This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 137.195.49.125

This content was downloaded on 21/10/2015 at 15:08

Please note that terms and conditions apply.

Seismic assessment of bridge structures isolated by a shape memory alloy/rubber-based

isolation system

View the table of contents for this issue, or go to the journal homepage for more

2011 Smart Mater. Struct. 20 015003

(http://iopscience.iop.org/0964-1726/20/1/015003)

Home Search Collections Journals About Contact us My IOPscience

IOP PUBLISHING SMART MATERIALS AND STRUCTURES

Smart Mater. Struct. 20 (2011) 015003 (12pp) doi:10.1088/0964-1726/20/1/015003

Seismic assessment of bridge structuresisolated by a shape memoryalloy/rubber-based isolation systemOsman E Ozbulut and Stefan Hurlebaus

Zachry Department of Civil Engineering, Texas A&M University,3136 TAMU College Station, TX 77843-3136, USA

E-mail: shurlebaus@civil.tamu.edu

Received 15 June 2010, in final form 2 November 2010Published 2 December 2010Online at stacks.iop.org/SMS/20/015003

AbstractThis paper explores the effectiveness of shape memory alloy (SMA)/rubber-based isolationsystems for seismic protection of bridges against near-field earthquakes by performing asensitivity analysis. The isolation system considered in this study consists of a laminated rubberbearing, which provides lateral flexibility while supplying high vertical load-carrying capacity,and an auxiliary device made of multiple loops of SMA wires. The SMA device offersadditional energy dissipating and re-centering capability. A three-span continuous bridge ismodeled with the SMA/rubber-based (SRB) isolation system. Numerical simulations of thebridge are conducted for various near-field ground motions that are spectrally matched to atarget design spectrum. The normalized forward transformation strength, forwardtransformation displacement and pre-strain level of the SMA device, ambient temperature andthe lateral stiffness of the rubber bearings are selected as parameters of the sensitivity study.The variation of the seismic response of the bridge with the considered parameters is assessed.Also, the performance of the SRB isolation system with optimal design parameters is comparedwith an SMA-based sliding isolation system. The results indicate that the SRB isolation systemcan successfully reduce the seismic response of highway bridges; however, a smart isolationsystem that combines sliding bearings together with an SMA device is more efficient than theSRB isolation system.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

The concept of seismic isolation has been considered anattractive strategy over the past decades to mitigate thedamaging effects of earthquakes on civil structures suchas multi-story buildings, bridges and nuclear power plants.Seismic isolation aims to shift the fundamental frequency ofa structure away from the frequency band of most commonearthquakes in order to reduce seismic loads applied to thestructure and to provide additional damping capacity to thestructure. Among various isolation systems that have beenproposed, rubber isolation systems have been widely studiedand used throughout the world [1]. Laminated rubber bearingshave considerable lateral flexibility, vertical load-carryingcapacity and restoring force capability. The commonly used

rubber isolation systems combine laminated rubber bearingsand some mechanical dampers such as hydraulic dampers,viscous dampers, steel bars or lead-plugs within the bearingitself. A laminated rubber bearing with a lead core, knownas a lead–rubber bearing, is the most popular rubber isolationsystem. Another widespread rubber isolator is a high-dampingrubber bearing which increases the damping of the isolationsystem by incorporating damping in the elastomer itself [2].

Although seismic isolation has become one of the mostpopular solutions for seismic protection of bridge structures,the performance of the isolated bridges against near-fieldearthquakes has been questioned by several researchers inrecent years [3–5]. Near-field earthquakes are characterizedby long period and large velocity pulses in the velocity timehistory. Since the period of these pulses usually coincides

0964-1726/11/015003+12$33.00 © 2011 IOP Publishing Ltd Printed in the UK & the USA1

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

with the period of isolated structures, ground motions withnear-field characteristics amplify the seismic response of theisolation system [6]. In particular, the isolation devicesexperience very large displacements under near-field groundmotions which may cause critical problems such as instabilityin the isolator, pounding as well as unseating of the deck. Forexample, a post-earthquake bridge performance investigationafter the 2008 Wenchuan earthquake revealed that manybridges were heavily damaged in the earthquake affectedregion. It was concluded that the rubber bearings cansuccessfully reduce the seismic hazards to bridges in the area,but further research is recommended to limit the excessivebearing deformations [7].

In recent years, several attempts have been made tocombine smart materials with rubber bearings [8, 9]. Onesuch material is shape memory alloy (SMA) that is a classof metallic alloy. SMAs can recover their original shapeafter experiencing large strains. SMAs owe this uniquecharacteristic to solid–solid phase transformations. Thesephase transformations can be mechanically induced, knownas the superelastic effect, or thermally induced, named asthe shape memory effect. Due to its re-centering and energydissipating capabilities, the superelastic behavior of SMAs hasattracted attention for vibration control of structures over thepast decades [10–13]. Superelastic SMAs are initially in theiraustenitic phase, and transform to martensite phase when astress is applied beyond a critical level. Upon removal ofloading, the material experiences a reverse transformation frommartensite back to austenite, resulting in a unique stress–straincurve. By exploiting the complete shape recovery ability andhysteretic behavior of superelastic SMAs, several researchershave proposed a smart isolation system [14–17].

