You are on page 1of 7
APPLICATION OF MAGNETORHEOLOGICAL DAMPERS TO SEISMICALLY EXCITED STRUCTURES Shirley J. Dyke', Fu Yit, Stephen Frecht, and J. David Carlson* ‘Department of Civil Engineering Washington University St.Louis, Missouri 63130 ABSTRACT ‘This paper presents the results of a recent experiment con- ducted to demonstrate the potential of magnetorheological (MA) devices for seismic control of civil engineering struc- tures. The study discussed herein focuses on the use of mul- tiple MR devices in a structure, extending the results of prior experiments with MR devices. Four parallel-plate, shear mode MR dampers are used to control a six-story structure. MR dovices are selected because they have features which make them quite promising for civil engineering applications ‘The experiment is conducted using the facilities in the Wash- ington University Structural Control and Earthquake Engi neering Laboratory NOMENCLATURE Av T+Bym. parameters of the Bouc-Wen model Chath parameters of the MR damper model 0a» Cob parameters of the MR damper model Sei desired control force for the ah device fi control force produced by the th device KG) controler transfor function matrix hy siitfness of the MF damper model Von maximum contro! voltage applied to device y control voltage to the ah control device x, state vector of the structural system iy ‘ground acceleration in absolute acceleration of the fh floor hysteretic variable of the Boue-Wen model Mechanical Products Division Lord Corporation Cary, North Carolina 27511 1 INTRODUCTION Since the concept of structural control was first proposed to the civil engineering community in 1972 [19], a wide variety of techniques have been developed and investigated [13]. In ‘July 1998, the Second World Conference on Structurat Con- trol was held in Kyoto, Japan, Over 400 researchers from 20, ‘countries participated. Many researchers at the conference {elt that semi-active structural control seems to offer quite a promising opportunity for widespread acceptance by the civil engineering community ‘The primary reason for this sentiment lies in the inherent characteristics of semi-active dovices. Semi-active devices are stable because they do not have the ability to input energy into the structural system (Including the control device and the structure). However, they offer the ability to dynami- cally vary their propertis, and are expacted to be quite etfec- tive for civil engineering applications. Adcitionally, they typically have low power requirements, eliminating the need for a large extemal power source. Furthermore, because they ‘are inherently stable, high authority controllers can be implo- ‘mented, resulting ina control system that may surpass that of comparable active systems, ‘A variety of semi-active devices have been proposed in recent years, including variable orifice dampers, variable fric- tion devices, adjustable tuned liquid dampers, variable stif- rnoss dampers, and dovices based on controllable fluids Magnetorheological (MR) dampers are classified as control lable fluid devices, and possess many features that make them attractive for civil engineering applications. They require minimal power, are able to respond quickly to changes in tho control input, are expected to be relatively inexpensive, and ‘are considered to be quite reliable because they have few moving parts, Moreover, recent tests on a 20-ton MR damper at the University of Notre Dame have demonstrated that those devices can provide forces of the magnitude required for full-scale structural control applications [17] 410 Bocause these devices are highly nonlinear, one of the main ‘challenges in the application of this technology i in the devel- ‘opment of suitable control algorithms. Analytical investiga- tions have demonstrated that the performance of a control systems based on MR dampers is highly dependent on the choice of algorithm employed [7]. A clipped-optimal control algorithm has been effectively ulized for controlling mutt slory test structure using a single MR damper [10-1]. Fur- ther, analytical investigations extended this approach to the ccase in which multiple dampers are employed (6), ‘The focus of this paper is to provide experimental verification of the efficacy of this control system when multiple MR dovices are used to