You are on page 1of 8

Engineering Structures 22 (2000) 222229 www.elsevier.

com/locate/engstruct

Base isolation system with shape memory alloy device for elevated highway bridges
Krzysztof Wilde
b

a,*

, Paolo Gardoni b, Yozo Fujino

a Department of Civil Engineering, University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113, Japan Department of Civil and Environmental Engineering, University of California, Berkeley, CA 94720, USA

Received 16 April 1998; accepted 6 August 1998

Abstract Base isolation provides a very effective passive method of protecting bridges from the hazard of earthquakes. The proposed smart isolation system combines a laminated rubber bearing with a device made of shape memory alloy (SMA). The smart base isolation utilizes the different responses of the SMA at different levels of strain to control the displacements of the rubber bearing at various excitation levels. At the same time the hysteresis of the alloy is used to increase the energy dissipation capacity. 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction In recent years, base isolation has become an accepted design alternative for earthquake hazard mitigation for various structures. Extensive reviews on historical developments and recent literature on isolation systems have been provided by Kelly [1] and Buckle and Mayes [2]. Various designs of base isolation have been proposed and many concepts have been incorporated into the seismic isolation framework. The most important common features of such systems are a shift in the natural frequency of the structure to a lower value, and an increase in structural damping. Innovative techniques for controlling structural response are searching for smart materials which can introduce new possibilities in earthquake protection methods. One class of such materials are metallic alloys known as shape memory alloys (SMAs). An SMA displays a thermoplastic martensitic phase transformation. As the SMA is cooled through its transformation temperature range its parent state (austenite) transforms to the lower temperature martensite phase. Such a transformation also occurs when stress is applied to the

* Corresponding author. Tel.: 81-3-5802-3312; Fax: 81-35689-7292; E-mail: kris@kyouryou.t.u-tokyo.ac.jp

material. Since stress-induced martensite is not thermally stable, as soon as the stress is removed the alloy can revert to the austenite. The inverse transformation recovers the induced strain, an effect referred to as superelasticity. The superelastic behavior of an SMA is observed only when the material is strained above the transformation temperature range, namely, above the austenite nish temperature. The superelasticity, freedom of shaping the hysteresis through material selection and various heat treatments, large ductility and long fatigue life of SMAs make them a particularly attractive material for control systems and devices designed to increase structural performances. Graesser and Cozzarelli [3] suggested the use of Nitinol (NiTi alloy) as a material for dampers. They proposed a one-dimensional constitutive law for superelastic behavior and experimentally veried the applicability of the mathematical model. The feasible application of an SMA damping device has been investigated by Clark et al. [4] and Higashino et al. [5]. The damper they proposed was built of multiple loops of Nitinol wire wrapped around cylindrical support posts. The device was experimentally tested over a range of strain amplitudes, loading frequencies and temperatures. Analytical studies of the damper, used as a bracing element, were performed on a two-dimensional computer model of a 6-storey building. The obtained response reduction of

0141-0296/00/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 1 - 0 2 9 6 ( 9 8 ) 0 0 0 9 7 - 2

K. Wilde et al. / Engineering Structures 22 (2000) 222229

223

displacements was of the order of 1/2, and the energy absorbed in a frame was 1/8 compared to a frame without an SMA device. However, in their simulations, the SMA hysteresis loop was modeled by a bi-linear law where neither the superelastic effect nor hardening of the Nitinol after the completion of transformation to martensite could be incorporated. The seismic isolation of highway bridges by a laminated rubber bearing with a lead core has been extensively studied [6,7]. An ordinary elastomeric bearing with a lead core increases the damping of the structure due to the hysteretic properties of the inelastic deformation of lead and gives a signicant reduction of pier shear force and pier displacement compared to a monolithic bridge. However, the behavior of lead rubber bearings under strong ground motions results in large deformations of the superstructure and an additional displacement restraining device is necessary. The most suitable device for isolation of the bridge deck from seismic input should have variable properties with respect to the intensity of external excitation. For loading caused by stopping of cars, wind action or small earthquakes, the bearing should behave as a rigid connector, so that minor external forces do not damage the expansion joints or other auxiliary elements of the bridge. For medium intensity loading, the isolation system should be exible and provide adequately high damping. At a very large ground motion the isolation device should be able to limit the relative displacement of the superstructure within the design range to prevent the collapse of the deck. The combination of an ordinary isolation system with a displacement restraining device is one way of varying the isolation property. Although variable properties can also be provided by active or semi-active devices [8], such systems are relatively complex since they require special hardware, sensors and constant maintenance. A simple bar of SMA combined with a laminated rubber bearing can provide a damper with the desired variable characteristics based solely on the material properties of the alloy. Furthermore, the proposed device has an inherent centering ability due to superelastic behavior of SMA. In this paper, the design of a SMA bar damper added to the laminated rubber bearing isolation system is presented. The optimization of the size of the SMA device is performed on the Kobe earthquake record (January, 1995) scaled to different magnitudes. The performance of the proposed smart isolation system is presented together with the performance of a conventional isolation system using lead laminated rubber bearings with an additional stopper device.

