STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. 2013; 20:890902 Published online 3 May 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1500

Semi-active vibration control device based on superelastic NiTi wires

F. P. Amarante dos Santos1,*,, C. Cismaiu1 and J. Pamies Teixeira2
Centro de Investigao em Estruturas e ConstruoUNIC, Faculdade de Cincias e Tecnologia, Universidade Nova de Lisboa, Quinta da Torre, 2829-516 Caparica, Portugal 2 UNIDEMIUnidade de Investigao e Desenvolvimento em Engenharia Mecnica e Industrial, Faculdade de Cincias e Tecnologia, Universidade Nova de Lisboa, Quinta da Torre, 2829-516 Caparica, Portugal
1

SUMMARY The present paper presents a physical prototype that simulates the response of a single degree of freedom dynamic system, equipped with a novel semi-active vibration control device. This device comprises two superelastic NiTi elements working in phase opposition and aims to prevent deck unseating in simply supported bridges, during a seismic excitation. The special design of this device allows to avoid problems related to stressrelaxation phenomena and material degradation because of cyclic loading that have been observed in similar passive dissipation devices. The proposed design uses a strategy that permits the continuous adapting of the accumulated stress in the NiTi wires, on the basis of the response of the device to external excitations. Although unloaded, the NiTi elements remain strain/stress free, preventing stressrelaxation phenomena. With the occurrence of a dynamic excitation, a cumulative strain/stress process in the superelastic wires is initiated, enabling higher martensite transformation ratios and therefore increasing the damping capacities of the system while keeping the stresses in the wires inside a narrower superelastic window. The strain/stress accumulation in the superelastic wires is a direct result of the motion of the structure itself, with no need for external energy input. Copyright 2012 John Wiley & Sons, Ltd. Received 13 May 2011; Revised 9 January 2012; Accepted 12 March 2012
KEY WORDS: semi-active vibration control; energy dissipation systems; shape-memory alloys; superelasticity; seismic devices; unseating

1. INTRODUCTION Past earthquakes have shown that the damage induced in bridges can assume several different forms, depending, among others, on factors such as the ground motion itself, conditions depending on the building site, the adopted bridge structural solution, and its specic detailing provisions. Unseating of the bridge superstructure at in-span hinges, or at simple supports, is one of the most severe forms of bridge seismic damage, leading to possible catastrophic consequences. In the case of simply supported bridges, unseating brings about the toppling of the spans from their supports, causing the structure to collapse. Bridges with inadequate seat lengths or restraints are more prone to span unseating, requiring adequate seismic retrotting to modify their dynamic response and control the deck displacement. This is traditionally achieved using restraining systems built up of steel cables that, if designed to remain elastic, lack the ability to dissipate energy and are responsible for the transmission of large seismic forces to other structural components. After yielding, these elements tend to accumulate plastic deformations in repeated loading cycles that can also result in unseating [1]. Several other unseating prevention devices for bridges have been presented in the past decades, namely uid-viscous
*Correspondence to: F. P. Amarante dos Santos, Centro de Investigao em Estruturas e ConstruoUNIC, Faculdade de Cincias e Tecnologia, Universidade Nova de Lisboa, Quinta da Torre, 2829516 Caparica, Portugal. E-mail: fpas@fct.unl.pt

Copyright 2012 John Wiley & Sons, Ltd.

