion. In a large majority of the casesin the United States, it is a publicsector agency. Furthermore, thecosts are most likely direct mon-etary costs to structurally retrofitor replace some components of thelifeline system. In the case studieswe will present, a governmentagency incurs the direct cost ofmitigation.

Step 3: Determine the Benefits ofMitigation Alternatives

Once the costs are estimated foreach mitigation alternative, oneneeds to specify the potential ben-efits that impact each of the inter-ested parties. In the case of seismicrisk, one considers either a scenarioearthquake event or a set of sce-nario earthquakes of different mag-nitudes, location, duration, andattenuation that can affect the sys-tem. With the specification of thevulnerability of the lifeline system,the damage to the various compo-nents of the system is then esti-mated for each alternative option.Then, overall system reliability isestimated.

With the status quo, there will beno benefits because no retrofit orreplacement scheme is character-ized. In other words, the status quois the damage to the system with-

out mitigation. In the other alter-natives, benefits will be estimatedfrom the change in damage to thesystem with the status quo anddamage to the system with mitiga-tion in place. Once the system reli-ability with each mitigationalternative has been specified, itshould be possible to quantify theeffects of serviceability to the in-terested parties by attaching a dol-lar value to them. The calculationfrom damage to loss is a compli-cated process and will differ fromone interested party to another(e.g., losses to industry from busi-ness interruption differs from resi-dential loss due to relocation).

Step 4: Calculate Attractiveness ofMitigation Alternatives

In order to calculate the attrac-tiveness of mitigation, the nature ofthe benefits to each of the inter-ested parties is estimated and com-pared to the upfront costs ofmitigation. With respect to lifelines,the alternatives involve a degree ofoutage or serviceability over a pre-scribed time horizon (T). One char-acterizes the impact on the keyinterested parties during the daysor weeks that the system will notbe fully functional. One then utilizesa societal discount rate to convertthe benefits and costs of the alter-native over time into a net presentvalue (NPV). If the NPV is greaterthan zero, then the alternative isconsidered attractive. Alternatively,one could calculate the ratio of thediscounted benefits to the upfrontcosts to determine the attractive-ness of the alternative. Wheneverthis ratio exceeds 1 the alternativeis viewed as desirable.

To illustrate, consider the case ofdamage to a water distribution sys-tem from a scenario earthquake

“Theattractivenessof mitigation toa lifelinesystem shouldbe viewed asbeneficial tosociety as awhole.”


event. There could be a period oftime where businesses may not beable to operate in a normal man-ner due to loss of service. If mitiga-tion were implemented (e.g.,underground pipelines were re-placed or retrofitted), the time torestore water to businesses may beshortened. Suppose that the miti-gation measure reduced the resto-ration time by 3 days following asevere earthquake. The resultingsavings in business interruptioncosts during this three-day time in-terval following the earthquake arethen discounted back to thepresent. These savings are multi-plied by the annual probability ofsuch an earthquake occurring tocompute the expected benefits ofmitigation. A similar calculationwould be made for any earthquakethat could occur in the area. Thesesavings are then summed up todetermine the total expected ben-efits which are then compared tothe upfront costs of mitigation todetermine the cost-effectiveness ofthe mitigation measure.

One must be careful, however,when considering the time that ser-vice is unavailable to the users ofcertain lifeline systems. Dependingon the type of system (e.g., trans-portation, water distribution, elec-tric power), days without serviceare, on average, less with electricpower systems than other types ofsystems. This is primarily due to theredundancy in these systems andthe critical nature of electric powersystems in emergency responseand coordination following anearthquake (Chung et al., 1995).

Step 5: Choose the Best AlternativeFinally, once the attractiveness of

each alternative is calculatedthrough a net present value calcu-

lation or a ratio of the benefits tothe costs, one can choose the alter-native with the highest NPV or ben-efit-cost ratio. This criterion is basedon the principle of allocating re-sources to its best possible use sothat one behaves in an economi-cally efficient manner.

Applying the Five-StepProcedure to aTransportation System

We now illustrate how the abovefive-step procedure of CBA can beutilized to evaluate mitigation for a life-line system through a simple example.In the next section, we will indicatethe types of data that are required toundertake different levels of analysesof more realistic problems.

Step 1: Specify Nature of the ProblemThe following question has been

posed to a public agency that ownsand operates a transportation net-work in an earthquake prone regionin the United States: Should the trans-portation agency seismically upgradetheir 2,200 highway bridges so thatthey perform adequately in a majorearthquake?

There are only two alternativesfor this problem:A1. Seismically retrofit the bridgesA2. Do not retrofit the bridges

(i.e., maintain status quo)There are a number of interested

parties who are affected by thisproblem. These include the resi-dents who use the bridges to com-mute to work, the businesses thatuse the transportation network tomove goods, the owner of the trans-portation network, police and fireteams that cannot move across thebridge, environmental groups con-cerned with the impact of a col-

130

lapsed bridge on residents of thesea. We will focus on just the ownerof the network to keep the analy-sis simple.

Step 2: Determine Direct Costs ofMitigation Alternatives

The direct cost associated withseismically retrofitting the bridgesis the material and labor cost tocomplete the retrofit scheme. Inthis case, we assume that, on aver-age (i.e., over 2,200 bridges), thiscost is approximately $30 persquare foot.

Step 3: Determine the Benefits ofMitigation Alternatives

The benefits of retrofitting thebridges depend on whether or notan earthquake occurs and thelength of time (T) that the bridgesare impaired after the earthquakewith and without the retrofitscheme in place. For simplicity, weconsider only one scenario event,with two possible outcomes: (1)the earthquake occurs with prob-ability, p, or (2) no earthquake oc-curs with probability, 1-p.

For this event, the benefits con-sist of several different impact cat-egories. We will consider onecategory for this analysis: the dam-age to the bridges. Damage and losswill be the same or lower if thebridges are retrofitted (A1) than ifthey are not (A2). In other words,the benefits from mitigation are thereduction in costs to repair thedamaged bridges. In this case, weassume that the cost to repair thedamaged bridges without a retro-fit scheme (i.e., replace the bridges)is $140 per square foot. With theretrofit scheme, the cost is zero.

Step 4: Calculate Attractiveness ofMitigation Alternatives

To determine the impact of ret-rofitting the bridges through a netpresent value calculation, three ad-ditional pieces of data are required:(1) the probability of the scenarioearthquake, p; (2) the life of thebridge, T, in years; and (3) the an-nual societal discount rate, d.

As each of these parameters isvaried, one will obtain a differentrelationship between the netpresent value (NPV) of the ex-pected benefits and costs of A1 andA2. To illustrate this point, supposethat one utilized the following pa-rameters to determine whether ornot retrofitting the bridges is ben-eficial to the transportation depart-ment. First, the probability of theearthquake occurring, p, is a 1 in a100-year event (i.e., p = 0.01). Thelife of the bridge, T, is 50 years, andthe discount rate, d, is 4%.

Suppose an earthquake occurredand the bridges were retrofitted.Then, the reduction in losses fromthis retrofit scheme is $140 persquare foot, and the expected ben-efits from mitigation would be cal-culated (1/100)(140) = 1.4. With d= 4%, the expected discounted ben-efits of retrofitting over the 50 yearlife of the average bridge would be$30 per square foot. As we varyeach of the parameters, the dis-counted expected benefits frommitigation will change. In general,higher values of d and/or smallervalues of p and T will cause ben-efits to decrease. If the cost of ret-rofitting the bridges were set at $30per square foot, then the netpresent value would be exactly 1.Any time that this cost was lessthan $30 it would be beneficial toretrofit the bridge. Of course, froma societal point of view, it would

“Developing ameaningfulcost-benefitanalysismethodologyrequiresbringingscientists,engineers andsocial scientiststogether toanalyze aproblem.”


be beneficial to retrofit the bridgeif the cost per square foot exceeded$30. There would be additionalbenefits to residents who use thebridges to commute to work or forpleasure, businesses using thebridges to move goods, and avoid-ance of business interruption dueto loss of the transportation net-work. These additional benefits ofmitigation are due to the reductionin time to repair the damagedbridges.

Step 5: Choose the Best AlternativeThe criterion used by CBA is to

maximize net present value (NPV) ofsocietal benefits or minimization ofthe total societal costs. In the aboveexample, with the expected costsequal to the benefits of $30 persquare foot, retrofitting the bridges(A2) would be preferable over main-taining the status quo (A1).

In addition to the maximization ofsocial benefits, there may be equityconsiderations that play a role in theevaluation of different alternatives.For example, if there were concernsby taxpayers on how much extrathey would have to pay for tolls overthe bridges to reflect the extra ex-penditures of retrofitting the bridges,then this may impact on the imple-mentation of A2 even if it wasdeemed cost-effective using maximi-zation of NPV as a criterion.

In essence, the choice of an opti-mal alternative is based on a set ofassumptions that need to be care-fully examined. In particular, onewill want to undertake a set of sen-sitivity analyses to determine howrobust the proposed solution is. Forexample, if the expected cost ofretrofitting the bridges were only$15 per square foot, then therewould be little doubt that mitiga-tion would be a cost effective one.

The reasoning is simple. The ben-efit-cost ratio for this problemwould be (30/15), which is 2. Evenif the estimates of the benefits ofmitigation were off by a factor of2, one would still want to retrofitthe bridges if the cost of this mea-sure is $15 per square foot.

Mitigation of Lifelinesin Tennessee andCalifornia

We now turn to the two case stud-ies of lifeline systems to illustrate howone would compute the benefits andcosts of mitigation: an electric powersystem in Shelby County, Tennesseeand a water distribution system inAlameda and Contra Costa counties,California. The impacts of an earth-quake on these lifeline systems in-clude a wide range of direct andindirect losses. Our focus is on twotypes of losses to two different stake-holders: (1) the impact of interrup-tion of service on businessoperations and the resulting lossesin gross regional product (to ShelbyCounty) and (2) loss of service to resi-dential customers and the need torelocate individuals and entire fami-lies to temporary shelters (inAlameda and Contra Costa counties).

In these analyses, we consideronly these two limited impacts ofearthquake loss. Other impacts,such as the costs associated withfire following an earthquake, arenot considered here. For the totalloss to the water distribution sys-tem in Alameda and Contra Costacounties, we refer the reader to thestudy completed for the East BayMunicipal Utility District (Homerand Goettel, 1994).

132

Mitigation of an Electric PowerSystem in Shelby County,Tennessee

As a continuation of an analysisdone by researchers at MCEER(Shinozuka et al., 1998), we devel-oped a framework to look at thecost-effectiveness of mitigation tothe electric power system of theMemphis Light, Gas, and Water Di-vision (MLGW) of the Memphis,Tennessee area. Loss to the grossregional product (GRP) of the dif-ferent business sectors in the re-gion due to loss of power isconsidered.

Figure 2 depicts a map of theMemphis/Shelby County area withindicated Modified Mercalli Inten-sity (MMI) impacts by census tractfor a M = 7.5 earthquake. TheShelby County area is at risk fromseismic activity in the New MadridSeismic Zone (NMSZ) of the Cen-tral United States. This seismic zonelies within the central MississippiValley, extending from northeast Ar-kansas, through southeast Missouri,western Tennessee, western Ken-

tucky to southern Illinois, andwhose center is located to thenorthwest of Memphis. The NMSZhas a history of earthquake activ-ity, with the largest earthquakesrecorded in 1811-1812, and thus, itis an interesting case study for po-tential earthquake damage.

For this analysis, we used a M =7.5 earthquake with an epicenterin Marked Tree, Arkansas, situatedfifty-five kilometers northwest ofdowntown Memphis. We lookedspecifically at mitigation that couldbe undertaken on electric powersystems, basing this study on a net-work of electric power serviced byMemphis Light, Gas and Water(MLGW). We examine the loss ofserviceability of the electric powersystem from the scenario M = 7.5earthquake, based on data devel-oped by Chang (See Shinozuka et.al, 1998). We should note that dam-age to the overall system reliabilitywas estimated using Monte Carlosimulation techniques, and withthis damage, loss of service to dif-ferent service zones was estimated.The focus of this research is howto use this output from the simula-tion analysis to look at the cost-ef-fectiveness of mitigation.

Electric power is absolutely cru-cial in maintaining the social sys-tems of a city, directly affectingbusinesses that cannot operatewithout it. The study of mitigationfor the gas, power and water distri-bution systems is not studied here.For an update on the damage andloss to the water distribution sys-tem, see Chang et al. (2000).

Mitigation in this case is a seis-mic retrofit of a component of theelectric power substation. Specifi-cally, the transformers, used withinthe substations of the network toconvert power, are retrofitted to

� Figure 2. Shelby County, Tennessee (NCEER Bulletin 1996)2


withstand lateral loading. Tyingdown the wheels on the track cansignificantly improve the slidingmovement and reduce the chancesthat these large, critical structureswill overturn.3

The estimation of loss to the busi-nesses in the area, due to serviceinterruption, is calculated using thedirect economic loss methodologyfrom ATC-25 and modified byChang to accommodate the impor-tance factors specific to the Mem-phis area. The initial loss, L

j, to

business sector j in the region di-rectly following earthquake is esti-mated by equation (1).

L a g Fj j j= − ⋅ ⋅( )1 (1)

In the above equation, a is theinitial availability to the area. Inother words, it is the percentage ofcustomers receiving service imme-diately after the earthquake. gj isthe gross regional product (GRP)in the region for the business sec-tors, j. In our case, j includes tenseparate sectors. Examples includeagriculture, construction, manufac-turing, and services. The GRP wasdeveloped using data from the U.S.Census Bureau. Finally, Fj is the im-portance factor of the business sec-tor, described as the percentage ofproduction in sector j that wouldbe lost if electric power servicewere completely disrupted (ATC-25, 1991).

For the scenario earthquakeevent, suppose restoration time isT days. So, the loss of productionover T days, LT, is addressed byequation (2).

L LT t

TT jt

T

=-

=

*( )

0

Σ (2)

In this simplified analysis, if thetransformers are seismically retro-fitted, then we assume the loss ofserviceability will be reduced, sincesome or all of the transformerswould be functional after the earth-quake. After knowing the differ-ence in the number of functionaltransformers before seismic retro-fit and after the retrofit scheme, onecan then compute the savings tothe business sector. In other words,the reduction in electric power out-age characterizes the benefits ofmitigation should the scenarioearthquake occur in the servicearea. The expected benefits of miti-gation are calculated by multiply-ing these benefits by theprobability of the earthquake (p)and discounting over the relevanttime horizon (T) that the trans-former is expected to be in use (i.e.,step 4 in the cost-benefit analysis).

To illustrate the type of CBA thatcould be undertaken, consider thefollowing illustrative example. As-sume that the cost of retrofittingeach transformer is $10,000. If all60 transformers in the electricpower system are retrofitted, thereis a one-time total cost of $600,000.Therefore, there are two alterna-tives: the status quo and retrofittingall the transformers for an upfrontcost of $600,000. The stakeholdersin the analysis are the businessesin the area needing electric powerfor production output. The benefitsof mitigation are the reduction inrestoration time due to more trans-formers functional following theearthquake event.

Additionally, we assume that theelectricity system will last for 50years and an 8% annual discountrate is used to compute the ex-pected benefits over the 50-yearhorizon from retrofitting the trans-

134

formers. We consider an earth-quake of M = 7.5 would occur inthe region once every 500 years, es-timated from a study done by theUSGS (Atkinson et al., 2000).

Figure 3 depicts the sensitivity ofthe CBA to restoration times of theelectric power system would bereduced by either 1, 3 or 5 days.The figure is designed to show howCBA can be utilized to evaluatewhether or not mitigation of cer-tain parts of the system is cost-ef-fective to certain stakeholders inthe analysis. In this example, retro-fitting transformers can be seen tobe beneficial (i.e., the benefit-costratio is greater than 1) for a timehorizon T greater than 13 yearswhen restoration of the power sys-tem is reduced by five days and forT greater than 23 years when res-toration is reduced by three days.Retrofitting the transformer is notcost-effective for any T when res-toration of the power system isonly one day. This is important tonote because after the Northridgeearthquake in 1994, electric power(LAPWD) was restored to ninety-

percent of the users within one day.But, in the Chi-Chi, Taiwan earth-quake in 1999, it took longer to re-store power, with rolling blackoutsto the northern part of the islandfor a week or more.

This analysis should be viewed asillustrative rather than definitive,since it is highly dependent on theassumptions made regarding thecosts of retrofitting and the busi-ness interruption savings as well asthe discount rate, d. For example, ifd were four percent, then retrofit-ting the transformers would beeven more attractive than underthe current analysis.

