In the Proceedings of the 3rd International Conference on Motion and

Vibration Control, September 16, Chiba, Japan, 1996, Vol. III, pp. 291296.

EXPERIMENTAL VERIFICATION OF SEMI-ACTIVE STRUCTURAL CONTROL

STRATEGIES USING ACCELERATION FEEDBACK
S.J. Dyke,* B.F. Spencer Jr.,* M.K. Sain and J.D. Carlson
*Dept.

of Civil Engineering and Geological Sciences, University of Notre Dame, Notre Dame, IN 46556, U.S.A.
Dept.

of Electrical Engineering, University of Notre Dame, Notre Dame, IN 46556, U.S.A.

Mechanical

Products Division, Lord Corporation, Cary, NC 27511, U.S.A.

ABSTRACT
Because semi-active control systems combine the best
features of passive and active control, they appear to have
significant potential to advance the acceptance of structural control as a viable means for dynamic hazard mitigation. This study demonstrates experimentally the efficacy
of recently proposed acceleration feedback control strategies for semi-actively controlled structures subjected to
earthquake excitation. A new type of semi-active control
device, known as a magnetorheological damper, is implemented in a three-story model building and tested on a
uniaxial shaking table. The results reported herein indicate
that this semi-active control system is quite effective at
seismic response reduction.

1. INTRODUCTION
Semi-active control devices have received a great deal of
attention in recent years, because they offer the adaptability of active control devices without requiring the associated large power sources. As defined here, a semi-active
control device is one that cannot increase the mechanical
energy in the controlled system (i.e., including both the
structure and the device), but has properties that can be
dynamically varied. Several such devices have been proposed for civil engineering applications (see, e.g., [1, 2]
and the references contained therein).
Because of its mechanical simplicity, low operating power
requirements, environmental robustness, and demonstrated potential for developing forces sufficient for full-scale
applications [36], one particularly promising class of
semi-active control devices for seismic response reduction
is found in magnetorheological (MR) dampers [69]. MR
dampers use MR fluids to produce controllable dampers.
MR fluids typically consist of micron-sized, magneticallypolarizable particles dispersed in a carrier medium such as
mineral or silicone oil. They are the magnetic analogs of

electrorheological (ER) fluids, and like ER fluids, the essential characteristic of the MR fluids is their ability to reversibly change from a free-flowing, linear viscous fluid
to a semi-solid in milliseconds when exposed to a magnetic field (or in the case of ER fluids, an electric field).
While there are many similarities between MR and ER
fluids, there are a number of important differences [3, 4].
For example, the achievable yield stress of MR fluids is an
order of magnitude greater than its ER counterpart, and
MR fluids are not sensitive to impurities such as are commonly encountered during manufacturing and usage.
Moreover, devices employing the MR fluid can be controlled with a low power (e.g., less than 50 watts), low
voltage (e.g., ~1224V), current-driven power supply outputting only ~12 amps. This level of power can be readily supplied by batteries.
Although MR dampers offer significant promise, they are
intrinsically nonlinear (as are most semi-active control devices), making it a challenging task to develop control
strategies that can optimally exploit their unique features.
Many of the approaches proposed in the literature to control semi-active systems either use only local information
in the control algorithm, or are based on state feedback
[1012]. While local controllers can be effective, their
ability to achieve global response reduction is limited.
Moreover, the use of state feedback is generally viewed as
being impracticable for civil engineering structures. Alternatively, control strategies based on acceleration measurements can be readily implemented in full-scale structures
[1315]. For semi-actively controlled structures, Dyke, et
al. [7, 8] have proposed a clipped-optimal control strategy
based on acceleration feedback and demonstrated via simulation its efficacy for seismic response reduction.
The focus of this paper is to present experimental verification of the acceleration feedback control strategies developed by Dyke, et al. [7, 8]. Following a description of the
experimental setup, the process of system identification

and control design are discussed. This control design is

then implemented in the three-story test structure configured with a single MR damper. The experimental results
reported herein indicate that high performance can be obtained with a semi-active control system using acceleration feedback control strategies.