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

891

dampers and metallic dampers [2]. Although these devices are able to dissipate energy, they lack the capacity for re-centering, which is a very important asset to control hinge opening in bridges during seismic actions. To overcome the limitations presented by these devices, alternative solutions for seismic retrotting of structures have been proposed, on the basis of shape-memory alloys (SMAs) [1], a unique class of metallic alloys that exhibit a peculiar thermo-mechanical property, called superelasticity. This property enables the material to withstand large cyclic deformations (up to 8%), without residual strains while developing a hysteretic loop. This high inherent damping, combined with repeatable re-centering capabilities and relatively high strength properties, has led several authors to study the retrot and rehabilitation of bridges using SMA restraining cables [16]. Their efcacy when used as seismic links has been conrmed, as they reduce the relative hinge displacements at the abutments much more effectively than conventional steel cable restrainers [46]. Taking advantage of the superelastic (SE) effect, SMA-based seismic damping devices are aimed to concentrate the energy dissipation in controlled locations. These energy dissipation capabilities can be enhanced by the introduction of an initial prestress in the SE wires. Prestress enables the system to work at higher martensite transformation ratios and, hence, to exhibit higher damping [7,8]. However, when the material ages at constant strain, in the coexistence domain, stressrelaxation occurs, leading to a stress step in the forward transformation plateau of the hysteretic cycle [9,10], decreasing the effectiveness of the original prestress. It is also known that SE elements exhibit permanent deformation because of progressive cumulative creep during cyclic loading, before the full stabilization of the SE hysteresis [11,12]. This causes the net strain produced by a given structural oscillation to be reduced, decreasing the energy dissipation capabilities of the material. Cumulative creep in NiTi alloys is related to dislocations and other lattice defects, generated at high stress during the preceding loading cycle [13,14], and because of the repeated motion of the parent-martensite interface [15]. Wang et al. have shown that with the same number of cycles, the residual strain increases with strain amplitude [16]. The novel semi-active vibration control device, proposed in the present paper, originates from the NiTi-based passive dissipation device reported by Dolce et al. [7]. As to increase its energy dissipation capabilities, the original device was made of sets of prestressed wires working in phase opposition. To avoid problems related to relaxation phenomena and cyclic loading, usually associated with similar passive control devices, the proposed design uses a strategy that allows the continuous adapting of the accumulated strain in the SE wires, on the basis of the response of the device to external excitations. Although unloaded, the NiTi elements remain strain/stress free, preventing the stressrelaxation phenomena. With the occurrence of a dynamic excitation, strain/stress starts to accumulate in the SE wires, enabling higher martensite transformation ratios and increasing the damping capacities of the system. The cumulative strain/stress in the SE wires is a direct result of the motion of the structure itself, with no need of external energy input. Moreover, the device is able to limit the stresses in the wires inside a narrower SE window, mitigating the degradation of the material because of cyclic loading. 2. PRESTRESS, STRESSRELAXATION, AND CYCLIC LOADING IN NITI WIRES Shape-memory alloys in practical use are almost limited to three types: NiTi alloys, Cu-based alloys, and ferrous alloys. For applications regarding the use of SMAs in bridge restrainers, under the direct action of external ambient factors (i.e., wet weather, direct rain), NiTi exhibits an excellent corrosion behavior, when compared with the moisture sensitive Cu-based alloys [17]. NiTi also presents good fatigue properties, high corrosion resistance, similar to stainless steel, and ductility [8]. For this reason, NiTi is used in most of the SE-based civil engineering applications. 2.1. Prestress in superelastic NiTi wires To analyze the inuence of prestress on the dissipation capabilities of a SE control system comprising two NiTi restraining wires working in phase opposition, one uses a simplied version of the physical prototype described in Section 3, corresponding to the conguration presented in Figure 1(a). The device uses 0.406 mm diameter NiTi Euroex SE508 wires (EUROFLEX GmbH, Pforzheim, Germany), with a chemical composition of 51 at% Ni. The device is subjected to a quasi-static harmonic displacement yielding the full extent of the martensitic transformation in the wires. The evolution of the
Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

892

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

36
Pre-stress F [N]

36 24 12 0 12
[mm] F [N]

24 12 0

NiTi SE508 12

24 36 20 15 10 5
F

24 20 36 20 15 10 5 36 0 5 10

[mm]

10

15

15

20

(b) Pre-stress = 0
36 24
F [N]

(c) Pre-stress = 50 MPa

24 12 0 12
[mm] F [N]

NiTi SE508

12 0 12

Pre-stress

24 36 20 15 10 5 0 5 10

24 20 36 20 15 10 5 0 5 10

[mm]

15

15

20

(a) Device

(d) Pre-stress = 100 MPa

(e) Pre-stress = 200 MPa

Figure 1. Inuence of prestress on superelastic hysteresis: (a) device; (b) prestress = 0; (c) prestress = 50 MPa; (d) prestress = 100 MPa; and (e) prestress = 200 MPa.

resulting boxforce-displacement diagrams and the shape of the corresponding SE hysteresis is presented in Figure 1(b)(e), as the prestress in the wires increases from 0 to 200 MPa. The capacity of a given system to mitigate vibrations is usually evaluated through the equivalent viscous damping, dened as zeq ED =4pES0 , where ED is the energy dissipated in a cycle of harmonic vibrations and ES0 is the maximum strain energy [18]. The behavior of the system with no prestress is represented in Figure 1(b), where the NiTi wires enable the development of a full tension/compression SE hysteresis, with an equivalent viscous damping of about 10%. When the NiTi wires are prestressed, Figure 1(c), (d), and (e), the equivalent viscous damping gradually increases as the hysteresis changes from two distinct SE areas to a single hysteretic curve. The equivalent viscous damping associated with the SE hysteresis obtained for a prestress of 200 MPa is 23%. This means that by prestressing the NiTi wires, the SE damping available in the system has doubled. 2.2. Stressrelaxation in shape-memory alloys Superelasticity is a property associated with a diffusionless phase transformation in solids called martensitic transformation. During this transformation, the atoms are cooperatively rearranged into a different crystalline structure with identical chemical composition, through a displacive distortion process [19]. In SMAs, the martensitic transformation changes the material from the parent phase, a hightemperature phase called austenite, to a low-temperature phase called martensite. Being a rst-order phase transition, parent and product phases coexist during the phase transformation [20]. In the stressfree state, an SMA is characterized by four transformation temperatures: Ms and Mf during cooling and As and Af during heating. The rst two (with Ms > Mf) indicate the temperatures at which the forward transformation starts and nishes, respectively. The last two (with As < Af) are the temperatures at which the inverse transformation starts and nishes, Af being the temperature above which the martensite becomes completely unstable [15]. The transformation temperatures of the Euroex SE508 NiTi wires are Mf = 45  C, Ms = 35  C, As = 15  C, and Af = 5  C. When an SE alloy is submitted to a given constraint (stress or strain), in its coexistence phase, the critical stress to induce martensite decreases by a certain amount (s). This decrease is due to residual stresses that promote the formation of martensite, which remains in the material once the applied constraint is removed. These residual stresses are a function of the intensity of the constraint and of its duration [10]. By using the ClausiusClapeyron coefcient (CCC), which translates the relation between temperature and the critical stress to induce the martensitic transformation, the variation of the transformations temperatures Ms and Mf can be expressed as a function of s, yielding Ms = Mf = s/CCC. Hence, the
Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