Mitigation of a WaterDistribution System in Alamedaand Contra Costa Counties,California

The East Bay Municipal UtilityDistrict (EBMUD) is the public util-ity that provides potable water ser-vice to Alameda and Contra CostaCounties in northern California. Weutilized information from theEBMUD Seismic Evaluation Pro-gram Final Report (Homer andGoettel, 1994), an intensive studyof the East Bay water system thatdeveloped mitigation initiativesthat will be completed during theirSeismic Improvement Program.Our analysis focuses on the mitiga-tion of 172 water storage tanks in120 of the 122 EBMUD pressurezones.

To conduct this analysis, four sce-nario earthquake events were cho-sen for analysis, based on thecompleted EBMUD report (Table1). These events represent the maxi-mum plausible earthquake eventthat could occur on any given fault,with the exception of the Hayward

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 5 10 15 20 25 30 35 40 45 50

Time period (years)

Be

ne

fit-

Co

st

Ra

tio

1 day

3 day

5 day

Baseline

� Figure 3. Illustrative Example of Effects of Electricity Lifeline Mitigation


Tank Size Seis. Upgrade Cost 1 Partial Loss: Cost to

make 'operational' after

EQ. No upgrade prior to

EQ" 2

Partial Loss: Repair time

for upgrade to make

'Operational' after EQ

Full Loss: Install Temp.

Bolted Steel Tank for

"Operational

Restoration" 3

New Tank Cost: Steel

Replacement 4

Full Loss: Replace Tank

with Equivalent size tank

Concrete

0.5 500,000$ $ 150,000 6 weeks 10 weeks 1,000,000$ 1 year

1 600,000$ $ 150,000 7 weeks 10 weeks 1,500,000$ 1 year

1.5 750,000$ $ 150,000 7 weeks 10 weeks 1,750,000$ 1 year

2 900,000$ $ 200,000 8 weeks 10 weeks 2,000,000$ 1 year

3 1,200,000$ $ 200,000 8 weeks 10 weeks 2,500,000$ 1 year

5 1,500,000$ $ 200,000 9 weeks 10 weeks 3,000,000$ 1 year

Steel

0.25 100,000.00$ $ 50,000 1 week 10 weeks 650,000$ 1 year

0.5 200,000.00$ $ 50,000 1 week 10 weeks 1,000,000$ 1 year

1 250,000.00$ $ 50,000 1 week 10 weeks 1,500,000$ 1 year

3 300,000.00$ $ 70,000 2 weeks 10 weeks 2,500,000$ 1 year

5 400,000.00$ $ 70,000 2 weeks 10 weeks 3,000,000$ 1 year

M = 6.0. For the Hayward fault,an earthquake of magnitude M= 6.0 is the most probable eventand M =7.0 is the maximum plau-sible event.

In this analysis, there are twoalternatives: the status quo andthe mitigation of the water stor-age tanks. Those affected by the dis-ruption of the water supply systemare the residents in the service areathat may be forced to relocate totemporary shelters for a period ofdays or even weeks. The direct costof retrofitting the water tanks isbased on data supplied by theEBMUD and given in Table 2.

Loss of Serviceability and Effectson Residential Displacement

The benefits of mitigation are thelosses resulting from loss of waterservice to residential customerswith and without the seismic up-grade of the tanks. When water ser-

vice is disconnected from a placeof residence, the individuals wholive in these buildings will beforced to find alternative shelteruntil water service is restored. Inthe event of an earthquake, thepotable water system can be dam-aged to an extent that water ser-vice may not be restored to an areauntil many weeks following theevent. Figure 4 details our method-ology for estimating the costs thatare associated with these displacedindividuals.

First, we determined the likeli-hood of tank failure in each pres-sure zone. EBMUD performed adetailed analysis of their system, us-

Fault Name M DescriptionHayward 7.0 event ruptures 50 to 60 km long segment of faultHayward 6.0 event ruptures an 8 to 13 km long segment of the faultCalaveras 6.75 event ruptures the northernmost part of the known active faultConcord 6.5 event occurs along this fault east of the EBMUD service area

� Table 1. Scenario Events

Footnotes:1 Includes non-seismic upgrades, which are approximately 15% of seismic costs.2 Partial loss for steel tank due to damaged valve pit piping. Time due to valve replacement. The tank structure assumes to either fully survive or fully

fail.3 The existing tanks may be out of service for this period of time; however, if there are multiple reservoirs in a pressure zone, the business interruption

will likely be a much lesser time than indicated. Assume temporary replacement with either a .28 MG or 0.4 MG tank.4 Cost includes: demolish existing tank, valve pit modification, metal appurtenances, miscellaneous site work, water quality piping, 15% construction

contingencies.

� Table 2. Direct Cost to Retrofit Water Tanks

136

ing their own system risk analysissoftware, SERA. Ten simulations ofeach of the four above earthquakes(i.e., forty simulations in all) wererun to determine the extent ofdamage that their system is ex-pected to face following an earth-quake. For each simulation, tankswere said to fully fail, partially fail,or not fail. Our analysis of expectedloss consists only of two of thesethree states – full failure and no fail-ure – since we have no means bywhich to accurately analyze thecomplex network effects of partialloss. Therefore, if a tank was saidto fully fail in 6 of 10 Hayward M7.0 scenarios, then we say there isa 60% chance of full tank failurefollowing a Hayward M 7.0.

Next, determination of likely wa-ter service within individual pres-sure zones was calculated. Somepressure zones are served by morethan one tank, and therefore, the

overall loss of service in a pressurezone is a combination of the effectsof individual tanks in the pressurezone failing. The loss to pressurezone one, L

PZ1, therefore, can by

defined by equation (3). In theequation, A

N1 is the probability of

full loss for a tank in pressure zoneone and N1 is the number of tanksin zone one.

LA

NPZN

t

N

11

0 1=

=∑ (3)

Although a pressure zone mayexperience loss of service due tofailure of its internal tanks, it mayalso experience service interrup-tion if pressure zones that pro-vide water to its tanks experiencefailure. To fully understand the to-tal loss within a zone, we musttake this into consideration as in-dicated by equation (4):

P X X X

X X X Ln

n PZ

1 1 2

1 2 11

= ⋅ ⋅ ⋅ ⋅ ⋅ +− ⋅ ⋅ ⋅ ⋅ ⋅ ⋅

( )

[ ] (4)

In equation (4), P1 is the percent-

age of customers in pressure zoneone that will be without servicefollowing an earthquake, X

1

through XN represent the probabil-

ity of failure in pressure zones thatfeed into pressure zone one, andL

PZ1 is the loss due to tank failure

in pressure zone one. If every pres-sure zone that provides water tothe tanks in pressure zone one fails,then no customers in pressurezone one will have service. Theprobability that this occurs is rep-resented by the product of theprobabilities that each individualprecedent pressure zone will fail.If any or all of these precedent pres-sure zones has service, however, wewill assume that the district will beable to provide enough water to its

Shelter Cost per

Person per Day

Damage to

Tank A

Damage to

Tank B

Damage to

Tank C

Damage within

Pressure Zone

Damage in

Zone A

Damage in

Zone B

Damage in

Zone C

Damage in

Upstream Zones

Total Service Losswithin Pressure

Zone

Number of

People Seeking

Public Shelter

Days until

Restoration of Service

Total Shelter Costs

� Figure 4. Cost to Shelter Displaced Residents from one Pressure Zone


tanks in pressure zone one so thatcustomers have their minimumwater demands met. Then, the onlyresidents without water servicewill be those represented by L

PZ1,

or those who do not have servicedue to tank failure within pressurezone one.

Once the overall service to eachpressure zone is calculated, thenumber of people displaced andseeking temporary shelter can bedetermined. Since we are analyzing120 pressure zones in the system,the number of people seeking tem-porary shelter following an earth-quake is as follows:

R u P n sS z z hz zz

= ⋅ ⋅ ⋅=

∑1

120

(5)

In equation (5), Rs represents the

number of residents seeking tem-porary public shelter following adisaster and u

z is the number of

occupied housing units in pressurezone z. This is found by overlayingU.S. Census data on the service dis-trict. P

z is the percent of service loss

in pressure zone z, or the percent-age of housing units without wa-ter service, and n

hz is the number

of people per housing unit in zonez. This is calculated by dividing thetotal population in zone z by thetotal number of occupied housingunits. Finally, s

z is percentage of dis-

placed residents in zone z who wellseek shelter in a public facility. Thisis determined by inputting U.S. Cen-sus tract demographic informationinto a methodology developed forHAZUS (NIBS, 1997)4. This analysisis run for each of the districts 120pressure zones considered, thenaggregated to determine the totalnumber of residents across the dis-trict who will seek shelter follow-ing a disaster.

To determine the total cost ofsheltering residents displaced byloss of potable water service totheir home following the earth-quake, we multiply R

s by $11.50.

This value is based on an AmericanRed Cross estimate of between $10and $13 to shelter a person for asingle day (Red Cross, 2000).

Finally, to determine the total ag-gregate displacement costs follow-ing an earthquake, we must firstconsider whether the per-day costswill remain constant over time. Weassume that if a tank is fully dam-aged it will not be able to supplywater to any of its customers untilit is fully repaired, or until a replace-ment tank can be acquired. There-fore, there will be no incrementalincrease in service over time andthe per-day total costs of shelter-ing residents will remain constantover that period (since people arenot able to return home). Based onthis assumption, the total cost toshelter displaced residents follow-ing an earthquake is easily calcu-lated by multiplying the total costper day by the total time until res-toration of service. In this case, weassume ten weeks or seventy days,based on information provided byEBMUD.

Note that the above analysis char-acterizes the impact on sheltercosts under the assumption thatthe status quo was being main-tained. The alternative option is toretrofit the water tanks in some ofthe pressure zones in Alameda andContra Costa counties. By undertak-ing this protective measure, onereduces the chances that one ormore of the tanks will fail andhence, disrupt water service forsome period of time. The benefitsof this mitigation measure will bedetermined by the reduction in res-

138

toration time for water to resi-dences in the area affected by theearthquake.

One also needs to take into ac-count whether individuals wereforced to evacuate their homes be-cause there was severe structuraldamage due to the earthquake. Sup-pose one is evaluating the benefitsof retrofitting water tanks to resi-dents in the area. Then one shouldeliminate any homes where evacu-ation would be required even if thewas no disruption of water. Other-wise, one would be overstating thebenefits from retrofitting the watertanks in the East Bay area.

This analysis of the residentialsector in the East Bay area illus-trates the type of calculations onewould make to evaluate the ex-pected benefits of mitigation. Turn-ing to business interruption lossesfrom an earthquake, a similar analy-sis to the one in Shelby County isbeing undertaken with EBMUD innorthern California to evaluate theimpact of mitigation on this sector.Combining both the residential dis-placement costs and business inter-ruption, one could undertake amore comprehensive CBA, deter-mining under what circumstancesthe proposed mitigation will bemost cost-effective.

Making CBA Useful forPolicy Analysis

The above examples are only il-lustrative as to how CBA can beused, rather than how it is actuallyapplied to either Shelby County,Tennessee or Alameda and ContraCosta counties, California. For CBAto be a useful tool for policy analy-sis in a specific region, one musthave the most accurate data avail-

able for the analysis and keep theinterested parties’ priorities inmind. For this reason, we have beenworking very closely with person-nel at EBMUD to ensure that ouranalysis of their water supply is ameaningful one.

Our intention is to undertake asufficiently rich analysis of howmitigation can be utilized for par-ticular lifeline systems, such as theEBMUD water distribution system.Unless key decision makers canappreciate the role CBA can playin determining whether to imple-ment specific mitigation measures,this methodology may have sometheoretical interest but no practi-cal importance.

The task of making CBA a usefulmethodology is a challenging one.It requires bringing together scien-tists and engineers with social sci-entists to analyze a problem. Itrequires one to articulate the na-ture of the uncertainties associatedwith the recurrence interval ofearthquakes of different magni-tudes, as well as the confidence in-tervals surrounding the expectedbenefits and costs of different al-ternative strategies. In a nutshell, itrequires the integration of sciencewith policy.

The data and techniques are nowavailable to undertake this type ofintegration with respect to earth-quake mitigation. The challenge isto present the analyses to decisionmakers so they are willing to de-fend the proposed recommenda-tion because it makes economicsense to them while satisfying theirpolitical concerns. Cost-benefitanalysis provides a framework foraccomplishing this important task.


ATC-25, 1991, Seismic Vulnerability and Impact of Disruption of Lifelines in theConterminous United States, Applied Technology Council, Redwood City, California.

Atkinson, G. et al., 2000, “Reassessing New Madrid.” 2000 New Madrid SourceWorkshop, http://www.ceri.memphis.edu/usgs/pubs.shtml.

Boardman, A., Greenberg, D., Vining, A., and Weimer, D., 2001. Cost-Benefit Analysis:Concepts and Practice (2nd Edition), Upper Saddle River, New Jersey: Prentice Hall.

Chang, S., Rose, A., Shinozuka, M., Svekla, W.D., and Tierney, K., 2000. ModelingEarthquake Impact on Urban Lifelines Systems: Advances and Integration.Research Report, Multidisciplinary Center for Earthquake Engineering Research,Buffalo, New York.

Chung, R.M., Jason, N.H., Mohraz, B., Mowrer, F.W, and Walton, W.D., (eds.), 1995. Post-Earthquake Fire and Lifelines Workshop: Long Beach California, January 30-31,1995 Proceedings, NIST Special Publication 889, U.S. Department of Commerce.

Homer, G. and Goettel, K., 1994. Economic Impacts of Scenario Earthquakes, FinalReport for East Bay Municipal Utility District.

Kleindorfer, P. and Kunreuther, H., 1999, “The Complementary Roles of Mitigation andInsurance in Managing Catastrophic Risks.” Risk Analysis, 19:727-38.

NIBS, 1997. HAZUS: Hazards U.S.: Earthquake Loss EstimationMethodology, NationalInstitute of Building Sciences, NIBS Document Number 5200.

Red Cross, 2000. Phone conversation with Garrett Nanninga, American Red CrossNational Headquarters, 28 July 2000.

Shinozuka, M., Rose, A. and Eguchi, R. (eds.), 1998, Engineering and SocioeconomicImpacts of Earthquakes: An Analysis of Electricity Lifeline Disruption in the NewMadrid Area, Monograph No. 2, Multidisciplinary Center for Earthquake EngineeringResearch, Buffalo, New York.

Shinozuka, M., Cheng, T.-C., Feng, M., and Mau, S.-T., 2000. Seismic Performance Analysisof Electric Power Systems, Research Report, Multidisciplinary Center for EarthquakeEngineering Research, Buffalo, New York.

Shinozuka, M., Grigoriu, M., Ingraffea, A.R., Billington, S.L., Feenstra, P., Soong, T.,Reinhorn, A., and Maragakis, E., 2000. Development of Fragility Information forStructures and Nonstructural Components, Research Report, Multidisciplinary Centerfor Earthquake Engineering Research, Buffalo, New York.

References

Endnotes1 Of course, the public utility or private sector organization operating the lifeline facility

may raise its rates to reflect the additional cost of the mitigation measure. In thissense, all the users of the facility bear the costs of loss prevention.

2 http://mceer.buffalo.edu/publications/bulletin/96/02/apr96nb.html.

3 This type of mitigation was chosen, based on discussions with Masanobu Shinozukaand insight from the annual MCEER conference in November 1999.

4 This methodology does consider that a proportion of displaced residents will staywith friends or family rather than seek publicly-funded shelters.

141

Research Objectives

This paper describes an approach used to develop retrofit strategiesfor hospitals and other critical facilities in low to moderate seismichazard zones, where strong earthquakes are infrequent, but if theyshould occur, the consequences would be high. Hospitals in New YorkState and other urban centers in the eastern U.S. fall into this category,where seismic retrofit requires information on the impact of losingmedical services after a destructive earthquake. A team of MCEER re-searchers is currently developing an approach to address this task. It isa truly multidisciplinary effort, with team members from a variety ofdisciplines including engineering, seismology, structural dynamics, riskand reliability analysis, manufacturing process engineering, computersimulation, urban and regional planning, and economics. When thisresearch task is completed, it will be united with MCEER’s generalhospital project to develop seismic retrofit strategies.