Wilde et al [18] compared the performance of twoisolation systems that couple a laminated rubber bearing witheither a lead-plug or an SMA device for seismic protection ofelevated highway bridges. Casciati et al [19] developed anisolation device that consists of a sliding system and inclinedCuAlBe SMA bars, and carried out extensive experimentaltests on a prototype of the device. In another study byCasciati et al [20], the performance of the developed SMAisolation device was investigated for a seismically excitedhighway bridge benchmark problem. Choi et al [21] proposedan SMA–rubber bearing composed of a laminated rubberbearing and pre-stressed superelastic NiTi wires wrapping thebearing in the longitudinal direction. Numerical simulationswere conducted on a multi-span continuous bridge in order toevaluate the effectiveness of the proposed smart bearing. Notethat the effect of temperature on the behavior of superelasticSMAs has not been considered in the above studies.

In this study, a sensitivity analysis is conducted in orderto investigate the effectiveness of an SMA/rubber-based (SRB)isolation system for protecting highway bridges against near-field earthquakes. The smart isolation system consists of alaminated rubber bearing that decouples the superstructurefrom the bridge piers and an SMA device that providesadditional energy dissipation and re-centering capacity. First,a neuro-fuzzy model that is capable of simulating thesuperelastic behavior of SMAs at various temperatures and

loading rates is briefly introduced. Then, a three-spancontinuous bridge is modeled together with laminated rubberbearings and an auxiliary SMA device. Nonlinear time-historyanalyses of the isolated bridge are performed for a total ofsix excitation cases. A time-domain method, which spectrallyadjusts time histories of historical ground motions to matcha target spectrum at multiple damping levels, is employedto generate artificial earthquakes that are used for dynamicanalyses. The variation of seismic response of the isolatedbridge with the normalized forward transformation strength ofthe SMA device Fo, the forward transformation displacementof the SMA device uy , the pre-strain level of the SMA wires,the lateral stiffness of the laminated rubber bearings kb andenvironmental temperature changes is investigated. The bridgeresponse quantities evaluated in the sensitivity analysis includepeak values of deck drift, deck acceleration, and normalizedbase shear.

2. A temperature- and rate-dependent model ofsuperelastic SMAs

In order to conduct numerical simulations of bridges isolatedby the SRB isolation systems, an effective model thatcharacterizes the hysteretic behavior of superelastic SMAsis needed. The mechanical response of SMAs is highlydependent on temperature and loading rate [22, 23]. WhenSMAs are used as an isolation system component for seismicprotection of bridges, they will experience both temperaturechanges and dynamic loads. Therefore, it is essential toconsider the degree to which the behavior of SMAs is affectedby variations of loading rate and temperature.

Figure 1 presents the results of the uniaxial tensile testsconducted on NiTi shape memory alloy wires using an MTS(Material Testing System) servo-hydraulic load frame. TheSMA wire has a diameter of 1.5 mm and is obtained fromSAES Smart Materials. The alloy chemical composition is55.8% nickel by weight and the balance titanium. The austenitestart and finish temperatures are specified by the manufactureras As = −10 ◦C and Af = 5 ◦C, respectively. The testconditions covered a loading frequency range of 0.05–2 Hz anda temperature range of 0–40 ◦C. The results of the tests that areconducted at different strain amplitudes, different temperaturesand different loading frequencies are given in each subplotof figure 1. It can be seen that the loading frequency andtemperature considerably affect the behavior of superelasticSMAs. In particular, an increase in either temperature orloading frequency shifts the hysteresis loop upward. Also,the area of the hysteresis loop, which indicates the energydissipation of superelastic SMAs, narrows with increasingtemperature or loading frequency. Note that the effect of thetemperature is more pronounced as compared to the influenceof loading frequency.

This study employs a neuro-fuzzy model to capturethe temperature- and rate-dependent behavior of NiTi shapememory alloys. A fuzzy inference system (FIS) is a simplescheme that maps an input space to an output space usingfuzzy logic. Here, an FIS which employs strain, strain rate andtemperature as input variables and predicts the stress as single

2

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 1. Experimental stress–strain curves at various loading conditions.

Figure 2. Stress–strain curves at various conditions for experimental results and the fuzzy model.

output is created. An adaptive neuro-fuzzy inference system(ANFIS) which employs neural network strategies to developa fuzzy model whose parameters cannot be predeterminedby the user’s knowledge is used to tune the parameters ofthe initial FIS. A thorough description of the model can befound in the work by Ozbulut and Hurlebaus [24]. Figure 2illustrates the stress–strain curves for experimental resultsand the prediction of the fuzzy model for different loadingconditions. It can be seen that the developed fuzzy modelsatisfactorily reproduces the hysteresis loops of superelasticSMAs at different temperatures and loading frequencies.

