Robust Broadband Vibration Control of A Flexible Structure Using An Electrical Dynamic Absorber

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Home Search Collections Journals About Contact us My IOPscience

Robust broadband vibration control of a flexible structure using an electrical dynamic absorber

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

2011 Smart Mater. Struct. 20 075002

(http://iopscience.iop.org/0964-1726/20/7/075002)

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

Download details:
IP Address: 203.237.44.229
The article was downloaded on 01/06/2011 at 07:50

Please note that terms and conditions apply.


IOP PUBLISHING SMART MATERIALS AND STRUCTURES
Smart Mater. Struct. 20 (2011) 075002 (9pp) doi:10.1088/0964-1726/20/7/075002

Robust broadband vibration control of a


flexible structure using an electrical
dynamic absorber
Sang-Myeong Kim1 , Semyung Wang1,3 and Michael J Brennan2
1
Mechatronics, Gwangju Institute of Science and Technology, 261 Cheomdan-gwagiro,
Buk-gu, Gwangju, 500-712, Korea
2
Department of Mechanical Engineering, UNESP, Ilha Solteira, 15385-000, SP, Brazil

E-mail: smkim123@hanmail.net, smwang@gist.ac.kr and mjbrennan0@btinternet.com

Received 21 December 2010, in final form 23 March 2011


Published 31 May 2011
Online at stacks.iop.org/SMS/20/075002

Abstract
This paper presents a simple but practical feedback control method to suppress the vibration of
a flexible structure in the frequency range between 10 Hz and 1 kHz. A dynamic vibration
absorber is designed for this, which has a natural frequency of 100 Hz and a normalized
bandwidth (twice the damping ratio) of 9.9. The absorber is realized electrically by feeding
back the structural acceleration at one position on the host structure to a collocated
piezoceramic patch actuator via an analog controller consisting of a second-order lowpass filter.
This absorber is equivalent to a single degree-of-freedom mechanical oscillator consisting of a
serially connected mass–spring–damper system. A first-order lowpass filter is additionally used
to improve stability at very high frequencies. Experiments were conducted on a free–free beam
embedded with a piezoceramic patch actuator and an accelerometer at its center. It is
demonstrated that the single absorber can simultaneously suppress multiple vibration modes
within the control bandwidth. It is further shown that the control system is robust to slight
changes in the plant. The method described can be applied to many other practical structures,
after retuning the absorber parameters for the structure under control.
(Some figures in this article are in colour only in the electronic version)

1. Introduction absolutely stable in theory. By increasing the feedback gain,


which corresponds to the skyhook damping coefficient, perfect
Broadband vibration control is very attractive if it can control may even be plausible. This, however, is unrealistic
suppress all the vibration modes within the control bandwidth, with a flexible structure because no transducer or electrical
regardless of any changes of the flexible structure under device has an infinite bandwidth that can deal with an infinite
control. If this is possible, the same controller can be number of modes of the flexible structure. Another problem
applied to any flexible structure, such as beams, plates, in relation to the force actuator is that it generally requires
and complex structures of many practical machines and a fixed ground to react against. Such an installation is not
vehicles: for example, flexible robot manipulators [1] and feasible for land, sea, air or space vehicles. Thus alternative
smart panels [2, 3]. Such a universal controller was proposed piezoelectric patch actuators have emerged as practical means
by Karnopp [4] and Balas [5], whose idea was an electrical of actuation. Researchers have subsequently investigated
realization of a skyhook damper by feeding a point velocity the applicability of CVF using piezoelectric transducers,
response back to the collocated point force actuator through a for example, with beams [6] and plates [7]. However, a
gain, i.e., collocated velocity feedback (CVF). When multiple problem with these distributed transducers is that, although,
pairs of collocated sensors and actuators are used, each pair is for example, a piezoelectric actuator and an accelerometer
equivalent to a skyhook damper and thus the control system is may appear collocated in a geometrical sense, they are not
collocated in a system analysis sense [8, 9], i.e., they are
3 Author to whom any correspondence should be addressed.
not dual. Nevertheless, the idea of using simple proportional

0964-1726/11/075002+09$33.00 1 © 2011 IOP Publishing Ltd Printed in the UK & the USA
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

