Seismic Design and Analysis of Underground Estructures - Hashash, Hook, Schmidt y Yao

.Tunnelling and Underground Space Technology 16 2001 247293

ITAAITES Accredited Material

Seismic design and analysis of underground structures

Youssef M.A. Hashasha,, Jeffrey J. Hooka, Birger Schmidtb,John I-Chiang Yaoa

aDepartment of Ciil and Enironmental Engineering, Uniersity of Illinois at Urbana-Champaign, 205 N. Mathews Aenue, MC-250,Urbana, IL 61801, USA

bParsons Brinckerhoff, San Francisco, CA, USA

Abstract

Underground facilities are an integral part of the infrastructure of modern society and are used for a wide range ofapplications, including subways and railways, highways, material storage, and sewage and water transport. Underground facilitiesbuilt in areas subject to earthquake activity must withstand both seismic and static loading. Historically, underground facilitieshave experienced a lower rate of damage than surface structures. Nevertheless, some underground structures have experiencedsignificant damage in recent large earthquakes, including the 1995 Kobe, Japan earthquake, the 1999 Chi-Chi, Taiwanearthquake and the 1999 Kocaeli, Turkey earthquake. This report presents a summary of the current state of seismic analysis anddesign for underground structures. This report describes approaches used by engineers in quantifying the seismic effect on anunderground structure. Deterministic and probabilistic seismic hazard analysis approaches are reviewed. The development ofappropriate ground motion parameters, including peak accelerations and velocities, target response spectra, and ground motiontime histories, is briefly described. In general, seismic design loads for underground structures are characterized in terms of thedeformations and strains imposed on the structure by the surrounding ground, often due to the interaction between the two. Incontrast, surface structures are designed for the inertial forces caused by ground accelerations. The simplest approach is to ignorethe interaction of the underground structure with the surrounding ground. The free-field ground deformations due to a seismicevent are estimated, and the underground structure is designed to accommodate these deformations. This approach is satisfactorywhen low levels of shaking are anticipated or the underground facility is in a stiff medium such as rock. Other approaches thataccount for the interaction between the structural supports and the surrounding ground are then described. In the pseudo-staticanalysis approach, the ground deformations are imposed as a static load and the soil-structure interaction does not includedynamic or wave propagation effects. In the dynamic analysis approach, a dynamic soil structure interaction is conducted usingnumerical analysis tools such as finite element or finite difference methods. The report discusses special design issues, includingthe design of tunnel segment joints and joints between tunnels and portal structures. Examples of seismic design used forunderground structures are included in an appendix at the end of the report. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Seismic design; Seismic analysis; Underground structures; Tunnels; Subways; Earthquake design

Corresponding author. Tel.: 1-217-333-6986; fax: 1-217-265-8041. .E-mail address: [email protected] Y.M.A. Hashash .

0886-779801$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. .PII: S 0 8 8 6 - 7 7 9 8 0 1 0 0 0 5 1 - 7

( )Y.M.A. Hashash et al.Tunnelling and Underground Space Technology 16 2001 247293248

Preface

This paper was developed as part of the activities of( )the International Tunnelling Association ITA Working

Group No 2: Research. The paper provides a state-of-the-art review of the design and analysis of tunnelssubject to earthquake shaking with particular focus onpractice in the United States of America. The Authorswish to acknowledge the important contribution ofWorking Group 2 members including Mr. Yann Leblais,Animateur, Yoshihiro Hiro Takano, Vice-Animateur,Barry New, Member, Henk J.C. Oud and Andres Assis,Tutor and Former Tutor, respectively, as well as theITA Executive Council for their review and approval ofthis document.

1. Introduction

Underground structures have features that maketheir seismic behavior distinct from most surface struc-

.tures, most notably 1 their complete enclosure in soil . .or rock, and 2 their significant length i.e. tunnels .

The design of underground facilities to withstandseismic loading thus, has aspects that are very differentfrom the seismic design of surface structures.

This report focuses on relatively large undergroundfacilities commonly used in urban areas. This includeslarge-diameter tunnels, cut-and-cover structures and

.portal structures Fig. 1 . This report does not discusspipelines or sewer lines, nor does it specifically discussissues related to deep chambers such as hydropowerplants, nuclear waste repositories, mine chambers, andprotective structures, though many of the design meth-ods and analyses described are applicable to the designof these deep chambers.

Large-diameter tunnels are linear undergroundstructures in which the length is much larger than thecross-sectional dimension. These structures can begrouped into three broad categories, each having dis-

.tinct design features and construction methods: 1 .bored or mined tunnels; 2 cut-and-cover tunnels; and

. .3 immersed tube tunnels Power et al., 1996 . Thesetunnels are commonly used for metro structures, high-way tunnels, and large water and sewage transportationducts.

Bored or mined tunnels are unique because they areconstructed without significantly affecting the soil orrock above the excavation. Tunnels excavated using

.tunnel-boring machines TBMs are usually circular;other tunnels maybe rectangular or horseshoe in shape.Situations where boring or mining may be preferable to

.cut-and-cover excavation include 1 significant excava- .tion depths, and 2 the existence of overlying struc-

tures.

.Fig. 1. Cross sections of tunnels after Power et al., 1996 .

Cut-and-cover structures are those in which an openexcavation is made, the structure is constructed, and fillis placed over the finished structure. This method istypically used for tunnels with rectangular cross-sec-

tions and only for relatively shallow tunnels 15 m of.overburden . Examples of these structures include sub-

way stations, portal structures and highway tunnels.Immersed tube tunnels are sometimes employed to

traverse a body of water. This method involves con-structing sections of the structure in a dry dock, thenmoving these sections, sinking them into position andballasting or anchoring the tubes in place.

This report is a synthesis of the current state ofknowledge in the area of seismic design and analysisfor underground structures. The report updates the

.work prepared by St. John and Zahrah 1987 , whichappeared in Tunneling Underground Space Technol. Thereport focuses on methods of analysis of undergroundstructures subjected to seismic motion due toearthquake activity, and provides examples of perfor-mance and damage to underground structures duringrecent major earthquakes. The report describes theoverall philosophy used in the design of undergroundstructures, and introduces basic concepts of seismichazard analysis and methods used in developing designearthquake motion parameters.

The report describes how ground deformations areestimated and how they are transmitted to an under-

( )Y.M.A. Hashash et al.Tunnelling and Underground Space Technology 16 2001 247293 249

ground structure, presenting methods used in the com-putation of strains, forces and moment in the structure.The report provides examples of the application ofthese methods for underground structures in Los Ange-les, Boston, and the San Francisco Bay Area.

This report does not cover issues related to staticdesign, although static design provisions for under-ground structures often provide sufficient seismic resis-tance under low levels of ground shaking. The reportdoes not discuss structural design details and reinforce-ment requirements in concrete or steel linings forunderground structures. The report briefly describesissues related to seismic design associated with groundfailure such as liquefaction, slope stability and faultcrossings, but does not provide a thorough treatment ofthese subjects. The reader is encouraged to reviewother literature on these topics to ensure that relevantdesign issues are adequately addressed.

2. Performance of underground facilities during seismicevents

Several studies have documented earthquake da- .mage to underground facilities. ASCE 1974 describes

the damage in the Los Angeles area as a result of the .1971 San Fernando Earthquake. JSCE 1988 describes

the performance of several underground structures,including an immersed tube tunnel during shaking in

. .Japan. Duke and Leeds 1959 , Stevens 1977 , Dowd- . .ing and Rozen 1978 , Owen and Scholl 1981 , Sharma

. .and Judd 1991 , Power et al. 1998 and Kaneshiro et .al. 2000 , all present summaries of case histories of

damage to underground facilities. Owen and Scholl .1981 have updated Dowding and Rozens work with

.127 case histories. Sharma and Judd 1991 generatedan extensive database of seismic damage to under-ground structures using 192 case histories. Power et al. .1998 provide a further update with 217 case histories.The following general observations can be made re-garding the seismic performance of underground struc-tures:

1. Underground structures suffer appreciably lessdamage than surface structures.

2. Reported damage decreases with increasing over-burden depth. Deep tunnels seem to be safer andless vulnerable to earthquake shaking than areshallow tunnels.

3. Underground facilities constructed in soils can beexpected to suffer more damage compared toopenings constructed in competent rock.

4. Lined and grouted tunnels are safer than unlinedtunnels in rock. Shaking damage can be reducedby stabilizing the ground around the tunnel and

by improving the contact between the lining andthe surrounding ground through grouting.

5. Tunnels are more stable under a symmetric load,which improves ground-lining interaction. Improv-ing the tunnel lining by placing thicker and stiffersections without stabilizing surrounding poorground may result in excess seismic forces in thelining. Backfilling with non-cyclically mobile mate-rial and rock-stabilizing measures may improvethe safety and stability of shallow tunnels.

6. Damage may be related to peak ground accelera-tion and velocity based on the magnitude andepicentral distance of the affected earthquake.

7. Duration of strong-motion shaking duringearthquakes is of utmost importance because itmay cause fatigue failure and therefore, largedeformations.

8. High frequency motions may explain the localspalling of rock or concrete along planes of weak-ness. These frequencies, which rapidly attenuatewith distance, may be expected mainly at smalldistances from the causative fault.