control a structure, Four paralle-plate, shear-mode MR dampers are employed to control a six story test structure subjected to uniaxial ground acceleration. Each (ofthe control devices can generate a maximum force of 20 N, which is approximately 1 98% the weight ofthe structure. Two MR control devices are located betwoon the base and first floor, and two are located between the first and second floors. Experiments are conducted at the Washington University Structural Control and Earthquake Engineering Laboratory. The results demonstrate that multiple MR dampers can be ctfectivaly used to reduce the seismic responses of a struc- tur. 2 EXPERIMENTAL SETUP Experimental investigations were performed in the Washing- ton University Structural Control and Earthquake Engineering Laboratory [5] (also see: wiw.seas wustl.edu/resoarch! (quake). This experimental facility houses a uniaxial seismic simulator and was established to provide a testbed for exper- imental verification of new seimic contro! techniques. The simulator consists of a 1.5 x LS m® (5x5 ft?) aluminum sliding table mounted on high-precision, low-tiction, linear bearings. The sliding table is driven by @ hydraulic system consisting of an 80 kN (18 kip) hydraulic actuator, two IST Systems 162M 15 GPM servo-valves, a 40 GPM hydraulic accessory manifold, a 43 GPM hydraulic power unit, and a '5910 digital controller. The operational frequency range of the instrument is 0-50 Hz, the stroke is 15 om (6 in), velocities of cover 75 cm'sec (30 inisec) are achievable, and accelerations of greater than 4 gs can be imparted to a 900 kg (1 ton) test load, The MR devices employed in this experiment are prototype dovices, shown schematically in Fig. 1. The tested devices ‘were obtained from the Lord Corporation for testing and eval- uation (see: wawmrfluid.com). The device consists of two steo! plates placed parallel to each other. The dimensions of the damper are 4.45x 1.92.5 om® (1.75 0.75% 1.0 in° The magnetic field produced in the device is goneratod by an electromagnet consisting of a coll at one end of the davice, Forces are generated wien the moving plate, coated with a thin foam saturated with MR fluid, sides between the two par- allel plates. 4il Direction of Motion. —_—_ MR Fluid ‘Satured Foam Coil Figure 1: Schematic Diagram of a Shear Mode MR Damper. ‘The outer plates of the MR device are 0.635 om (0.25 in) apart, and the force capacity of the device is dependent on the strength of the fluid and on the size of the gap between the side plates and the center plate. In this experiment, a con- tor plate with a thickness of 0.495 cm (0.195 in) is selected resulting in a gap of 0.071 cm (0.028 in). Power is supplied to the device by a regulated voltage power supply driving a DC to pulse-width modulator (PWM). Tho PWM supplies voltage pulses to the MR damper at a fro- quency greater than 20 kHz, and the command voltage to the ‘cuit controls the duty cycle ofthe individual pulses. The cir- cuit has been calibrated such that a SV command signal cor- responds to 100% duty cycle. The test structure used inthis experimentis a six-story, single bay, stee! frame (Fig. 2). The structure is 180 om (74 In) tall and has a mass of 147 kg (325 Ib) which is distributed uni- formly among the floors. Parallel-plate, shear-mode MR dampers are placed between the ground and first floor, and between the first and second floors of the structure. Two dampers are used in each location. ‘Sonsors are installed in the model building for use in deter: ‘mining the control action. Accelerometers located on the first, second, fourth and sixth floors provide measurements of the absolute accelerations, and a force transducer is placed in series with one MR damper on each floor to measure the control force f being applied to the structure, Note that only these six measurements are used in the control algorithm. A ‘schematic diagram of the experimental setup is shown in Fig, 3, The laboratory has a DSPT Siglab 20-42 dynamic signal and system analyzer from DSP Technology for data acquisition This is a DSP-based instrument that can measure four input ‘channels and provide two output signals, has digitally imple- ‘mented anti-aliasing fiters, zoam ospabilties, and a dynamic range of 90 dB. Features include a function generator, oscillo- ‘scope, spectrum analyzer, etc Implementation of the diserete controler is performed using a realtime control implementation sysiom developed. by ‘Space, Inc.. The system is modular for easy expansion, and Figure 2: Photo of Experimental Structure, consists of a 82 channel A/D board, a 6 channel D/A board, a digital signal processing (DSP) board based on the BOMHz ‘TMS320C40 chip. Controllers developed in MATLAB [14] can readily be downloaded to the OSP board using the Real-Time Workshop. 