Wires
Wirestoto
Electromagnet
Electromagnet

AA
AA
AAA
Bearing & Seal
AAAA
AAAA
AA
AAAA
AA
MR Fluid AAAA
AA
AAAA
AAA
AA
AA
AA
Coil
AAAA
AAA
AA
AA
A
AAAA
AAA
Diaphragm AAAA
AA
AAAAAAA
AAAA
AA
AAA
Accumulator AAAA
AAA
AAAA
AAA
AAAA
Bearing & Seal

2. EXPERIMENTAL SETUP

MR Fluid

Figure 1 is a diagram of the semi-actively controlled,

three-story, model building at the Structural Dynamics
and Control/Earthquake Engineering Laboratory at the
University of Notre Dame (http://www.nd.edu/~quake/).
The test structure used in this experiment is designed to be
a scale model of the prototype building discussed in
Chung, et al. [16] and is subject to one-dimensional
ground motion. The building frame is constructed of steel,
with a height of 158 cm. The floor masses of the model
weigh a total of 227 kg, distributed evenly between the
three floors. The time scale factor is 0.2, making the natural frequencies of the model approximately five times
those of the prototype.

Annular
Orifice

Coil

Diaphragm

Accumulator

Fig. 2. Schematic of MR Damper.

plied voltage. The peak power required is less than 10
watts. The damper has fast response time, typically reaching rheological equilibrium in less than 10 msec after applying the magnetic field.

A single magnetorheological (MR) damper is installed between the ground and the first floor, as shown in Fig. 1.
The MR damper employed here is a prototype device,
shown schematically in Fig. 2, obtained from the Lord
Corporation for testing and evaluation [6] (see also http://
www.rheonetic.com/mrfluid/). The damper is 21.5 cm long
in its extended position and has a 2.5 cm stroke. The
main cylinder is 3.8 cm in diameter and houses the piston,
the magnetic circuit, an accumulator and 50 ml of MR fluid. The magnetic field produced in the device is generated
by a small electromagnet in the piston head. The current
for the electromagnet is supplied by a linear current driver
which generates a current that is proportional to the ap-

Sensors are installed in the model building for use in determining the control action. Accelerometers located on
each of the three floors provide measurements the absolute accelerations xa1 xa2 xa3 , an LVDT (linear variable
differential transformer) measures the displacement x d of
the MR damper, and a force transducer is placed in series
with the MR damper to measure the control force f being
applied to the structure. Note that only these five measurements are used in the control algorithm. However, to evaluate the performance of the control strategies, LVDTs are
attached to the base and to each floor of the structure to
measure the relative displacements of the structure.
Implementation of the discrete controller was performed
using the Spectrum Signal Processing Real-Time Digital
Signal Processor (DSP) System. The specific capabilities
of this board which make it suitable for use in structural
control systems are described in Quast, et al. [17].

xa3

xa2
Current
Driver

3. SYSTEM IDENTIFICATION
One of the most important and challenging tasks in control
synthesis and analysis is development of an accurate mathematical model of the structural system under consideration, including both the structure and the associated
control devices. A block diagram of the system to be identified, including the MR damper and the test structure, is
shown in Fig. 3, where xg is the ground acceleration, v is
the command signal to the MR damper, and y is the vector of measured structural outputs. Note that the response
of the MR damper is dependent on the relative displacement x d of the structure at the point where they are con-

xa1, x d
f
xg

Control
Computer
Fig. 1. Diagram of MR Damper Implementation.
2

xd
v

MR
Damper

xg

where z is an evolutionary variable that accounts for the

history dependence of the response. The model parameters depend on the voltage v to the current driver as follows

xd
xa1
y
xa2
xa3

f
Primary
Structure

= a + b u ,

Fig. 3. Block Diagram of the Integrated

Structural System.