9. Ground motion may be amplified upon incidencewith a tunnel if wavelengths are between one andfour times the tunnel diameter.

10. Damage at and near tunnel portals may be sig-nificant due to slope instability.

The following is a brief discussion of recent casehistories of seismic performance of underground struc-tures.

2.1. Underground structures in the United States

( )2.1.1. Bay Area rapid transit BART system, SanFrancisco, CA, USA

The BART system was one of the first undergroundfacilities to be designed with considerations for seismic

.loading Kuesel, 1969 . On the San Francisco side, thesystem consists of underground stations and tunnels infill and soft Bay Mud deposits, and it is connected toOakland via the transbay-immersed tube tunnel.

During the 1989 Loma Prieta Earthquake, the BARTfacilities sustained no damage and, in fact, operated ona 24-h basis after the earthquake. This is primarilybecause the system was designed under stringentseismic design considerations. Special seismic joints .Bickel and Tanner, 1982 were designed to accommo-date differential movements at ventilation buildings.The system had been designed to support earth andwater loads while maintaining watertight connectionsand not exceeding allowable differential movements.No damage was observed at these flexible joints, thoughit is not exactly known how far the joints moved during

.the earthquake PB, 1991 .


.Fig. 2. Section sketch of damage to Daikai subway station Iida et al., 1996 .

2.1.2. Alameda Tubes, Oakland-Alameda, CA, USAThe Alameda Tubes are a pair of immersed-tube

tunnels that connect Alameda Island to Oakland in theSan Francisco Bay Area. These were some of theearliest immersed tube tunnels built in 1927 and 1963without seismic design considerations. During the LomaPrieta Earthquake, the ventilation buildings experi-enced some structural cracking. Limited water leakageinto the tunnels was also observed, as was liquefactionof loose deposits above the tube at the Alameda portal.Peak horizontal ground accelerations measured in the

.area ranged between 0.1 and 0.25 g EERI, 1990 . Thetunnels, however, are prone to floatation due to poten-

tial liquefaction of the backfill Schmidt and Hashash,.1998 .

2.1.3. L.A. Metro, Los Angeles, CA, USAThe Los Angeles Metro is being constructed in sev-

eral phases, some of which were operational during the1994 Northridge Earthquake. The concrete lining ofthe bored tunnels remained intact after the earthquake.While there was damage to water pipelines, highwaybridges and buildings, the earthquake caused no da-mage to the Metro system. Peak horizontal groundaccelerations measured near the tunnels rangedbetween 0.1 and 0.25 g, with vertical ground accelera-

.tions typically two-thirds as large EERI, 1995 .

2.2. Underground structures in Kobe, Japan

The 1995 Hyogoken-Nambu Earthquake caused amajor collapse of the Daikai subway station in Kobe,

.Japan Nakamura et al., 1996 . The station design in1962 did not include specific seismic provisions. Itrepresents the first modern underground structure tofail during a seismic event. Fig. 2 shows the collapseexperienced by the center columns of the station, whichwas accompanied by the collapse of the ceiling slab andthe settlement of the soil cover by more than 2.5 m.

During the earthquake, transverse walls at the ends

of the station and at areas where the station changedwidth acted as shear walls in resisting collapse of the

.structure Iida et al., 1996 . These walls suffered sig-nificant cracking, but the interior columns in theseregions did not suffer as much damage under thehorizontal shaking. In regions with no transverse walls,collapse of the center columns caused the ceiling slabto kink and cracks 150250-mm wide appeared in thelongitudinal direction. There was also significant sepa-ration at some construction joints, and correspondingwater leakage through cracks. Few cracks, if any, wereobserved in the base slab.

Center columns that were designed with very light .transverse shear reinforcement relative to the main

.bending reinforcement suffered damage ranging fromcracking to complete collapse. Center columns withzigzag reinforcement in addition to the hoop steel, as inFig. 3, did not buckle as much as those without thisreinforcement.

.According to Iida et al. 1996 , it is likely that therelative displacement between the base and ceilinglevels due to subsoil movement created the destructive

Fig. 3. Reinforcing steel arrangement in center columns Iida et al.,.1996 .


horizontal force. This type of movement may haveminor effect in a small structure, but in a large onesuch as a subway station it can be significant. Thenon-linear behavior of the subsoil profile may also besignificant. It is further hypothesized that the thicknessof the overburden soil affected the extent of damagebetween sections of the station by adding inertial forceto the structure. Others attribute the failure to highlevels of vertical acceleration.

.EQE 1995 made further observations about DaikaiStation: Excessive deflection of the roof slab would

.normally be resisted by: 1 diaphragm action of the .slab, supported by the end walls of the station; and 2

passive earth pressure of the surrounding soils,mobilized as the tube racks. Diaphragm action was lessthan anticipated, however, due to the length of the

station. The method of construction cut-and-cover,involving a sheet pile wall supported excavation withnarrow clearance between the sheet pile wall and the

.tube wall made compaction of backfill difficult toimpossible, resulting in the tubes inability to mobilizepassive earth pressures. In effect, the tube behavedalmost as a freestanding structure with little or no extrasupport from passive earth pressure. However, it is notcertain that good compaction would have prevented thestructural failure of the column. Shear failure of sup-porting columns caused similar damage to the Shinkan-

.sen Tunnel through Rokko Mountain NCEER, 1995 .Several key elements may have helped in limiting the

damage to the station structure and possibly preventedcomplete collapse. Transverse walls at the ends of thestation and at areas where the station changed widthprovided resistance to dynamic forces in the horizontaldirection. Center columns with relatively heavy trans-

.verse shear reinforcement suffered less damage andhelped to maintain the integrity of the structure. The

Fig. 4. Slope Failure at Tunnel Portal, Chi-Chi Earthquake, CentralTaiwan.

Fig. 5. Bolu Tunnel, re-mining of Bench Pilot Tunnels, showingtypical floor heave and buckled steel rib and shotcrete shell Menkiti,

.2001 .

fact that the structure was underground instead ofbeing a surface structure may have reduced the amountof related damage.

.A number of large diameter 2.02.4 m concretesewer pipes suffered longitudinal cracking during theKobe Earthquake, indicating racking andor compres-

.sive failures in the cross-sections Tohda, 1996 . Thesecracks were observed in pipelines constructed by both

the jacking method and open-excavation cut-and-.cover methods. Once cracked, the pipes behaved as

four-hinged arches and allowed significant water leak-age.

2.3. Underground structures in Taiwan

Several highway tunnels were located within the zoneheavily affected by the September 21, 1999 Chi Chi

.earthquake M 7.3 in central Taiwan. These areLlarge horseshoe shaped tunnels in rock. All the tunnelsinspected by the first author were intact without anyvisible signs of damage. The main damage occurred attunnel portals because of slope instability as illustratedin Fig. 4. Minor cracking and spalling was observed insome tunnel lining. One tunnel passing through theChelungpu fault was shut down because of a 4-m fault

.movement Ueng et al., 2001 . No damage was reportedin the Taipei subway, which is located over 100 kmfrom the ruptured fault zone.

2.4. Bolu Tunnel, Turkey

The twin tunnels are part of a 1.5 billion dollarproject that aims at improving transportation in themountainous terrain to the west of Bolu between Istan-

.bul and Ankara http:geoinfo.usc.edugees . Eachtunnel was constructed using the New Austrian Tunnel-

.ing Method NATM where continuous monitoring ofprimary liner convergence is performed and support


elements are added until a stable system is established.The tunnel has an excavated arch section 15 m tall by16 m wide. Construction has been unusually challeng-ing because the alignment crosses several minor faultsparallel to the North Anatolian Fault. The August 17,1999 Koceali earthquake was reported to have hadminimal impact on the Bolu tunnel. The closure rate ofone monitoring station was reported to have temporar-ily increased for a period of approximately 1 week, thenbecame stable again. Additionally, several hairlinecracks, which had previously been observed in the finallining, were being continuously monitored for additio-nal movement and showed no movement due to theearthquake. The November 12, 1999 earthquake causedthe collapse of both tunnels 300 m from their easternportal. At the time of the earthquake, a 800-m sectionhad been excavated, and a 300-m section of unrein-forced concrete lining had been completed. The col-lapse took place in clay gauge material in the unfin-ished section of the tunnel. The section was covered

.with shotcrete sprayed concrete and had bolt anchors.Fig. 5 shows a section of the collapsed tunnel after ithas been re-excavated. Several mechanisms have beenproposed for explaining the collapse of the tunnel.These mechanisms include strong ground motion, dis-placement across the gauge material, and landslide.

.ORourke et al. 2001 present a detailed description ofthe tunnel performance.

2.5. Summary of seismic performance of undergroundstructures

The Daikai subway station collapse was the firstcollapse of an urban underground structure due toearthquake forces, rather than ground instability. Un-derground structures in the US have experiencedlimited damage during the Loma Prieta and Northridgeearthquakes, but the shaking levels have been muchlower than the maximum anticipated events. Greaterlevels of damage can be expected during these maxi-mum events. Station collapse and anticipated strongmotions in major US urban areas raise great concernsregarding the performance of underground structures.It is therefore necessary to explicitly account for seismicloading in the design of underground structures.