3. Sensitivity analysis

3.1. Modeling of an isolated bridge

A three-span continuous bridge is selected for the sensitivityanalysis [25]. The deck of the bridge has a mass of 771.1 ×103 kg, and the mass of each pier is 39.3 × 103 kg. Thebridge has a total length of 90 m, and each pier is 8 m tall.The moment of inertia of piers and the Young’s modulusof elasticity are given as 0.64 m4 and 20.67 × 109 N m−2,respectively. The fundamental period of the non-isolatedbridge is 0.45 s. As shown in figure 3, the isolated bridgeis modeled as a two-degrees-of-freedom system with the SRBisolation system which consists of a laminated rubber bearingand an SMA device. Since the isolation systems installed at theabutment and pier have similar characteristics and, therefore,the seismic response of the bridge at the abutment and pierhave the same trend, only an internal span is modeled. The

equations of motion are given as

m1u1(t) + c1u1(t) + k1u1(t)

−FIS(u1, u1, u2, u2, t) = −m1ug(t)

m2u2(t) + FIS(u1, u1, u2, u2, t) = −m2ug(t)

(1)

where m1, m2 and u1, u2 are the masses and displacements ofpier and deck, respectively, c1 and k1 represent the coefficientof damping and stiffness of piers, and ug is the groundacceleration. FIS denotes the sum of the restoring force of thelaminated rubber bearings and SMA device. Laminated rubberbearings are modeled by linear spring and dashpot elements.The coefficient of damping and stiffness of rubber bearings aredenoted as c2 and k2, respectively, in figure 3. The equivalentdamping ratio of the bearings is selected to be 2%. The fuzzymodel described above is used to compute the instantaneousforce from the SMA elements.

3.2. Selection of input time histories

A variety of methods have been proposed to modify a historicaltime history so that its response spectrum is compatible with agiven target spectrum [26, 27]. One approach that is commonlyused for generating response spectrum compatible groundmotions is to adjust Fourier amplitude spectra in frequencydomain. Although it is a straightforward method and providesa close match to the target spectrum, it also has significantpotential problems such as distorting the energy characteristicsof accelerograms and producing very unrealistic seismicdemands [28]. Hancock et al [29] proposed an alternative

3

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

LaminatedRubber Bearing

SMAdevice

m2

m1

k2

u2

u1

k1

c2

c1

Figure 3. Model of an isolated bridge with an SMA/rubber isolation system.

approach that performs spectral matching in time domainusing wavelets. The method, known as RspMatch2005, cansimultaneously match response spectra at multiple dampingvalues, while preserving the non-stationary character ofthe reference time history. Unlike the spectral matchingin frequency domain, RspMatch2005 does not corrupt thevelocity and displacement time histories, and avoids creatingground motions with unrealistic energy content.

In this study, the program RspMatch2005 is used togenerate spectrum compatible ground motions that are usedin dynamic time-history analyses of the isolated bridge. Atotal of six historical near-field earthquake records are selectedas seed accelerograms to investigate the effectiveness ofthe SRB isolation system under near-field ground motions.Table 1 gives the characteristics of the ground motions suchas magnitude, the closest distance to the fault plane, peakground acceleration and velocity, and significant duration.A response spectrum constructed as per the InternationalBuilding Code [30] for a site in southern California, assumingfirm rock conditions, is selected as target spectrum [31]. Theleft subplot of figure 4 shows the target response spectrum usedin the analysis and response spectra of selected ground motionsfor 5% damping level. The selected seed accelerogramsare adjusted using RspMatch2005 in order to simultaneouslymatch 5%-, 10%- and 25%-damped response spectra. Theright subplot of figure 4 shows the spectrally matched responsespectra of the selected records for 5% damping level. It canbe seen that the spectral misfit is reduced significantly after themodification of the accelerograms by RspMatch2005.

3.3. Sensitivity analysis

The cost of SMA material has been one of the impedimentsto actual implementation even though it has considerablydecreased in the past decade [32]. However, economicallyfeasible solutions can be achieved with NiTi-based SMAs ifthey are used in small devices or applied to selected regionsof structures [33]. The SMA device considered in this studysimply consists of multiple loops of superelastic NiTi wireswrapped around two wheels. The simple configuration of thedevice avoids extra fabrication costs.

Several key design parameters for the SMA device areshown in figure 5 on an idealized force–deformation curve.

Table 1. Description of the ground motions used in the analyses.

EarthquakeMagnitude(Mw)

Distance(km)

PGA(g)

PGV(cm s−1)

Duration(s)

1979 ImperialValley

6.5 1.0 0.44 109.8 8.5

1986 N. PalmSprings

6.0 8.2 0.59 73.3 4.5

1994 Sylmar 6.7 6.2 0.90 102.8 9.01971 SanFernando

6.6 2.8 1.22 112.5 3.8

1992 Landers 7.3 1.1 0.72 97.6 13.11989 LomaPrieta

6.9 6.1 0.56 94.8 10.2

In the figure, Fy and uy represent forward transformationforce and displacement of the SMA device, respectively; Fd

and ud respectively denote design force and displacementcorresponding to the limit of superelastic force–displacementrelationship of the SMA device; kSMA and αkSMA denote initiallateral stiffness and post-transformation stiffness of the device,respectively. For the NiTi wire considered in this study, α,which represents the ratio of post-transformation stiffness andinitial stiffness of the SMA device, is observed to be 0.1; theforward transformation strain of the SMA wire εy is about 1%and the maximum recovery strain of the SMA wire is about6%. Here, uy is selected as analysis parameter while ud canbe computed directly for the given uy . Another parameterfor the sensitivity analysis is selected to be the normalizedforward transformation strength of the SMA device Fo, whichis defined as Fo = Fy/Wd, where Wd is the weight of thebridge deck. Note that once uy and Fy are given, the geometricdimensions of the SMA elements can be computed from

uy = εy LSMA

kSMA = Fy

uy= ASMA ESMA

LSMA

(2)

where ESMA, ASMA and LSMA are the Young’s modulus, cross-sectional area and length of the SMA wires, respectively.