c 0 = c 0a + c 0b u

u = ( u v )

Spencer and Dyke [9] proposed a control-oriented approach to system identification for semi-actively controlled structures that will be employed in this study. Here
the problem is simplified by decoupling the identification
of the nonlinear semi-active device from that of the primary structure. If the semi-active controller is assumed to be
adequate to keep the response of the primary structure in
the linear range, then standard linear system identification
techniques can be used to develop a model for the primary
structure. Nonlinear identification means must still be employed to identify the semi-active control device. The approach consists of four steps: (i) identification of a model
for the semi-active control device, (ii) identification of a
model for the primary structure, (iii) integration and optimization of the device and structural models, and (iv) validation of the integrated model of the system. These steps
are briefly described in the subsequent paragraphs.

n1

The frequency domain approach to linear system identification presented in Dyke, et al. [1315] is used in the second step of the process, development of a model of the test
structure. As shown in Fig. 3, the two inputs to the structure are the ground excitation xg and the applied control
force f . The four measured system outputs include the
displacement x d of the structure at the attachment point
of the MR damper, and the absolute accelerations, xa1 ,
xa2 , xa3 , of the three floors of the test structure (i.e.,
y = [ x d xa1 xa2 xa3 ] ). Thus, a 4 2 transfer function matrix must be identified to describe the characteristics of the structure in Fig. 3. The eight transfer functions
are determined by independently exciting each of the inputs of the structure with a random input and measuring
the structural responses. Each of the transfer functions is
then modeled as a ratio of two polynomials in the Laplace
variable s. This task was accomplished via a least squares
fit of the ratio polynomials, evaluated on the j axis, to
the experimentally obtained transfer functions.

(1)
n

( xd y) z + A ( xd y)

1
y = ---------------- { z + c 0 xd + k 0 ( x d y ) }
c0 + c1

(2)

The system is then assembled in a state space form using

the analytical representation of the transfer functions (i.e.,
the poles, zeros and gain). Because the system under consideration is a multi-input/multi-output system (MIMO),
such a construction is not straightforward. Two separate
systems are formed, each with a single input corresponding to one of the inputs to the structure. The MIMO system is formed by stacking the states of the two individual
systems and performing a model reduction. The state
space representation of the reduced model is then given by

(3)

y
xd

Bouc-Wen
c1

k0
c0
k1

(5)

Eq. (5) is necessary to model the dynamics involved in the

damper reaching rheological equilibrium. A nominal set
of parameters that were used to initialize the identification
of the integrated system model, presented subsequently,
was obtained based on the response of the MR damper in
a series of displacement-controlled tests [6].

The first step in the process is to identify a model for the

MR damper. The simple mechanical idealization of the
MR damper depicted in Fig. 4 has been shown to accurately predict the behavior of the prototype MR damper
over a broad range of inputs [6]. The equations governing
the force f predicted by this model are given by

z = xd y z z

(4)

where u is given as the output of the first-order filter

nected as indicated by the feedback interaction path in the

diagram.

f = c 1 y + k 1 ( x d x 0 )

c 1 = c 1a + c 1b u ,

x r = Ax r + Bf + Exg ,

y = C y x r + D y f + v,

(6)

where v represents the measurement noise vector. Additional details regarding this frequency domain identification approach can be found in Dyke, et al. [1315].

Fig. 4. Simple Mechanical Model of the MR Damper.