The data show that in general, damage to tunnels isgreatly reduced with increased overburden, and da-mage is greater in soils than in competent rock. Da-

.mage to pipelines buckling, flotation was greater thanto rail or highway tunnels in both Kobe and Northridge.The major reason for this difference seems to havebeen the greater thickness of the lining of transporta-tion tunnels. Experience has further shown that cut-and-cover tunnels are more vulnerable to earthquakedamage than are circular bored tunnels.

3. Engineering approach to seismic analysis and design

Earthquake effects on underground structures can .be grouped into two categories: 1 ground shaking;

.and 2 ground failure such as liquefaction, fault dis-placement, and slope instability. Ground shaking, whichis the primary focus of this report, refers to the defor-mation of the ground produced by seismic waves propa-gating through the earths crust. The major factors

.influencing shaking damage include: 1 the shape, .dimensions and depth of the structure; 2 the proper-

.ties of the surrounding soil or rock; 3 the properties .of the structure; and 4 the severity of the ground

shaking Dowding and Rozen, 1978; St. John and.Zahrah, 1987 .

Seismic design of underground structures is uniquein several ways. For most underground structures, theinertia of the surrounding soil is large relative to theinertia of the structure. Measurements made by Oka-

.moto et al. 1973 of the seismic response of animmersed tube tunnel during several earthquakes showthat the response of a tunnel is dominated by thesurrounding ground response and not the inertialproperties of the tunnel structure itself. The focus ofunderground seismic design, therefore, is on the free-field deformation of the ground and its interaction withthe structure. The emphasis on displacement is in starkcontrast to the design of surface structures, whichfocuses on inertial effects of the structure itself. Thisled to the development of design methods such as theSeismic Deformation Method that explicitly considersthe seismic deformation of the ground. For example,

.Kawashima, 1999 presents a review on the seismicbehavior and design of underground structures in softground with an emphasis on the development of theSeismic Deformation Method.

The behavior of a tunnel is sometimes approximatedto that of an elastic beam subject to deformationsimposed by the surrounding ground. Three types of

.deformations Owen and Scholl, 1981 express the re-sponse of underground structures to seismic motions: . . .1 axial compression and extension Fig. 6a,b ; 2

. .longitudinal bending Fig. 6c,d ; and 3 ovalingrack- .ing Fig. 6e,f . Axial deformations in tunnels are gener-

ated by the components of seismic waves that producemotions parallel to the axis of the tunnel and causealternating compression and tension. Bending deforma-tions are caused by the components of seismic wavesproducing particle motions perpendicular to the longi-tudinal axis. Design considerations for axial and bend-ing deformations are generally in the direction along

.the tunnel axis Wang, 1993 .Ovaling or racking deformations in a tunnel struc-

ture develop when shear waves propagate normal or


.Fig. 6. Deformation modes of tunnels due to seismic waves after Owen and Scholl, 1981 .

nearly normal to the tunnel axis, resulting in a distor-tion of the cross-sectional shape of the tunnel lining.Design considerations for this type of deformation arein the transverse direction. The general behavior of thelining may be simulated as a buried structure subject toground deformations under a two-dimensional plane-strain condition.

Diagonally propagating waves subject different partsof the structure to out-of-phase displacements Fig.

.6d , resulting in a longitudinal compressionrarefac-tion wave traveling along the structure. In general,larger displacement amplitudes are associated withlonger wavelengths, while maximum curvatures areproduced by shorter wavelengths with relatively small

.displacement amplitudes Kuesel, 1969 .The assessment of underground structure seismic

response, therefore, requires an understanding of theanticipated ground shaking as well as an evaluation of

the response of the ground and the structure to suchshaking. Table 1 summarizes a systematic approach forevaluating the seismic response of underground struc-tures. This approach consists of three major steps:

1. Definition of the seismic environment and develop-ment of the seismic parameters for analysis.

2. Evaluation of ground response to shaking, whichincludes ground failure and ground deformations.

3. Assessment of structure behavior due to seismic .shaking including a development of seismic de-

.sign loading criteria, b underground structure re- .sponse to ground deformations, and c special

seismic design issues.

Steps 1 and 2 are described in Sections 4 and 5,respectively. Sections 68 provide the details of Steps3a, 3b and 3c.


.Fig. 7. Deterministic seismic hazard analysis procedure after Reiter, 1990 .

4. Definition of seismic environment

The goal of earthquake-resistant design for under-ground structures is to develop a facility that canwithstand a given level of seismic motion with damagenot exceeding a pre-defined acceptable level. The de-sign level of shaking is typically defined by a designground motion, which is characterized by the ampli-tudes and characteristics of expected ground motions

.and their expected return frequency Kramer, 1996 . Aseismic hazard analysis is used to define the level of

.shaking and the design earthquake s for an under-ground facility.

A seismic hazard analysis typically characterizes thepotential for strong ground motions by examining theextent of active faulting in a region, the potential forfault motion, and the frequency with which the faultsrelease stored energy. This examination may be dif-

.ficult in some regions e.g. Eastern USA where fault-ing is not readily detectable. There are two methods of

.analysis: a the deterministic seismic hazard analysis . .DSHA ; and b the probabilistic seismic hazard analy-

.sis PSHA . A deterministic seismic hazard analysisdevelops one or more earthquake motions for a site,for which the designers then design and evaluate theunderground structure. The more recent probabilisticseismic hazard analysis, which explicitly quantifies theuncertainties in the analysis, develops a range of ex-pected ground motions and their probabilities of occur-rence. These probabilities can then be used to de-termine the level of seismic protection in a design.

( )4.1. Deterministic seismic hazard analysis DSHA

A deterministic seismic hazard analysis involves thedevelopment of a particular seismic scenario to sum-

marize the ground motion hazard at a site. This sce-nario requires the postulated occurrence of a particu-lar size of earthquake at a particular location. Reiter . 1990 outlined the following four-step process see Fig..7 :

1. Identification and characterization of all earth-quake sources capable of producing significantground motion at the site, including definition ofthe geometry and earthquake potential of each.The most obvious feature delineating a seismiczone is typically the presence of faulting. Reiter .1990 generated a comprehensive list of featuresthat may suggest faulting in a given region. How-ever, the mere presence of a fault does not neces-sarily signify a potential earthquake hazard thefault must be active to present a risk. There hasbeen considerable disagreement over the criteriafor declaring a fault active or inactive. Rather thanusing the term active, the US Nuclear Regulatory

.Commission Code of Federal Regulations, 1978coined the term capable fault to indicate a faultthat has shown activity within the past35 000500 000 years. For non-nuclear civil infras-tructure, shorter timeframes would be used.

2. Selection of a source-to-site distance parameter foreach source, typically the shortest epicentralhypo-central distance or the distance to the closest rup-tured portion of the fault. Closest distance to rup-tured fault is more meaningful than epicentral dis-tance especially for large earthquakes where theruptured fault extends over distances exceeding 50km.

3. Selection of a controlling earthquake i.e. thatwhich produces the strongest shaking level at the

.site , generally expressed in terms of a ground


motion parameter at the site. Attenuation relation-ships are typically used to determine these site-specific parameters from data recorded at nearbylocations. Several studies have attempted to corre-late earthquake magnitudes, most commonly mo-ment magnitudes, with observed fault deformationcharacteristics, such as rupture length and area,and have found a strong correlation. However, theunavailability of fault displacement measurementsover the entire rupture surface severely limits ourability to measure these characteristics. Instead,researchers have tried to correlate the maximumsurface displacement with magnitude to varyingresults. Empirically based relationships, such as

.those developed by Wells and Coppersmith 1994 ,can be utilized to estimate these correlations. An-other, more basic way to evaluate the potential forseismic activity in a region is through examinationof historical records. These records allow engineersto outline and track active faults and their releaseof seismic potential energy. The evaluation of fore-and aftershocks can also help delineate seismic

.zones Kramer, 1996 . In addition to the examina-tion of historical records, a study of geologic recordof past seismic activities called paleo-seismologycan be used to evaluate the occurrence and size of

earthquakes in the region. Geomorphic surface.landform and trench studies may reveal the num-

ber of past seismic events, slip per event, andtiming of the events at a specific fault. In some

14 .cases, radiocarbon C dating of carbonized roots,animal bone fossils or soil horizons near the fea-tures of paleoseismic evidence can be utilized toapproximate ages of the events.

4. Formal definition of the seismic hazard at the sitein terms of the peak acceleration, velocity and

displacement, response spectrum ordinates, andground motion time history of the maximum credi-ble earthquake. Design fault displacements shouldalso be defined, if applicable.

A DSHA provides a straightforward framework forthe evaluation of worst-case scenarios at a site. How-ever, it provides no information about the likelihood orfrequency of occurrence of the controlling earthquake.If such information is required, a probabilistic ap-proach must be undertaken to better quantify theseismic hazard.