Retrofit Strategies For Hospitals inthe Eastern United States

by George C. Lee, Mai Tong and Yasuhide Okuyama

National Science Foundation,Earthquake EngineeringReseach Centers Program

George C. Lee, Director,Multidisciplinary Centerfor EarthquakeEngineering Research, andSamuel P. Capen Professorof Engineering, ApostolosPapageorgiou, Professor,and Mai Tong, SeniorResearch Scientist,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Li Lin, Associate Professor,Department of IndustrialEngineering, University atBuffalo

Yasuhide Okuyama,Assistant Professor andErnest Sternberg,Associate Professor,Department of Architectureand Planning, University atBuffalo

Masanobu Shinozuka, FredChampion Chair in CivilEngineering, Departmentof Civil Engineering,University of SouthernCalifornia

A major MCEER research thrust is the development of retrofit strat-egies for critical facilities. By fostering team efforts, the research is

focused on comprehensive protection of emergency medical service func-tion of hospitals in the event of a destructive earthquake. This researchrequires system integrated studies involving earthquake hazards, fragili-ties of all structural and nonstructural components and systems, as well ashuman services provided by medical and support staff, and impact andbenefit-cost analyses. Experiences and approaches developed from thehospital projects will then be extended to seismic retrofit of other criticalfacilities such as communication centers or manufacturing complexes.

MCEER’s hospital retrofit research program addresses two specific typesof seismic hazard locations. The first is for hospitals located in regionswith frequent and/or high seismic hazard levels such as many communi-ties in California where retrofit is required by law. The second type ofhospitals are those located in regions with low seismic hazard levels whereearthquakes have a very long return period but the structures and con-tents are likely to be damaged when earthquakes do occur.

142

For the first type of hospitals (e.g.,those in California) a considerableamount of engineering and socialscience studies are being carriedout by MCEER and other research-ers (for example, see Johnson et al.,1999). Relatively little informationis available on how to approachhospitals located in low seismichazard regions but having high risks(e.g., those located in eastern U.S.urban centers). This article brieflysummarizes MCEER’s approach indeveloping retrofit strategies for thelatter, with an emphasis on estab-lishing an evaluation system for ret-rofit strategies.

Different QuestionsAsked for Californiaand New YorkHospitals

Because of California Law SB1953(Alquist Act), California hospitaladministrators and code writingauthorities are required to considerthe nature of functional design forcritical care facilities. OSHPD (Cali-fornia Office of Statewide HealthPlanning and Development), whichis directed by SB1953 to addressand implement the legal require-ments, now requires that by Janu-ary 1, 2030, all hospital buildingswill meet the seismic standards ofthe Hospital Act. Also, OSHPD is

in the stage of writing the imple-mentation procedures required bySB1953 for the nonstructural pro-visions. Under such legal require-ments, the challenges for Californiaare largely focused on engineeringtasks.

In areas of the eastern U.S. suchas New York City, hospital retrofitdecisions are made based on differ-ent considerations. In particular,given that only limited financial re-sources are available for protectionfrom various natural hazards of ap-proximately the same level of prob-ability of occurrence, retrofitdecisions become an optimal riskmanagement issue. For these twodifferent conditions, we may thusbegin by asking two different ques-tions for the MCEER hospitalproject.

For California hospitals:• How can the requirements to

retrofit be met cost-effectively?For New York hospitals:

• Should resources be allocated forthe seismic retrofit of hospitals?

MCEER’s hospital project is di-vided into two separate aspects intheir initial phase. For the moregeneral situation (represented byCalifornia hospitals), major effortsare devoted to engineering activi-ties to establish fragility informationfor the physical components andsystems, and identify critical prob-lem areas in structures, nonstructural

• State of New YorkDepartment of Health,Bureau of Architecturaland Engineering FacilityPlanning, Troy, New York

• Columbia PresbyterianMedical Center, New York,New York

• Erie County MedicalCenter, Buffalo, New York

Hospital administrators, building owners and otherstakeholders in regions of minor to moderate seismicitycan use the evaluation system for retrofit strategies tomake optimal risk management decisions. Resources forhazard mitigation of all types are limited, and a decision-making method based on solid cost-benefit principles willbe a valuable tool.

143Retrofit Strategies For Hospitals in the Eastern U.S.

components, equipment, etc. thatrequire seismic retrofit. For thesecond situation (represented byhospitals in the eastern U.S. ), weconcentrate on establishing a deci-sion-making method which canprovide information on the impactto the community if medical ser-vice function is lost after an earth-quake due to different levels ofdamage scenarios to the variousrequired service functions of thehospital. Once a decision is madeto perform seismic retrofit, the pro-cess will be merged with that de-veloped for the California hospitals.At that point, we consider impactto the community when there aremultiple hospitals, followed by ben-efit-cost analyses for different pos-sible retrofit options.

System Evaluation forHospitals in New York

We envision a five-step decision-making process for the seismic ret-rofit of hospitals. These steps are:1. Establish earthquake hazard2. Develop fragility informa-

tion and identify criticalproblem areas in the physi-cal system

3. Establish an analysis tool(hospital operation model)to carry out evaluation of se-lected seismic scenarios todetermine their impacts tomedical services of a givenhospital

4. Carry out community im-pact analyses (multiple hos-pitals/health care facilities)

5. Perform benefit-cost analysisand determine retrofit op-tions.

For a free-will decision, the thirdstep is crucial because the retrofitbenefit has to be evaluated in com-parison to those of other compet-ing projects for limited resources.

Being aware of this importantlink, MCEER is concentrating on thethird step by working with severalhospitals in New York State. Thesehospitals are located in SeismicZone C (Z = 0.15), where earth-quakes with magnitude ≥4.5 or in-tensity ≥ VI have been experiencedhistorically.

For the purpose of evaluatingvarious natural hazard reductionschemes, we start to develop a hos-pital operation model based on pa-tient flow as shown in Figure 1.

If hospital services are consideredas a process, the key element in thispatient-flow model is the centerblock that describes the process ofhow patients receive their medicalservices. The services are supportedby two types of resources: human


� Figure 1. The Patient Flow Model of a Hospital

Hospital Services

Power System

Water System

Transportation System

Administration Support Staff

NursesDoctors

HVAC

Information System

Medical Equipment

etc.

Patients In Patients OutHospital Services

Power System

Water System

Transportation System

Administration Support Staff

NursesDoctors

HVAC

Information System

Medical Equipment

etc.

Patients In Patients Out

144

and material. Since our target is toevaluate the benefits of structural/nonstructural retrofit, the emphasisis given to material resources,which typically include power sys-tems, water systems, informationsystems, medical systems, transpor-tation systems, HVAC and others.Depending on the designated func-tion of hospitals (trauma center,general hospitals, special medicalcare, etc.), the center block will in-volve different service units. Ourcurrent emphasis is on modelingthe emergency medical service,which is illustrated in Figure 2.

Each of the service units has in-ternal structures and are intercon-nected. Therefore, modeling of thehospital operation has to considertwo layers of relationships.

With the emergency medical ser-vice unit of a hospital configurationdescribed in Figure 2, a key step ofmodeling is the internal relation-ship between various service units(departments) and the delivery ofemergency medical services. In par-ticular, some factors such as season-ality, abnormality, and patient

distribution, a critical disaster eventmay have a different impact tothese relationships. Also, to evalu-ate different retrofit schemes, thelevel of detail of the model may vary.In Figure 2, only some typical ser-vice units are indicated for the pur-pose of illustration. The arrows onlyprovide examples of possible onedirection patient flow. The total sys-tem would be too complicated forillustration of the concept. In gen-eral, an all-purpose comprehensivemodel may not be a good approach,since too many factors introducetoo many uncertainties, whichwould eventually lead to an unreli-able model.

Similar to modeling the necessarycomponents of providing emer-gency medicine, the utility systemsuch as water supply, as expressedin Figure 3, can be modeled so thatthe relationship among the variousunits can be examined (e.g., effectof damaged water pumps on thewater supply). This utility modelcan be linked to the emergencymedical service model, where con-sumption of water is required. Here,

Patient Source

Primary Care

X-Ray

/ MRI

Lab

ICU

OR

ER

Specialty Care

Out patient

In patient

Patient Source

Primary Care

X-Ray

/ MRI

Lab

ICU

OR

ER

Specialty Care

Out patient

In patient

� Figure 2. Hospital Medical Service Units


it is important to understand thatthe same utility system may bemodeled differently for applica-tion to various physical problemsand retrofit treatments.

Forrester NetworkModel

As mentioned earlier, the pur-pose of modeling hospital opera-tion is to evaluate the benefit ofseismic retrofit. For this reason,we need a quantitative model.This requirement can be satisfiedby a Forrester type of networkmodel. The essential steps of aForrester model are to break downthe physical units and their rela-tions into standard input/outputunits and networked relations be-tween these basic units. Then therelations are modeled by differencefunctions including differentiation,integration, and other elementary

functions; or they could be repre-sented by an empirical function orsome logical relations.

Figure 4 shows an example ofForrester-type systems model forthe emergency room in a large hos-pital. Incoming patients are classi-fied into three categories: minor,moderate and major injuries/illness.Patients classified as minor andmoderate are attended to first, andcomplete an information sheet in

Water

Supply

Water tanks

OutletsMajor Pipelines

Water Pumps

Distribution

pipelines

Service

Units

Water

Supply

Water tanks

OutletsMajor Pipelines

Water Pumps

Distribution

pipelines

Service

Units

� Figure 3. Water System Units

� Figure 4. Forrester Type Systems Model for Emergency Room Operation

146

the reception area; they then pro-ceed to the triage rooms for treat-ment. The time spent in this process(reception to triage) depends onthe degree of injuries/illness (pa-tients classified as “moderate” spendmore time than those classified as“minor”) and on the capacity of thetriage rooms. Then, after the triageroom, the patients proceed to thewaiting room before receiving treat-ment. The time spent in this pro-cess (waiting room to treatmentroom) also depends on the degreeof injuries/illness and the capacityof treatment rooms for minor andmoderately injured patients. On theother hand, a severely injured pa-tient skips the triage process, dueto the urgency of their injuries/ill-ness, and the information on theirinjuries is provided by paramedicstaff. The time spent from receptionto treatment for a patient classifiedas “major” depends on the capacityof the treatment rooms. Becausethis model represents only the op-eration of an emergency room, thepatients are moved out of themodel after the treatment rooms.

Once an internal disaster occurs(such as fire, water pipe broken,power outage, and so forth), theprocess times of the triage roomand treatment room are affectedand become longer depending onthe damage. In order to simulate aninternal disaster, the material re-sources of the hospital are modeledas shown in Figure 5. Internal di-saster in this model can be damageto either water pipes, electric lines,medical gas pipes, emergency roomstructure, structure in the other ar-eas, medical supplies in the emer-gency room, the inventory ofmedical supplies in the hospital, orany combination of these situations.The damage can be defined as a

time variant function (either con-tinuous or discrete). In general, life-line facilities, medical gas, andmedical supplies are modeled suchthat under normal circumstances,these resources are consumed ona per patient basis, and the inven-tory is filled either when consumed(water, electricity, medical gas) orwhen additional supplies are pur-chased (medical supplies). Once aninternal disaster occurs, one ormore of these resources sustaindamage. For water, electricity, andmedical gas, the piping may be dam-aged and they cannot be suppliedas usual (reduced amount or totalcut off). Then, emergency transferfrom the reserve tanks (for waterand medical gas) or from the hos-pital power generators (for electric-ity) compensates for the reducedsupply. The capacity of these re-serve emergency measures can bespecified by resource. For medicalsupplies, when an internal disasterlimits their capacity in the emer-gency room or damages them di-rectly, reserve supplies can betransferred from the hospital’s in-ventory. In conjunction with dam-age to lifelines and supplies, relatedlaboratories (such as X-ray, bloodtest, and so on) and facilities (oper-ating rooms, intensive care units,and so forth) in the hospital are alsorestricted by the internal disaster.

The impacts on these materialresources are then fed back to theoperation model discussed above.Due to the internal disaster and re-lated damage, the process times ofboth triage and treatment roomsare affected, and become longer.Consequently, the number of pa-tients treated in this emergencyroom may be reduced in a complexmanner. The degree of impact toeach patient (classified as minor,


moderate or major) may vary dueto the different requirements fortreatment (for example, a patientwith major injuries requires moreelectricity, medical gas, and medi-cal supplies than less seriously clas-sified patients).

This model can be tailored for aspecific hospital with informationregarding the capacity of the emer-gency room and hospital, and thesize of its emergency reserve of re-sources. The model can simulatepatient flows through the emer-gency room under any scenario ofinternal disaster, external disaster,or a combination of both. Themodel can also provide statistics,such as average treatment time, av-erage time spent in the waitingroom, and so forth.

Transfer Function and State-Space Models

A Forrester type of networkmodel is convenient for modelers,but it is not easy to carry out sys-tem analysis and evaluation. Fortu-nately, we have several otheranalytical models which are equiva-lent to the time domain networkmodel. Two of the most usefulequivalent models are the transferfunction model and state-spacemodel. For these models, manyavailable control theoretical analy-ses can be applied. For instance, thezero-pole analysis for transfer func-tion analysis may help us to under-stand the frequency domaincharacteristics of the modeled sys-tem. Several impact response pa-

� Figure 5. Sub-Model of Material Resources for Internal Disaster

148

rameters such as arising time, ad-justing time, peak response timeand PO% (percentage overshoot)may provide a quantitative measureof the hospital performance undera sudden hazard event.

For instance, consider an earth-quake event which results in a sud-den increase of in-flow patient rateand simultaneous damage to someutility systems. With the shortage ofmedical staff, water and power sup-ply, the medical services of the hos-pital will be reduced while patientsare arriving at a much higher rate.The rising time indicates howquickly the hospital capacity will besaturated. Adjusting time revealshow long the services delay prob-lem will last. The peak responsetime indicates when the worst casewill be and PO% describes howbad the situation could be. Withthese evaluations, we may be able

References

Chong, W.H. and Soong, T.T., Sliding Fragility of Unrestrained Equipment in CriticalFacilities, Technical Report MCEER-00-0005, University at Buffalo, July 2000.

Coyle, R.G., Management System Dynamics, John Wiley & Sons, 1977.

Forrester, J.W., Industrial Dynamics, The MIT Press, 1980.

Johnson, G.S., Sheppard, R., Quilici, M.D., Eder, S.J., and Scawthorn, C. R., SeismicReliability Assessment of Critical Facilities: A Handbook, SupportingDocumentation, and Model Code Provisions, Technical Report MCEER-99-0008,University at Buffalo, April 1999.

Roth, C., “Logic Tree Analysis of Secondary Nonstructural Systems with IndependentComponents,” Earthquake Spectra, Vol. 15, No. 3, 1999.

Shinozuka, M., Grigoriu, M. (Coordinating Author), Ingraffea, A.R., Billington, S.L., Feenstra,P., Soong,T.T., Reinhorn, A.M. and Maragakis, E., “Development of Fragility Informationfor Structures and Nonstructural Components,” MCEER Research Progress andAccomplishments, May 2000.

Soong, T.T. (Coordinating Author), Yao, G.C., and Lin, C.C., “Damage to Critical FacilitiesFollowing the 921 Chi-Chi, Taiwan Earthquake,” MCEER Research Progress andAccomplishments, May 2000.

Zhu, Z. and Soong, T.T., “Toppling Fragility of Unrestrained Equipment,” EarthquakeSpectra, Vol. 14, No. 4, 1998.

to determine how much serviceloss will be from the event. Conse-quently, the value of a retrofit willbe evaluated against the chance ofthe risk and the associated poten-tial loss.

ConclusionBased on information supplied to

us from the hospitals in New York,we are in the process of establish-ing these models. It is our inten-tion to examine the dynamicbehavior of the emergency medi-cine unit of the hospitals under pre-scribed hazard conditions anddamage scenarios, and eventually todevelop a cost vs. risk evaluationprocedure leading to decision-mak-ing support for seismic retrofit.

Some pertinent references con-cerning retrofit strategies for hos-pitals are provided.

149

Research Objectives

Passive site remediation is a new concept proposed for non-dis-ruptive mitigation of liquefaction risk at developed sites susceptible toliquefaction. It is based on the concept of slow injection of stabilizingmaterials at the edge of a site and delivery of the stabilizer to the targetlocation using the natural groundwater flow. Stabilizer candidates need tohave long controllable gel times and low viscosities so they can flow intoa liquefiable formation slowly over a fairly long period of time. Colloidalsilica is a potential stabilizer for passive site remediation because at lowconcentrations it has a low viscosity and a wide range of controllable geltimes of up to about 200 days. Loose sands treated with colloidal silicagrout had significantly higher deformation resistance to cyclic loading thanuntreated sands. Groundwater and stabilizer transport modeling were doneto determine the range of conditions where passive site remediation mightbe feasible. For a 200-foot by 200-foot treatment area with a single line ofinjection wells, it was found that passive site remediation could be feasiblein formations with hydraulic conductivity values of 0.05 cm/s or more andhydraulic gradients of 0.005 and above.