3

The next step in identifying a model of the integrated

structural system is to optimize the set of parameters for
the MR damper model (Eqs. 15) for the case when it is
installed in the test structure, and combine the models of
the device and structure to form the integrated system
model (shown in Fig. 3). Updating the parameters of the
MR damper model is necessary because the MR damper
may function at a somewhat different operating point
when installed in the test structure than in the initial tests
in which the damper was driven with a hydraulic actuator.
A series of tests was conducted in which the structure was
excited at the base, while various commanded voltages v
were applied to the MR damper. A least-squares outputerror method was employed in conjunction with a constrained nonlinear optimization to obtain the 14 model parameters in Eqs. (15) [6]. The optimization was
performed using the sequential quadratic programming algorithm available in MATLAB [18]. Optimized parameters were determined to fit the generalized model of the
MR damper to the experimental data. The resulting parameters are: c 0a = 8 N sec/cm, c 0b = 6 N sec/cm/V, k 0
= 50 N/cm, c 1a = 290 N sec/cm, c 1b = 5N sec/cm/V, k 1
= 12 N/cm, x 0 = 14.3 cm, a =100, b = 450 V1, =
363 cm2, = 363 cm2, A = 301, n = 2, = 190 sec1.

tude of the desired optimal force and f f c is positive, 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 resulting control
law is graphically represented in Fig. 5 and can be concisely stated as
v = Vmax H ( { f c f } f )

where V max is the voltage to the current driver associated

with saturation of the MR effect in the tested device, and
H ( ) is the Heaviside step function. A block diagram of
this semi-active control system is shown in Fig. 6.
One of the attractive features of this control strategy is that
the feedback for the controller is based on readily obtainable acceleration measurements, thus making them quite
implementable. In addition, the proposed control design
does not require a model for the MR damper, although the
model of the damper is important to system analysis.

The integrated model is then formed by connecting the

models of the MR damper (Eqs. 15) and structure (Eq. 6)
as shown in Fig. 3. This model of the integrated system
was validated in [9] and will be employed to design acceleration feedback controllers for the test structure, as discussed in the next section.

fc
v = V max
v = 0

v = 0
f

4. CONTROL DESIGN

v = 0

The clipped-optimal control approach proposed by Dyke,

et al. [7, 8] is to design a linear optimal controller K c ( s )
that determines a desired control force f c based on the
measured structural responses y and the measured force
f applied to the structure, i.e.,
y
1
f c = L K c ( s )L
f

(8)

v = 0
v = V max

Fig. 5. Graphical Representation of Algorithm for

Selecting the Command Signal

(7)
xg

where L { } is the Laplace transform. Because the force

generated in the MR damper is dependent on the responses of the structural system, the desired optimal control
force f c cannot always be produced by the MR damper.
Only the voltage v applied to the current driver for the
MR damper can be directly controlled. A force feedback
loop is incorporated to induce the MR damper to generate
approximately the desired optimal control force f c . To
this end, the command signal v is selected as follows.
When the MR damper is providing the desired optimal
force (i.e., f = f c ), the voltage applied to the damper
should remain at the present level. If the magnitude of the
force produced by the damper is smaller than the magni-

MR
Damper

f
Structure
xd
f

Eq. (8)
Control Law

f
fc

Kc ( s )

Fig. 6. Block Diagram of the Semi-Active

Control System.
4

5. EXPERIMENTAL RESULTS
(cm)

1.3

Uncontrolled
Controlled

x3

To evaluate the performance of the semi-active control

system employing the MR damper, controllers were implemented in the laboratory as discussed in the previous
sections. Acceleration feedback control strategies based
on H 2 /LQG methods, developed and experimentally verified in [1315], were employed to design the optimal controller K c ( s ) . Two semi-active control designs, denoted A
and B, are considered herein. Controller A was designed
by placing a high weighting on the third floor relative displacement. Controller B was designed by placing a high
weighting on the third floor acceleration. The three-story
model structure was subjected to a scaled version of the
N-S component of the 1940 El Centro earthquake and the
measured responses were recorded. Figure 7 shows the
uncontrolled (i.e., without the MR damper attached) and
semi-actively controlled (using Controller A) responses
for the tested structure. The effectiveness of the proposed
control strategy is clearly seen, with peak third floor displacement being reduced by 74.5% and the peak third
floor acceleration being reduced by 47.6%.

xa3

(cm/sec2)

1.3
0

0.5

1.5

0.5

1.5

2.5

3.5

4.5

2.5

3.5

4.5

1500

1500
0

time (sec)

Fig. 7. Controlled and Uncontrolled Structural

Responses due to El Centro Earthquake.