The effects of the environmental temperature changes onthe seismic response of the isolated bridge are also evaluatedin this study. Lastly, the pre-strain level of the SMA wires

4

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 4. Target response spectrum compared to response spectra of selected ground motions before and after RspMatch2005 modification.

Table 2. Parameters considered in the sensitivity analysis and their levels.

Normalized forwardtransformation strengthFo

Normalized forwardtransformationdisplacement uy (mm) Temperature (◦C) Pre-strain level (%)

Lateral stiffness ofrubber bearing(kN mm−1)

0.15, 0.17, 0.190.21, 0.23, 0.250.27, 0.29, 0.32 20, 30, 40, 50 0, 10, 20, 30, 40 0, 0.5, 1, 1.5, 2, 2.5 11, 16, 25, 32, 45, 1010.34, 0.36, 0.38, 0.40

Figure 5. Analysis parameters on an idealized force–deformationcurve.

and the lateral stiffness of the laminated rubber bearings kb areconsidered as other parameters for the sensitivity study. Table 2tabulates the parameters and the levels of these parametersconsidered in the sensitivity study.

Numerical simulations of the bridge isolated by the SRBisolation system are conducted in order to assess the influenceof the above-described parameters on the seismic responseof the isolated bridge. The six spectrally matched historicalground motion records are used as external excitations.The response quantities evaluated here are peak relative

displacement of the deck, peak absolute acceleration of thedeck, and peak base shear normalized by the weight of thedeck.

Figure 6 shows the variation of the peak responsequantities with the normalized forward transformation strengthof the SMA device. The results are obtained for uy = 50 mm,T = 20 ◦C, kb = 16 kN cm−1 and without any pre-strain inthe SMA wires. It can be seen that initially the peak deckdrift almost continuously decreases for the increasing valuesof Fo, yet the rate of this decrease becomes smaller or turnseven to a slight increase when Fo is over 0.30. The peakdeck acceleration and the peak normalized base shear taketheir minimum values in the vicinity of Fo = 0.20–0.25 formost of the excitation cases and then they start to increasealmost constantly for the higher values of Fo. It can beconcluded from these observations that the optimal value ofFo which effectively reduces the deck drift and simultaneouslycontrols the superstructure acceleration and demands on thesubstructure is around 0.25.

The variation of the peak deck drift, deck accelerationand normalized base shear with the forward transformationdisplacement of the SMA device is given in figure 7 for Fo =0.25, T = 20 ◦C, and kb = 16 kN cm−1. No pre-strain ispresent on the SMA wires. It is observed that the peak relativedisplacement of the deck does not change significantly for thedifferent values of uy . Since the smaller values of uy implyshorter length of the SMA wires, it is preferred to choose asmall uy . However, note that the seismic demand on the piersincreases with decreasing values of uy . Specifically, there isan average increase of 39% in the peak base shear for thesix excitation cases when uy is changed from 50 to 20 mm.

5

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 6. Variation of various peak response quantities with the normalized forward transformation strength of the SMA device.

Figure 7. Variation of various peak response quantities with the forward transformation displacement of the SMA device.

Also, since the stiffness of the SMA device increases whenthe length of the SMA wires used for the device shortens, thesuperstructure acceleration increases. In particular, when uy isdecreased from 50 to 20 mm, the deck acceleration amplifiesby an average factor of 1.3 for all considered cases. It can beconcluded from figures 6 and 7 that the SRB isolation systemamplifies the peak deck acceleration and peak normalized baseshear for large values of Fo and small values of uy . Therefore,one should make a careful selection for these two parametersin order to mitigate the displacement response of the deck andat the same time limit the deck acceleration and base shear.

In order to investigate the effect of temperature changeson the performance of the SRB isolation system, a set of time-history analyses is conducted for an environmental temperaturerange of 0–40 ◦C. The simulations are performed for Fo =0.25, uy = 40 mm, and kb = 16 kN cm−1. No pre-strain ispresent on the SMA wires. The variations of peak responsequantities with temperature are illustrated in figure 8. Theforce–displacement relationships of the SRB isolation systemfor the Imperial Valley, Landers and Loma Prieta earthquakesare also shown in figure 9. It can be seen that there is areduction in the peak deck drift with the increasing temperaturefor all excitation cases except the Imperial Valley earthquake.As shown in figure 9, the forward transformation takes places