Passive Site Remediation forMitigation of Liquefaction Risk

by Patricia M. Gallagher and James K. Mitchell


Patricia M. Gallagher,Assistant Professor,Department of Civil andArchitectural Engineering,Drexel University, FormerGraduate Assistant,Department of CivilEngineering, VirginiaPolytechnic Institute andState Universtity

James K. Mitchell,University DistinguishedProfessor Emeritus,Department of CivilEngineering, VirginiaPolytechnic Institute andState University

At many sites susceptible to liquefaction, the simplest way to mitigatethe liquefaction risk is to densify the soil. For large, open and undevel-

oped sites, the easiest and cheapest methods for densification are by “tra-ditional” procedures such as deep dynamic compaction, explosivecompaction, or vibrocompaction. However, at constrained or developedsites, ground improvement by densification may not be possible due tothe presence of structures sensitive to deformation or vibration. Addition-ally, access to the site could be limited and normal site use activities couldinterfere with mitigation activities. At these sites, the most common meth-ods for remediation are grouting or underpinning. Passive site remediationis a new concept proposed for non-disruptive improvement of developedsites susceptible to liquefaction. Passive site remediation is based on theconcept of the slow injection of stabilizing materials at the up gradientedge of a site and delivery of the stabilizer to the target location using the

150

• Mr. Chris Gause ofMaster Builderscontributed materials forlaboratory testing ofmicrofine cement groutsand expertise on groutingtechnology.

• Dupont provided some ofthe colloidal silica used inthe experimental program.

• Dr. Ernst Ahrens, SandiaNational Laboratory andMr. Henry Pringer, BlueCircle, contributed cementfor testing.

natural or augmented groundwaterflow. The concept is illustrated inFigure 1.

The set time of the stabilizerwould be controlled so there wouldbe adequate time for it to reach thedesired location beneath the siteprior to gelling or setting. If thenatural groundwater flow were in-adequate to deliver the stabilizer tothe right place at the right time, itcould be augmented by use of low-head injection wells ordowngradient extraction wells.Once the stabilizer reached thedesired location beneath the site, itwould gel or set to stabilize the for-mation.

Passive site remediation tech-niques could have broad applica-tion for developed sites wheremore traditional methods of groundimprovement are difficult or impos-sible to implement. It would be lessdisruptive to existing infrastructureand facilities than existing groundimprovement methods. Additionally,access to the entire site would beunnecessary using this technology,and normal site use activities wouldprobably not need to be disrupted.Finally, excessive deformation anddisturbance of the ground aroundand beneath existing structurescould be avoided.

The objective of this study wasto establish the feasibility of passivesite remediation. The work includedidentification of stabilizing materi-als, a study of how to adapt or de-sign groundwater flow patterns todeliver the stabilizers to the rightplace at the right time, and an evalu-ation of potential time require-ments and costs.

Performance Criteriaand Identification ofPotential Stabilizers

For a stabilizer to work in thisapplication, it should have a lowviscosity and a long induction pe-riod between mixing and the on-set of gelation. Once gelation starts,it should proceed rapidly. The sta-bilizer should also be permanent,nontoxic and cost-effective. Mate-rials evaluated as potential stabiliz-ers included colloidal silica,microfine cement grouts, chemicalgrouts, zero-valent iron, andultramicrobacteria. Colloidal silicawas selected as a potentially suit-able stabilizing material because ithas a wide range of gel times and alow viscosity. Colloidal silica is anaqueous suspension of tiny silicaparticles that can be made to gel

Passive site remediation is a new technology for mitiga-tion of liquefaction and ground failure risk at developedand inaccessible sites where more traditional methods forground improvement are not suitable. Slow permeation ofchemical soil stabilizer beneath and around foundations onand in potentially liquefiable soils should be especially at-tractive to owners, engineers, and planners who are chargedwith assuring seismic safety of existing infrastructure.

151Passive Site Remediation for Mitigation of Liquefaction Risk

by adjusting the pH or the salt con-centration of the solution. Gel timesof more than 200 days have beenmeasured in laboratory tests. Addi-tionally, the initial viscosities of di-lute solutions of colloidal silica areabout 2 centipoise (water=1 cP)and the viscosities remain very lowfor most of the induction period.

Microfine cement grout waseliminated because its viscosity istoo high to meet the necessary re-quirements for passive siteremediation. Additionally, since ce-ment grouts are particulate suspen-sions, the particles tend to settle inthe suspension and further increasethe viscosity. Numerous chemicalgrouts were considered. All wereeliminated as potentially suitablestabilizers, but for different reasons.Sodium silicate was eliminated be-cause gel time is not well controlledat long gel times. Additionally, thechemical durability of sodium sili-cate formulations with long gel

times is questionable. Acrylamide isa neurotoxin in powdered form, soit was eliminated due to environ-mental, safety, and handling con-cerns. Additionally, it is veryexpensive. Acrylate was eliminateddue to durability concerns. Epoxyand polysiloxane were rejectedbecause they are very expensive.Zero-valent iron is extremely sensi-tive to oxidation and reduction, soit would be difficult to treat a largearea and the minerals precipitatedwould probably not be chemicallydurable. Ultramicrobacteria mightbe able to clog the pores of a for-mation with a biofilm, but biofilmscan be dissolved by strong oxidantssuch as bleach, so there are dura-bility concerns.

FeasibilityThe feasibility of passive site

remediation depends on the an-swers to the following questions:

� Figure 1. Passive treatment for mitigation of liquefaction risk.

152

1. Will the colloidal silica grout ad-equately stabilize the soil?

2. Can the stabilizer be deliveredto the liquefiable formation andachieve adequate coveragewithin the induction period ofthe grout?

3. How much will it cost?Strength testing of stabilized

sands was done to address the firstissue. Groundwater and stabilizertransport modeling were done todetermine if the stabilizer could bedelivered to the formation withinthe induction period of the grout.Finally, a preliminary cost analysiswas done to address the final issue.

Strength Testing ofStabilized Sands

Cyclic triaxial tests were done onMonterey No. 0/30 sand samplestreated with colloidal silica grout toinvestigate the influence of colloi-dal silica grout on the deformationproperties of loose sand (relativedensity, Dr = 22%). The grain sizedistribution of Monterey No. 0/30sand is shown in Figure 2. Distinctlydifferent deformation propertieswere observed between groutedand ungrouted samples. Untreatedsamples developed very little axialstrain after a few cycles of loadingand prior to the onset of liquefac-tion. However, once liquefactionwas triggered, large strains oc-curred rapidly and the samples col-lapsed within a few additionalcycles. In contrast, grouted sandsamples experienced very littlestrain during cyclic loading. Whatstrain accumulated did so uniformlythroughout loading and thesamples remained intact after cyclicloading.

An example is shown in Figure 3for two samples at a relative den-sity of 22 percent that were testedat a cyclic stress ratio of 0.27. Thecyclic stress ratio is defined as theratio of the maximum cyclic shearstress to the initial effective confin-ing stress. The untreated samplestrained 1 percent in 11 cycles andcollapsed in 13 cycles. The sampletreated with 10 weight percentcolloidal silica was tested for 400cycles. It strained less than abouthalf a percent in 11 cycles, about 8percent in 400 cycles, and nevercollapsed. Only the first 40 load-ing cycles are shown in Figure 3.These results are typical forsamples treated with 10 percentcolloidal silica by weight. For com-parison, a magnitude 7.5 earth-quake would be expected togenerate about 15 uniform stresscycles.

Samples stabilized with concen-trations of 15 and 20 weight per-cent colloidal silica experiencedvery little (less than two percent)strain during cyclic loading. Sandsstabilized with 10 weight percentcolloidal silica resisted cyclic load-ing well, but experienced slightlymore (up to eight percent) strain.Overall, treatment with colloidalsilica grout significantly increasedthe deformation resistance of loosesand to cyclic loading.

Groundwater andStabilizer TransportModeling

Stabilizer delivery is the main fea-sibility issue with respect to passivesite remediation. Preliminarygroundwater and solute transportmodeling were done using thecodes MODFLOW, MODPATH, and


• Task 2.3, GeotechnicalRehabilitation Site andFoundation Remediation

• Task 2.6, MEDAT-2Workshop on Mitigation ofEarthquake Disaster byAdvanced Technologies


MT3DMS for a generic liquefiableformation. A “numerical experi-ment” was done to determine theranges of hydraulic conductivityand hydraulic gradient where pas-sive site remediation might be fea-sible. For a 200-foot by 200-foottreatment area, with single lines ofinjection and extraction wells,travel times through the treatmentarea will be about 100 days or lessif a formation has a hydraulic con-ductivity greater than about 0.05cm/s and a hydraulic gradienthigher than about 0.005. Based onthe possible gel times, this timeframe is considered feasible. Extrac-tion wells will increase the speedof delivery and help control thedown gradient extent of stabilizermovement.

The results of solute transportmodeling indicate that stabilizerdelivery will vary throughout thetreatment area. A typical stabilizercontour plot for a hypothetical for-mation with a uniform hydraulicconductivity of 0.05 cm/s and ahydraulic gradient of 0.005 isshown in Figure 4. A stabilizer con-centration of 100 g/l would be de-livered through an infiltrationtrench for 100 days. The best cov-erage would be achieved close tothe source of the stabilizer. Concen-trations would decrease laterallyaway from the source and downgradient of the source. If the mini-mum amount of stabilizer requiredfor adequate stabilization could bedelivered to the majority of thetreatment area, it is likely that theformation would be stable enoughto withstand seismic loading. How-ever, there could be some differen-tial or variable response across thesite. It may be necessary to delivera higher concentration at the upgradient edge of the treatment area

0

20

40

60

80

100

0.00010.0010.010.1110

Particle Size (mm)

Per

cen

tF

iner

by

Wei

gh

t SILT CLAYSANDG

Monterey #0/30 sand

Most liquefiable soils

Untreated Monterey Sand (Dr = 22%)

-4.0-2.00.02.04.06.08.0

0 10 20 30 40

Cycles

Axia

l S

train

(%

)

Monterey Sand with 10% Colloidal Silica

(Dr = 22%)

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

0 10 20 30 40

Cycles

Axia

l S

train

(%

)

Untreated Monterey Sand (Dr = 22%)

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

0 10 20 30 40

Cycles

Axia

l S

train

(%

)

� Figure 3. Axial deformation during cyclic loading (CSR=0.27) for treated anduntreated sand.

� Figure 2. Gradation curve for Monterey No. 0/30 sand.

154

in order to get an adequate concen-tration at the down gradient edge.

Heterogeneity in the formationwill actually control how well thestabilizer can be delivered. If the for-mation is highly variable, then thestabilizer concentration will vary

from point to point within the for-mation. An example stabilizer con-tour profile through a treatmentarea with a variable hydraulic con-ductivity is shown in Figure 5. Inthis case, the hydraulic conductiv-ity was varied slightly in each layeras shown for a total variationthroughout the layer of about oneorder of magnitude. The remainderof the simulation is the same as theprevious case. The layers withhigher hydraulic conductivity havea higher concentration at the downgradient edge. These layers wouldprobably be more stable than lay-ers with lower hydraulic conduc-tivity that receive a lowerconcentration of grout during thetreatment period. However, even ifthe regions of lower hydraulic con-ductivity liquefy, the presence ofvery stable seams will likely lessenthe severity of the overall deforma-tion. Accurate characterization ofthe hydraulic conductivity through-out the treatment area will be es-sential for successful treatment bypassive site remediation.

0 100

k (cm/s)

0 100Horizontal and vertical

� Figure 4. Stabilizer contours for 200 ft. by 200 ft. treatment area (outlinedin black) after 100 days of treatment. Stabilizer delivered through infiltra-tion trench at concentration of 100 g/l. Two extraction wells at the downgradient edge withdraw a total of 7500 cfd. Contour intervals are 10 g/l.Concentration at extraction wells is 60 g/l. Travel paths for individual wa-ter particles are superimposed over the treatment area in 10-day incre-ments. Particle travel times are about 75 to 80 days.

� Figure 5. Stabilizer profile through centerline of 200 ft. by 200 ft. treatment area after 100 days of treatment. Stabilizer de-livered through infiltration trench at concentration of 100 g/l. Extraction wells at the down gradient edge withdraw a total of7500 cfd. Contour intervals are 10 g/l. Concentration at extraction wells is about 70 g/l in lower 30 ft. Travel paths for indi-vidual water particles are superimposed over the treatment area in 10-day increments. Particle travel times range fromabout 40 to 420 days.


CostThe cost of passive site

remediation is expected to be com-parable to other methods of chemi-cal grouting. It is likely that a 10weight percent concentration ofcolloidal silica will be adequate tostabilize a liquefiable formation. Itis possible that lower concentra-tions could be used. Based on a 10percent concentration, it is ex-pected that materials costs wouldbe in the range of $120 to $180 percubic meter of treated soil. Thesecosts are competitive with othermethods of chemical grouting.

ConclusionBased on the feasibility analysis,

passive site remediation appears tobe a promising new concept formitigation of liquefaction risk. Atthis time, a minimum concentrationof 10 percent colloidal silica ap-pears to be suitable for stabilizingliquefiable sands. Additional testingis being done with concentrationsof 5 weight percent to determineif the level of strain during cyclicloading would be acceptable.

Delivery of the stabilizer is thecentral feasibility issue with respectto passive site remediation. For a200-foot by 200-foot treatment areawith a single line of injection wells,it was found that passive siteremediation could be feasible informations with hydraulic conduc-tivity values of 0.05 cm/s or moreand hydraulic gradients of 0.005and above. However, the actual con-centration profile across the sitewill depend on the variation in hy-draulic conductivity throughoutthe formation.

The anticipated final outcome ofthis work is a new technology formitigation of liquefaction and groundfailure risk. Passive site remediationtechnology will be less disruptive toexisting infrastructure and facilitiesthan existing methods. It is expectedthat passive site remediation will becost-competitive with other methodsof chemical grouting. Model testingof both the injection method and theperformance of grouted ground isplanned as the next step in the evalu-ation of this new technology. It willbe done using a geotechnical centri-fuge equipped with a shake table.

157

Research Objectives

The objectives of the research are to: 1) develop regressions betweenbuilding damage and various seismic parameters to improve loss estima-tion, 2) identify the most reliable seismic parameter for estimating build-ing losses, and 3) develop GIS-based pattern recognition algorithms forthe identification of locations with the most intense post-earthquake build-ing damage. This latter objective is coupled to the goal of improving emer-gency response as part of the MCEER vision of creating earthquake resilientcommunities. GIS-based technologies for visualizing damage patterns pro-vide a framework for rapidly screening remote sensing data and dispatch-ing emergency services to the locations of greatest need.

Advanced GIS for Loss Estimation andRapid Post-Earthquake Assessment ofBuilding Damage

by Thomas D. O’Rourke, Sang-Soo Jeon, Ronald T. Eguchi and Charles K. Huyck


Lifeline Damage

MCEER researchers working on the seismic retrofit and rehabilitationof lifelines (Program 1) and on seismic response and recovery (Pro-

gram 3) have collaborated on a project organized to apply advanced GIStechniques for the rapid identification of locations with most intense build-ing damage. This collaboration has resulted in the development of regres-sions between building damage and various seismic parameters, as well asthe application of GIS-based recognition algorithms with the potential forscreening remote sensing data to identify areas of highest post-earthquakedamage intensity.

Using GIS technology, Cornell researchers developed the largest U.S.database ever assembled for spatially distributed transient and permanentground deformation in conjunction with earthquake damage to water sup-ply lifelines (O’Rourke et al., 1999). This research has helped to delineatelocal geotechnical and seismological hazards in the Los Angeles regionthat are shown by zones of concentrated pipeline damage after theNorthridge earthquake. The research has resulted in regressions betweenrepair rates for different types of trunk and distribution pipelines andvarious seismic parameters. The regressions are statistically reliable andhave improved predictive capabilities compared with the default relation-ships currently used in loss estimation programs. They will be referencedin the next version of HAZUS software that implements the National Loss

Thomas D. O’Rourke,Thomas R. Briggs Professorof Engineering and Sang-Soo Jeon, GraduateResearch Assistant,Department of Civil andEnvironmentalEngineering, CornellUniversity

Ronald T. Eguchi, Presidentand Charles K. Huyck,Vice President, ImageCat,Inc.