6. CONCLUSION
The efficacy of a new clipped-optimal control strategy
based on acceleration feedback has been experimentally
verified on a semi-actively controlled structure employing
a magnetorheological (MR) damper. This control strategy
produced excellent results, achieving significant improvements over the passive control results. Moreover, the capabilities of the MR damper have been shown to mesh
well with the requirements and constraints associated with
the seismic response reduction in civil engineering structures. Tests on recently designed full-scale MR dampers is
currently underway [5].

Table 1 provides a summary of the peak structural responses. In addition to the results for Controllers A and B,
two passive cases are reported in Table 1. Passive-off and
passive-on refer to the cases in which the voltage to the
MR damper is held at a constant value of V = 0 and
V = Vmax = 2.25 V, respectively. Here, x i is the displacement of the i th floor relative to the ground, d i is the
interstory drift (i.e., x i x i 1 ), xai is the absolute acceleration of the i th floor, and f is the measured control
force.

Note that algorithms that explicitly incorporate actuator

dynamics and control-structure interaction into the control
design process may offer additional controlled performance gains [19]. Efforts are currently underway to investigate this possibility.

The results indicate that both of the passive systems are

able to achieve a reasonable level of performance. In both
tests, the passive-on controller is able to achieve a larger
reduction in the third floor displacement than the passiveoff case. However, notice that the passive-on controller increases the maximum floor acceleration by 14.4% over
that of the passive-off case.

ACKNOWLEDGMENT
This research is supported in part by National Science
Foundation Grant Nos. CMS 9301584, CMS 9500301
and CMS 9528083. In addition, the authors from Notre
Dame would like to express their appreciation to Lord
Corporation of Cary, North Carolina for providing the
prototype magnetorheological damper.

The semi-active control systems perform significantly better than both of the passive systems. Controller A achieves
a 24.3% reduction in the peak third floor displacement and
a 29.1% reduction in the maximum interstory displacement over the best passive case. Moreover, these results
were obtained while also achieving a modest reduction in
the maximum acceleration over that of the passive response. Additional reduction in the peak third floor relative displacement over the best passive case is achieved
with Controller B (33.3%), although at an increase in the
maximum floor acceleration. Notice that for both semi-actively controlled systems, these performance gains are
achieved while requiring smaller control forces than are
required in the passive-on case.

REFERENCES
[1] Fujino, Y., Soong, T.T. and Spencer Jr., B.F.: Structural Control: Basic Concepts and Applications, Proc.
ASCE Struct. Cong., Chicago, Illinois, pp. pp. 361
370 (1996).
[2] Spencer Jr., B.F.: Recent Trends in Vibration Control
in the U.S.A., Proc., 3rd Int. Conf. on Motion and Vibr. Control, Chiba, Japan (1996).
5

Table 1: Experimental Peak Responses due to the El Centro Earthquake.

Control
Strategy
xi

(cm)
di

(cm)
xai

(cm/sec2)
f (N)

Uncontrolled

Passive-Off

Passive-On

Clipped-Optimal
Controller A

Clipped-Optimal
Controller B

0.710
1.068
1.249
0.710
0.362
0.205
879
1110
1500

0.236
0.362
0.436
0.236
0.167
0.106
666
714
804
258

0.126
0.312
0.420
0.126
0.196
0.110
920
808
897
1030

0.127
0.229
0.318
0.127
0.139
0.092
711
642
786
696

0.151
0.213
0.280
0.151
0.123
0.087
957
859
748
754

[3] Carlson, J.D.: The Promise of Controllable Fluids,

Proc. of Actuator 94 (H. Borgmann and K. Lenz,
Eds.), AXON Technologie Consult GmbH, pp. 266
270 (1994).