( )4.2. Probabilistic seismic hazard analysis PSHA

A probabilistic seismic hazard analysis provides aframework in which uncertainties in the size, location,and recurrence rate of earthquakes can be identified,quantified, and combined in a rational manner. Suchan analysis provides designers with a more completedescription of the seismic hazard at a site, where varia-tions in ground motion characteristics can be explicitlyconsidered. For this type of analysis, future earthquakeevents are assumed spatially and temporally indepen-

.dent. Reiter 1990 outlined the four major steps in- .volved in PSHA see Fig. 8 :

1. Identification and characterization of earthquakesources, including the probability distribution ofpotential rupture locations within the source zone.These distributions are then combined with thesource geometry to obtain the probability distribu-tion of source-to-site distances. In many regionsthroughout the world, including the USA, specificactive fault zones often cannot be identified. In

.Fig. 8. Probabilistic seismic hazard analysis procedure after Reiter, 1990 .


these cases, seismic history and geological con-siderations become critical for hazard analyses.

2. Characterization of the seismicity or temporal dis-tribution of earthquake recurrence. Informationobtained from historical data and paleoseismologi-cal studies can help to develop a recurrence rela-tionship that describes the average rate at which anearthquake of certain size will be exceeded.

3. Determination of the ground motion produced atthe site by any size earthquake occurring at anysource zone using attenuation relationships. Theuncertainty inherent in the predictive relationshipis also considered.

4. Combination of these uncertainties to obtain theprobability that a given ground motion parameterwill be exceeded during a given time period.

The probabilistic approach incorporates the uncer-tainties in source-to-site distance, magnitude, rate ofrecurrence and the variation of ground motion charac-teristics into the analyses. In areas where no activefaults can be readily identified it may be necessary torely on a purely statistical analysis of historicearthquakes in the region. The details of this proce-dure are beyond the scope of this report.

4.3. Design earthquakes criteria

Once the seismic hazard at the site is characterized,the level of design earthquake or seismicity has to bedefined. For example, in PSHA, the designer mustselect the probability of exceedance for the sets ofground motion parameters. Current seismic design phi-

losophy for many critical facilities requires dual two-.level design criteria, with a higher design level

earthquake aimed at life safety and a lower design levelearthquake intended for economic risk exposure. Thetwo design levels are commonly defined as maximum

.design earthquake or safety evaluation earthquakeand operational design earthquake or function eval-

.uation earthquake , and have been employed in manyrecent transportation tunnel projects, including the LosAngeles Metro, Taipei Metro, Seattle Metro, and Bos-ton Central ArteryThird Harbor Tunnels.

4.3.1. Maximum Design Earthquake .The Maximum Design Earthquake MDE is defined

in a DSHA as the maximum level of shaking that canbe experienced at the site. In a PSHA, the MDE isdefined as an event with a small probability of ex-

.ceedance during the life of the facility e.g. 35% . TheMDE design goal is that public safety shall be main-tained during and after the design event, meaning thatthe required structural capacity under an MDE loadingmust consider the worst-case combination of live, dead,

and earthquake loads. Recently, some owners e.g. San

.Francisco BART have begun requiring their facilities,identified as lifelines, to remain operational after MDElevel shaking.

4.3.2. Operating Design Earthquake .The Operating Design Earthquake ODE is an

earthquake event that can be reasonably expected tooccur at least once during the design life of the facilitye.g. an event with probability of exceedence between

.40 and 50% . In an ODE analysis, the seismic designloading depends on the structural performance re-quirements of the structural members. Since the ODEdesign goal is that the overall system shall continueoperating during and after an ODE and experiencelittle or no damage, inelastic deformations must bekept to a minimum. The response of the undergroundfacility should therefore remain within the elastic range.

4.4. Ground motion parameters

Once an MDE or ODE is defined, sets of groundmotion parameters are required to characterize thedesign events. The choice of these parameters is re-lated to the type of analysis method used in design. Ata particular point in the ground or on a structure,ground motions can be described by three translationalcomponents and three rotational components, thoughrotational components are typically ignored. A groundmotion component is characterized by a time history ofacceleration, velocity or displacement with three sig-nificant parameters: amplitude; frequency content; andduration of strong ground motion.

4.4.1. Acceleration, elocity, and displacement amplitudesMaximum values of ground motion such as peak

ground acceleration, velocity and displacement arecommonly used in defining the MDE and ODE devel-oped through seismic hazard analysis. However, experi-ence has shown that effective, rather than peak, groundmotion parameters tend to be better indicators ofstructural response, as they are more representative ofthe damage potential of a given ground motion. This isespecially true for large earthquakes. The effectivevalue is sometimes defined as the sustained level ofshaking, and computed as the third or fifth highest

.value of the parameter Nuttli, 1979 . Earthquake da-mage to underground structures has also proven to bebetter correlated with particle velocity and displace-ment than acceleration. Attenuation relationships aregenerally available for estimating peak ground surfaceaccelerations, but are also available for estimating peakvelocities and displacements. Tables 2 and 3 can beused to relate the known peak ground acceleration toestimates of peak ground velocity and displacement,respectively, in the absence of site-specific data.


Table 2Ratios of peak ground velocity to peak ground acceleration at surface

.in rock and soil after Power et al., 1996

.Ratio of peak ground velocity cmsMoment .to peak ground acceleration gmagnitude

.M .Source-to-site distance kmw

020 2050 50100

aRock6.5 66 76 867.5 97 109 978.5 127 140 152

aStiff soil6.5 94 102 1097.5 140 127 1558.5 180 188 193

aSoft soil6.5 140 132 1427.5 208 165 2018.5 269 244 251

a In this table, the sediment types represent the following shearwave velocity ranges: rock 750 ms; stiff soil is 200750 ms; andsoft soil 200 ms. The relationship between peak ground velocityand peak ground acceleration is less certain in soft soils.

4.4.2. Target response spectra and motion time historyThe most common way to express the parameters of

a design ground motion is through acceleration re-sponse spectra, which represents the response of adamped single degree of freedom system to groundmotion. Once a target response spectrum has been

Table 3Ratios of peak ground displacement to peak ground acceleration at

.surface in rock and soil after Power et al., 1996

.Ratio of peak ground displacement cmMoment .to peak ground acceleration gmagnitude

.M .Source-to-site distance kmw

020 2050 50100

aRock6.5 18 23 307.5 43 56 698.5 81 99 119

aStiff soil6.5 35 41 487.5 89 99 1128.5 165 178 191

aSoft soil6.5 71 74 767.5 178 178 1788.5 330 320 305

a In this table, the sediment types represent the following shearwave velocity ranges: rock 750 ms; stiff soil is 200750 ms; andsoft soil 200 ms. The relationship between peak ground velocityand peak ground acceleration is less certain in soft soils.

chosen, one or more ground motion time histories maybe developed that match the design response spectra.These time histories can be either synthetic or basedon actual recordings of earthquakes with similar char-acteristics.

While the response spectrum is a useful tool for the .designer, it should not be used if 1 the soil-structure

.system response is highly non-linear, or 2 the struc-ture is sufficiently long that the motion could varysignificantly in amplitude and phase along its length. In

.these cases, time histories St. John and Zahrah, 1987combined with local site response analysis are typicallymore useful.

4.4.3. Spatial incoherence of ground motionFor many engineering structures, the longest dimen-

sion of the structure is small enough that the groundmotion at one end is virtually the same as that at theother end. However, for long structures such as bridgesor tunnels, different ground motions may be encoun-tered by different parts of the structure and traveling

wave effects must be considered Hwang and Lysmer,.1981 . This spatial incoherence may have a significant

impact on the response of the structure. There are four .major factors that may cause spatial incoherence: 1

. .wave-passage effects; 2 extended source effects; 3ray-path effects caused by inhomogeneities along the

.travel path; and 4 local soil site effects. The reader .should refer to Hwang and Lysmer 1981 for details on

these factors. Recorded ground motions have shownthat spatial coherency decreases with increasing dis-

.tance and frequency Kramer, 1996 . The generation ofground motion time histories with appropriate spatialincoherence is a critical task if the designer is tocompute differential strains and force buildup along atunnel length. The designer will have to work closelywith an engineering seismologist to identify the rele-vant factors contributing to ground motion incoherenceat a specific site and to generate appropriate ground

.motion time histories. Hashash et al. 1998 show howthe use of time histories with spatial incoherence af-fects the estimation of axial force development in atunnel and can lead to significant longitudinal push-pulland other effects.

4.5. Wae propagation and site-specific response analysis

Research has shown that transverse shear wavestransmit the greatest proportion of the earthquakesenergy, and amplitudes in the vertical plane have beentypically estimated to be a half to two-thirds as great asthose in the horizontal plane. However, in recentearthquakes such as Northridge and Kobe, measuredvertical accelerations were equal to and sometimeslarger than horizontal accelerations. Vertical compo-


nent of ground motion has become an important issuein seismic designs.

Ample strong ground motion data are generally notavailable at the depths of concern for undergroundstructures, so the development of design ground mo-tions needs to incorporate depth-dependent attenua-tion effects. Popular analytical procedures use one-dimensional site response techniques, although theseanalyses ignore the effects of all but vertically propa-gating body waves. One method, discussed by Schnabel,

.et al. 1972 , applies a deconvolution procedure to asurface input motion in order to evaluate the motion atdepth. A second method involves applying ground mo-tions at various depths to find the scale factors neces-sary to match the input motion. Both of these proce-dures are repeated for a collection of soil propertiesand ground motions to develop a ground motion spec-

.trum for the site St. John and Zahrah, 1987 . Linear, .equivalent linear SHAKE, Schnabel et al., 1972 or

non-linear Hashash and Park, 2001; Borja et al., 1999,D-MOD, Matasovic and Vucetic, 1995, Cyberquake,

.BRGM, 1998, Desra, Finn et al., 1977 one-dimen-sional wave propagation methods are commonly usedto propagate waves through soft soil deposits. Ground

motions generally decrease with depth e.g. Chang et.al., 1986 . Performing a wave propagation analysis is

needed as the amplitude and period of vibration of theground motion shift as the shear wave passes throughsoft soil deposits. In the absence of more accurate .numerical methods or data, Table 4 can be used todetermine the relationship between ground motion atdepth and that at the ground surface.