158

Estimation Methodology spon-sored by FEMA. The regressions andstatistical databases are being incor-porated in pre-standards for esti-mating water supply lossesdeveloped by the American LifelineAlliance through ASCE under con-tract with FEMA.

The research has led to the discov-ery of a relationship for visualizingdamage patterns by linking the twodimensional representation of localdamage with the grid size used in GISto analyze the spatial distribution ofdata. This concept is illustrated in Fig-ure 1, which shows the relationshipdeveloped for visualizing post-earth-quake pipeline damage (O’Rourke etal., 1999).

With GIS, the spatial distribution ofdamage can be analyzed by dividingany map into squares, or cells, each ofwhich is n by n in plan. A repair rate isdefined as the number of repairs di-vided by the total distance of pipelinesin each cell. Contours of equal repairrate, or damage rate, can then bedrawn using the grid of equi-dimen-sional, n-sized cells. If the contour in-terval is chosen as the average repairrate for the entire system or portionof the system covered by the map,then the area in the contours repre-sents the zones of highest (greaterthan average) earthquake intensity asreflected in pipeline damage.

• California Office ofEmergency Services(OES)

• Los Angeles Departmentof Water and Power

• Federal EmergencyManagement Agency

The users of this research include two main groups: 1)water distribution companies, such as the Los Angeles De-partment of Water and Power (LADWP) and the East BayMunicipal Utility District (EBMUD) and 2) those who ben-efit from improved loss estimation, such as insurance com-panies, the Federal Emergency Management Agency, stateand municipal planners, and emergency service providers.

0.1 0.3 0.5 0.70.0 0.2 0.4 0.6 0.8Dimensionless Grid Size, DGS

0.1

0.3

0.5

0.0

0.2

0.4

0.6

Thre

shold

Are

aC

ove

rage,T

AC

Fit Curve

All LA

All SFV

Section of SFV

Sherman Oaks

North of LA Basin

San Francisco

TAC = DGS / ( 1.68 * DGS + 0.11)

� Figure 1. Hyperbolic Fit for Threshold Area Coverage and Dimensionless Grid Size forPipeline Damage Patterns (Toprak, et al., 1999)

TAC=(A1+A2)/A

C

DGS =SQRT(N

2

/A

T

)

159Advanced GIS for Loss Estimation and Rapid Post-Earthquake Assessment of Building Damage

A hyperbolic relationship wasshown to exist between the thresh-old area coverage (TAC) [in thiscase, the fraction of the total maparea with damage exceeding theoverall average repair rate] and thedimensionless grid size, defined asthe square root of n2, the area of anindividual cell, divided by the totalmap area, A

T. This relationship is il-

lustrated in Figure 1, for which aschematic of the parameters is pro-vided by the inset diagram. The re-lationship was found to be validover a wide range of different mapscales spanning 1200 km2 for theentire Los Angeles water distribu-tion system affected by theNorthridge earthquake to 1 km2 ofthe San Francisco water distribu-tion system in the Marina affectedby the Loma Prieta earthquake(Toprak et al., 1999)

As explained by O’Rourke et al.(1999), the hyperbolic relationshipcan be used for damage pattern rec-ognition, and for computer "zoom-ing" from the largest to smallestscales to identify zones of concen-trated disruption. If such a relation-ship can be shown to be valid forbuilding damage, then remote sens-ing data acquired to characterizebuilding damage can be evaluatedrapidly for the locations of highestloss intensity and thereupon tar-geted for emergency response.

Building DamageInspection records available

through the California Office ofEmergency Services (OES) were ob-tained for 62,020 buildings that wereinvestigated after the Northridgeearthquake. Of these, 48,702 build-ings were associated with one andtwo-story timber frame structures,

principally single and multiple fam-ily residences. The inspection recordsinclude location and estimate of dam-age as a percentage of replacementcost. Figure 2 shows the types of tim-ber frame structures included in thedatabase with examples of damage,as presented by the NAHB ResearchCenter (1994). The database did notinclude specific information aboutthe type of damage to each structure.

The number of one to two storytimber frame structures affected bythe Northridge earthquake was esti-mated from tax assessor records sup-plied through OES. The numbers ofstructures determined in this waywas 278,662. As a check, an alternatenumber was estimated from 1990census block data (Wessex, 1996) byevaluating the number of buildingsassociated with detached and at-tached single housing units in com-bination with the buildings neededto accommodate two and three tofour housing units. The resulting num-ber was 267,868, which is in excel-lent agreement with the 278,662units estimated from tax assessorrecords.

Program 1: Seismic Evaluationand Retrofit of LifelineNetworks

Program 3: EarthquakeResponse and Recovery

a) Foundation Cripple Wall Collapse

b) Racking of Walls c) Roof Framing Damge

� Figure 2. Typical Damage to 1-2 Story Timber Frame Buildings after theNorthridge Earthquake (after NAHB, 1994)

160

Two damage parameters, referredto as damage ratio and damage fac-tor, were calculated with the data-base. Damage ratio is the fractionor % of existing structures withdamage equal to or exceeding aparticular damage factor. Damagefactor is damage expressed as a %of building replacement cost.

Loss EstimationRegressions

A GIS grid composed of 2,106cells was created, and the num-ber of relevant structureswithin each 0.42 km2 cell wascalculated and geocoded at thecell center. Figure 3 shows theGIS grid developed with tax as-sessor records superimposedon the spatial variation of peakground velocity determinedfrom more than 240 strong mo-tion records geocoded as partof this study. By superimposinga grid in which each cell is char-acterized by various damage ra-tios linked with damage factors,linear regressions can be per-

formed to quantify damage vs. seis-mic excitation for loss estimationpurposes.

Figures 4a and b show the linearregression of damage ratio (DR) vs.peak ground velocity (PGV) andSpectrum Intensity (SI) for variousdamage factors (DF). Most regres-sions for PGV show excellent fits

2 4 6 8 2

10 100SI (cm/sec)

2

4

6

8

2

4

6

8

2

4

6

0.01

0.10

1.00

Dam

age

Ratio

(%)

DF 10 % (r2 = 0.94)

DF 30 % (r2 = 0.93)

DF 50 % (r2 = 0.93)

DF 70 % (r2 = 0.86)

30 %

50 %

70 %

10 %

2 4 6 8 2

10 100PGV (cm/sec)

2

4

6

8

2

4

6

8

2

4

6

0.01

0.10

1.00

Dam

age

Ratio

(%)

DF 10 % (r2 = 0.95)

DF 30 % (r2 = 0.92)

DF 50 % (r2 = 0.87)

DF 70 % (r2 = 0.68)

30 %

50 %

70 %

10 %

Dam

age

Rat

io (

%)

Dam

age

Rat

io (

%)

� Figure 3. GIS Grid with Tax Assessor Data Superimposed on Peak GroundVelocities

� Figure 4. Damage Ratio Regression with Peak Ground Velocity (PGV) and Spectrum Intensity(SI) for 1-2 Story Timber Frame Buildings


as indicated by high r2, althoughthe "goodness" of fit is relativelylow for DF ≥ 70%. In contrast, allregressions for SI have high r2 andexcellent characteristics with re-spect to residuals. The relation-ships between damage andvarious seismic parameters wereprobed in this way to determinewhich parameters were statisti-cally most significant as damagepredictors. The seismic param-eters investigated include the mea-sured peak ground acceleration,velocity, and displacement; spec-tral acceleration and velocity forperiods of 0.3 and 1.0 s; Arias In-tensity; and SI.

SI is defined as the area underthe pseudo-velocity, SV, responsespectra curve for a damping ratio,ξ of 20 % between periods, T, of 0.1and 2.5 s:

SI SV T DT= ∫ ( , ).

.ξ

0 1

2 5

(1)

SI was first proposed by Housner(1959) as a measure of the maxi-mum stresses that would be in-duced in elastic structures byground motion. Katayama et al.(1988) found that house damagecorrelated more strongly with SIthan with peak acceleration, andrecommended that SI calculationsbe performed for ξ = 20%.

An examination of Figure 4 re-veals that DR, DF, and the seismicparameter, SP, are interrelated in aconsistent way. Using multiple lin-ear regression techniques, this in-terrelationship can be expressed as

Log DR Log K Log SP Log DF = + - α β(2)

in which K, α, and β are con-stants. (2) can be rewritten as

DR = K(SP / DFβ α α/ ) (3)

in which (SP/DFβ/α) is the scaledseismic parameter.

Using (3), the data in Figure 4aand b are replotted in Figures 5aand b as linear regressions that ac-count for DR, DF, and SP. As evincedby high r2 and excellent character-istics with respect to residuals, therelationships in Figures 5a and b arestatistically significant.

Seismic parameters normalizedwith respect to DF combine theeffects of strong motion and dam-age level in a convenient way thatfacilitates loss estimation. For ex-ample, consider an area for whichthe predicted SI is 30 cm/sec. The% of timber structures in that areathat would have damage equal toor exceeding 10% of the buildingreplacement cost is calculated asDR = 0.92(30/101.37)1.051, or 1.19%.For damage equal to or exceeding50% replacement cost, DR =0.92(30/501.37)1.051, or 0.11%.

8 2 4 6 8 2 4 6 8 2 4 6 8

0.1 1.0 10.0 100.0

PGV/DF1.16, ([cm/sec]/%1.16 )

6

8

2

4

6

8

2

4

6

8

2

4

6

8

0.01

0.10

1.00

10.00

Dam

age

Ratio

(%)

Fit Equation:

DR (%) = 0.40 (PGV/DF1.16)1.268

r2 = 0.93

2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8

0.1 1.0 10.0 100.0

SI/DF1.37, ([cm/sec]/%1.37 )

2

4

6

8

2

4

6

8

2

4

6

8

0.01

0.10

1.00

10.00

Dam

age

Ratio

(%)

Fit Equation:

DR (%) = 0.92 (SI/DF1.37)1.051

r2 = 0.96

Dam

age

Rat

io (

%)

Dam

age

Rat

io (

%)

� Figure 5. Damage Ratio Regression for Scaled PGV and SI for 1-2 Story TimberFrame Buildings

162

Damage PatternRecognition

Using the same concepts thatwere applied to the spatial distri-bution of pipeline damage, investi-gations were undertaken to definea relationship between TAC (de-fined for buildings as the fractionof the total map area with DR ex-ceeding the overall average DR)and the dimensionless grid size. Thegrid containing tax assessor datawas sufficiently refined to evaluatethe effects of progressively largergrid sizes by developing grids withcell sizes of n by n, 2n by 2n, 4n by4n, and 8n by 8n. The data gener-ated for these grids and various DFsare plotted in Figure 6. All relation-ships are well described by hyper-bolic functions, similar to the onefor pipeline damage in Figure 1. Be-cause building damage is character-ized by DF, it has an additionaldimension when compared to thesingle hyperbolic curve for pipe-lines. In fact,Figure 6 is actually a

hyperbolic surface for which theinitial slopes and asymptotes varyas a function of DF.

Experience has shown that theideal TAC for visualization is about0.33. Figure 6 indicates that the di-mensionless grid size to achievethis TAC varies significantly, de-pending on DF.

Figure 7 illustrates the applica-tion of Figure 6 for damage inten-sity recognition and computer"zooming." The top map shows theentire San Fernando Valley and ad-jacent areas. The figure shows acascade of different computerscreens that "zoom" on increasinglysmaller areas from top to bottom.Each screen shows as blue dots theactual locations of one to two storytimber frame structures with DF ≥70% after the Northridge earth-quake. The red contour lines indi-cate areas of highest damageintensity.

The contour lines in the top mapwere generated by analyzing thedata with a GIS grid size chosen forTAC = 0.33 from Figure 6. The areaoutlined in red in the top map wasidentified for more detailed assess-ment. The grid size for this assess-ment was again determined for TAC= 0.33 from Figure 6. The map inthe center of the figure shows anexpanded view of the area outlinedabove after computer analysis. Thezones of highest damage intensityfor this new map are surroundedby red contour lines.

A similar procedure was followedfor the center map in which an areafor more detailed assessment wasidentified within the blue outline.The map at the bottom of the fig-ure shows an expanded view of thisnew area in which the areas ofgreatest damage intensity are againsurrounded by red contour lines.

“By combiningan advancedGIS withadvancedremote sensing,MCEER isdeveloping theenablingtechnologiesfor a newgeneration ofemergencyresponse andrapid decisionsupportsystems.”

TAC=(A1+A2)/A

C

DGS =SQRT(N

2

/A

T

)

� Figure 6 Hyperbolic Fit for Threshold Area Coverage and Dimensionless Grid Sizefor 1-2 Story Timber Frame Buildings


Increasing density of contour linesreflects increasing intensity of dam-age. The scale embodied in the bot-tom map allows the viewer todiscriminate damage patterns onvirtually a block-by-block basis.

Because the relationships in Fig-ure 6 are independent of scale, thesame algorithm for a specific DFcan be used for increasingly smallerportions of the original map to"zoom" on areas of most intensedamage. The visualization algorithm

Area ofMore Local,DetailedEvaluationShown inNext FigureBelow

Areaof MoreLocal,DetailedEvaluationShown inNext FigureBelow

Zones ofContoursIndicateHighestIntensity ofDamage

� Figure 7. GIS Evaluation of Building Damage for DF ≥ 70 %

164

allows personnel who are not spe-cifically knowledgeable aboutstructures or trained in pattern rec-ognition to identify the locations ofmost severe damage for allocationof aid and emergency services. Theentire process is easy to computer-ize, and personnel would be ableto outline any part of a map with a"mouse" and click on the area sodefined. In each instance of defin-ing a smaller area for evaluation, theaverage DR is recalculated for thenew, smaller-sized map. In this way,the average is calibrated to eachnew map area.

When the damage pattern recog-nition algorithms are combined withregional data rapidly acquired by ad-vanced remote sensing technologies,the potential exists for acceleratedmanagement of data and quick de-ployment of life and property savingservices. By combining an advancedGIS with advanced remote sensing,MCEER is developing the enablingtechnologies for a new generation ofemergency response and rapid deci-sion support systems.

SummaryGIS research on visualizing dam-

age patterns in pipeline networkshas been extended to buildings.Algorithms developed for pipelineshave been modified and validatedto chose optimal GIS mesh dimen-sions and contour intervals for vi-sualizing post-earthquake damagepatterns in buildings. This work hasbeen performed by a collaborationwith researchers working in Pro-grams 1 and 3 on advanced tech-nologies for loss estimation andreal-time decision support systems.Robust and statistically significantregressions have been developedbetween the fraction of existingtimber frame buildings at any dam-age state and the magnitudes ofvarious seismic parameters. Suchwork improves loss estimation sig-nificantly and also creates advancedtechnology to visualize post-earth-quake damage patterns in buildingsfor rapid decision support and de-ployment of emergency services.

References

Housner, G. W., (1959), “Behavior of Structures During Earthquakes,” Journal ofEngineering Mechanics Division, ASCE, Vol. 85, No. EM4, pp. 109-129.

Katayama, T., Sato, N., and Saito, K., (1988), “SI-Sensor for the Identification ofDestructive Earthquake Ground Motion,” Proceedings, 9th World Conference onEarthquake Engineering, Vol. VII, Tokyo-Kyoto, Japan, August, pp. 667-672.

NAHB Research Center, (1994), Assessment of Damage to Residential BuildingsCaused by the Northridge Earthquake, Reported Prepared for U.S. Department ofHousing and Urban Development by NAHB Research Center, Upper Marlboro, MD.

O’Rourke, Thomas D., Toprak, Selcuk, and Jeon, Sang-Soo, (1999), “GIS Characterizationof the Los Angeles Water Supply, Earthquake Effects, and Pipeline Damage,” ResearchProgress and Accomplishments 1997-1999, Multidisciplinary Center for EarthquakeEngineering Research, University at Buffalo, July, pp. 45-54.

Toprak, S., O’Rourke, T.D., and Tutuncu, I., (1999), “GIS Characterization of SpatiallyDistributed Lifeline Damage,” Proceedings, 5th US Conference on LifelineEarthquake Engineering, Seattle, WA, ASCE, Reston, VA, August, pp. 110-119.

Wessex (1996), U.S. Demographics, Wessex, Inc., Winnetka, IL.