at considerably lower force level at 0 ◦C and the criticalforce for the forward transformation increases with increasingtemperature. The larger hysteretic force generated in the SRBisolation system decreases the displacement response of thedeck at higher temperatures. It is also observed that themaximum variation of the peak deck drift as the temperatureincreases by 20 ◦C compared to the reference temperature of20 ◦C is only about 9%. However, there is an increase of62% in the peak deck drift for the N. Palm Springs earthquakewhen the temperature decreases by 20 ◦C compared to thereference temperature. Furthermore, the average variation ofthe peak deck drift for the six excitation cases is only 6%when the temperature increases to 40 ◦C, while it is 25% whenthe temperature decreases to 0 ◦C. This implies that the deckresponse is more sensitive to a decrease than an increase intemperature. This can be attributed to the fact that there isa larger change in the forward transformation force when thetemperature decreases to 0 ◦C as compared to its increase to40 ◦C as can be seen in figure 9. In addition, as illustrated infigure 1, the initial stiffness of the SMA wires remains almostconstant when the temperature increases by 20 ◦C comparedto the reference temperature but it somewhat decreases astemperature decreases to 0 ◦C. As a consequence of the largerSMA force at higher temperatures, the peak deck acceleration

6

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 8. Variation of various peak response quantities with environmental temperature changes.

Figure 9. Force–displacement relationship of the SRB isolation system at different temperatures.

and peak normalized base shear increase with the increasingtemperature. Specifically, the maximum variation of peakdeck acceleration and normalized base shear is ±27% whentemperature differs by ±20 ◦C from the reference temperatureof 20 ◦C.

The effect of pre-strain on the SMA wires that are used inthe auxiliary SMA device of the isolation system is investigatedby changing the pre-strain level of the SMA wires from 0 to2.5%. The corresponding stress values for 0.5, 1, 1.5, 2 and2.5 % pre-strain are 150, 247, 288, 310, and 326 MPa. Thepre-strain levels are selected such that the maximum strainexperienced by the SMA wires is within the 6% recovery strainlimit. Figure 10 presents the variation of the mean of thepeak response quantities with the pre-strain level of the SMAwires for the isolated bridge subjected to different earthquakes.The simulations are conducted for Fo = 0.15, 0.20, 0.25 anduy = 40 mm, T = 20 ◦C, and kb = 16 kN cm−1. Thedifferent values of Fo are considered to evaluate any interactionbetween the normalized yield strength and the pre-strain level.It is observed that when the SMA wires are pre-strained toabout 1%, the relative deck displacement decreases comparedto the case without any pre-strain on the wires for all values ofFo. Since the initial behavior of the superelastic SMA wiresis almost linear elastic and the forward transformation startsat over 1% strain, a larger hysteresis loop, i.e. an increase in

dissipated energy, is available when the wires are pre-strained.This causes a reduction in the peak deck drift. However,when the pre-strain level is increased more than 1%, thecorresponding decrease in deck drift is not significant for mostof the cases. It is also observed that applying a pre-strain over1% tends to increase the peak deck acceleration. Furthermore,the variations in the peak shear force transferred to the piersare small when the pre-strain level is increased from 1% to2.5%. Overall, these observations imply that applying an initialtensile force to the SMA wires that corresponds to a strainof 1–1.5% increases the effectiveness of the SMA device inreducing the seismic response of the bridges isolated by theSRB isolation system.

Finally, the effect of the rubber stiffness on theperformance of the SMA/rubber isolation system is evaluated.Here, the values chosen for the lateral stiffness of the laminatedrubber are given in table 2. These values of the rubber stiffnesscorrespond to isolation periods between 1.0 and 3.0 s forthe bridge isolated by laminated rubber bearings. Figure 11shows the variation of the mean of the response quantities asa function of kb for the isolated bridge subjected to differentearthquakes. The results are given for uy = 40 mm, T =20 ◦C and Fo = 0.10, 0.15 and 0.25. For different designvalues of Fo for the SMA component of the SMA/rubberisolation system, increasing the stiffness of the rubber bearing

7

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 10. Variation of the mean of the peak response quantities with the pre-strain level of the SMA wires.

Figure 11. Variation of the mean of the peak response quantities with the stiffness of the rubber bearings.

decreases the peak deck drift but augments the peak deckacceleration and base shear, which indicates a loss of thepotential advantages of seismic isolation. Also, since verylow values of kb result in excessive isolator deformations,kb = 16 or 25 kN cm−1 corresponding to isolation periods of2.5 and 2.0, respectively, provides the best performance for theconsidered bridge structure.

3.4. A comparative study

In this section, the performance of the SMA/rubber isolationsystem is compared with an SMA-based sliding isolationsystem that is studied by Ozbulut and Hurlebaus [17]. Thesliding isolation system, named as a superelastic-friction baseisolator (S-FBI), combines the superelastic shape memoryalloys with a flat steel–Teflon bearing rather than the laminatedrubber bearing considered in this study.

Based on the results of section 3.3, an optimal SRBisolation system has the design parameters of uy = 40 mm,Fo = 0.25, kb = 25 kN cm−1 and a pre-strain level of1.5%. The optimum design parameters for the S-FBI systemare adopted from [17]. For the S-FBI system, the flat slidingbearings have a friction coefficient of 0.10 and the parametersof the SMA device are uy = 30 mm and Fo = 0.10. No

pre-strain is applied on the SMA wires. Note that for the abovedesign parameters, the volume of the SMA wires used in the S-FBI system is 71% less than the volume of the SMA materialemployed in the SRB isolation system. Also, the resultsfrom the simulations of the bridge isolated by natural rubberbearings (NRBs) with a lateral stiffness of kb = 25 kN cm−1

and 2% viscous damping and pure-friction (P-F) bearings withfriction coefficient of 0.10 are given to serve as a benchmarkin the performance evaluation of the SMA-based isolationsystem.