3. SYSTEM IDENTIFICATION In this experiment a variation of the approach used by Spen- ‘cor and Dyke [15] was used to identify a model of the system, The procedure used to model these systems is described in the following sections. 3.1. Parallel-Plate, Shear Mode MR Dampers ‘The first step in developing a model of the MR device was to ‘obtain experimental responses of the MR damper. A load frame was constructed to obtain this data. The load frame employs a 2 kip Shore Western hydraulic actuator to apply forces to the MR device, which is placed in series with a force transducer made by PCB, Ine. A photo of the load frame is provided in Fig. 4 The response of this device to a sinusoidal displacement input for a set of constant voltages is provided in Fig. §. From these results, the maximum force generated in a device is, approximately 20 N. The dynamic range (defined as the ratio. of the peak force with a maximum control input to the peak force with a OV input) of the device used in this experiment is approximately 7. The MR effect begins to saturate at approxi- 412 Computer Controller Figure 3: Diagram of MR Controller implementation. Figure 4: Photograph of the Load Frame used to Characterize the Shear Mode MR Damper, mately 35 V. The rise time of the system, primarily deter- rmined by the resistance and inductance of the col itself, is ‘approximately 40 msec. Spencer et al, [16] recently developed a phenomenological ‘model of an MR dovice. The damper used in this study used IMR fluid in valve-mode, and consisted of a piston device. The two chambers on either side of the piston were filed with MR fluid. Forces are devoloped in this device by forcing fluid through an annular orifice inthe piston, and the electromag- ‘not was positioned to generate the magnetic field across the office. The device used an accumulator to prevent cavitation of the fluid. The proposed model of the device captured the salient features of the device. A spring was included in the ‘mode! to capture the behavior due fo the accumulator as well {as a dashpot to account for leakage around the piston. Force (Ibf) Displacement (in) ~ Figure 5: Response of MR Devices to Sinusoidal Displacement at Various Constant Command Levels. ‘The MA devices used in this study are paralietlate, shoar- mode dampers. Based on the slightly diferent characteristics ‘observed in the experimentally measured responses of the damper, the earlier model was not suitable for modeling this dovice. Thus, the phenomenological model used in the prior studies was modified by removing the elements correspond- ing to the accumulator and the leakage past the piston. The resulting mechanical model is shown in Fig. 6. The force in this system is given by F = cgithgr+oz a where the evolutionary variable < is governed by ~vldelel"""—Balel"+ az 2) By adjusting the parameters of the model y, Bn, and A, ‘one can control the linearity in the unloading and the smooth: ress of the transition from the pre-yield to the post-yield region. To use the device for control purposes, a model is required that is capable of predicting the behavior of the MR damper for a time-varying command input. Thus, the functional dependence of the parameters on the command voltage were determined, The folowing relations are proposed a(u) = 4, +044 and eg ou) ° jouc-Wen t Figure 6: Mechanical Model of the MR Damper. 