165

Research Objectives

MCEER was part of a team charged with the seismic upgrade ofIstanbul's Ataturk International Airport Terminal Building. The projectinvolved analysis of post-earthquake damage to the Terminal Buildingfollowing the 1999 Marmara, Turkey earthquake, and development of aretrofit scheme. Nonlinear static and dynamic analyses were conductedusing procedures set forth in FEMA 273 (including numerous contri-butions from MCEER researchers). Pushover analyses were conductedusing IDARC (developed with MCEER support). Dynamic analyses ofthe roof-isolated structure considering inelastic frame behavior wereconducted using computer software based on modifications of pro-gram 3D-BASIS (developed with MCEER support). Several retrofitschemes were investigated, including conventional methods and ma-terials, and new technologies. The scheme selected included isolationof the roof trusses, addition of shock transmission units, and selectedretrofit and strengthening of reinforced concrete construction.

The Turkish build-operate-transfer consortium, TEPE-AKFEN-VIE (TAV)and its advisor, New York-based Turner International, approved the evalu-ation and retrofit scheme. Members of the team were LZA Technology,a division of Thornton-Tomasetti Group (leader), Michael C.Constantinou and Andrew S. Whittaker (formerly of PEER), both of theUniversity at Buffalo, and Tuncel Engineering and FondsiyuonMuhendislik Insaatvetic Ltd., both of Istanbul. The project is an excel-lent example of bringing research results to implementation.

Seismic Evaluation and Retrofit ofthe Ataturk International AirportTerminal Building

by Michael C. Constantinou, Andrew S. Whittaker and Emmanuel Velivasakis

At the time of the August 19, 1999, Izmit earthquake, the new AtaturkInternational Airport Terminal building was nearing completion. The

airport, which is located 25 km from the center of Istanbul, was shakenand damaged by the earthquake. (The airport is located approximately 70km from the fault rupture plane.) The new Terminal building is a three-story reinforced concrete building with a space-frame roof. The plan foot-print is approximately 240 m by 168 m. A view of the terminal building ispresented in Figure 1.

LZA Technology, Thornton-Tomasetti Group


Michael C. Constantinou,Professor, Andrew S.Whittaker, AssociateProfessor, Oscar Ramirez,Graduate ResearchAssistant, Ani NataliSigaher, GraduateResearch Assistant, EricWolff, Graduate ResearchAssistant, and Tony Yang,Undergraduate Student,Department of Civil,Structural andEnvironmentalEngineering, University atBuffalo

Emmanuel Velivasakis,Senior Vice President, LZATechnology, New York, NewYork

166

The lowest story of the buildingprovided mechanical and baggagehandling services. The second andthird stories of the building housedthe Arrivals and Departures Halls,respectively. A plan view of thebuilding is shown in Figure 2a. Asshown in Figure 2a, the 240 m by168 m building is composed of 20pods (or independent frames); thetypical pod dimension is 48 m by48 m. A typical pod is shownshaded in this figure. The pods wereseparated by 50 mm wide expan-sion joints (EJs). Figure 2b showsthe framing in the third and secondstories of a typical pod; the fram-ing in the first story was similar tothat in the second story. Figure 2cis a cross-section through the build-

ing showing typicalframing. The building isframed in reinforcedconcrete. Above thethird floor level, cantile-ver columns on 24 mcenters supported athree-dimensional steelspace-frame roof struc-ture. The space-frameroof was equipped withsleeved movementjoints to permit thermalexpansion and contrac-

tion of the roof. These movementjoints did not align with the expan-sion joints in the reinforced con-crete framing. At the third floor andbelow, gravity loads were sup-ported by reinforced concretewaffle slabs and columns at 12 mon center. Lateral loads were re-sisted by waffle-slab moment-frameconstruction in each direction.Solid beams of a depth equal to thatof the waffle slab spanned betweenthe columns. Around the perimeterof each pod, the column sizes weresubstantially reduced from those inthe interior of the pod. At the cor-ners of each pod, the four columnswere approximately square butwith dimensions one-half of thoseof the interior columns (and called

• Eric Stovner andJohn Abbruzzo, LZATechnology

• Anoop Mokha, VicePresident, EarthquakeProtection Systems

• Douglas Taylor, TaylorDevices

• Faruk Tuncel, Engineerof Record, TuncelEngineering andConstruction Co., Ltd.,Turkey

• Thomas J. McCool,Project Engineer andConstruction Consultant,Turner SteinerInternational, New York

• Sani Sener, CEO, andGokhan Ozber, DeputyGeneral Manager, TepeAkfen Vie (TAV),Investment Constructionand Operation Inc., Turkey(Airport Owner)

� Figure 1. New Ataturk International Airport TerminalBuilding

It is anticipated that this project will be of interest tobuilding owners in seismic regions throughout the world.Three different retrofit approaches were taken, tostrengthen various sections of the building. The perfor-mance of the seismic isolation and lock-up devices in fu-ture seismic events will provide additional data on theirreliability and effectiveness for earthquake hazard miti-gation. The entire evaluation and subsequent installationof the retrofit scheme was accomplished in less than fourmonths.

167Seismic Evaluation and Retrofit of the Ataturk International Airport Terminal Building

quarter columns in this paper).Along each edge of each pod, andbetween the corner quarter col-umns, the columns were approxi-mately rectangular and half the areaof the typical interior columns (andcalled half columns in this paper.)

During the August 19, 1999 earth-quake, the Terminal building wassubjected to modest earthquakeshaking. The maximum recordedhorizontal ground acceleration re-corded at the Airport was approxi-mately 0.l g; the maximum vertical

acceleration was less than 0.05g. In-vestigations by LZA engineers im-mediately following the earthquakeidentified damage to parts of thebuilding, including spalling of thecover concrete and buckling of thelongitudinal rebar at the base of thethird story columns (showing alsolap splices and little transverserebar in the hinge zones), loss ofconcrete at the underside of theroof truss-cantilever column con-nections and slippage of the roof-truss baseplates atop the columns,

• The seismic evaluationprocedures used in thisproject were developed withfull or partial MCEERsupport, including IDARC4.0, 3D-BASIS, FEMA 273/274and NEHRP 2000.

240 m

168 m

24 m

48 m

48 m 50 mm EJ (typ.)

a. Plan view of the Terminal building

Full co

Half col.

24 m48 m

48 m

Quarter col.

Third story

12 m12 m

12 m

Second story

EJ in roofSpace frame

roof

3rd Floor

2nd Floor

b. Plan view of framing in one pod c. Cross-section showing typicalframing

� Figure 2. Construction of the Terminal Building

168

and splitting cracks and spalledconcrete in the beam-column jointsat the third floor level (indicatingno transverse reinforcement in thejoints). Photographs of damage tothe building are shown in Figure 3.

Seismic Evaluation ofTerminal Building

One frame in one of the typical48 m by 48 m pods was selectedfor evaluation by nonlinear staticanalysis. The objectives of the analy-sis of the existing building were to(a) correlate the locations of theobserved damage and that pre-dicted by analysis, (b) to estimatethe displacement capacity of theexisting framing system, and (c) toprovide guidance to the designteam on plausible retrofit schemes.The lightly shaded zones in Figure2b indicate the location and widthof the sample frame. The central bayof framing was selected because itincluded the third story cantilevercolumns that were damaged duringthe earthquake. This frame was 24

m wide above the 3rd floor leveland 12 m wide at that level andbelow.

The nonlinear static (or push-over) analysis was conducted usingthe procedures set forth in FEMA273 (FEMA 1997). The existing fram-ing was modeled using the as-builtconstruction drawings. The tribu-tary widths of the waffle-slab orbeam framing for stiffness calcula-tions was set equal to 12 m; thestrength of the beam framing wasbased on the reinforcement in thesolid segments of the waffle slabsbetween the column. The third-story columns were linked at theroof level by rigid axial elements tosimulate the effect of the space-frame roof structure. The deforma-tion capacities of the reinforcedconcrete components were setequal to the values listed in Chap-ter 6 of FEMA 273 for reinforcedconcrete columns, beams, andbeam-column joints. Because littletransverse reinforcement was pro-vided in the critical or hinging re-gions, the deformation capacities ofall components were established

Overall Project Information:LZA Technology:http://www.lzagroup.com

Hardware Information:Earthquake ProtectionSystems:http://earthquakeprotection.comTaylor Devices:http://www.taylordevices.com

� Figure 3. Observed Damage to the Terminal Building; base of third story column (left) and base ofroof truss connection to column (right)


assuming non-conforming trans-verse reinforcement. A sample cal-culation for the third-story columnshowed a maximum plastic rota-tion of 0.01 radian for the perfor-mance level of collapse preventionbecause the axial load ratio was lessthan 0.1, the transverse reinforce-ment was non-conforming, and theshear force ratio was less than 3. Atarget displacement for the push-over analysis was established basedon revised criteria established bythe owner after the August 19 earth-quake, namely, a spectrum with or-dinates equal to 150 percent of theelastic spectrum set forth in the1997 Turkish seismic code for thesite of the airport. The resulting tar-get displacement at the roof levelwas 230 mm for an elastic periodof 1.25 seconds.

The pushover analysis was ac-complished using IDARC 4.0 (Valleset al., 1996). The existing buildingwas analyzed using two lateral-loadprofiles; second-order effects wereautomatically included. For thisframe, both a uniform pattern anda modal pattern were used. Thecollapse mechanism involved hing-ing of the third story cantilever col-umns for both loading profiles.Figure 4 presents the base shear-roof displacement relationships forthe two lateral-load profiles. The sig-nificant difference between thetwo curves is due to the large dif-ferences between the weights atthe roof and lower levels and theresulting differences between theloading profiles. Nearly all of thebuilding deformation occurs in thethird story of the building with themodal load pattern and the fram-ing below the third story does notyield. Such a distribution of pre-dicted damage is completely con-sistent with the observed damage

to the reinforced framing with oneexception: the beam-column jointdamage was not predicted becausethese joints were assumed to berigid for the analysis. For informa-tion, the third-story drift corre-sponding to a plastic rotation of0.01 radian in the third-story col-umn was 80 mm. Clearly, the defor-mation capacity of the third-storycolumns would be exhausted wellbefore the target displacement atthe roof level was achieved. Furtherevaluation of the existing buildingshowed that the moment-resistingframes were undesirable weak col-umn-strong beam frames.

Conventional andProtective SystemsRetrofit Concepts

Retrofit schemes for the Terminalbuilding were developed using con-ventional methods and materialsand new technologies. The conven-tional retrofit options consideredby the design team all made use ofa new ductile lateral-force-resistingsystem, including steel bracedframes, special reinforced concreteshear walls, and special reinforcedconcrete moment frames. All of theconventional retrofit schemes in-

0

500

1000

1500

2000

2500

3000

3500

4000

0 100 200 300 400

Displacement at top of 3rd story column (mm)

Base

sh

ear

(kN

)

1

2

2

1

Collapse mechanism

forms in the 3rd story

Modal Pattern

Uniform Pattern

Point 1: Center 3rd story column forms a plastic hinge.

Point 2: End 3rd story columns form plastic hinges.

� Figure 4. Pushover Analysis of a Frame in the Original Building

170

cluded new foundations under thenew lateral-force-resisting compo-nents, repair and reconstruction ofthe third-story column to roof-trussframing, elimination of the expan-sion joints between the pods, join-ing and jacketing adjacent quarterand half columns, and jacketing ofthe interior third-story columns.The addition of a new lateral-force-resisting system was not consid-ered feasible because substantialnew vertical elements (walls orbraced frames) could not be in-stalled in or below the first story.

New technologies in the form ofseismic isolation and supplementaldamping were evaluated for theretrofit of the Terminal building.Because the addition of supplemen-tal damping devices would haveinvolved the addition of braces orwall panels in the lower two sto-ries of the building, a detailed ret-rofit design using supplementaldampers was not prepared. Twoseismic isolation options were con-sidered: base isolation of the entirebuilding, and isolation of the rooftrusses. The building isolation op-tion was the preferred option of thetwo, but was rejected by the ownerbecause of the advanced state ofconstruction of the building. Theinstallation of an isolation systemimmediately above the foundationwould have required the demoli-tion and reconstruction of theground floor of the Terminal build-ing, and the removal and reinstalla-tion of the mechanical and baggagehandling systems located in the firststory of the building. The secondisolation option was studied in de-tail, and was selected for the retro-fit of the building by the owner.Information on this retrofit schemeis presented below.

Retrofit (Upgrade) ofthe Terminal Building

This scheme selected for the ret-rofit (or upgrade) of the Terminalbuilding involved the isolation ofthe roof trusses to reduce the de-mand on the third story columnsand the framing at the lower lev-els, the addition of shock transmis-sion units to the roof trusses to lockthe space-frame truss pods togetherduring earthquake shaking so thatthe space frame would act as a dia-phragm, and selected retrofit andstrengthening of the reinforcedconcrete construction as summa-rized below.

Because of architectural con-straints, the size of the third storycolumns could not be increasedsubstantially, so the existing flexuralstrength of these columns as canti-lever elements dictated the inertialforce that could be developed atthe roof level. Fuses in the form ofFriction Pendulum (FP) isolationbearings were used to limit the lat-eral forces that could be imposedon the third-story columns. (FPbearings were used because suchbearings can isolate light compo-nents and structures.) Preliminarycalculations called for an isolatedperiod of 3.00 seconds (based onthe radius of the sliding surface), adesign friction coefficient of 0.09,and a displacement capacity of ap-proximately 260 mm.

The quarter and half columnsaround the perimeter of each podwere joined to the adjacent quar-ter and half columns respectivelyusing reinforced concrete. Addi-tional vertical reinforcement wasplaced in the joints between thepart columns and around the pe-rimeter of the joined columns to


further strengthen the columns.The objective of such strengthen-ing was to eliminate the weak col-umn-strong beam framing. Thesecolumns and the modestly strength-ened interior square columns werethen jacketed with circular steelcasings to substantially increase theshear strength of all the columnsand to provide confinement in po-tential hinge regions. Such columnstrengthening was implemented inthe second and third stories. In ad-dition, the expansion joints be-tween the pods at the second andthird floor levels were eliminatedby tieing the pods together withreinforced concrete components tosubstantially increase the redun-dancy of the lateral-force-resistingsystem. No beams at either the sec-ond or third floors were strength-ened.

The performance of the retrofit-ted building was checked by non-linear static analysis of the framedescribed above. The resultingpushover curve for the modal loadpattern is presented below in Fig-ure 5 together with a sketch show-ing the sequence of plastic hingeformation (1 through 20). Note thatsome hinges form simulta-neously. For a roof displace-ment less than 500 mm,hinges form in the beams, atthe base of the columnsabove the foundation, and inthe two-story columns at theunderside of the third storyonly.

The retrofit design was fur-ther evaluated by nonlineardynamic analysis using 20ground motion records thatmatched on average the re-vised design spectrum de-scribed above. The meanmaximum displacement at

the roof level of 190 mm is indi-cated by the open-ended arrow onthe pushover curve of Figure 5. Thedeformation demands on thebeams and columns in the retrofit-ted building frame at this roof dis-placement were consideredacceptable for the performancelevel of collapse prevention.

Two photographs of the retrofitwork showing an installed FP bear-ing (before release) and a jacketedcolumn (covered by an architec-tural treatment) are presented in

0

500

1000

1500

2000

2500

3000

3500

4000

0 100 200 300 400 500 600 700 800 900 1000

Displacement at top of 3rd story column (mm)

Base

sh

ear

(kN

)

1

57

1213

141

5

17 19 20

1

3

1 1

2 0

1 4

1 3

1 2

1 9

1 0

9

1 1

4

8

21 7 1 0 1 8 1 2

1 5 5 7 6 1 6

� Figure 5. Pushover Analysis of the Retrofitted Frame

� Figure 6. Isolation and strengthen-ing of the Terminal building; FPisolators atop jacketed columns(above) and jacketed third-storycolumn (left)

172

FEMA, (1997), NEHRP Guidelines for the Seismic Rehabilitation of Buildings, ReportNo. FEMA 273, Federal Emergency Management Agency, Washington D.C. October.

Valles, R.E., Reinhom, A. M., Kunnath, S. K., Li, C and Madan, A., (1996), IDARC2DVersion 4.0: A Computer Program for the Inelastic Damage Analysis of Buildings,Report No. NCEER-96-0010, National Center for Earthquake Engineering Research,University at Buffalo, June.

References

Figure 6. No retrofit work was un-dertaken in either the first story orto the foundations. The nonlinearstatic analysis and the nonlinear dy-namic analysis indicated no inelas-tic action in the first story abovethe footings, in part because thecolumns in this story were substan-tially stronger and taller than thecolumns in the stories above. Ret-rofit of the footings beneath the ex-isting columns was not feasiblegiven the advanced state of the con-struction at the time the retrofitscheme was developed.