893

f i transformation temperatures increase when an SE alloy ages under loading Ms > Ms and Mff > Mfi . The CCC of the Euroex SE508 NiTi wires is 6.5 MPaK 1. If the material ages at constant strain in the coexistence domain, a stress step occurs, leading to the stressrelaxation of the material. Figure 2 represents the simplied stresstemperature phase diagram path associated with aging at constant strain in the coexistence domain, together with the variation in the corresponding isothermal hysteresis. Three major regions can be identied in the stresstemperature phase diagram: Md region, where only detweened martensite exists, A region, where only austenite exists, and Mt, dA region, the coexistence domain where all phases can coexist. One can also identify the two transformation strips,[M] and [A], for the direct and inverse transformations, respectively. If an SE cycle is interrupted after describing the rst initial path (o-a-b), the martensitic transformation slowly continues to develop, and strain tends to increase. Because the strain is blocked, a set of auto balanced compressive forces develops within the SE specimen, causing the stress to decrease. If, after a certain period of time, the mechanical tensile cycle is continued, the SE hysteresis deviates from the original path. The stress step reduction is a function of time and austenitic fraction of the material during aging. Estimates of the time constants and amplitudes of changes in the transformation temperatures for NiTi SE alloys are very difcult to obtain because of the large time scales involved in the aging process. In Cu-based alloys, given the smaller time scales involved, variations of Ms are easier to obtain and have been reported to reach about 15% [21]. 2.3. Cyclic loading in NiTi alloys To evaluate the stabilization process, a 2.40 mm diameter Euroex SE508 NiTi wire sample is subjected to an experimental program, consisting of 60 consecutive loading cycles, performed at ambient temperature (20  C) and with four different strain rates, ranging from quasi-static conditions up to a strain rate of 0.333%/s. The tests are carried out using a Zwick/Roell Z050 (Zwick GmbH & Co. KG, Ulm, Germany) testing machine, and the resulting stressstrain diagrams are presented in Figure 3.

Figure 2. Aging under constant strain in the coexistence zone: phase diagram path and correspondent hysteresis.

720 600 480 360 240 120 0

1

= 0.008%

720 600 480 360

1

= 0.067%

60 [%]

240 120

60

1
1

5
= 0.250%

0 720 600 480

[%]

1
1

5
= 0.333%

720 600 480 360 240 120 0

60

360
60

240 120
[%] [%]

Figure 3. Cyclic tensile tests at increasing strain rates: stressstrain diagrams.

Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

894

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

It can be seen that during the cyclic loading, the starting point of the stressstrain diagrams tends to drift away from the origin because of cumulative creep deformation. This drifting is explained by the accumulation of slip deformations during stress-induced transformation, resulting in residual strains, which increase during the rst cycles and tend to stabilize with saturation. The slip deformations induce internal stresses that can assist the formation of stress-induced martensite and, as a consequence, decrease the critical stress needed to induce martensite [13]. Conversely, the level of the reverse transformation stresses changes very little. As a consequence, the width and size of the hysteretic loop is signicantly reduced upon cycling [9] as well as the damping capabilities of the material. As the number of cycles continues to grow, the slip deformations increase the density of dislocations within the crystal structure. These dislocations obstruct the formation of martensite in a way similar to strain hardening in plasticity. As a consequence, the slope of the strainstress curve increases during loading [13]. The key parameters dening the extent of the hysteretic loop, namely the cumulative creep deformation (ep), the variation of the critical stress needed to induce martensite (s), and the equivalent viscous damping (xeq), are quantied and plotted in Figure 4. It can be seen that the variation of these parameters with the number of cycles is rather important. However, an asymptotic behavior is observed, meaning that they tend to gradually stabilize. It can also be seen that, for a given number of cycles, changes of these parameters increase with the strain rate.