[12] Inaudi, J.A., Hayen, J.C. and Iwan, W.D.: A Semi-Active Damping Brace System, J. of Engrg. Mech.,
ASCE, submitted.
[13] Dyke, S.J., Spencer Jr., B.F., Quast, P., Sain, M.K.,
Kaspari Jr., D.C. and Soong, T.T.: Experimental Verification of Acceleration Feedback Control Strategies
for An Active Tendon System, Nat. Center for Earthquake Engrg. Res., Tech. Report NCEER-94-0024
(1994).

[4] Carlson, J.D. and Weiss, K.D.: A Growing Attraction

to Magnetic Fluids, Machine Design, Aug., pp. 6164
(1994).
[5] Carlson, J.D. and Spencer Jr., B.F.: Magneto-Rheological Fluid Dampers for Semi-Active Seismic Control, Proc., 3rd Int. Conf. on Motion and Vibr.
Control, Chiba, Japan (1996).

[14] Dyke, S.J., Spencer Jr., B.F., Quast, P., Sain, M.K.,
Kaspari Jr., D.C. and Soong, T.T.: Acceleration Feedback Control of MDOF Structures, J. of Engrg. Mech,
ASCE (in press).

[6] Spencer Jr., B.F., Dyke, S.J., Sain, M.K. and Carlson,
J.D.: Phenomenological Model of a Magnetorheological Damper, J. Engrg. Mech., ASCE, in press.

[15] Dyke, S.J., Spencer Jr., B.F., Quast, P., Kaspari Jr. and
Sain, M.K.: Implementation of an AMD Using Acceleration Feedback Control, Microcomputers in Civil
Engrg., in press.

[7] Dyke, S.J., Spencer Jr., B.F., Sain, M.K. and Carlson,
J.D.: Seismic Response Reduction Using Magnetorheological Dampers, Proc., IFAC World Cong.,
San Francisco, California (1996).

[16] Chung, L.L., Lin, R.C., Soong, T.T. and Reinhorn,

A.M.: Experiments on Active Control for MDOF
Seismic Structures, J. of Engrg. Mech., ASCE, Vol.
115, No. 8, pp. 160927 (1989).

[8] Dyke, S.J., Spencer Jr., B.F., Sain, M.K. and Carlson,
J.D.: Modeling and Control of Magnetorheological
Dampers for Seismic Response Reduction. Smart
Structures and Materials, to appear.

[17] Quast, P., Sain, M.K., Spencer Jr., B.F. and Dyke,
S.J.: Microcomputer Implementations of Digital Control Strategies for Structural Response Reduction, Microcomputers in Civil Engrg., Vol. 10, pp. 1325,
(1995).

[9] Spencer Jr., B.F. and Dyke, S.J.: Semi-Active Structural Control: System Identification for Synthesis and
Analysis, Proc. of the First Europ. Conf. on Struct.
Control, Barcelona, Spain, May 2931 (1996).

[18] MATLAB: The Math Works, Inc. Natick, Massachusetts (1994).

[10] Leitmann, G., and Reithmeier, E.: Semiactive Control

of a Vibrating System by Means of Electrorheological
Fluids, Dynamics and Control, Vol. 3, pp. 733
(1993).

[19] Dyke, S.J., Spencer Jr., B.F., Quast, P., and Sain,
M.K.: The Role of Control-Structure Interaction in
Protective System Design, J. of Engrg. Mech, ASCE,
Vol. 121, No. 2, pp. 32238 (1995).

[11] McClamroch, N.H. and Gavin, H.P.: Closed Loop

Structural Control Using Electrorheological Dampers, Proc., American Control Conf., Seattle, Washington, pp. 41734177 (1995).