2. Model of an elevated highway bridge with damping devices The considered elevated highway bridge, shown in Fig. 1, is a three-span-continuous concrete box girder of 14 m in width. The reinforced concrete piers have a height of 11.5 m and the distance between them is 40 m. Two systems of isolation of the superstructure from the pier are considered: a laminated rubber bearing with an additional SMA device, referred to as an SMA system, and a laminated rubber bearing with a lead core and displacement restrainer, referred to as an NZ system. The considered elevated highway bridge is modeled as a two-degree-of-freedom system (Fig. 1). The equations of motion are m dx d(t) NF(xd,x d,xp,x p,t) mdx g(t),mpx p(t) c px p(t) kpxp(t) NF(xd,x d,xp,x p,t) mp x g(t) where xd, xp and md, mp are the displacements and masses of the deck and pier, respectively, and x g is the ground acceleration. The coefcients of damping and stiffness of the pier are denoted by cp and kp. F(xd,x d,xp,x p,t) is the hysteretic force generated by the isolation system and N is the number of bearings supporting the superstructure. The parameters of the two-degree-of-freedom model are estimated by the method given by Kawashima [9]. 2.1. Model of laminated rubber bearing with SMA device A comprehensive constitutive law for Shape Memory Alloys (SMAs) in a thermodynamic formulation has been proposed by Achenbach [10]. The model, derived for the case of plane strain, denes the inelastic strain rate in terms of stress, phase fractions and several internal variables in a manner analogous to inelastic formulation for creep and viscoplasticity. However, for the preliminary verication of engineering applications of SMAs, such a model is far too complex, so a relatively simple law proposed by Graesser and Cozzarelli [3] is herein adopted. This model describes the uniaxial behavior of superelastic material. A schematic plot of the curve of a superelastic material is given in Fig. 2. The constitutive law due to Graesser and Cozzarelli modies the expression of the backstress such that twinning hysteretic and/or superelastic behavior can be modeled. The stress, and strain, , are given as follows: (1)


Y
n

(2a) (2b)

Ein ftcerf(a)[u( )]

224

K. Wilde et al. / Engineering Structures 22 (2000) 222229

Fig. 1. Model of the elevated highway bridge.

u(x)

1 x0

0 x0

(4b)

The coefcients ft, c, a are material constants controlling the recovery of the elastic strain during unloading. Since SMA is intended to be used over a wide range of strain, the GraesserCozzarelli model is extended to represent the hardening of the SMA after the transition to martensite is completed. As the load increases, the pure martensite follows the elastic response with modulus Em. The modied model is of the form: E


Y
n

uII() uI() Em

(5a)

2 2a2sign() a3 )uIII() (3a1


Fig. 2. Schematic stressstrain relations of the extended hysteretic model of shape memory alloy.

Ein f1cerf(a)[u( )]
where the functions uI(), uII() and uIII() are:

(5b)

where denotes one-dimensional backstress, Y is the yield stress, or in the case of SMA the beginning of the stress-induced transition from austenite to martensite, n is the overstress power. Ey/(E Ey) is a constant controlling the slope of , where E is the elastic modulus of austenite and Ey is the slope after yielding. Inelastic strain, in, is given by:

uI() (1 uII() uIII()) uII()

(6a) (6b)

in E

uIII()

1 m

0 otherwise

> 0 and 1 m 1 otherwise

(6c)

(3) uII() in Eq. (5a) describes the elastic The term Em behavior of martensite, which is activated when the strain is higher than m. Strain, m, denes the point when the transformation of SMA from austenite to martensite is completed. The smooth transition from the curve of slope Ey to slope Em is obtained by adding the last term in Eq. (5a) which is evaluated only during loading and for strain 1 m. The constants a1, a2 and

The error function, erf(x), and the unit step function, u(x), are dened as: 2 2 erf(x) et dt
0

(4a)