5. Evaluation of ground response to shaking

The evaluation of ground response to shaking can be . .divided into two groups: 1 ground failure; and 2

ground shaking and deformation. This report focuseson ground shaking and deformation, which assumesthat the ground does not undergo large permanentdisplacements. A brief overview of issues related toground failure are also presented.

5.1. Ground failure

Ground failure as a result of seismic shaking in-cludes liquefaction, slope instability, and fault displace-ment. Ground failure is particularly prevalent at tunnelportals and in shallow tunnels. Special design consider-ations are required for cases where ground failure isinvolved, and are discussed in Section 8.

5.1.1. LiquefactionLiquefaction is a term associated with a host of

different, but related phenomena. It is used to describe

Table 4Ratios of ground motion at depth to motion at ground surface after

.Power et al., 1996

Tunnel Ratio of ground motiondepth at tunnel depth to .m motion at ground surface

6 1.0615 0.91530 0.830 0.7

the phenomena associated with increase of pore waterpressure and reduction in effective stresses in saturatedcohesionless soils. The rise in pore pressure can resultin generation of sand boils, loss of shear strength,lateral spreading and slope failure. The phenomena aremore prevalent in relatively loose sands and artificialfill deposits.

Tunnels located below the groundwater table in liq- .uefiable deposits can experience a increased lateral

. .pressure, b a loss of lateral passive resistance, c .flotation or sinking in the liquefied soil, d lateral

displacements if the ground experiences lateral spread- .ing, and e permanent settlement and compression

and tension failure after the dissipation of pore pres-sure and consolidation of the soil.

5.1.2. Slope instabilityLandsliding as a result of ground shaking is a com-

mon phenomena. Landsliding across a tunnel can re-sult in concentrated shearing displacements and col-lapse of the cross section. Landslide potential is great-est when a pre-existing landslide mass intersects thetunnel. The hazard of landsliding is greatest in shal-lower parts of a tunnel alignment and at tunnel portals.

At tunnel portals, the primary failure mode tends tobe slope failures. Particular caution must be taken if

the portal also acts as a retaining wall St. John and.Zahrah, 1987 . During the September 21, 1999 Chi Chi

earthquake in Taiwan slope instability at tunnel portalswas very common, e.g. Fig. 4.

5.1.3. Fault displacementAn underground structure may have to be con-

structed across a fault zone as it is not always possibleto avoid crossing active faults. In these situations, theunderground structure must tolerate the expected faultdisplacements, and allow only minor damages. All faultsmust be identified to limit the length of special designsection, and a risk-cost analysis should be run to de-termine if the design should be pursued.

5.2. Ground shaking and deformation

In the absence of ground failure that results in large


permanent deformation, the design focus shifts to thetransient ground deformation induced by seismic wavepassage. The deformation can be quite complex due tothe interaction of seismic waves with surficial soft de-posits and the generation of surface waves. For engi-neering design purposes, these complex deformationmodes are simplified into their primary modes. Under-ground structures can be assumed to undergo threeprimary modes of deformation during seismic shaking: . .1 compressionextension; 2 longitudinal bending;

. .and 3 ovallingracking Fig. 6 . The simplest mode toconsider is that of a compression wave propagatingparallel to the axis of a subsurface excavation. Thatcase is illustrated in the figure, where the wave isshown inducing longitudinal compression and tension.The case of an underground structure subjected to anaxially propagating wave is slightly more complex sincethere will be some interaction between the structureand the ground. This interaction becomes more impor-tant if the ground is soft and shear stress transferbetween the ground and the structure is limited by theinterface shear strength. For the case of a wave propa-gating normal or transverse to the tunnel axis, thestress induces shear deformations of the cross sectioncalled racking or ovaling. In the more general case, thewave may induce curvature in the structure, inducingalternate regions of compression and tension along thetunnel. The beam-like structure of the tunnel liningwill then experience tension and compression on oppo-site sides.

6. Seismic design loading criteria

Design loading criteria for underground structureshas to incorporate the additional loading imposed byground shaking and deformation. Once the groundmotion parameters for the maximum and operationaldesign earthquakes have been determined, load criteriaare developed for the underground structure using theload factor design method. This section presents the

.seismic design loading criteria Wang, 1993 for MDEand ODE.

6.1. Loading criteria for maximum design earthquake,MDE

Given the performance goals of the MDE Section.4.3.1 , the recommended seismic loading combinations

using the load factor design method are as follows:

6.1.1. For cut-and-coer tunnel structures

.UDLE1E2EQ 1

where Urequired structural strength capacity, D

effects due to dead loads of structural components,Leffects due to live loads, E1effects due to verti-cal loads of earth and water, E2effects due to hori-zontal loads of earth and water and EQeffects dueto design earthquake motion.

( )6.1.2. For bored or mined circular tunnel lining

.UDLEXHEQ 2

.where U, D, L and EQ are as defined in Eq. 1 ,EXeffects of static loads due to excavation e.g.

.ORourke, 1984 , and Heffects due to hydrostaticwater pressure.

6.1.3. Comments on loading combinations for MDE

The structure should first be designed with ade-quate strength capacity under static loading condi-tions.

The structure should then be checked in terms ofductility its allowable deformation vs. maximum

.deformation imposed by earthquake as well asstrength when earthquake effects, EQ, are con-sidered. The EQ term for conventional surfacestructure design reflects primarily the inertial effecton the structures. For tunnel structures, theearthquake effect is governed not so much by aforce or stress, but rather by the deformation im-posed by the ground.

In checking the strength capacity, the effects ofearthquake loading should be expressed in terms ofinternal moments and forces, which can be calcu-lated according to the lining deformations imposedby the surrounding ground. If the strength criteria

. .expressed by Eq. 1 or Eq. 2 can be satisfiedbased on elastic structural analysis, no furtherprovisions under the MDE are required. Generally,the strength criteria can easily be met when the

earthquake loading intensity is low i.e. in low.seismic risk areas andor the ground is very stiff.

If the flexural strength of the structure lining, using . .elastic analysis and Eq. 1 or Eq. 2 , is found to be

exceeded e.g. at certain joints of a cut-and-cover.tunnel frame , one of the following two design

procedures should be followed:1. Provide sufficient ductility using appropriate de-

.tailing procedure at the critical locations of thestructure to accommodate the deformations im-posed by the ground in addition to those caused by

. ..other loading effects see Eqs. 1 and 2 . Theintent is to ensure that the structural strength doesnot degrade as a result of inelastic deformationsand the damage can be controlled at an acceptablelevel.

In general, the more ductility that is provided,the more reduction in earthquake forces the EQ


.term can be made in evaluating the requiredstrength, U. As a rule of thumb, the force reduction

factor can be assumed equal to the ductility fac-.tor provided. This reduction factor is similar by

definition to the response modification factor used .in bridge design code AASHTO, 1991 .

Note, however, that since an inelastic sheardeformation may result in strength degradation, itshould always be prevented by providing sufficientshear strengths in structure members, particularlyin the cut-and-cover rectangular frame. The use ofductility factors for shear forces may not be ap-propriate.

2. Re-analyze the structure response by assuming theformation of plastic hinges at the joints that arestrained into inelastic action. Based on the plastic-hinge analysis, a redistribution of moments andinternal forces will result.

If new plastic hinges are developed based on theresults, the analysis is re-run by incorporating the

.new hinges i.e. an iterative procedure until allpotential plastic hinges are properly accounted for.Proper detailing at the hinges is then carried out toprovide adequate ductility. The structural design in

. ..terms of required strength Eqs. 1 and 2 canthen be based on the results from the plastic-hingeanalysis.

As discussed earlier, the overall stability of thestructure during and after the MDE must be main-tained. Realizing that the structures also must have

.sufficient capacity besides the earthquake effectto carry static loads e.g. D, L, E1, E2 and H

.terms , the potential modes of instability due to thedevelopment of plastic hinges or regions of inelas-

.tic deformation should be identified and prevented .Monsees and Merritt, 1991 .

For cut-and-cover tunnel structures, the evaluation .of capacity using Eq. 1 should consider the uncer-

tainties associated with the loads E1 and E2, andtheir worst combination. For mined circular tunnels ..Eq. 2 , similar consideration should be given tothe loads EX and H.

In many cases, the absence of live load, L, maypresent a more critical condition than when a fulllive load is considered. Therefore, a live load equalto zero should also be used in checking the struc-

. .tural strength capacity using Eq. 1 and Eq. 2 .