Plots illustrating the peak response quantities for differentisolation systems subjected to the six different earthquakes aregiven in figures 12–14. The means of the results for theseexcitations are also presented in the same plots. It is clearthat both SMA-based isolation systems successfully reducethe deck drift for all excitation cases. As can be seen fromfigure 12, the results for peak deck drift for the SRB isolationsystem and S-FBI system are very close for the individualearthquake cases and the means of the results for all excitationsare the same. However, the SRB isolation system produceshigher peak deck accelerations and base shears than the S-FBIsystem. In particular, for the S-FBI system, the mean of thepeak deck acceleration for the considered excitations is as lowas 53% of that of the SRB isolation system and the mean of the

8

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 12. Peak deck drift for the various isolation systems subjected to near-field earthquakes.

Figure 13. Peak deck acceleration for the various isolation systems subjected to near-field earthquakes.

Figure 14. Peak normalized base shear for the various isolation systems subjected to near-field earthquakes.

peak normalized base shear is 35% lower than that of the SRBisolation system.

In order to further compare the performance of the twodifferent SMA-based isolation systems, figure 15 displaysthe time histories of the deck drift, deck acceleration andnormalized base shear for the Imperial Valley earthquake. Thetime-history results for the NRB and P-F isolation systems arealso provided in the figure.

It can be seen that the use of either the SRB isolationsystem or the S-FBI system significantly reduces the deck drift.Also, note that there is no residual displacement for both SMA-based isolation systems at the end of the motion. On the otherhand, considerable residual deformations are present in the

P-F system which lacks re-centering force capability. It can bealso seen that the NRB system damps out the vibrations over amuch longer time.

It can be observed that the SRB isolation system produceshigher deck acceleration and base shear response than the S-FBI system. However, the results for the SRB isolation systemare still comparable to those of the NRB system. It should bealso noted that the P-F system limits the maximum accelerationtransmitted to the superstructure to a certain level that is afunction of the friction coefficient. As compared to the P-Fsystem, the S-FBI system to some extent increases the deckacceleration and base shear as a result of the increased stiffnessdue to the SMA device. However, it can be seen that the

9

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 15. Time histories of various response quantities of theisolated bridge subjected to the Imperial Valley earthquake.

responses of the S-FBI system are comparable to those of theP-F system.

The increases in the deck acceleration response and pierbase shear for the SRB isolation system as compared to the S-FBI system can be better explained by comparing the hystereticforces generated in both isolation systems. Figure 16 illustratesthe force–deformation curves of the SRB isolation system andthe S-FBI system for the Imperial Valley earthquake. It canbe seen that the SRB isolation system has a higher stiffnessthan the S-FBI system and the hysteretic force generated in theSRB isolation system is considerably larger than that of the S-FBI system. This higher isolator force that is transmitted to thepiers from the deck results in larger base shears for the SRBisolation system.

Figure 17. Time histories of absolute input energy for thenon-isolated and isolated bridge structures subjected to the ImperialValley earthquake.

Figure 17 displays the absolute input energy to the non-isolated bridge and isolated bridge with various isolationsystems. It is clear that the input energy decreases when thebridge is isolated. This decrease is more noticeable for theP-F system and the S-FBI system. It should be also notedthat the S-FBI system has the minimum energy accumulationat the end of the motion for both absolute and relative energyformulations.

Figure 18 shows the time history of recoverable energy(kinetic energy + strain energy) transmitted to the bridgestructure isolated by various isolation systems. It can be seenthat there is a substantial reduction in recoverable energy,which is the cause of damage in the structure, for the P-Fsystem and S-FBI system in comparison with that of the NRBsystem and SRB system. This is due to the fact that theforce transmitted to the superstructure by the P-F or the S-FBI system is considerably smaller than that by the NRB orthe SRB system. When the SMA-based isolation systems arecompared, it can be noticed that the recoverable energy of thebridge structure isolated by the S-FBI system is 60% smallerthan that of the SRB system.

Time histories of the absolute input energy and energyabsorbed by the subcomponents of SMA-based isolation

Figure 16. Force–deformation curves of the SRB isolation system and the S-FBI system subjected to the Imperial Valley earthquake.

10

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

Figure 18. Time histories of recoverable energy for various isolationsystems subjected to the Imperial Valley earthquake.

systems, i.e., the rubber and steel–Teflon bearings and theSMA device are plotted in figure 19 for the S-FBI systemand the SRB isolation system subjected to the Imperial Valleyearthquake. It can be seen that the energy is dissipated mainlyby the SMA device for the SRB isolation system whereas theSMA device serves as a re-centering component in the S-FBIsystem and the energy is dissipated through friction in thesliding surface for the S-FBI system.