K. Wilde et al. / Engineering Structures 22 (2000) 222229

225

a3 control the curvature of the transition. The constants are selected so that the slopes of the function dened by the last term at points 1 and m are consistent with the slopes of SMA plastic behavior and martensite elastic response. The smoothness of the transition is governed by the selection of slope at strain 2. Fig. 2 explains the introduced constants in a graphical form. In this study, the simplest conguration of the SMA device is chosen, i.e. a set of two bars working in tension and compression attached to the pier and superstructure. The hysteretic force generated by the SMA bars is FSMA A (7)

3. Response analysis The two isolation systems are analyzed for three excitation levels: Kobe NorthSouth component (1995), scaled to 0.6 g and to 0.4 g (Fig. 3), and a sinusoidal excitation with frequency g 6.28 rad/s and amplitude 0.2 g. The sine wave is chosen to represent the case of a low excitation level corresponding to wind load, trafc load and small earthquake. The parameters of the two-degrees-of-freedom model are found to be: md 8 105 kg, mp 1.4 105 kg, kp 3.2256 108 N/m and p 5%. The material selection for the alloy as well as the thermomechanical treatment provide a wide selection of SMA properties [12]. In these simulations, the elastic modulus of SMA austenite state is selected as E 9.8325 1010 N/m2. The elastic modulus of pure martensite is Em 7.3744 1010 N/m2. The limit of the superelastic range is m 0.08 and the points dening the transition from the slope Ey to elastic response of pure martensite are 1 m 0.03 and 2 (m 1)/2. The yield stress is assumed to be Y 14.49 107 N/m2. The parameters controlling the shape of the superelastic loop are: n 1, 0.019, c 0.001, a 900 and ft 0.08. The selected parameters of SMA correspond to an alloy with perfect superelastic behavior. The length and the cross-sectional area of the SMA bars is optimized such that the stress in the SMA bars reaches the elastic response of pure martensite during the strongest considered ground motion. For this condition the total cross-sectional area of the SMA bars was found to be A 0.0175 m2 and the length of the bars is l 2 m. The SMA enters the hardening due to the elastic behavior of the martensite at a displacement of 0.16 m. The laminated rubber bearings [9] have a cross-sectional area of Ar 0.8881 m. The height of the rubber layers, i.e. summation of the thickness of all the rubber layers in a single bearing, is t 0.154 m. The stiffness of the rubber is krubber ArG/t 3.8723 106 N/m2, where G is the shear modulus of the rubber equal to 7.848 105 N/m2. The equivalent damping ratio of the bearing is rubber 1%. Each lead rubber bearing has an area of Ar, height t and a lead core of diameter 13.5 cm. The parameters of the model of the NZ isolation system are: Fy 143 590 kN, Y 0.007 m, 0.22, 0.9, 1.4, 1.0 and 1 [9,11]. The displacement restrainer [9] has a stiffness kstopper 1.8104 108 N/m, and is designed to be activated when the relative displacement between pier and superstructure is equal to 15 cm. The time responses of the bridge with isolation systems are obtained by the RungeKutta method with variable time step. The time histories of the relative displacement

where A denotes the total area of the bars and is the stress in the bars. It is assumed that the SMA response in tension and compression is identical. The bridge deck is supported on two laminated rubber bearings designed according to Design example of a highway bridge based on the Manual for Menshin Design of Highway Bridges [9]. The hysteretic force, F(xd,x d,xp,x p,t), appearing in Eq. (1), represents the summation of the forces in SMA bars and laminated rubber bearings. 2.2. Model of laminated rubber bearing with lead core and displacement restraining device The NZ isolation system consists of two laminated rubber bearings with a lead core, N 2. The hysteretic restoring force in the NZ system, FNZ, is modeled by Wens model [11] as FNZ(s,s ) Fy s (1 )FyZ Y (8)

where the relative displacement, s, is s xd xp. Z is the dimensionless displacement governed by a rst-order differential equation of the form s YZ ZZ 1 s Z s (9)

Fy and Y are the force and the yield displacement, respectively and , , are dimensionless parameters. The smoothness of the transition from the elastic to the plastic region is governed by , and is the ratio of the stiffness of plastic and elastic responses. The isolation system has a displacement restraining device which prevents excessive relative displacement between the pier and the superstructure. A rubber is placed on the restrainer to lessen the impact force of a collision. The restrainer is modeled by a linear spring and is activated at the selected maximum allowable displacement of the deck with respect to the pier cap [9].

226

K. Wilde et al. / Engineering Structures 22 (2000) 222229

Fig. 3.