6.2. Loading criteria for operating design earthquake, ODE

.For the ODE Section 4.3.2 , the seismic designloading combination depends on the performance re-quirements of the structural members. Generally

speaking, if the members are to experience little to no .damage during the lower-level event ODE , the inelas-

tic deformations in the structure members should bekept low. The following loading criteria, based on loadfactor design, are recommended:

6.2.1. For cut-and-coer tunnel structures

. .U1.05D1.3L E1E2 1.3EQ 31

Where D, L, El, E2, EQ and U are as defined in Eq. .1 , 1.05 if extreme loads are assumed for E1 and1E2 with little uncertainty. Otherwise, use 1.3.1

( )6.2.2. For bored or mined circular tunnel lining

. .U1.05D1.3L EXH 1.3EQ 42

where D, L, EX, H, EQ and U are as defined in Eq. .2 , 1.05 if extreme loads are assumed for EX and2H with little uncertainty. Otherwise, use 1.3 for2EX only, as H is usually well defined.

The load factors used in these two equations havebeen the subject of a lot of discussion. The final selec-tion depends on the project-specific performance re-quirements. For example, a factor of 1.3 is used for

. .dead load in the Central Artery I-93 Tunnel I-90Project Central Artery Project Design Criteria, Bech-

.telParsons Brinckerhoff, 1992 .

6.2.3. Comments on loading combinations for ODE

The structure should first be designed with ade-quate strength capacity under static loading condi-tions.

For cut-and-cover tunnel structures, the evaluation .of capacity using Eq. 3 should consider the uncer-

tainties associated with the loads E1 and E2, andtheir worst combination. For mined circular tunnels ..Eq. 4 , similar consideration should be given tothe loads EX and H.

When the extreme loads are used for design, asmaller load factor is recommended to avoid unnec-essary conservatism. Note that an extreme load maybe a maximum load or a minimum load, dependingon the most critical case of the loading combina-

.tions. Use Eq. 4 as an example. For a deep circu-lar tunnel lining, it is very likely that the mostcritical loading condition occurs when the maximumexcavation loading, EX, is combined with the mini-

mum hydrostatic water pressure, H unless EX is.unsymmetrical . For a cut-and-cover tunnel, the

most critical seismic condition may often be foundwhen the maximum lateral earth pressure, E2, iscombined with the minimum vertical earth load,E1. If a very conservative lateral earth pressure


coefficient is assumed in calculating the E2, thesmaller load factor 1.05 should be used.1

. Redistribution of moments e.g. ACI 318, 1999 forcut-and-cover concrete frames is recommended toachieve a more efficient design.

. If the strength criteria expressed by Eq. 3 or Eq. .4 can be satisfied based on elastic structural analy-sis, no further provisions under the ODE are re-quired.

If the flexural strength of the structure, using elastic . .analysis and Eq. 3 or Eq. 4 , is found to be

exceeded, the structure should be checked for its.ductility to ensure that the resulting inelastic de-

formations, if any, are small. If necessary, the struc-ture should be redesigned to ensure the intendedperformance goals during the ODE.

. Zero live load condition i.e. L0 should also be . .evaluated in Eq. 3 and Eq. 4 .

7. Underground structure response to grounddeformations

In this section, the term EQ effects due to design.earthquake introduced in Section 6 is quantified. The

development of the EQ term requires an understand-ing of the deformations induced by seismic waves in theground and the interaction of the underground struc-ture with the ground.

This section describes procedures used to computedeformations and forces corresponding to the three

deformation modes compression-extension, longitudi-.nal bending and ovallingracking presented in Section

5.2. A brief summary of design approaches is providedin Table 6.

7.1. Free field deformation approach

The term free-field deformations describes groundstrains caused by seismic waves in the absence ofstructures or excavations. These deformations ignorethe interaction between the underground structure andthe surrounding ground, but can provide a first-orderestimate of the anticipated deformation of the struc-ture. A designer may choose to impose these deforma-tions directly on the structure. This approach mayoverestimate or underestimate structure deformationsdepending on the rigidity of the structure relative tothe ground.

7.1.1. Closed form elastic solutionsSimplified, closed-form solutions are useful for de-

veloping initial estimates of strains and deformations ina tunnel. These simplified methods assume the seismicwave field to be that of plane waves with the same

amplitudes at all locations along the tunnel, differingonly in their arrival time. Wave scattering and complexthree-dimensional wave propagation, which can lead todifferences in wave amplitudes along the tunnel are

neglected, although ground motion incoherence Sec-.tion 4.4.3 tends to increase the strains and stresses in

the longitudinal direction. Results of analyses based onplane wave assumptions should be interpreted with

.care Power et al., 1996 . . .Newmark 1968 and Kuesel 1969 proposed a sim-

plified method for calculating free-field ground strainscaused by a harmonic wave propagating at a givenangle of incidence in a homogeneous, isotropic, elastic

.medium Fig. 9 . The most critical incidence angleyielding maximum strain, is typically used as a safetymeasure against the uncertainties of earthquake pre-diction. Newmarks approach provides an order of mag-nitude estimate of wave-induced strains while requiringa minimal input, making it useful as both an initial

design tool and a method of design verification Wang,.1993 .

.St. John and Zahrah 1987 used Newmarks ap-proach to develop solutions for free-field axial andcurvature strains due to compression, shear andRayleigh waves. Solutions for all three wave types areshown in Table 5, though S-waves are typically associ-ated with peak particle accelerations and velocities .Power et al., 1996 . The seismic waves causing thestrains are shown in Fig. 10. It is often difficult todetermine which type of wave will dominate a design.Strains produced by Rayleigh waves tend to governonly in shallow structures and at sites far from the

.seismic source Wang, 1993 .Combined axial and curvature deformations can be

obtained by treating the tunnel as an elastic beam. ab.Using beam theory, total free-field axial strains,

are found by combining the longitudinal strains gener-ated by axial and bending deformations Power et al.,

.1996 :

V aP Pab 2 2 cos r sincos 2C .5CP Pfor Pwaves

V aS Sab 3 sincos r cos 2C .6CS Sfor Swaves

V aR Rab 2 2 cos r sincos 2C C .R 7R .for Rayleighwaves compressional component


.Fig. 9. Simple harmonic wave and tunnel after Wang, 1993 .

Where:

r : radius of circular tunnel or half height of a rectan-gular tunnel

a : peak particle acceleration associated with P-wavePa : peak particle acceleration associated with S-waveSa : peak particle acceleration associated with RayleighR

wave: angle of incidence of wave with respect to tunnel

axis : Poissons ratio of tunnel lining materiallV : peak particle velocity associated with P-wavepC : apparent velocity of P-wave propagationpV : peak particle velocity associated with S-wavesC : apparent velocity of S-wave propagationsV : peak particle velocity associated with RayleighR

WaveC : apparent velocity of Rayleigh wave propagationR

As the radius of the tunnel increases, the contribu-tion of curvature deformation to axial strain increases.However, calculations using the free-field equations ofTable 5 indicate that the bending component of strainis, in general, relatively small compared to axial strainsfor tunnels under seismic loading. The cyclic nature ofthe axial strains should also be noted although atunnel lining may crack in tension, this cracking is

usually transient due to the cyclic nature of the inci-dent waves. The reinforcing steel in the lining will closethese cracks at the end of the shaking, provided there

is no permanent ground deformation and the steel has.not yielded . Even unreinforced concrete linings are

considered adequate as long as the cracks are small,uniformly distributed, and do not adversely affect the

.performance of the lining Wang, 1993 .It should be noted that the apparent P- and S-wave

velocities used in these equations may be closer tothose of seismic wave propagation through deep rocksrather than the shallow soil or rock in which a tunnel

may be located based on data from Abrahamson 1985,.1992, 1995 . The apparent S-wave velocities fall in the

range of 24 kms while apparent P-wave velocities .fall in the range of 48 kms Power et al., 1996 .

7.1.2. Oaling deformation of circular tunnelsOvaling deformations develop when waves propagate

perpendicular to the tunnel axis and are therefore,designed for in the transverse direction typically under.two-dimensional, plane-strain conditions . Studies have

suggested that, while ovaling may be caused by wavespropagating horizontally or obliquely, vertically propa-gating shear waves are the predominant form ofearthquake loading that causes these types of deforma-

.tions Wang, 1993 .

()

Y.M

.A.H

ashashetal.