Summary andConclusions

An innovative retrofit and up-grade scheme was developed and

implemented for the new AtaturkInternational Airport Terminalbuilding, which was damaged dur-ing the August 19, 1999, Izmit earth-quake. The retrofit scheme involvedthe use of conventional strengthen-ing and seismic isolation hardwareto avoid building collapse in theevent of a maximum earthquake.The efficacy of the retrofit schemewas demonstrated by nonlineardynamic and static analysis. Thestudies of the existing building andthe development of the retrofitschemes commenced in late Sep-tember 1999, and the retrofit con-struction work was completed bythe end of December 1999: a pe-riod of approximately 12 weeks.

173

Research Objectives

The goal of the project is to use the federally sponsored loss estimationsoftware, HAZUS (Hazards–U.S.), to project the magnitude of potentiallosses that might be experienced by the metropolitan New York City areaas a consequence of a damaging earthquake. After modification of thedefault HAZUS datasets for soil and building characteristics, a more cred-ible estimation of losses will emerge and be useful to metropolitan emer-gency personnel, as well as to other public and private stakeholders. It ishoped that the results will contribute to improved disaster mitigation andemergency response plans throughout the area.

Estimating Earthquake Losses for theGreater New York City Area

by Andrea S. Dargush (Coordinating Author), Michael Augustyniak, George Deodatis, Klaus H. Jacob,Laura McGinty, George Mylonakis, Guy J.P. Nordenson, Daniel O’Brien, Scott Stanford, Bruce Swiren,

Michael W. Tantala and Sam Wear


Federal EmergencyManagement Agency

Federal EmergencyManagement AgencyRegion II

New Jersey Office ofEmergency Management

New York State EmergencyManagement Organization

The metropolitan New York-New Jersey region is vulnerable to severalpotential natural disasters. Ice storms, snow, and other severe weather

events such as hurricanes, associated storm surges, and flooding, foot-and-mouth, and the West Nile virus are among those of current concern andprominent public attention. After adding such man-made disruptions asterrorism and hazardous materials releases, and the hazards posed by de-caying infrastructure and aging buildings, earthquakes seem a dim pros-pect. But however unrecognized or unacknowledged the threat fromearthquakes may be in New York City, seismic events have occurred and,until recently, no codes have existed to mandate earthquake strengthen-ing of structures. Statistics indicate that potential losses to a large urbanarea such as New York would be considerable. According to Scawthornand Harris (1989), the economic impact of a damaging earthquake (e.g.,M6.0) in New York would be in the billions of dollars due to direct struc-tural and architectural damage, and does not reflect additional impacts onbuilding contents, business continuity, fire suppression and human safety.An earthquake in the greater New York area would thus be considered alow-probability, yet high consequence event. Credible estimates of futureloss can be effective tools to encourage area stakeholders to mitigate againstthe possible future damaging consequences of earthquakes.

Hazards U.S. (HAZUS) is a standardized, nationally-applicable loss esti-mation tool, developed by the Federal Emergency Management Agency(FEMA) in cooperation with the National Institute of Building Sciences.The HAZUS software utilizes geographic information systems, such as

George Mylonakis, WalterFish and Paul Spiteri,The City College of the CityUniversity of New York

Bruce Swiren, FederalEmergency ManagementAgency, Region II

Klaus Jacob, NoahEdelblum and JonathanArnold, Lamont DohertyEarth Observatory

Andrea Dargush, MichaelKukla and MontreePolyium, MCEER

Scott Stanford, New JerseyGeological Survey

Michael Augustyniak, NewJersey State Police Office ofEmergency Management

Research Teamcontinued on page 2

174

The users of this research include emergency responders, cityplanners, utility managers, building owners, design and construc-tion engineers, critical facility managers, school and hospitaladministrators, the public at large and other researchers.

Dan O’Brien, New York StateEmergency ManagementOffice

Dan Cadwell, New York StateGeological Survey

George Deodatis, GuyNordenson, MichaelTantala and AmandaKumpf, PrincetonUniversity

Sam Wear, Laura McGintyand Andrea Albert,Westchester CountyGeographic InformationSystems, Department ofInformation Technology

ArcView and MapInfo, to producedetailed maps and analytical re-ports that describe a community’spotential losses. The current ver-sion applies a uniform engineering-based loss estimation approach toquantify damages, economic lossesand casualties resulting from earth-quakes. Future adaptations are in-tended to carry out similar analysesfor hazards such as flooding andhigh wind.

Similar loss modeling studieshave been conducted in the NewYork area in past years to examinethe impact of flooding and hurri-canes, and have been used to guideemergency plans. Recognizing thepotential collective benefits of a re-gional study, mitigation specialistsat FEMA Region II, the New YorkState Emergency ManagementAgency (NYSEMO), the New JerseyState Police Office of EmergencyManagement (NYOEM) and theNew Jersey Geological Surveyjoined together to develop a simi-lar loss estimate for earthquakes.With NYSEMO and NJOEM, aFEMA-supported project was initi-ated in 1998 to apply HAZUS to thegreater New York City area. A re-gional consortium has been formed- the New York City-area Consor-tium for Earthquake-loss Mitigation(NYCEM). MCEER was selected toprovide general coordination of theactivities and to conduct outreachactivities to promote the outcomesof the project. The purpose of theconsortium is to help develop thenecessary databases to effectively

utilize the loss estimation programHAZUS to identify potential eco-nomic loss to the greater New YorkCity area as a consequence of adamaging earthquake. In the latterstages of Year 2, increased attentionwas given to inclusion of informa-tion on critical facilities, to assesspost-event functionality. The con-sortium consists of public agencyofficials, business owners, emer-gency managers, engineers and ar-chitects, utility owners and otherarea stakeholders. Researchers atLamont–Doherty Earth Observa-tory, Princeton University, and theCity College of New York work to-gether to develop soils and build-ing stock inventories which can beused to refine the default data con-tained within HAZUS. The data willbe used in future executions ofHAZUS and will assist researchersin refining their ultimate loss esti-mations for New York.

These studies are being comple-mented by information generatedby similar studies for WestchesterCounty, New York and for a num-ber of counties in New Jersey. Ageneralized HAZUS analysis willprovide a regional picture of poten-tial earthquake damage and loss.

These assessments will ultimatelybe used to encourage and promoteearthquake mitigation action at thelocal, regional and statewide levels.

Technical SummaryAfter preliminary explorations of

available data and its transferrability

175Estimating Earthquake Losses for the Greater New York City Area

• MetropolitanTransportation Authority/New York Transit Authority

• New York City Departmentof Design andConstruction

• New York City Departmentof Finance and OperationsResearch

• New York State GeologicalSurvey

NYCEM Technical AdvisoryCommittee

Dr. Carl J. Costantino,Professor Emeritus,

Department of CivilEngineering, City Collegeof New York

Mr. Ronald T. Eguchi, CEO,ImageCat

Mr. Ramon Gilsanz,Gilsanz, Murray andSteficek

Dr. Jeremy Isenberg, CEO,Weidlinger Associates, Inc.

Dr. Stephanie A. King,Stanford University

Mr. Charles A. Kircher,Principal, Charles Kircherand Associates

Dr. Joanne M. Nigg,Professor of Sociology andCo-Director, DisasterResearch Center, Universityof Delaware

Dr. Thomas O'Rourke,Professor, School of Civiland EnvironmentalEngineering, CornellUniversity

Mr. Maurice S. Power, VicePresident, GeomatrixConsultants, Inc.

Mr. Thomas Z. Scarangello,Vice President, Thornton-Tomasetti Engineers

to a HAZUS format, the NYCEMteam prioritized its efforts, focus-ing on the development of a data-base of geologic and buildinginformation for Manhattan below59th Street. As the project pro-gressed, additional stores of databecame available that allowed ex-pansion of the study to the entireborough. In the current Year 3 ofthe project, results of the Manhat-tan study will be merged withthose of parallel studies being car-ried out in New Jersey andWestchester County, New York, forextrapolation to a larger region cov-ering about 31 counties in NewYork and New Jersey. The respec-tive activities of the team membersand the parallel studies are de-scribed in the following section.Priorities, methodologies andprogress of the NYCEM team havebenefited from periodic technicaladvisement from a panel of expertsin earthquake engineering and lossmodeling.

Lamont Doherty EarthObservatory (LDEO)

Based on extensive prior re-search, the LDEO team, led by KlausJacob, realized from the outset thatthe default soils data within HAZUSdid not accurately reflect the sub-surface conditions of the Manhat-tan area. Near surface geologicdifferences can introduce localvariations in shaking that can influ-ence both the amplitude and spec-tral composition of groundmotions, thus impacting structuresand lifelines.

It would be a challenge to theteam to upgrade the data to pro-duce the needed NEHRPgeotechnical site classes at an indi-

vidual Manhattan census tract level(Jacob, 1999). Pre-existing geologi-cal studies were geographically lim-ited and not specific enough toderive important information ondepth to bedrock. Other subsurfacedata collected in the form of bore-hole logs, were collected by privatedevelopers or other entities, hold-ing the data as proprietary.

Using information provided bythe New York City Department ofDesign and Construction, data from150 geotechnical borings for lowerManhattan were studied in the firstyear of the project. Using casinginformation and results from thestandard penetration tests (SPT)conducted, it was possible to de-rive shear wave velocity profiles asa function of depth at each loca-tion, then translated into the appro-priate NEHRP site class. Some 200other older borings were available,but only offered information ondepth to bedrock.

To classify the overlying soils, ashear wave velocity vs. depth func-tion was derived from a subset ofthe initial 150 borings and appliedto generate point values at the 200+sites. The resultant classificationswere primarily NEHRP soil types Dand C. Surprisingly, no class A (veryhard rock) sites emerged, althoughnumerous B (firm rock) sites wereidentified. Several individualborings indicated class E condi-tions. However, the conditionswere not sufficiently continuousthroughout the census tract tomerit assignment of E to the tract.

This may be a pervasive problemassociated with the census tractapproach to geological classifica-tion. These differences in soil typefrom the default assignments re-sulted in apparent reductions of thetotal loss estimations for buildings

176

• The NYCEM project isprimarily linked to MCEERoutreach efforts toincrease awareness andacknowledgment ofearthquake risk and thevalue of loss estimationmethodologies asmitigation tools.

• Data collected from theproject may benefitongoing projects focusingon earthquake risk tocritical facilities in areasof low-to-moderate hazardbut with high collateraldamage.

• Future phases of theproject will likely benefitfrom MCEER lifelineresearch which has led toimprovements in HAZUStreatment of buriedpipeline systems.

by factors of 0.7, 0.72 and 0.8 forscenario events of M

w 5.0, 6.0, and

7.0, respectively. Additional refine-ment of the shear wave velocityfunctions made possible withadded calibration borings may bereflected in future loss estimates.

In Year 2, additional refinementof the lower Manhattan study areawas made possible by the inclusionof additional data provided by theMetropolitan Transportation Au-thority/New York City Transit Au-thority. The study area was alsoexpanded to include the entire bor-ough of Manhattan. In spite of theadditional information, data densitywas greatest for midtown andlower Manhattan and sparse forareas north of 137th St..

Modifications were made to themethodology developed in Year 1to reflect consideration of theNEHRP 10-foot rule, which assignssite category E to any soil profilethat contains a continuous ten-foot thick layer of very soft soils(Jacob, 2000). Additional studyalso led to the increase in the cal-culated shear wave velocity forbedrock at the bedrock/soil inter-face, better reflecting the prevail-ing rock formations in the area.Site classes for all DDC borings inYear 1 were recalculated underthese assumptions.

A primary product in Year 2 is acensus-tract based soil map for theentire borough, which can beused in validating results, gener-ated by HAZUS. Figure 1 describesthe distribution of soil/rock typesacross the island, with higher el-evations in uptown Manhattan onstiff rock/soil (Class B), interme-diate type C in Midtown, and softsoil (Class D) in lower Manhattan,along the identified fault zones inUpper Manhattan and in low-ly-

ing, coastal areas. An anticipatedproduct of this effort will be thedevelopment of a contoured soilsmap for Manhattan. This will allowmore accurate extrapolation of cen-sus-tract specific geotechnical char-acteristics.

An important Year 2 activity wasthe coordination of approach toclassification of soils among theNew York City, New Jersey andWestchester County study areas.With input from members of theNEHRP Seismic ProvisionsGeotechnical Subcommittee andthe Building Seismic Safety Coun-cil, a uniform methodology was de-rived to assure that assumptionsused to assign NEHRP soil classtypes are, to the extent possible inagreement with the objectives ofthe NEHRP guidelines. There was

� Figure 1. NEHRP Site Classes for Manhattanby Census Tracts


New York City-areaConsortium forEarthquake HazardMitigation:http://www.nycem.org

Federal EmergencyManagement AgencyHAZUS:http:www.fema.gov/hazus

Westchester County GIS:http://giswww.co.westchester. ny.us

general consensus that the informa-tion needed to accurately assignNEHRP classes is often unavailableand that in view of other inherentuncertainties within the HAZUS al-gorithm, it should not unduly im-pact results.

The HAZUS algorithm is de-signed to execute loss estimationsusing both deterministic andprobabilistic earthquake inputs.Using historical local seismicevents, the LDEO team also pro-vided the epicentral locations forscenario earthquake events to beused in the deterministic HAZUSexecutions.

Princeton University

New York City has one of thehighest concentrations of high risebuildings per square foot than anyother city. Unreinforced masonrystructures housing businesses, ar-chitectural treasures, masonry firestations, theaters and art galleries,apartment buildings, and more,stand shoulder to shoulder withthese skyscrapers.

Many New York structures holdhistorical status and many morewere constructed without the guid-ance of a seismic building code. Be-fore 1996, earthquakes were not adesign consideration although asconstruction progressed in the20th century, wind load factorswere being incorporated, thus ad-dressing (albeit coincidentally)some of the horizontal displace-ments that might be experiencedduring an earthquake.

Even today, seismic statutes ap-ply to new construction only. Con-sensus opinion is that retrofittingthousands of New York buildingsto meet seismic standards is im-practical and economically unreal-

istic. It is therefore even more im-portant to accurately identify areasof highest potential vulnerability toearthquake ground shaking so thatmitigation, emergency responseand recovery approaches can bestrengthened.

In HAZUS, the default building in-ventory is categorized by occupancy(commercial, residential, industrial,governmental, educational, religious,agricultural), model building type(e.g., steel, reinforced concrete, wood,masonry), structural configurationand height. HAZUS also makes otherassumptions about typical percent-age distribution of buildings withina census tract by age, quality (infe-rior, built to code, superior) seismicdesign level (low, moderate and high)and height. The NYCEM team atPrinceton University (Guy J. P.Nordenson, George Deodatis,Michael Tantala and Amanda Kumpff)recognized the obvious disparitiesbetween the typical suburban build-ing inventory presented in HAZUSand the actual structural characteris-tics of New York. Such disparitieswould ultimately affect the accuracyof damage and loss scenarios to begenerated by the program.

One of the objectives of the stud-ies being carried out at Princetonis to develop a credible buildinginventory for the New York Cityarea, which will be more represen-tative of its wide mix of structures.Various damage and loss scenarioswould then be carried out, usingboth the improved building dataand the soils data provided by theLDEO team.

In Year 1 of the study, the teamused Sanborn fire insurance mapdata (building height, size, location,occupancy and type) and visual in-spection to modify building inven-tories for two representative

178

census tracts – Wall Street (prima-rily commercial) and Kips Bay (pri-marily residential) (Nordenson etal., 2000). Broader regional studiesexamined New York below 59thStreet, and the 31 county, New York-New Jersey area. HAZUS runs werethen carried out to assess the sen-sitivity of the methodology tochanges in building inventory data.Dramatic differences in loss esti-

mates emerged, particularly whenusing smaller magnitude earth-quakes. Total loss estimates in themodified runs differed significantlyfrom those of the default (in somecases more than a factor of ten).Based on the runs, it appears thatthe most sensitive building informa-tion for refined loss estimationsmight be square footage and height,and not structural type, as previ-

� Figure 2. Building Count by Structural Type and Neighborhood


ously assumed. This will require fur-ther validation.

With assistance from Dr. JackEichenbaum, New York City Asses-sor, the Princeton team acquiredvaluable building data from theNew York City Department of Fi-nance. The Mass Appraisal System(MAS) data consists of over 2 mil-lion property entries for the entirecity, including such information asparcel square footage, address,height, use, building footprint di-mensions, construction year, qual-ity, and architectural style (Tantalaet al., 2001).