Accelerogram of Kobe earthquake NS component scaled to 0.4 g.

Fig. 4. Response of the bridge with SMA and NZ isolation system to harmonic excitation of amplitude 0.2 g.

Fig. 5.

Response of the bridge with SMA and NZ isolation system to Kobe earthquake scaled to 0.4 g.

Fig. 6.

Response of the bridge with SMA and NZ isolation system to Kobe earthquake scaled to 0.6 g.

between the pier and the superstructure of the bridge with the SMA and NZ systems are plotted in Figs. 46. Relative displacement between deck and pier of the system with an SMA device under 0.2 g excitation amplitude is very small compared to the NZ system, since the alloy remains in its elastic range, without going into the

formation of stress-induced martensite. For this excitation amplitude, the SMA device works like a xed connection between deck and pier, while with the NZ system, even for this small excitation amplitude, the relative displacement reaches 12.9 cm. For an excitation amplitude of 0.4 g, the SMA bars enter the low stiffness range

K. Wilde et al. / Engineering Structures 22 (2000) 222229

227

due to martensitic phase transformation and the relative displacements of the two systems are comparable. In bridge response based on Kobe earthquake scaled to 0.6 g, both systems have similar maximum values of relative displacements since both systems control the maximum displacement. The SMA system enters the elastic response of martensite once at 4.7 s, and after that the relative displacement does not exceed 6 cm. The displacement restrainer in the NZ system is activated ve times. The hysteretic forces generated in the isolation systems plotted against the relative displacements show the variable response of the SMA system (Fig. 7). It can be seen that hysteretic responses of the SMA system due to Kobe records are large and provide considerable

amount of damping. However, the force in the SMA isolation system is about three times larger than in the NZ system. Since the SMA system has larger stiffness than the NZ system the force transmitted to the pier from the deck is larger, resulting in a larger shear force in the pier. The time history of shear force based on the Kobe record scaled to 0.6 g is shown in Fig. 8. The maximum shear force in the pier for the bridge with an SMA system coincides with the SMA entering the martensite state, and is 6.906 106 N. The maximum pier shear force for the bridge equipped with an NZ system is 8.948 106 N, and occurs when the displacement restraining device is activated. One drawback of adding a damper to isolation by laminated rubber bearings is an increase in the acceleration response. The nonlinear characteristics of SMAs, particularly during the transition from plastic response to the elastic one, result in a sudden jump in acceleration. For all the loading levels, the peak acceleration of the deck with an SMA device is greater than the acceleration response of the NZ system. This difference is the most pronounced under 0.6 g amplitude excitation, as shown in Fig. 9. The performance of the proposed base isolation system is analyzed by comparing the energy time histories of the structure with an SMA and with an NZ system. The input energy is dened as the total energy related to inertia forces induced by the ground motion. The kinetic energy of pier and deck plus the elastic energy of the pier and laminated rubber bearings is referred to here as the structure damage energy. The mathematical formulations of the input energy, energy absorbed by SMA/NZ system and structure damage energy are as follows:

Input energy md u gx ddt mp u gx pdt


0 0

t0

t0

Energy absorbed by SMA/NZ system N FNZ/SMA(x d x p)dt kr (xd xp)(x d x p)dt


0 0

t0

t0

Structure damage energy kr (xd xp)(x d x p)dt kp xpx pdt


0 0

t0

t0

Fig. 7. Forcedisplacement hysteresis loops of bridge with SMA and NZ isolation system. (a) Due to harmonic excitation of amplitude 0.2 g; (b) due to Kobe earthquake scaled to 0.4 g; (c) due to Kobe earthquake scaled to 0.6 g.

md x dx ddt mp x px pdt
0 0

t0

t0

(10)

The energy time histories of the structure with SMA

228

K. Wilde et al. / Engineering Structures 22 (2000) 222229

Fig. 8.

Pier shear force of the bridge with SMA and NZ isolation system to Kobe earthquake scaled to 0.6 g.

Fig. 9.

Acceleration response of the deck of the bridge with SMA and NZ isolation system to Kobe earthquake scaled to 0.6 g.

and NZ isolation systems based on the Kobe record scaled to 0.6 g are shown in Fig. 10. The damage energy of the bridge with the SMA system is high only in the time interval from the 4th to the 5th second, corresponding to the largest acceleration level in the Kobe earthquake. The maximum damage energy of the bridge with the SMA system is 14% smaller than the damage energy in the case of the bridge with the NZ system. However, the input energy to the bridge with the SMA system is 69% larger than the input energy to the bridge with the NZ system. There is a trade-off in designing the variable isolation with SMA device. Since the device is attached to provide a variable response for the isolation system, the force in the SMA must be large at the large displacement level to control the relative displacement between the deck and the pier. This condition results in a stiffer connection and worse isolation of the superstructure from the pier. Since the fundamental frequency of the bridge is closer to the predominant frequency of the earthquake, the energy input to the structure is larger.