3. PHYSICAL PROTOTYPE 3.1. General description of the prototype The physical prototype presented in this section simulates the response of a single degree of freedom dynamic system, equipped with a novel semi-active vibration control device made up of two SE NiTi wire elements working in phase opposition. The objective of the proposed semi-active control device is to limit the hinge opening during a dynamic event, while keeping the stress in the wires inside a prescribed SE window, to mitigate the degradation of the material because of cyclic loading but guaranteeing a minimal threshold to their stress level. As the strain is self-adjusting, the wires are strain/stress-free when not subjected to a dynamic excitation, limiting aging problems. The stress in the SE wires is adjusted by controlling the displacements of the wires at their supports. The two wire supports can assume, independently, two congurations, locked or unlocked. By default, the supports are locked, restraining the SE wires. If the system needs to compensate for an excessively low or high stress in a given wire, it momentarily unlocks the wire through a controlled velocity process and without introducing additional forces into the system. The semi-active system bounds the stresses in the SE wires between an upper and a lower stress limit. During the dynamic oscillations, if the stress in a given wire reaches one of these values, the corresponding support is unlocked until the displacement of the mass is reversed. The process of unlocking and subsequent locking of an SE wire allows the system either to increase their cumulative stress, when compensating for excessively low stresses, or to decrease them, when compensating for excessively high stresses. As the compensation for lower stresses is larger and usually more frequent than the compensation for higher stresses, the net accumulated stress in the SE wires is positive. The proposed semi-active device uses a control strategy that allows the continuous adapting of the accumulated stress in the wires, on the basis of the response of the device to external excitations. As in
0 0.2 0.4 0.6 0.8 1.0 1.2
Cycles [%]

0
0.008%/s 0.067%/s 0.250%/s 0.333%/s

20 40 60 80 100 120 140 160

[MPa]

10.0
0.008%/s 0.067%/s 0.250%/s 0.333%/s

[%]

9.0 8.0 7.0 6.0

0.008%/s 0.067%/s 0.250%/s 0.333%/s

Cycles

10

20

30

40

50

60

10

20

30

40

50

60

5.0

10

20

30

40

50

60

(a)

(b)

(c)

Figure 4. Inuence of cycling on superelasticity: (a) cumulative creep; (b) critical stress to induce martensite; and (c) equivalent viscous damping.
Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

895

an active control system, a controller monitors the feedback measurements and generates appropriate command signals for the device and, as in a passive control system, the control forces are developed as a result of the motion of the structure itself, with no need of external energy input. As the control forces act as to oppose the motion of the structural system, they promote the global stability of the structure [2226]. The prototype is used to simulate the dynamic response of a simple supported bridge with SE restraining cables in the deckabutment interfaces, as illustrated in Figure 5. For the system to be able to continuously read and adjust the stress values in the SE wires, each wire is equipped with a load cell and a linear actuator placed at opposite endings of the wire. The design and manufacturing of the physical prototype was carried out using an educational version of SolidWorks. Its kernel elements are shown in Figure 6. The moving mass module (MMM), which accommodates the load cells, is mounted on rails, enabling the displacement of the module in the direction of the dynamic loading. This displacement and the corresponding acceleration are monitored by a Solartron DC25 (Solartron Metrology, Bognor Regis, United Kingdom) long-stroke linear variable differential transformer displacement transducer and an accelerometer. The rails are xed to a set of bars that also support both of the linear actuators through thick end plates. The linear actuators comprise a clamping device that constrains the movement of the remaining extremity of the SE wires. The bars are rigidly connected to a shake-table by means of four brackets. The base of the MMM (10 20 20 30 cm) is made of steel and can accommodate a series of additional steel plates, with different dimensions and weights, which are xed by two metallic pins. The MMM is provided with a set of four steel beveled shape wheels, with RHP 618/8 bearings (ARN Trading Ltd t/a AHR International, St Albans, UK) and is mounted on a set of two 40 10 300 mm aluminum bars, with wedge-shaped rails. As the angle presented by the wedge-shaped rails is wider than the one shown by beveled wheels, the contact area between the two pieces is minimized, reducing the mechanical friction developed in this interface. 3.2. Description of the miniature load cells To cope with the limited dimensions of the MMM, two miniature load cells were designed, one for each SE wire. The load cells clamp the SE wires into thin aluminum plates, which are laterally xed
Deck Abutment Abutment

Superelastic restrainer cables

Superelastic restrainer cables

Figure 5. Simple supported bridge with a superelastic restraining system.

Linear actuator Moving mass module Pin Load cell Wedged rail End plate Beveled wheel Bracket Clamp Bar

Figure 6. General design concept for the physical prototype.

Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

896

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

to the MMM. These aluminum plates are equipped with a set of four strain gauges disposed in a Wheatstone-bridge conguration to enable accurate force readings. The Wheatstone-bridge conguration is used to help the measurement of small variations in the resistance that the four strain gauges produce, corresponding to a change in the specimens strain eld. This bridge conguration is usually called a full-bridge because it has four active strain gauges. In this particular case, two are mounted in the direction of the bending strain, with one on the maximum strain zone of the plate and the other on the minimum strain zone. The other two acts as Poisson gauges and are mounted transversely to the principal axis of strain, adjacent to the other two strain gauges. To build this bridge, two HBM T-rosettes with two measuring grids, model 1-XY13-1.5/350, were used. Figure 7 represents a schematic drawing showing the position of the two T-rosettes on the force sensor, together with the corresponding full-bridge circuit diagram. In this circuit diagram, VCH is the measured voltage of the signal and VEX is the excitation voltage. An Finite element (FE) model of the force sensor was analyzed in ANSYS (ANSYS, Inc., Canonsburg, Pennsylvania, USA) to obtain the strain eld corresponding to the loading of the SE wire. This FE model enables not only to dene the positioning of the T-rosettes on the force sensor but also to calibrate the thickness of the aluminum plate to obtain an adequate signal from the bridge. The force used in the FE simulation corresponds to a stress level of about 300 MPa in a NiTi SE508 wire with 0.406 mm diameter, which corresponds to half of the maximum stress associated with the loading plateau. The resulting dimensions for the aluminum plate were 60 17 2 mm. The corresponding strain eld in the longitudinal direction of the plate and the displacements along the direction of the SE wire is presented in Figure 8(a) and (b), respectively. The maximum absolute strain in the plate yields 0.25 me, corresponding to a stress of about 18 MPa. The maximum displacement in the plate along the wires axis yielded 0.013 mm. To guarantee that the strain gauge is able to perform its task properly, the strain to be measured has to be transferred faultlessly and free of loss. To achieve this purpose, an intimate connection is required between the strain gauge and the aluminum plate. The surface of the metal was rst cleaned, and a special cyanoacrilate cold curing adhesive (Z70) was used to guarantee the required bonding between the strain gauge and the plate.
T-rosette 1 T-rosette 2 R 1 ( ) R 4 (+ )

R 2 (+ R 3 ( )

R 4 (+ ) R 1 ( ) V EX

V CH

Aluminum plate

R 2 (+

R 3 ( )

Figure 7. Schematics of the load cell. Full-bridge circuit diagram.

(a) Strain eld

(b) Displacement eld

Figure 8. Finite element simulation of the aluminum plate: (a) strain eld and (b) displacement eld.
Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

897

The strain gauges were connected to a National Instruments SCXI-1520 8-Channel Universal Strain Gage Input Module (National Instruments Corporation, Austin, Texas, USA). A virtual instrument (VI) was created using a LabView framework, enabling the signal acquisition from the force sensor and, once calibrated, yielding the desired force readings. A DAQ Assistant Express VI (National Instruments Corporation, Austin, Texas, USA, using NI-DAQmx Software), was used to create, edit, and run the analog input corresponding to the voltage measurement task. By using an averaging process, a sample compression of the data points was performed to attenuate the noise derived from the readings. To calibrate the force-sensor VI, strain readings associated with reference weights were scaled and offset. 3.3. Description of the linear actuators Two Bosh Rexroth EMC 32-125 L100 OF01 precision rolled ball screw actuators were used to adjust the stress level in the SE wires. They are coupled with two SIGMA II Omron-Yaskawa cylindrical servo-motors (Bosch Rexroth AG, Lohr am Main, Germany), model SGMAH-02AAA61D-OY. A SIGMA-II Omron-Yaskawa SGDH-02AE-OY servo-drive is used to control the servo-motor, allowing for analog and pulse inputs for speed, torque, and position control. A special clamping system was designed, allowing the connection of SE the wires to the linear actuators. The clamps are cylindrical-shaped elements made of steel, with 20 mm diameter. They are built up of two main components that when pressed together constrain the SE wire from moving. One of these pieces displays a narrow, longitudinal indentation that helps to x the wire into place. The linear actuator modules are supported by two aluminum plates, measuring 135 80 20 mm. These plates are rigidly xed to a pair of 720 mm long, load bearing aluminum Bosch-Rexroth 30 30 strut proles weighing 0.80 kg/m. These bars also accommodate the rails for the MMM. The Bosch-Rexroth proles are xed on the top stage of the Quanser Shake-Table II, as illustrated in Figure 9, where a view of the complete prototype is shown. 3.4. Description of the control system The control system of the physical prototype is presented next. The control variable of the system is the analog voltage level for the servo-drives speed reference, ranging from zero to 6 V, which corresponds to the highest allowable rotating speed for continuous operation (3000 rpm). The implemented controller, combining proportional, integral, and derivative control actions, can assume two possible states: PID-OFF and PID-ON, which correspond to the unlocked and locked support conditions. Two different setpoints are possible for PID-ON state, corresponding to the upper and lower force limits dened for SE wire. The controller chooses the setpoint depending on the force level in the SE wire at a given instant. This state is activated above or below the two prescribed force thresholds, the high and low PID thresholds, respectively. Between these two force delimiters, the controller sets the PID-OFF state. The differential gap existing between the force thresholds that activate the PID controller, and the corresponding force setpoints, works as an anticipatory feature of the system. It provides the controller with a reference force input while the force in the SE wire is still approaching this value. The controller transfer function Gc(s) comprehends three terms, each one associated with the corresponding control action,

Figure 9. Physical prototype.

Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

898

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

  Ki 1 Td s Gc s Kp Kd s Kp 1 s Ti s

(1)

where s is a complex variable, Kp, Ki, and Kd are constants called gains, Ti = Kp/Ki is the integral time, and Td = Kd /Kp is the derivative time. Being a closed-loop control system, the stress output signal C(s) is fed back to the summing point, where it is compared with the reference stress input R(s), yielding the actuating error signal E(s). The output signal of the controller, U(s), is the analog speed reference input for the servo-drive. The feedback-path transfer function, H(s), corresponds to the force-sensor, which measures the output variable to make it comparable with the reference input signal, resulting in the feedback stress signal, B(s). The controlled process is dened by the transfer function Gcp(s). The implementation was performed using the LabView PID Control Toolset Applications, on the basis of a basic PID algorithm, having as main inputs the setpoint, which is the reference input stress, R(s), the process variable, B(s), the PID gains, and the output range. The PID gains consist of a cluster of three values, corresponding to Kp, Ti, and Td. The over-travel limit function is used to prevent the SE wire from failure by limiting the maximum allowing stress to a prescribed value. The resulting block diagram is presented in Figure 10. The main controller is divided into two independent controllers, each one responsible for one of the SE wires. Each of these controllers has two reference stress inputs, delimiting the effective SE window, and yield a reference speed output

Figure 10. Block diagram of the superelastic vibration control system with two restraining elements.
Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

899

for the corresponding servo-drive. The global output signal from the controller comprehends the reference speeds for both servo-drives, U(s) = U1(s) + U2(s). In the design process of a control loop, to meet given performance specications, the constants Kp, Ti, and Td in Eq. (1) have to be adequately tuned. When the mathematical model of the process cannot be easily obtained, an analytical approach to this tuning is not possible, and experimental approaches have to be used. To improve the performance of the controller, a tuning algorithm based on the ZieglerNichols [27] stability boundary rule was implemented. In this tuning method, the Ki and Kd gains are rst set to zero, and the Kp gain is increased until it reaches the ultimate gain, Ku, at which the output of the loop starts to oscillate. The corresponding period of oscillation, Pu is called the ultimate period. For the proposed prototype, the ultimate gain yields Ku = 0.30 and the ultimate period Pu = 0.7 s. To obtain a quarter decay ratio, which corresponds to a decay of 25% on the transient response in one period, yielding a good compromise between quick response and adequate stability margins [28], the PID gains were set to Kp = 0.18, Ti = 0.35, and Td = 0.084.

4. EXPERIMENTAL RESULTS To assess the performance of the proposed control system, the physical prototype is subjected to a harmonic excitation with a frequency of 1 Hz and an amplitude of 5 cm, using a shake-table. The low and high setpoints of the PID controller are set to 11% and 22% of the maximum force (Fmax = 60 N), and the PID threshold limits are set to 13% and 24% of the maximum force, respectively, resulting in a gap of 2%. In spite of this anticipatory feature of the controller, a small overshoot is observed for both the upper and lower stress limits. To analyze the response of the system to the external excitation, the force time history in one of the SE wires is presented in Figure 11. At the beginning of the dynamic loading, the SE wires are stress/ strain-free. When the MMM starts to move, and during the rst 10 s of the experiment, the control system is switched off. As the stresses induced in the SE wires are in phase opposition, the wires are strained alternately. After the controller is switched on, the system rapidly starts to compensate for low stresses in the SE wires, and the corresponding force readings are shifted upwards. When the MMM moves in one direction and the stress in one of the SE wires decreases below a given threshold, the corresponding linear actuator shifts in the same direction, as it tries to accommodate the displacement introduced by the external excitation. During the controlled phase, the force values in the SE wires are delimited by the two reference stress inputs, also limiting the force which is transmitted to the supports. These values correspond to the low and high setpoints of the PID controller, respectively. The force accumulation in the SE wires enhances the damping capability of the system because the extent of the martensitic transformation is increased. This can be seen in the corresponding force-displacement diagram depicted in Figure 12. When the controller is switched off and the SE wires work alternately, they remain in their austenitic form. This is translated by a linear
CONTROL-OFF CONTROL-ON
Setpoint high PID threshold high PID threshold low Setpoint low Force

30 25 20

PID-ON

PID-OFF

15 10
Accumulated force
PID-ON

5 0

10

20

30

40

50

60

70

80

90

Figure 11. Force time history in one of the superelastic wires.

Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

900

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

20
Initial cycle

15 10 5 0 -5 -10 -5 0

2 1

Final cycle

[mm]

10

15

Figure 12. Force-displacement diagram evolution.

CONTROL-OFF 0.4

CONTROL-ON

0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 0 10 20 30 40 50 60 70 80 90

Figure 13. Acceleration time history.

relation between the total force yielded by the system and the corresponding displacements of the MMM, the initial cycle in Figure 12. When the control system is switched on and the SE wires start working simultaneously, as they begin to accumulate force, the overall stiffness of the system is increased. This increase is represented in Figure 12 by step 1, during which the slope of the line representing the force-displacement relation becomes steeper because both of the austenitic SE wires start contributing to the systems stiffness. After this initial stiffening, as the SE wires continue to accumulate force, stress-induced martensite starts to develop and the system becomes gradually more exible. This process is described by step 2 and leads to the appearance of a stabilized SE hysteretic loop. The shifting of the center of the loop to the negative displacement region means that the system has sustained a rigid-body movement to the left, with respect to its original position. The acceleration time history of the MMM, as monitored during the experimental test, is shown in Figure 13. According to this register, the acceleration levels in the MMM are decreased when the control system is switched on, which is consistent with the additional damping provided to the system by the SE wires. When the external excitation ceases to exist, the acceleration rapidly tends to zero. 5. CONCLUSIONS A physical prototype that simulates the response of a controlled single degree of freedom dynamic system allows to illustrate the potential of a novel semi-active vibration control device on the basis of SE NiTi wires. The design of the semi-active vibration control device is based on two main guidelines: 1. 2. SE wires are stress-free until the beginning of the dynamic excitation to prevent stressrelaxation; Control of the stress levels in the SE wires during the dynamic excitation. For low stress levels in the wires, a process of stress accumulation enables higher martensite ratios and therefore higher dissipation capabilities. In the case of high stress levels, a stress relieve process prevents
Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

Copyright 2012 John Wiley & Sons, Ltd.