413 In addition, as mentioned previously, the resistance and Inductance in the circuit introduces dynamics to the system, These dynamics are observed as a time lag in the response (of the device to changes in the command input. These dynamics are accounted for through the first order Fier = -n(u-») “ where v is the command voltage applied to the PWM circuit A constrained nonlinear optimization was used to obtain the ten parameters{16). The optimization was performed using a sequential quadratic programming algorithm available. in MATLAB [14]. The optimized parameters that were deter- mined to best fit the data in a varoty of representative tests are: a, = 226Niom, a, = 246Néom-V, co, = 0.032N-sec! om, “tag = OO2N-seclem-V, hy = 0.026Niem, n= 2, A = 100, y = 300cm™, B = 300 cm?, and n = 80sec". A Tepreseniative comparison between the experimental response and predicted results is shown in Fig. 7 3.2 Test Structure ‘To identity a model of the structure being controlled, a varia tion of the method developed by Spencer and Dyke [15] was used, In this approach an external device is required to impart external inputs to the structure corresponding to each ‘of the control inputs as well as the ground excitation. How- ‘over, using an actuator to excite the structure between the second and third floors of the structure would add mass to the structure, altering the transfer functions of the system, ‘Thus, some modification in the system identification proce- ure was required The stops in the modified approach are: i) experimental paring the transfer functions obtained experimentally to those of the lumped mass model, we found that the transfer func- tions were quite similar, and the poles of the experimental structure were within approximately 10% of those found in the lumped mass model. Thus, the oigenvalues of the lumped mass model were modified using the system transformation matrix, To verity the model, a comparison was then ps formed in the time domain using experimental data and pre- dicted structural responses to a measured ground acceleration. 4 CLIPPED OPTIMAL CONTROL DESIGN ‘One of the main challenges in semi-active control is the development of an appropriate control algorithm that can take advantage of key features of the control device. An important requirement of the control algorithm is that it be implement- able in full-scale applications. To be implementable, the algo- rithm should use available measurements, “such as accelerations, in determining the control action. Dyke, et al [7, 10, 11] have proposed a clipped-optimal control siratogy based on acceleration feedback for the MR damper. Analyti- cal and experimental studies have demonstrated that the MA ‘damper, used in conjunction with the clipped-optimal control algorithm, was effective for controling a multi-story structure with a single MA damper [10,11]. Dyke and Spencer [7] then extended the control algorithm to control muttple MR devices, and performed a numerical study demonstrating the efficacy of the control algorithm. This extension of the control algorithm is used in this experiment, and is summarized herein Inthe clipped-optimal controller, the approach is to append n force feedback loops to induce each MR damper to produce ‘approximately a desired control force. The desired control force of the ah MR damper is denoted /,. A linear optimal controller K,(s) is designed that caicilates a vector of desired control forces, f. = [fy 4-2 + fey|’» based on the measured structural respons vecior y ard the measured control force vector f, <1 y) L {-scoxft}h 6 whore [(.)'s the Laplace transform. Although the contrllor K.,(3) can be obtained from a variety of synthosis mothods, HY LOG strategies are advocated herein because of the stochastic nature of earthquake ground motions and because of their successful application in other cv engineering struc- tural contol applications (8-1) To discuss the algorithm used for determining the contro! action, consider the th MR damper used to control the struc ture. Because the response of the MR damper is dependent (on the relative structural displacements and velocities at the point of attachment of the MR damper, the force generated by the MR damper cannot be commanded; only the voltage », applied to the current driver of the ah MR damper can be direcily controlled. To induce the MR damper to generate approximately the corresponding desired optimal control force f,,, the command signal v, is selected as follows. When the ah MR damper is providing the desired optimal force (ie. f, = f,), the voltage applied to the damper should remain at the present level. If the magnitude ofthe force pro- duced by the damper is smaller than the magnitude of the desired optimal force and the two forces have the