4. Concluding remarks In this paper a variable isolation system for elevated highway bridges, consisting of laminated rubber bearings and SMA bars is proposed. To provide the variable characteristic of the proposed isolation system, the nonlinear response of SMA is used. The uniaxial model of a superelastic material is extended to incorporate the hardening of SMA after the completeness of the phase transformation from austenite to martensite.

Fig. 10. Energy time histories due to Kobe earthquake. (a) Bridge with SMA system; (b) bridge with NZ system.

K. Wilde et al. / Engineering Structures 22 (2000) 222229

229

The SMA isolation system provides stiff connection between the pier and the deck for small external loading. For a medium size earthquake, the SMA bars increase the damping capacity of the isolation due to stressinduced martensitic transformation of the alloy. For the largest considered earthquake, the SMA bars provide hysteretic damping and, in addition, act as a displacement controlling device due to hardening of the alloys after completeness of the phase transformation. The damage energy of the bridge with SMA isolation system is small, although the input energy to the structure is large compared to a bridge with lead rubber bearings. The SMA isolation system has inherent centering ability due to the superelastic response of the alloy. The application of a long SMA bar working in tension and compression requires some additional method to prevent buckling. An alternative way of using an SMA bar as an additional damper is to use it in bending. Experimental studies on a bending type SMA damper made from Nitinol are currently being undertaken at the University of Tokyo [13].

[3]

[4]

[5]

[6]

[7]

[8]

[9]

Acknowledgements The authors would like to thank Stefano Besseghini from ITM-CNR, Italy, for discussions and valuable comments.
[10]

[11]

[12]

References
[13] [1] Kelly JM. A seismic base isolation: a review and bibliography. Soil Dynamics and Earthquake Engineering 1986;5:20216. [2] Buckle JG, Mayes RL. Seismic isolation: history, application and

performancea world overview. Earthquake Spectra 1990;6(2):161202. Graesser EJ, Cozzarelli FA. Shape memory alloys as new materials for aseismic isolation. J Eng Mech ASCE 1991;117(11):2590608. Clark PW, Aiken ID, Kelly JM, Higashino M, Krumme RC. Experimental and analytical studies of shape memory alloy damper for structural control. In: Proceedings of Passive Damping, San Diego, CA, 1995. Higashino M, Aizawa S, Clark PW, Whittaker AS, Aiken ID, Kelly JM. Experimental and analytical studies of structural control system using shape memory alloy. In: Proceedings of 2nd International Workshop on Structural Control, Hong Kong, December, 1996. New Zealand Ministry of Work and Development. Design of leadrubber bridges bearings, Civil Division Publication 818/A. Wellington, New Zealand: New Zealand Ministry of Works and Development, 1983. Turkington H, Carr AJ, Cooke N, Moss PJ. Seismic design of bridges on leadrubber bearings. Journal of Structural Engineering ASCE 1989;115(12):300016. Kawashima K, Unjoh S, Iida H, Mukai H. Effectiveness of the variable damper for reducing seismic response of highway bridges. Recent Selected Publications at Earthquake Engineering Division, Public Works Research Institute, 1993, (No. 2), pp. 13348. Kawashima K, Okado M, Horikawa M. Design example of a highway bridge based on the Manual for Menshin Design of Highway Bridges. Recent Selected Publications at Earthquake Engineering Division, Public Works Research Institute, 1993, (No. 2), pp. 191208. Achenbach M, Atanochovic T, Muller I. A model for memory alloys in plane strain. International Journal of Solids and Structures 1986;22(2):17193. Constantinou C, Tadjbakhsh IG. Hysteretic dampers in base isolation: random approach. Journal of Structural Engineering ASCE 1985;111(4):70521. Duerig TW, Melton KN, Stockel D, Wayman CM. Engineering aspects of shape memory alloys. Oxford: Butterworth-Heinemann, 1990. Wilde K, Zheng Yi, Gardoni P, Fujino Y. Experimental and analytical study on shape memory alloy damper. In: Proceedings of Smart Systems for Bridges, Structures, and Highways, San Diego, CA, March 1998. SPIE Vol. 3325, pp. 18291.

You might also like