4. Conclusions

As an alternative to conventional rubber isolators such ashigh-damping rubber bearings and lead–rubber bearings, smartrubber bearing systems with shape memory alloys (SMAs)have been proposed in recent years. As a class of smartmaterials, shape memory alloys show excellent re-centeringand considerable damping capabilities which can be exploitedto obtain an efficient seismic isolation system. This studyinvestigates the sensitivity of the seismic response of a multi-span continuous bridge isolated by an SRB isolation system.The smart isolation system consists of a laminated rubberbearing and an additional re-centering and energy dissipatingdevice made of NiTi superelastic wires. A temperature- and

rate-dependent model is used to characterize the behavior ofthe SMA device. Six historical ground motion records areadjusted using the program RspMatch2005 to match a targetdesign spectrum and employed as the external excitation insimulations. The parameters for the sensitivity analysis arechosen to be the normalized forward transformation strength ofthe SMA device Fo, the forward transformation displacementof the SMA device uy , the pre-strain level of the SMAwires, the lateral stiffness of the laminated rubber bearings kb

and environmental temperature changes. A large number oftime-history analyses of the isolated bridge are performed toassess the effects of these parameters on the various responsequantities of isolated bridges. In particular, the variation ofpeak deck drift, deck acceleration, and normalized base shearwith analysis parameters are evaluated.

It is found that there is a trade-off between thedisplacement response of the deck and the deck acceleration,as well as the base shear for the increasing values of Fo.The optimum value of Fo is said to be in the vicinity of0.25. It is also observed that the variation of uy in the rangeof 20–50 mm does not significantly change the peak deckdisplacement response of the isolated bridge. Yet, since alower value of uy implies shorter wire lengths for a fixed 1%forward transformation strain, the stiffness of the SMA deviceincreases when uy decreases and, as a consequence, the deckacceleration and normalized base shear increase by about 30%and 39%, respectively. The variation of the seismic responseof the isolated bridge with environmental temperature is alsoevaluated. It is found that the effects of temperature change aremore prominent in the case of a decrease in the temperature.Specifically, as the temperature decreases by 20 ◦C comparedto the reference temperature of 20 ◦C, the peak deck driftexperiences an average of 25% increase with a maximum of62%, and the peak deck acceleration and normalized base shearvary by a maximum of 25% and 27%, respectively, for thesix excitation cases considered here. On the other hand, peakdeck drift increases by an average of 6% with a maximumof 9%, while there is a maximum increase of 27% in bothpeak deck acceleration and normalized base shear for all theconsidered cases when the temperature increases to 40 ◦C fromthe reference temperature. Therefore, the effect of temperaturechange cannot be neglected during design of the isolation

Figure 19. Time histories of the absolute input energy and energy absorbed by the subcomponents of the S-FBI system and SRB isolationsystem subjected to the Imperial Valley earthquake.

11

Smart Mater. Struct. 20 (2011) 015003 O E Ozbulut and S Hurlebaus

system since it affects the seismic response of the isolatedbridge considerably. It is also observed that when the NiTiwires are pre-strained so that they will have an initial strainin the range of 1–1.5%, the effectiveness of the SRB isolationsystem improves. The effect of rubber stiffness on the seismicresponse of the bridge is also analyzed.

Finally, the performance of the SRB isolation systems iscompared with an S-FBI system that combines a flat slidingbearing with an SMA device. It is found that bridge structuresisolated by either the SRB isolation system or the S-FBI systemhave very similar results for the peak deck drift response for theconsidered excitations. However, it is noted that the peak deckacceleration and peak base shear exhibit higher values in thecase of the SRB isolation system.

It is also observed that the S-FBI system attracts smallerquantities of input energy than the S-RBI system. It is shownthat the energy is mainly dissipated by the SMA device forthe SRB isolation system. On the other hand, the energydissipation in the S-FBI system is through friction, whilethe SMA component of the isolation system serves as a re-centering device. Since the energy dissipation in the SRBisolation system almost solely relies on the hysteretic behaviorof SMAs, a larger amount of SMA material is required for theSRB isolation system. Noting that the high cost of the SMA ismostly cited as one of the main barriers that precludes the useof SMAs in a full-scale seismic application, and consideringthe superior structural response of the S-FBI system, it canbe concluded that the S-FBI system which combines SMAswith flat sliding bearings has more favorable properties thanthe SRB isolation system which consists of a laminated rubberbearing and an SMA device.