same sign, the voltage applied to the current driver is increased to the ‘maximum level so as to increase the force produced by the damper to match the desired control force. Otherwise, the ‘commanded voltage is set to zero. The algorithm for selecting the command signal for the th MR damper is graphically rep resented in Fig. & and can be concisely stated as Via Hes LR) © Where Vins: is the maximum voltage to the PWM citcuit, and 11(-)is the Heaviside step function. A block diagram of this, semi-active control system is shown in Fig. 9. In the block iagram, the dependence of the MR damper forces on the structural responses is indicates by the link feeding back the vectors x, and X,, which contain the relative structural dis- Figure 8: Graphical Representation of Algorithm for Selecting the Command Signal. 414 placements and velocities atthe attachment points of the MRL damper. One of the attractive features of this control strategy is that the feedback for the controler is based on readily obtainable acceleration measurements, thus making them quite imple- ‘mentable. In addition, the proposed control design does not require @ model for the MR damper, although the model ofthe damper is important to system analysis. y | £ Ma A ett 7 [secure t + Fa.6 Ks) Control Law © Figure 9: Block Diagram of the ‘Semi-Active Control System. 5 EXPERIMENTAL RESULTS To intially evaluate the efficacy of the control system, sinusoi- dal tests near the frst two modos of the structure were con- ducted. Thiis approach tests the capabilities of the controller ina worst case scenario. In the first set of tests the excitation frequency was chosen to be 1.38 Hz, which is nearly the first ‘mode of the structure. The ground acceleration amplitude was chosen to produce the largest displacements possible without causing damage to the structure, Because the exclta- tion frequency is quite close to the fundamental made of the structure, the uncontrolled test was not conducted to avoid damaging the structure. However, the performance of the ‘semi-actively controlled structure was compared to that ofthe ppassive-on configuration, in which both dampers are supplied with a constant SV command input. Here, the semi-actively control system achieved an additional 22% reduction in sixth floor absolute acceleration over that of the passive-on sys- ‘tem, Fig. 10 shows the resulting absolute acceleration responses of the sixth floor ofthe structure. In the second set of tests the excitation frequency was cho- ‘sen to be 4.2 Hz, nearly the second mode of the structure, In these tests the controled system achioved a 62% reduction in the sixth floor absolute acceleration over the uncontrolled system (with no MR devices present). This result represents ‘an improvement of 20% over the passive-on response at this frequency. Further, these results demonstrate that the semi- actively controled structure has the ability to significantly reduce the structural responses in more than one mode of 415 vibration. Fig. 11 shows the experimentally measured abso- lute acceleration responses of the sixth floor ofthe structure. 6 CONCLUSION In this experiment the efficacy of a clipped-optimal control algorithm developed for use with multiple magnetarheological ‘dampers was investigated using the facilities in the Washing- ton University Structural Control and Earthquake Engineering Laboratory. The results experimentally demonstrated that the ‘control systom, which employs four magnetorheological MR ‘dampers, is affective for the control of a six story test struc- ture. Each of the control devices is capable of supplying a maximum force of 20 KN, which is only 1.36% the weight of the entire structure, Moreover, the performance af the semi: active controller surpassed that ofthe passive-on control sys- ‘tem at the first two modes of vibration, Video clips that visually document the results of this experi- ment are available at www.soas.wustl edu/resoarch’quako/. Further information can be obtained by contacting Dr. S. J Dyke at sdyko @cive.wusil edu, Sixth Floor Absolute Accelerations (cm/sec*) Time (sec) Figure 10: Passive-On and Semi-Active System Responses to 1.38 Hz Sine Wave (First Mode). i,itit Sixth Floor Absolute , Accelerations (cm/sec?) i Time (sec) Figure 11: Uncontrolled, Passive-On, and Semi-Active ‘System Responses to 4.2 Hz Sine Wave (Second Mode). ACKNOWLEDGMENTS This research is supported in part by the National Science Foundation Grant No, CMS 97-33272. The authors would like lo thank Lord Corporation for supplying the prototype MRL dampers used for this study. Additionally, the first author ‘would like to acknowledge the Washington University Schoo! ‘of Enginoering and Applied Sciencas for fuly supporting M! development of the Washington University Structural Control and Earthquake Engineering Laboratory REFERENCES [1] Carlson, J.D. (1984). “The Promise of Controllable Fls ids," Proc. of Actuator 94 (H. Borgmann and K. Lenz, Eds,), AXON Technologie Consult GmbH, pp. 266-270. {2} Carlson, J.D. and Weiss, K.D. (1994). "A Growing Attrac tion to Magnetic Fluids,” Machine Design, Aug., pp. 61 64, {3} Carlson, J.D, Catanzarite, DM, and St. Clair, KA (1996). "Commercial Magneto-Rheological Fluid Devie- 8," International Journal of Modern Physics, Vol. 10, Nos. 23 & 24, pp. 2857-2865. [4] Carlson, 0. and Spencer Je, BF (1996). "Magneto- Rheological Fluid Dampers for Semi-Active Seismic Con: tol” Proc. of the 3rd Int. Cont. on Motion and Vibr. Con- trol, Chiba, Japan. Dyke, SJ. (1998). "Design and Development of tho Washington University Seismic Simulator Facility,” Pro- ‘coodings, 12th ASCE Engineering Mechanics Specialty Conference, pp. 762-768, La Jolla, Calfornia, May 17~ 20, 1998, Dyke, Su., and Spencer Jr, BF (1996). "Seismic Re- sponse Control Using Mutiple MR Dampers.” Proc. of tho 2nd International Workshop on Struc. Control, Hong ong, December. i} 8) [7] Dyke, S.J. and Spencer Jr., BF (1997). “A Comparison (of Semi-Active Control Strategies for the MR Damper,” Proc. ofthe international Conf. on Inteligent Information ‘Systems, IASTED, pp. 680-584, Bahamas, December e106 [8] Dyke, Su., Spencer Jr, B.F, Quast, P, Sain, MiK., Kas- pari Jr, D.C. and Soong, TT. (1996). "Acceleration Feed- bback Control of MDOF Structures,” J, of Engrg. Mech., ASCE, Vol. 122, No. 9, pp. 907-918 {9} Dyke, Sw, Spencer Jr, BF, Quast, P, Kaspari Jr. and Sain, M.K (1996). Microcomputers in Civil Engineering: Spocial Issuo on Active and Hybrid Structural Control, Vol. 11, 1996, pp. 905-323 416 [10] Dyke, S.J, Spencer Je, BF, Sain, MK. and Carlson, J.D. (1996). "Experimental Verification of Semi-Active Struc tural Control Strategies Using Acceloration Feedback, Proc. of the 3rd International Conf. on Motion and Vibr. Control, Chiba, JAPAN, September, 1996, Vol. 3, pp. 291-296, [11] Dyke, $1, Spencer Jr, BF, Sain, MK. and Carlson, J.D. (1998). “An Experimental Study of MR Dampers for Seis mic Protection," Smart Materials and Structures: Special Issue on Large Civil Structures (in press) [12] Fujine, ¥,, Soong, TT. and Spencer Jc, BF. (1996). "Structural Control: Basic Concepts and Applications. Proc, of he ASCE Structures Congress XIV, Chicago, ll no's, pp. 1277-1287. [13] Housner, G.W., Masti, S.F, and Chassiakos, A.G., Eds. (1994), Proc. of the First World Cont. on Struc. Control, International Association for Structural Control, Los An: geles. [14] MATLAB (1997). The Math Works, Inc. Natick, Massa chusets, [15] Spencor Jc, BF and Dyke, S.J. (1996). “Semi-Active Structural Control: System identification for Synthesis and Analysis," Proc. of the 1st European Cont. on Struc. Control, Barcelona, [16] Spencer Jr, BF, Dyke, S.J, and Carlson J.D. (1997). “Phenomenological Model for Magnetorheological Dampers." Journal of Engineering Mechanics, ASCE, Vol. 128, No. 3, pp. 230-238. [17] Spencer Jr. BF, Carlson, J.D. Sain, MK., and Yang, G. (1997). "On the Current Status of Magnetorheological Dampers: Seismic Protection of Full-Scale Structures,” Proc, of tho Amer. Control Cont, pp. 458-62. [18] Spencer Jr, BF and Sain, M.K. (1997). "Controting Buildings: A New Frontier in Feedback," IEEE Control ‘Systems Magazino: Special issue on Emerging Techno!- ogies (Tariq Samad Guest Ed), Vol. 17, No. 6, 9p. 19-35. [19] Yao, J.P. (1972). "Concept of Structural Control,” Jour- ral of the Structural Division, ASCE, 98(ST7), 1567— 1874, [201 Vi, F, Dyke, S.J., Frech, 8. and Carlson, JD. (1988). “in- vestigation of Magnetorheological Dampers for Earth- quake Hazard Mitigation" Proc. of the Second World Cont. on Structural Control, Kyoto, JAPAN, June 30-July 2, 1998,

You might also like