Tunnelling

andU

ndergroundSpace

Technology

162001

247293

264

Table 5 .Strain and curvature due to body and surface waves after St. John and Zahrah, 1987

Wave type Longitudinal strain Normal strain Shear strain Curvature

V V V a1P P P P2 2 2P-wae cos sin sincos sincos l n 2C C C CP P P PV V V a1P P P P

for 0 for 90 for 45 0.385 for 3516lm lm m 2C C 2C CP P P max P

V V V aS S S S2 3S-wae sincos sincos cos K cos l n 2C C C CS S S SV V V aS S S S

for 45 for 45 for 0 K for 0lm n m m m 22C 2C C CS S S S

V V V aRayleigh wae R P R P R P R P2 2 2 cos sin sincos K sincos l n 2Compressional C C C CR R R R

component V V V aR P R P P R P

for 0 for 90 for 45 K 0.385 for 3516lm n m m m 2C C 2C CR R R R

V V aShear R S R P R S 2 sin cos K cos n 2component C C CR R R

V V aR S R S R S for 90 for 0 K for 0n m m m 2C C CR R R

2 .C C 21 p sThe Poissons ratio and dynamic modulus of a soil deposit can be computed from measured P- and S-wave propagation velocities in an elastic medium: or Cm P22 .C C 1p s

. . .2 1 1 12m m m2 2 C ; E C ; and G C , respectively.S m P m S( . .1 1m m


Table 6 .Seismic racking design approaches after Wang, 1993

Approaches Advantages Disadvantages Applicability

Dynamic earth pressure 1. Used with reasonable 1. Lack of rigorous For tunnels with minimalmethods results in the past theoretical basis soil cover thickness

2. Require minimal 2. Resulting in excessiveparameters and racking deformationscomputation error for tunnels with3. Serve as additional significant burialsafety measures 3. Use limited to certainagainst seismic types of groundloading properties

Free-field racking 1. Conservative for 1. Non-conservative for For tunnel structures withdeformation method tunnel structure stiffer tunnel structure more equal stiffness to ground

than ground flexible than ground2. Comparatively easy to 2. Overly conservative forformulate tunnel structures3. Used with reasonable significantly stiffer thanresults in the past ground

3. Less precision withhighly variable groundconditions

Soilstructure interaction 1. Best representation of 1. Requires complex and All conditionsfinite-element analysis soilstructure system time consuming

2. Best accuracy in computer analysisdetermining structure 2. Uncertainty of designresponse seismic input3. Capable of solving parameters may beproblems with several times thecomplicated tunnel uncertainty of thegeometry and ground analysisconditions

Simplified frame analysis 1. Good approximation of 1. Less precision with All conditions except formodel soilstructure interaction highly variable ground compacted subsurface

2. Comparatively easy to ground profilesformulate3. Reasonable accuracyin determiningstructure response

Ground shear distortions can be defined in two ways,as shown in Fig. 11. In the non-perforated ground, themaximum diametric strain is a function of maximumfree-field shear strain only:

d max . . 8d 2

The diametric strain in a perforated ground is fur-ther related to the Poissons ratio of the medium:

d . .2 1 . 9max md

Both of these equations assume the absence of thelining, therefore ignoring tunnelground interaction. Inthe free-field, the perforated ground would yield amuch greater distortion than the non-perforated,

sometimes by a factor of two or three. This provides areasonable distortion criterion for a lining with littlestiffness relative to the surrounding soil, while the

Fig. 10. Seismic waves causing longitudinal axial and bending strains .Power et al., 1996 .


Fig. 11. Free-field shear distortion of perforated and non-perforated .ground, circular shape after Wang, 1993 .

non-perforated deformation equation will be appropri-ate when the lining stiffness is equal to that of themedium. A lining with large relative stiffness shouldexperience distortions even less than those given by Eq. . .8 Wang, 1993 .

7.1.3. Racking deformations of rectangular tunnelsWhen subjected to shear distortions during an

earthquake, a rectangular box structure will undergo .transverse racking deformations Fig. 12 . The racking

deformations can be computed from shear strains inthe soil such as those given in Table 5.

7.1.4. Numerical analysisNumerical analysis may be necessary to estimate the

free-field shear distortions, particularly if the sitestratigraphy is variable. Many computer programs areavailable for such analyses such as 1-D wave propaga-tion programs listed in Section 4.5, as well as FLUSH . .Lysmer et al., 1975 , and LINOS Bardet, 1991 . Mostprograms model the site geology as a horizontally lay-ered system and derive a solution using one-dimen-

.sional wave propagation theory Schnabel et al., 1972 . .Navarro 1992 presents numerical computations for

ground deformations and pressures as a result of body .shear and compression wave as well as surface .Rayleigh and Love waves. The resulting free-fieldshear distortion can then be expressed as a shear straindistribution or shear deformation profile with depth.

7.1.5. Applicability of free field deformation approachThe free-field racking deformation method has been

used on many significant projects, including the San .Francisco BART stations and tunnels Kuesel, 1969

and the Los Angeles Metro Monsees and Merritt,.1991 . Kuesel found that, in most cases, if a structure

can absorb free-field soil distortions elastically, no spe-cial seismic provisions are necessary. Monsees and

.Merritt 1991 further specified that joints strained intoplastic hinges can be allowed under the Maximum

.Design Earthquake MDE , provided no plastic hingecombinations are formed that could lead to a collapsemechanism, as shown in Fig. 13.

The free-field deformation method is a simple andeffective design tool when seismically-induced ground

distortions are small i.e. low shaking intensity, verystiff ground, or the structure is flexible compared to the

.surrounding medium . However, in many cases, espe-cially in soft soils, the method gives overly conservativedesigns because free-field ground distortions in softsoils are generally large. For example, rectangular boxstructures in soft soils are typically designed with stiffconfigurations to resist static loads and are therefore,

.Fig. 12. Typical free-field racking deformation imposed on a buried rectangular frame after Wang, 1993 .


Fig. 13. Structure stability for buried rectangular frames after Wang,.1993 .

less tolerant to racking distortions Hwang and Lysmer,.1981; TARTS, 1989 . Soilstructure interaction effects

have to be included for the design of such structures .Wang, 1993 . A comparison of the free field deforma-tion approach with other methods for seismic rackingdesign is given in Table 6.

7.2. Soil structure interaction approach

The presence of an underground structure modifiesthe free field ground deformations. The following para-graphs describe procedures that model soil structureinteraction.

7.2.1. Closed form elastic solutions for circular tunnels,axial force and moment

In this class of solutions the beam-on-elastic founda- .tion approach is used to model quasi-static soil-struc-

ture interaction effects. The solutions ignore dynamic .inertial interaction effects. Under seismic loading, thecross-section of a tunnel will experience axial bendingand shear strains due to free field axial, curvature, andshear deformations. The maximum structural strains

.are after St. John and Zahrah, 1987 :

The maximum axial strain, caused by a 45 incidentshear wave, Fig. 9, is:

2A / fLLa . 10max 2 4E AE A l c2l c2 /K La

Where

L wavelength of an ideal sinusoidal shear wave ..see Eq. 15

Fig. 14. Induced forces and moments caused by seismic waves Power. .et al., 1996 , a Induced forces and moments caused by waves

.propagating along tunnel axis, b Induced circumferential forces andmoments caused by waves propagating perpendicular to tunnel axis.


K longitudinal spring coefficient of mediumain force per unit deformation per unit length

..of tunnel, see Eq. 14A free-field displacement response amplitude of

.an ideal sinusoidal shear wave see Eqs. 17 ..and 18

A cross-sectional area of tunnel liningcE elastic modulus of the tunnel liningl

.f ultimate friction force per unit length betweentunnel and surrounding soil

The forces and moments in the tunnel lining causedby seismic waves propagating along the tunnel axis areillustrated in Fig. 14a. The maximum frictional forcesthat can be developed between the lining and thesurrounding soils limit the axial strain in the lining.

.This maximum frictional force, Q , can be esti-max fmated as the ultimate frictional force per unit length

.times one-quarter the wave length, as shown in Eq. 10 .Sakurai and Takahashi, 1969 .

The maximum bending strain, caused by a 0 incidentshear wave, is:

22A /Lb . r 11max 4E I 2l c1 /K Lt

Where

I moment of inertia of the tunnel sectioncK transverse spring coefficient of the mediumt

in force per unit deformation per unit length of ..tunnel see Eq. 14

r radius of circular tunnel or half height of arectangular tunnel

Since both the liner and the medium are assumed tobe linear elastic, these strains may be superimposed.

Since earthquake loading is cyclic, both extremes posi-.tive and negative must be evaluated. The maximum

shear force acting on a tunnel cross-section can bewritten as a function of this maximum bending strain:

32E I Al c / 2L

V Mmax max4 /LE I 2l c1 /K LtE I b2 l c max . 12 / /L r

A conservative estimate of the total axial strain andstress is obtained by combining the strains from the

axial and bending forces modified from Power et al.,.1996 :

ab a b . . 13max max

Again, these equations are necessary only for struc-tures built in soft ground, as structures in rock or stiffsoils can be designed using free-field deformations. Itshould be further noted that increasing the structuralstiffness and the strength capacity of the tunnel maynot result in reduced forces the structure mayactually attract more force. Instead, a more flexibleconfiguration with adequate ductile reinforcement or

.flexible joints may be more efficient Wang, 1993 .7.2.1.1. Spring coefficients. Other expressions of maxi-

mum sectional forces exist in the literature SFBART,.1960; Kuribayashi et al., 1974; JSCE, 1975 , with the

major differences involving the maximization of forcesand displacements with respect to wavelength. JSCE .1975 suggests substituting the values of wavelengththat will maximize the stresses back into each respec-tive equation to yield maximum sectional forces. St.

.John and Zahrah 1987 suggest a maximization method .similar to the JSCE 1975 approach, except that the

spring coefficients K and K are considered functionsa tof the incident wavelength:

.16G 1 dm m .K K 14t a L .34m

where G , shear modulus and Poissons ratio ofm mthe medium, ddiameter of circular tunnel or height

.of rectangular structure . .These spring constants represent 1 the ratio of

.pressure between the tunnel and the medium, and 2the reduced displacement of the medium when thetunnel is present. The springs differ from those of aconventional beam analysis on an elastic foundation.Not only must the coefficients be representative of thedynamic modulus of the ground, but the derivation ofthese constants must consider the fact that the seismicloading is alternately positive and negative due to the

.assumed sinusoidal wave Wang, 1993 . When usingthese equations to calculate the forces and momentsfor tunnels located at shallow depths, the soil springresistance values are limited by the depth of cover andlateral passive soil resistance.