References

[1] Tyler R G 1991 Rubber bearings in base-isolated structures—asummary paper Bull. N. Z. Natl Soc. Earthq. Eng. 24 251–74

[2] Bozorgnia Y and Bertero V 2004 Earthquake Engineering:From Engineering Seismology to Performance-BasedEngineering (Florida: CRC Press)

[3] Shen J, Tsai M H, Chang K C and Lee G C 2004 Performanceof a seismically isolated bridge under near-fault earthquakeground motions J. Struct. Eng. 130 861–8

[4] Liao W I, Loh C H and Lee B H 2004 Comparison of dynamicresponse of isolated and non-isolated continuous girderbridges subjected to near-fault ground motions Eng. Struct.26 2173–83

[5] Park S W, Ghasemi H, Shen J, Somerville P G, Yen W P andYashinsky M 2004 Simulation of the seismic performance ofthe Bolu viaduct subjected to near-fault ground motionsEarthq. Eng. Struct. Dyn. 33 1249–70

[6] Deb S K 2004 Seismic base isolation–an overview Curr. Sci. 871426–30

[7] Qiang H, Xiuli D, Jingbo L, Zhongxian L, Lyun L andJianfeng Z 2009 Seismic damage of highway bridges duringthe 2008 Wenchuan earthquake Earthq. Eng. Eng. Vib.8 263–73

[8] Yoshioka H, Ramallo J C and Spencer B F 2002 Smart baseisolation strategies employing magnetorheological dampersJ. Eng. Mech. 128 540–51

[9] Usman M, Sung S H, Jang D D, Jung H J and Koo J H 2009Numerical investigation of smart base isolation systememploying MR elastomer J. Phys.: Conf. Ser. 149 012099

[10] DesRoches R and Smith B 2004 Shape memory alloys inseismic resistant design and retrofit: a critical review of theirpotential and limitations J. Earthq. Eng. 8 415–29

[11] Hurlebaus S and Gaul L 2006 Smart structure dynamics Mech.Syst. Signal Process. 20 255–81

[12] Song M, Ma N and Li H N 2006 Applications of shape memoryalloys in civil structures Eng. Struct. 28 1266–74

[13] Ozbulut O E, Roschke P N, Lin P Y and Loh C H 2010GA-based optimum design of a shape memory alloy devicefor seismic response mitigation Smart Mater. Struct.19 065004

[14] Cardone D, Dolce M and Ponzo F C 2006 The behaviour ofSMA isolation systems based on a full-scale release testJ. Earthq. Eng. 10 815–42

[15] Attanasi G, Auricchio F and Fenves G 2009 Feasibilityassessment of an innovative isolation bearing system withshape memory alloys J. Earthq. Eng. 13 18–39

[16] Zhang Y, Hu X and Zhu S 2009 Seismic performance ofbenchmark base isolated bridges with superelasticCu–Al–Be restraining damping device Struct. ControlHealth Monit. 16 668–85

[17] Ozbulut O E and Hurlebaus S 2010 Evaluation of theperformance of a sliding-type base isolation system with aNiTi shape memory alloy device considering temperatureeffects Eng. Struct. 32 238–49

[18] Wilde K, Gardoni P and Fujino Y 2000 Base isolation systemwith shape memory alloy device for elevated highwaybridges Eng. Struct. 22 222–9

[19] Casciati F, Faravelli L and Hamdaoui K 2007 Performance of abase isolator with shape memory alloy bars Earthq. Eng.Eng. Vib. 64 401–8

[20] Casciati F, Faravelli L and Al Saleh R 2009 An SMA passivedevice proposed within the highway bridge benchmarkStruct. Control Health Monit. 16 657–67

[21] Choi E, Nam T H, Oh J T and Cho B S 2006 An isolationbearing for highway bridges using shape memory alloysMater. Sci. Eng. A 438–440 1081–4

[22] Chen W and Song B 2006 Temperature dependence of a NiTishape memory alloy’s superelastic behavior at a high strainrate J. Mech. Mater. Struct. 12 339–56

[23] Schmidt I and Lammering R 2004 The damping behavior ofsuperelastic NiTi components Mater. Sci. Eng. A 378 70–5

[24] Ozbulut O E and Hurlebaus S 2010 Neuro-fuzzy modeling oftemperature-and strain-rate-dependent behavior of NiTishape memory alloys for seismic applications J. Intell.Mater. Syst. Struct. 21 837–49

[25] Wang Y P, Chung L L and Liao W H 1998 Seismic responseanalysis of bridges isolated with friction pendulum bearingsEarthq. Eng. Struct. Dyn. 27 1069–93

[26] Preumont A 1984 The generation of spectrum compatibleaccelerograms for the design of nuclear power plants Earthq.Eng. Struct. Dyn. 12 481–97

[27] Bommer J J and Acevedo A B 2004 The use of real earthquakeaccelerograms as input to dynamic analysis J. Earthq. Eng.8 43–91

[28] Naeim F and Lew M 1994 Deficiencies of design-spectrumcompatible accelerograms Struct. Des. Tall Build. 3 275–83

[29] Hancock J, Watson-Lamprey J, Abrahamson N A, Bommer J J,Markatis A, McCoy E and Mendis R 2006 An improvedmethod of matching response spectra of recorded earthquakeground motion using wavelets J. Earthq. Eng. 10 7–89

[30] IBC 2000 International Building Code (Falls Church, VA:International Code Council)

[31] Malhotra P K 2003 Strong-motion records for site-specificanalysis Earthq. Spectra 19 557–78

[32] Ocel J, DesRoches R, Leon R T, Hess G, Krummes R,Hayes J R and Sweeney S 2004 Steel beam-columnconnections using shape memory alloys J. Struct. Eng.130 732–40

[33] Janke L, Czaderski M, Motavalli M and Ruth J 2005Application of shape memory alloys in civil engineeringstructures–overview, limits and new ideas Mater. Struct. 38578–92

12