SEMI-ACTIVE VIBRATION CONTROL DEVICE BASED ON SUPERELASTIC NITI WIRES

901

the wires from yielding while also limiting the force transmitted to the adjacent structural elements. By limiting the stress levels in the SE wires, one also helps to mitigate the degradation effects in the SE material because of cyclic loading. A thorough description of the kernel parts of the prototype provides an extended insight into the developed approach to meet these guidelines. The integration of the control algorithms into the prototype is also addressed. The control algorithms, on the basis of a PID approach, are implemented using a LabView platform. Several considerations regarding this type of control systems are made, including relevant aspects related to controller tuning. To assess the performances of the proposed semi-active vibration control device, the physical prototype was subjected to an external harmonic excitation, induced by a shake-table. Analyzing the results of these experimental tests, one can conclude that the control device performed adequately. It promotes the force accumulation in the SE wires, enabling the development of a wide SE hysteretic loop and also controls their maximum force and decreases the acceleration of the MMM. The main drawback of the presented mechanism consists in the fact that the stress control in the SE NiTi wires is achieved through the use of electromechanical actuators. The compensation either for high or low stresses in the SE wires is achieved by the movement of these actuators, which are placed in series with the wires, in the force ow path. This means that the actuators have to be able to sustain the forces developed in the SE wires during the dynamic loading, which, in the case of civil engineering structures, are considerable high. To comply with these forces, it would be necessary to use high bearing capacity electromechanical actuators, with high costs and high energy demands, undesirable to a restraining mechanism. To meet the requirements of fast response and low power consumption, promising alternatives are currently being studied.
REFERENCES 1. Padgett JE, DesRoches R, Ehlinger R. Experimental response modication of a four-span bridge retrot with shape memory alloys. Structural Control and Health Monitoring 2009; DOI: 10.1002/stc.351 2. DesRoches R. Application of shape memory alloys in seismic rehabilitation of bridges. Technical Report NCHRP-91, IDEA Program, Transportation Research Board, National Research Council, 2005. 3. Johnson R, Padgett JE, Maragakis ME, DesRoches R, Saiidi MS. Large scale testing of Nitinol shape-memory alloy devices for retrotting of bridges. Smart Materials and Structures 2008; 17(3). DOI:10.1088/0964-1726/17/3/035018 4. Andrawes B, DesRoches R. Comparison between shape memory alloy restrainers and other bridge retrot devices. ASCE Journal of Bridge Engineering 2007; 12(6):700709. 5. DesRoches R, Delemont M. Seismic retrot of simply supported bridges using shape memory alloys. Engineering Structures 2002; 24:325332. 6. DesRoches R, Pfeifer T, Leon RT, Lam T. Full-scale tests of seismic cable restrainer retrots for simply supported bridges. Journal of Bridge Engineering 2003; 8(4):191198. 7. Dolce M, Cardone D, Marnetto R. Implementation and testing of passive control devices based on shape memory alloys. Earthquake Engineering and Structural Dynamics 2000; 29:945968. 8. Zhang Y, Zu S. A shape memory alloy-based reusable hysteretic damper for seismic hazard mitigation. Smart Materials and Structures 2007; 16:16031623. 9. Isalgue A, Torra V, Yawny A, Lovey FC. Metastable effects on martensitic transformation in SMA. Part VI. The ClausiusClapeyron relationship. Journal of Thermal Analysis and Calorimetry 2008; 91(3):991998. 10. Torra V, Pelegrina JL, Isalgue A, Lovey FC. Metastable effects on martensitic transformation in SMA (I): Recoverable effects by the action of thermodynamic forces in parent phase. Journal of Thermal Analysis and Calorimetry 2005; 81:131135. 11. Torra V, Isalgue A, Martorell F, Terriault P, Lovey FC. Built in dampers for family homes via SMA: An ANSYS computation scheme based on mesoscopic and microscopic experimental analyses. Engineering Structures 2007; 29:18891902. 12. Carreras G, Casciati F, Casciati S, Isalgue A, Marzi A, Torra V. Fatigue laboratory tests toward the design of SMA porticobraces. Smart Structures and Systems 2011; 7(1):4157. 13. Moumni Z, Van Herpen A, Riberty P. Fatigue analysis of shape memory alloys: energy approach. Smart Materials and Structures 2005; 14:S287S292. 14. Miyazaki S, Imai T, Otsuka K. Effect of Cyclic Deformation on the Pseudoelasticity Characteristics of Ti-Ni Alloys. Metallurgical and Materials Transactions A 1986; 17(1):115120. 15. Otsuka K, Wayman CM (eds.). Shape memory materials. Cambridge University Press: Cambridge, United Kingdom, 1998. 16. Wang X, Xu B, Yue Z. Phase transformation behavior of pseudoelastic NiTi shape memory alloys under large strain. Journal of Alloys and Compounds 2008; 463:417422. 17. Auguet C, Isalgue A, Lovey F, Martorell F, Torra V. Metastable effects on martensitic transformation in SMA Part 4. Thermomechanical properties of CuAlBe and NiTi observations for dampers in family houses. Journal of Thermal Analysis and Calorimetry 2007; 88(2):537548. 18. Chopra AK. Dynamics of Structures: theory and applications to earthquake engineering (2nd edn). Prentice-Hall, Inc.: New Jersey, USA, 2001. Copyright 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc

902

F. P. AMARANTE DOS SANTOS, C. CISMAIU AND J. PAMIES TEIXEIRA

19. Cismasiu C, Santos FPA. Numerical simulation of superelastic shape memory alloys subjected to dynamic loads. Smart Materials and Structures 2008; 17(2):2536. 20. Patoor E, Lagoudas DC, Entchev P, Brinson C, Gao X. Shape memory alloys, Part I: General properties and modeling of single crystals. Mechanics of Materials 2006; 38:391429. 21. Auguet C, Isalgue A, Torra V, Lovey FC, Pelegrina JL. Metastable effects on martensitic transformation in SMA Part VII. Aging problems in NiTi. Journal of Thermal Analysis and Calorimetry 2008; 92(1):6371. 22. Housner GW, Bergman LA, Caughey TK, Chassiakos AG, Claus RO, Masri SF, Skelton RE, Soong TT, Spencer BF, Yao JTP. Structural control. past, present, and future. Journal of Engineering Mechanics 1997; 123(9):897971. 23. Soong TT, Spencer BF. Supplemental energy dissipation: state-of-the-art and state-of-the-practice. Engineering Structures 2002; 24:243259. 24. Spencer BF, Nagarajaiah S. State of the art of Structural Control. Journal of Structural Engineering 2009; 129(7):845856. 25. Spencer BF, Sain MK. Controlling buildings: A new frontier in feedback. Special Issue of the IEEE Control Systems Magazine on Emerging Technology 1997; 17(6):1935. 26. Symans MD, Constantinou MC. Semi-active control systems for seismic protection of structures: a state-of-the-art review. Engineering Structures 1999; 21:469487. 27. National Instruments Corporation. PID Control Toolset User Manual, 2001. 28. Franklin GF, Powell JD, Emami-Naeini A. Feedback Control of Dynamic Systems (3rd edn). Addison-Wesley: New Jersey, USA, 1994.

Copyright 2012 John Wiley & Sons, Ltd.

Struct. Control Health Monit. 2013; 20:890902 DOI: 10.1002/stc