7.2.1.2. Idealized sinusoidal free field wae parametersfor use in soilstructure interaction analysis. Matsubara

.et al. 1995 provide a discussion of input wavelengthsfor underground structure design. The incident wave-length of a ground motion may be estimated as:

.LT C 15s


.Fig. 15. Lining response coefficient vs. flexibility ratio, full-slip interface, and circular tunnel Wang, 1993 .

where T is the predominant natural period of a shearwave in the soil deposit, the natural period of the siteitself, or the period at which maximum displacements

.occur Dobry et al., 1976; Power et al., 1996 . .Idriss and Seed 1968 recommend that:

4h .T , h is the thickness of the soil deposit 16CS

if ground motion can be attributed primarily to shearwaves and the medium is assumed to consist of a

uniform soft soil layer overlying a stiff layer St. John.and Zahrah, 1987 .

The ground displacement response amplitude, A,represents the spatial variations of ground motionsalong a horizontal alignment and should be derived bysite-specific subsurface conditions. Generally, the dis-


placement amplitude increases with increasing wave- .length SFBART, 1960 . Assuming a sinusoidal wave

with a displacement amplitude A and a wavelength L,A can be calculated from the following equations:

For free-field axial strains:

V2A s . sincos. 17L CS

For free-field bending strains:

2 a4 A s 3 . cos . 182 CL S

7.2.2. Oaling deformations of circular tunnelsIn early studies of racking deformations, Peck et al.

.1972 , based on earlier work by Burns and Richard . .1964 and Hoeg 1968 , proposed closed-form solu-tions in terms of thrusts, bending moments, and dis-placements under external loading conditions. The re-sponse of a tunnel lining is a function of the compress-ibility and flexibility ratios of the structure, and the

.in-situ overburden pressure h and at-rest coeffi-t .cient of earth pressure K of the soil. To adapt to0

seismic loadings caused by shear waves, the free-fieldshear stress replaces the in-situ overburden pressureand the at-rest coefficient of earth pressure is assigned

.a value of 1 to simulate the field simple shearcondition. The shear stress can be further expressed asa function of shear strain.

The stiffness of a tunnel relative to the surroundingground is quantified by the compressibility and flexibil-

.ity ratios C and F , which are measures of the exten-sional stiffness and the flexural stiffness resistance to

.ovaling , respectively, of the medium relative to the .lining Merritt et al., 1985 :

2 .E 1 rm l .C 19 . .E t 1 12l m m

2 . 3E 1 Rm l .F 20 .6E I 1l m

where E modulus of elasticity of the medium, Im .moment of inertia of the tunnel lining per unit width

for circular lining R, and tradius and thickness ofthe tunnel lining.

Assuming full-slip conditions, without normal sepa-ration and therefore, no tangential shear force, thediametric strain, the maximum thrust, and bending

.moment can be expressed as Wang, 1993 :

d 1 . K F 211 maxd 3

E1 m .T K r 22max 1 max6 .1m

E1 m 2 .M K r 23max 1 max6 .1m

where

.12 1m .K . 241 2 F56m

These forces and moments are illustrated in Fig. 14b.The relationship between the full-slip lining response

.coefficient K and flexibility ratio is shown in Fig. 15.1According to various studies, slip at the interface is

only possible for tunnels in soft soils or cases of severeseismic loading intensity. For most tunnels, the inter-face condition is between full-slip and no-slip, so bothcases should be investigated for critical lining forcesand deformations. However, full-slip assumptions un-der simple shear may cause significant underestimationof the maximum thrust, so it has been recommendedthat the no-slip assumption of complete soil continuity

be made in assessing the lining thrust response Hoeg,.1968; Schwartz and Einstein, 1980 :

Em .T K rK r 25max 2 max 2 max .2 1m

where

. . F 12 12 Cm m1 2 . 12 2m2 .K 1 . 262 . . F 32 12 Cm m

5 2C 8 6 68m m m2

As Fig. 16 shows, seismically-induced thrusts in-crease with decreasing compressibility and flexibilityratios when the Poissons ratio of the surroundingground is less than 0.5. As Poissons ratio approaches

.0.5 i.e. saturated undrained clay , the thrust responseis independent of compressibility because the soil is

.considered incompressible Wang, 1993 .The normalized lining deflection provides an indica-

tion of the importance of the flexibility ratio in lining .response, and is defined as Wang, 1993 :

d 2lining . K F . 271d 3freefield

According to this equation and Fig. 17, a tunnellining will deform less than the free field when the

flexibility ratio is less than one i.e. stiff lining in soft


.soil . As the flexibility ratio increases, the lining de-flects more than the free field and may reach an upperlimit equal to the perforated ground deformations. Thiscondition continues as the flexibility ratio becomes

.infinitely large i.e. perfectly flexible lining . .Penzien and Wu 1998 developed similar closed-form

elastic solutions for thrust, shear, and moment in thetunnel lining due to racking deformations. Penzien .2000 provided an analytical procedure for evaluatingracking deformations of rectangular and circular tun-nels that supplemented the previous publication.

In order to estimate the distortion of the structure, alining-soil racking ratio is defined as:

structure .R . 28 freefield

In the case of circular tunnel, R is the ratio of liningdiametric deflection and free-field diametric deflection.Assuming full slip condition, solutions for thrust, mo-ment, and shear in circular tunnel linings caused bysoil-structure interaction during a seismic event are

.expressed as Penzien, 2000 :

n n .d R d 29lining freefield

12 E Idn l lining . .T cos2 30 /3 2 4 .d 1l6E Idn l lining . .M cos2 31 /2 2 4 .d 1l24E Idn l lining . .V sin2 32 /3 2 4 .d 1l

The lining-soil racking ratio under normal loadingonly is defined as:

.4 1mn .R 33n . 1

.12 E I 56l mn . . 343 2 .d G 1m l

The sign convention for the above force componentsin circular lining is shown in Fig. 18. In the case of noslip condition, the formulations are presented as:

.d Rd 35lining freefield

24E Id l lining . .T cos2 36 /3 2 4 .d 1l

.Fig. 16. Lining thrust response coefficient vs. compressibility ratio, .no-slip interface, and circular tunnel Wang, 1993 .

6E Id l lining . .M cos2 37 /2 2 4 .d 1l24E Id l lining . .V sin2 38 /3 2 4 .d 1l

where

.4 1m .R 39 .1

.24E I 34l m . . 403 2 .d G 1m l


.Fig. 17. Normalized lining deflection vs. flexibility ratio, full slip interface, and circular lining Wang, 1993 .

.The solutions of Penzien 2000 result in values ofthrust and moment that are very close to those of

.Wang 1993 for full-slip condition. However, value ofthrust obtained from Wang is much higher compared

to the value given by Penzien in the case of no slip see.example 3 in Appendix B . This observation was also .noted by Power et al. 1996 . The reason for the

difference is still under investigation.

7.2.3. Racking deformations of rectangular tunnelsShallow transportation tunnels are usually box shaped

cut-and-cover method structures. These tunnels haveseismic characteristics very different from circular tun-nels. A box frame does not transmit static loads asefficiently as a circular lining, so the walls and slabs ofthe cut-and-cover frame need to be thicker, and there-fore stiffer. The design of cut-and-cover structures re-quires careful consideration of soil-structure interac-tion effects because of this increased structural stiff-ness and the potential for larger ground deformationsdue to shallow burial. Seismic ground deformationstend to be greater at shallow depths for two reasons:


.1 the decreased stiffness of the surrounding soils due .to lower overburden pressures; and 2 the site ampli-

fication effect. The soil backfill may also consist ofcompacted material with different properties from thein-situ soil, resulting in a different seismic response .Wang, 1993 .

The structural rigidity of box structures significantlyreduces computed strains, often making it overly con-servative to design these structures based on free-field

.strains Hwang and Lysmer, 1981 . While closed-formsolutions for tunnel-ground interaction problems areavailable for circular tunnels, they are not available forrectangular tunnels because of the highly variable geo-metric characteristics associated with these structures.For ease of design, simple and practical procedureshave been developed to account for dynamic soil-struc-

.ture interaction effects Wang, 1993 .A number of factors contribute to the soil-structure

interaction effect, including the relative stiffnessbetween soil and structure, structure geometry, inputearthquake motions, and tunnel embedment depth.The most important factor is the stiffness in simpleshear of the soil relative to the structure that replaces

.it, the flexibility ratio Wang, 1993 .Consider a rectangular soil element in a soil column

under simple shear condition, as shown in Fig. 19.When subjected to simple shear stress the shear strain,or angular distortion, of the soil element is given by .Wang, 1993 :

. . 41s H Gm

After rearranging this equation, the shear or

Seismic Design and Analysis of Underground Estructures - Hashash, Hook, Schmidt y Yao

Documents

Transcript of Seismic Design and Analysis of Underground Estructures - Hashash, Hook, Schmidt y Yao