The year 2 studies used this in-formation to further refine thebuilding inventory, althoughgeocoding of entries was neededto match information to the appro-priate census tract. In addition therichness of the MAS data requiredsimplification to fit into the HAZUScategorization scheme.

In contrast, the MAS constructiontype data required addition of morespecific information. Field surveyswere carried out with the City Col-lege team and were used to validatesome of the existing data as wellas some of the associated assump-tions being made. Additional datasources from local engineers, Dunand Bradstreet business databases,aerial photography, existing struc-tural drawings and other sourceshelped to further validate the newdatabase.

After thorough modification, theresultant MAS database includedrecords for more than 37,000 build-ings, or over 2.2 billion square feetof area. Figure 2 shows the relativedistribution of the number of build-ings in each of four category typesfor twelve Manhattan neighbor-hood districts, illustrating thatunreinforced masonry is the pre-

dominant building type. This is not,however, reflected in the amountof square footage occupied byunreinforced masonry, which isperhaps a more important variable.

In the end, the Princeton teamwas able to establish the buildinginventory for the entire island ofManhattan, with HAZUS-requiredinformation at the individual build-ing level. This is a unique accom-plishment for HAZUS applications.With the assistance of Princetonstudent Evan Schwimmer, the da-tabase has been further enhancedwith the addition of critical facili-ties data (hospitals, schools, andpolice and fire stations). Severalpreliminary probabilistic and deter-ministic HAZUS runs have beencarried out using the critical facili-ties data to generate projectedfunctionalities of schools, hospitals,and police and fire stations, to as-sess their relative capabilities toprovide shelter, provide emergencycare and control conflagration.Casualties, population displace-ment and short- and long-term eco-nomic changes have also beenprojected. These are described indetail in the Year 2 report, whichmay be found at the NYCEMwebsite.

These subsequent HAZUS (1999and SR-16 versions) runs used theimproved building inventory andfurther verified the need for im-proved data sources for buildingsand soils to derive credible damage,loss and casualty estimates.

Westchester County, New York

Under sponsorship of the NewYork State Emergency ManagementOrganization, a collaborative studywas carried out in WestchesterCounty, New York. The study evalu-

180

ated the applicability of the HAZUSdefault datasets, and made appro-priate modifications to the extentpossible.

Considerable effort was dedi-cated to data collection and to theupdating of the HAZUS datasets, inparticular, additions of informationon essential facilities (McGinty andWear, 2001). Important new cover-ages were included for such facili-ties as day care centers and seniorhousing. Additional efforts were fo-cused on the inclusion of informa-tion on essential facilities such ashospitals, schools, police and firestations, and emergency medicalservices.

This led to the identification ofshortfalls in county data availabil-ity and development of a strategyfor future data creation. In the ab-sence of borehole data, scientistsat the New York State GeologicalSurvey assisted by conducting seis-mic shear wave velocity tests onsurficial soils within the county.Tests were conducted in locationsof different NRC-defined soil typesso that a larger scale surficial soilsmap could be developed. Data wasthen extrapolated to a census tractlevel for necessary seismic soil clas-sifications.

The Westchester study culmi-nated in a HAZUS execution basedon three earthquake scenarios. Thegoal of the study is to provide lossestimation data to the sponsor toenable development of a statewidemitigation approach for potentialfuture earthquakes.

New Jersey

Under the oversight of MichaelAugustyniak at the New Jersey Of-fice of Emergency Management,studies were carried out by the

New Jersey Geological Survey todefine areas of seismic hazard.Building inventory data was col-lected for selected areas to upgradeHAZUS inventory data.

Geologic data were acquired andanalyzed in order to compile mapsof seismic soil class and liquefac-tion susceptibility for HudsonCounty, New Jersey (Stanford,1999). The soil class and liquefac-tion susceptibility data were en-tered into the HAZUS model foreach census tract in the county.

The HAZUS model was run withthe upgraded geologic data andwith the default geologic data forearthquake magnitudes of 5, 5.5, 6,6.5, and 7. The upgraded buildinginventory data has not yet been in-cluded in the runs but is presentlybeing processed by the Princetonteam.

The upgraded geologic informa-tion produced significant changesin both the spatial distribution ofdamage and the total damage esti-mates. This increased building dam-age was particularly notable in theHudson waterfront and HackensackMeadowlands areas of the county,where soils are softer and more liq-uefiable than HAZUS default soil.

Building damage was less on thePalisades Ridge and on uplands inKearny and Secaucus, where soilsare stronger than the default. Be-cause most building constructionin the county is concentrated onthese ridges, the total estimatedbuilding damage is somewhat lesswith the upgraded geologic datathan with the default data at allmagnitudes. Additional mappingand scenario executions have alsobeen carried out in Bergen county,with Essex County underway.

In addition to the HAZUS data up-grades and scenarios executions,

“It is importantto accuratelyidentify areasof highpotentialvulnerability toearthquakes sothat mitigation,emergencyresponse andrecoveryapproaches canbe strengthened.


shear-wave velocities were mea-sured on the three softest soil typesat nine locations. These were con-ducted to validate the soil class as-signments used in the scenarioexecutions, which use test-drillingdata as a proxy for shear-wave ve-locities. The measured velocitiesconfirmed the assignments. An ad-ditional 13 shear wave velocitymeasurements completed inBergen County on four soil typesindicate once again that the SPTdata accurately proxy for shearwave velocities, except in gravelrich areas. These have faster veloci-ties than SPT values suggest. Soilclasses were adjusted accordinglyin these areas.

City College of New York

To support the ongoing activitiesand to add some additional valida-tion checks, MCEER provided sup-port to George Mylonakis, CUNY,and his team of undergraduate en-gineering students to conduct fieldsurveys of building inventory data.Their objectives were to assess theaccuracy of both the defaultHAZUS building inventory data andto verify and benchmark the MASdata for two of three representativecensus tracts.

After initial training in HAZUS,the team carried out walking sur-veys of the Rockefeller Plaza cen-sus tract and a portion of the KipsBay census tract. Buildings over a10-block area were photographed,with data collected on frame type,exterior wall type, basement grade,construction quality, exterior con-dition, existing damage, HAZUSbuilding code, and occupancy.These categories correspond to therequired data fields within HAZUS(Mylonakis, et al., 1999). A total of

more than 600 buildings were sur-veyed by the CUNY and Princetonteams.

Survey data was input into an AC-CESS database and comparedagainst the default data. The differ-ences between the default and ac-tual data were quite evident for thenumber of buildings and for thepercentage of residential buildings.There is also a significant differencebetween the actual data and defaultdata for the number of commercialbuildings. The HAZUS dataset de-picts a greater variety of buildingtypes within a census tract thanactually exists. These led to furtherdifferences in loss estimations car-ried out for these areas.

To further validate the data sen-sitivity, the CUNY team generatedHAZUS loss estimates for Kips Bay,Rockefeller Center and Wall Streetcensus tracts (Mylonakis, 2000). TheWall Street tract data were col-lected by the Princeton team. Us-ing a fixed location (1884Rockaway Beach) scenario earth-quake at varying magnitudes, thedefault and modified datasets wereused to generate costs of structuraldamage. The 1999 version of theHAZUS software was used. ThePrinceton group conducted similaranalyses using HAZUS version ’97.Good agreement existed betweenthe results generated by differentversions of HAZUS.

OutreachSeveral years after the passage of

seatbelt legislation, drivers and pas-sengers still disregard the need forsimple protective actions. Convinc-ing the public, the government andthe private stakeholders in thenortheastern United States of the

182

need to prepare for an earthquakeis a challenge not easily met. It hasbeen an important objective of thisproject to convey that a low-prob-ability event is a potential reality,carrying with it consequences forwhich the metropolitan area maybe ill prepared.

The NYCEM project has taken ad-vantage of numerous media oppor-tunities to disseminate thisinformation and to explain the im-portance of accurate inventoriesand loss characterizations to the de-velopment of mitigation strategies.The work has been featured in sev-eral articles in the New York Timesand other area newspapers, hasbeen the subject of a news focuspiece on WNBC-News and has re-cently been prominently broadcastin the Discovery Channel program,“Earthquakes in New York?” Theawareness campaign was furtheraugmented by a minor tremor thatstruck midtown Manhattan, Janu-ary 17, 2001 (see Figure 3). This at-tention, combined with theleadership of Bruce Swiren ofFEMA Region II and Dan O’Brienof New York State Emergency Man-

agement, has helped to open dia-logues with City and metropolitanarea emergency managers aboutthe importance of earthquake haz-ard mitigation.

In addition, conventional out-reach activities have been carriedout, targeted toward regional stake-holders who can both assist in re-fining the model through dataprovision, and can benefit from theenhanced loss estimation model.Workshops and briefings have in-creased their interest in the projectand have led to sharing of informa-tion and data necessary to refinethe default HAZUS data. Metropoli-tan transportation engineers areinterested in being part of possiblefuture studies that would includetransportation lifelines in themodel.

Project results are made widelyavailable. A NYCEM website hasbeen established by MCEER, whichserves as a repository for technicalreports, data and maps generatedby the team. NYCEM results are re-ported regularly through theMCEER Bulletin and on the web,and several additional informationbriefings will be featured through-out the final year of the program.At the completion of the program,all data will be publicly accessiblethrough a central repository.

Conclusions andFuture Research

The study has revealed variancesin HAZUS default data that would,left uncorrected, compromise itsability to accurately project poten-tial losses due to a damaging earth-quake. Loss quantificationsgenerated by HAZUS have beenfound to be sensitive to variations

“The goal of theWestchesterCounty studywas to provideloss estimationdata to enabledevelopment ofa state-wideapproach topotential futureearthquakes.”

� Figure 3. Epicenter of January 17, 2001 Manhattan Earthquake(courtesy of Dan O’Brien)


in building inventory data, earth-quake magnitude and local soilconditions.

In some cases, the damage and lossvalues generated were notablyhigher using default data than withthe upgraded building and soils in-formation. The data collection andmodification activities have substan-tially enhanced the program’s abilityto become a credible forecast instru-ment of economic impact.

These activities have also signifi-cantly improved the general knowl-edge base about New York areageology and have led to the devel-opment of a more comprehensivebuilding inventory that can be usedby others. In fact, it is notably the firstdatabase where the building inven-tory has been described in a build-ing-by-building manner, which is notonly unique, but quite valuable tofuture regional studies. The ability ofthe NYCEM team to build this infor-mation base is notable, since availabledata for such a complex and largeurban area was difficult to identifyand obtain. Surmounting this initialobstacle has been an importantachievement of the project. Recon-ciling this information to the census-tract based approach of HAZUS hasalso been an intensive and challeng-ing effort.

In an area of such vast and vitaleconomic activity, it is paramountthat any estimation of potentialeconomic impact be based in well-founded data and analysis. While

both the methodology, its inherentuncertainties, and its associateddatasets have come under question,HAZUS is being utilized in otherlarge urban areas with higher lev-els of seismic hazard as a tool toassist area planners, responders,and other stakeholders.

The studies conducted have es-tablished a foundation for morecomplete loss estimation studiesfor the greater New York metropoli-tan area. Additional data collectionand study of regional lifeline sys-tems, including water, gas and high-way systems, will significantlyenhance the risk characterizationsthat HAZUS can provide for theNew York City area.

Further involvement by emer-gency responders, planners, builders,and health and human services offi-cials will help improve the effective-ness of these studies. Accurate andcomplete fingerprints of risk, enabledby more widespread participation byother stakeholders, will ideally leadto the development of an acceptabledamage and loss indicator that canbe embraced by responders, build-ers and the public.

This goal has been encouraged asthis study has progressed into its thirdyear, as several presentations of re-sults to municipal agencies has ledto greater coordination, in particularwith the New York City Office ofEmergency Management and withofficials in Westchester County.

“The NYCEMproject hastakenadvantage ofnumerousmediaopportunities todisseminateinformationand explain theimportance ofaccurateinventories andloss characteri-zations.”

Acknowledgements

The author would like to recognize and acknowledge the efforts of the followingindividuals: Bruce Swiren, FEMA Region II, and Dan O’Brien, New York State EmergencyManagement Office, for their sponsorship and leadership; Elizabeth Lemersal, FEMA, foroverall support of the program; and Dr. Jack Eichenbaum, City Assessor, New York CityDepartment of Finance, and Dr. A. N. Shah, Metropolitan Transportation Authority/NewYork City Transit Agency, for their generous assistance.

184

References

Cadwell, D. and Nottis, G., (2000), “Seismic Hazard Assessment Data, Westchester County,New York,” New York State Museum, Albany, New York.

HAZUS 99: Estimated Annualized Earthquake Losses for the United States, FederalEmergency Management Agency Mitigation Directorate, FEMA 366, February 2001, 33pages.

Jacob, K. H., (1999), “Site Conditions Affecting Earthquake Loss Estimates for New YorkCity,” Lamont-Doherty Earth Observatory of Columbia University, Palisades, NY, NewYork City Consortium for Earthquake Loss Mitigation (NYCEM): Earth Science Tasks,First Year Technical Report May 6, 1999, 35 pages.

Jacob, Klaus H., Edelblum, Noah, Arnold, Jonathan, (2000), “NEHRP Site Classes forCensus Tracts in Manhattan, New York City,” Lamont-Doherty Earth Observatory ofColumbia University, Palisades, NY, New York City Consortium for Earthquake LossMitigation (NYCEM): Earth Science Tasks, Second Year Technical Report, December31, 2000, 23 pages.

Jarvis, Hugh, (1999), “Disaster Damage: Cost Issues for the New York City Area,”Annotated Bibliography, MCEER Information Service.

McGinty, Laura and Wear, Sam, (2001), “Initial Earthquake Loss Estimation Analysis forWestchester County, New York,” Report to New York Emergency ManagementOrganization, Westchester Counter Department of Information Technology, GeographicInformation Systems, February, 2001, 53 pages.

Mylonakis, George et al., (1999), “Earthquake Loss Estimation for the New/York CityArea,” NYCEM-MCEER Progress Report.

Mylonakis, George, Fish, Walter, and Spiteri, Paul, (2000), “Development of a BuildingInventory for Manhattan Region,” NYCEM Preliminary Technical Report.

Nordenson, G., Deodatis, G., Tantala, M. and Kumpf, Amanda, (2000), “Earthquake LossEstimation for the New York City Area,” NYCEM First Year Technical Report.

Scawthorn, C. and Harris, S. K., (1989), “Estimation of Earthquake Losses for a LargeEastern Urban Center: Scenario Events for New York City,” Annals of the New YorkAcademy of Sciences, Volume 558, June 16, 1989, pages 435-451.

Stanford, S. D., Pristas, R. S., Hall, D. W., and Waldner, J. S., (2000), “Earthquake LossEstimation Study for Bergen County, New Jersey: Geologic Component,” prepared forthe New Jersey State Police Office of Emergency Management by the New JerseyGeological Survey, 8 p. plus appendices and plates.

Stanford, Scott, Pristas, Ronald S., Hall, David W. and Waldner, Jeffrey S., (1999), “GeologicComponent of the Earthquake Loss Estimation Study for Hudson County, New Jersey,”Final Report.

Stanford, S. D., Pristas, R. S., (1998), “Earthquake Loss Estimation Study for Newark, NewJersey: Geologic Component,” prepared for the New Jersey State Police Office ofEmergency Management by the New Jersey Geological Survey, 2 p. plus appendicesand plates.

Tantala, Michael W., Nordenson, Guy J.P., and Deodatis, George, (2001), “Earthquake LossEstimation Study for the New York City Area,” NYCEM Year 2 Technical Report, 1999-2000.

Production Staff

Jane E. Stoyle, EditorMichelle A. Zuppa, Layout and DesignHector A. Velasco, Illustration

AcknowledgementsThis report was prepared by the Multidisciplinary Center for Earthquake Engineering Research through

a grant from the National Science Foundation Earthquake Engineering Research Centers Program, acontract from the Federal Highway Administration, and funding from New York State and other spon-sors.

The material herein is based upon work supported in whole or in part by the National ScienceFoundation, Federal Highway Administration, New York State and other sponsors. Opinions, findings,conclusions or recommendations expressed in this publication do not necessarily reflect the views ofthese sponsors or the Research Foundation of the State University of New York.

H e a d q u a r t e r e d a t t h e U n i v e r s i t y a t B u f f a l o

Research Progress and Accomplishments: 2000-2001

Documents

Transcript of Research Progress and Accomplishments: 2000-2001