control is still attractive even if the allowable controller gain


is limited and has to be retuned for every plant (i.e., it is no
longer a universal controller). Efforts have thus been made to
maximize both the gain and the control bandwidth of a CVF
controller for a given plant, by ingenious mechanical designs
of the structure and the transducers [6–9]. An alternative
approach is taken in this paper, focusing on the control method.
The aim of this paper is to present a simple, practical
method to control the vibration of a flexible structure over a
broad frequency range. An electrical dynamic absorber (EDA)
is used for this purpose, which is realized by feeding back
the structural acceleration at one position on the host structure
to a collocated piezoelectric patch actuator via a second-order
Figure 1. The experimental control setup. The plant was a free–free
lowpass filter. This absorber is equivalent to a single degree-of- aluminum beam (600 mm × 25 mm × 3 mm) fitted with a PZT
freedom (SDOF) mechanical oscillator consisting of a serially ceramic patch actuator (50 mm × 25 mm × 3 mm) and an
connected mass–spring–damper system. Conventionally, accelerometer at its center. The lumped masses at both ends were not
mechanical dynamic absorbers (MDAs) are known to be part of the original plant, but were added later when testing the
narrow-band damping devices. EDAs have also been robustness of the controller.
successful for this purpose, such as for the vibration isolation
of an item of sensitive equipment [10], for the active damping
of a piezoelectric proof mass actuator [11], and for the 2. Theory
multimodal vibration control of flexible structures [12, 13].
It is demonstrated in this paper that the EDA can also be 2.1. Plant and experimental setup
conveniently used as a broadband damping device. This is The plant under control in the experimental work discussed
feasible because, when realized electrically, the effective mass in this paper was a free–free aluminum beam (600 mm ×
of the absorber can be made very large without increasing
25 mm × 3 mm) fitted with an accelerometer (B&K 4393) and
the actual total weight of the structure. High damping can
a PZT ceramic patch actuator (Fuji Ceramics, C-9, 50 mm ×
also be achieved simply by changing a parameter of the
25 mm × 3 mm) at its center, as shown in figure 1. A
control filter. The EDA is constructed using analog circuits,
signal conditioning amplifier with gain K s (NEXUS, pass
and an analog loop shaping filter is additionally used to
bandwidth: 0.1 Hz–100 kHz) and a voltage amplifier with
improve stability at very high frequencies. Feedback control
gain K a = 30 (PCB 790A01) were used in the feedback loop
experiments were conducted using a free–free beam as the
together with an analog controller that was manufactured in-
host structure, and an accelerometer and a PZT (lead zirconate
house. Constrained layer passive damping treatment was first
titanate) ceramic patch actuator were located at the center of
applied to the aluminum beam to obtain a passive damping
the beam for control purposes. It is demonstrated that the EDA
ratio of about 1% at low frequencies [13]. Double-sided sticky
can simultaneously suppress multiple vibration modes within
tape and copper tape were used for this purpose. The beam
the control bandwidth. The control system is also robust to
was suspended using a thin cord and impacts were applied
slight changes in the plant.
using an instrumented hammer (Endevco 2302-10) on top of
It should be noted that the papers cited in this section
the accelerometer, as shown in figure 1. The responses were
are those which are closely related to the work presented here
measured both when the controller was off and on in order to
and involve the vibration control of flexible structures using
collocated sensors and actuators. A general review on practical assess the control performance. The impact position at the
applications of active vibration control is given in [14], center of the beam was chosen merely for convenience and
particularly for civil structures. For potential applications to does not affect the performance and robustness of the feedback
automobiles and airplanes, one may refer to [15], as most control system. The lumped masses at both ends of the beam
passive treatments can, in theory, be replaced by active means. shown in figure 1 (photo) were not part of the original plant, but
Apart from CVF and the EDA presented here, there are also were added later when testing the robustness of the controller
different feedback control methods for broadband vibration to changes in the plant.
suppression. For example, H∞ control may alternatively The plant can be characterized by (Va /Vf ), where Va is the
be employed for flexible structures with the assumptions of voltage due to the measured acceleration of the beam, and Vf
structured and unstructured uncertainty in collocated [16] is the voltage supplied to the piezoelectric patch actuator. It
and non-collocated configurations [17]. A modified form of is more convenient here to analyze the generalized mobility
integral control was employed as an attempt to reduce low (i.e., velocity response) instead of accelerance, and this is
frequency vibrations of a cantilever beam using a strain-type obtained simply √by dividing the original plant response by
sensor [18]. The major advantages of the EDA presented in jω, where j = −1 and ω is the angular frequency. This
this paper are that it is easy to design, is effective, and is is shown in figures 2(a) and (b) in terms of the frequency
very stable [10–13]. This is because the EDA is, in principle, response amplitude and a Nyquist plot, respectively. Three
an electrical realization of a mechanical dynamic vibration different line types are used to display the responses in different
absorber. frequency bandwidths, as detailed in the caption of the figure.

2
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

Figure 3. Single-input–single-output disturbance rejection control of


the plant P(jω) by a controller C(jω).

2.2. Controller design


In this section the design of the controller is described. The
system is shown as a block diagram in figure 3, where d is
the response (displacement, velocity, or acceleration) before
control, e is the response after control, P(jω) is the plant
[V /V ], and C(jω) is the controller [V /V ]. The plant response
after control is given by

e = d[1 + L(jω)]−1 , (1)

where L(jω) = P(jω)C(jω) is the open-loop frequency


response function (FRF) of the control system. The control
performance can be described in terms of the vibration
reduction ratio S(jω) = e/d , which is written in decibels as

RR (dB) = 20 log10 |S(jω)|, (2)

where S(jω) = [1 + L(jω)]−1 is also called the sensitivity


function. In this paper, a control system is defined to be stable
and robust with a degree of r if and only if its open-loop FRF
Figure 2. Measured generalized mobility [V /V ] of the plant in locus does not enclose or cross the circle of radius r centered at
figure 1: the time integral of Va to an excitation voltage Vf . Three the instability point (−1, 0). This Nyquist robustness criterion
line types display the response: below 4 kHz (solid line); between 4
and 12 kHz (dashed); and above 12 kHz (dash–dot). The target
can be transcribed as [10]
bandwidth is between 10 Hz and 1 kHz (shown by dotted vertical
lines). (a) Amplitude, (b) Nyquist plot. RR (dB)  (−20 log10 r ) (at all frequencies), (3)

where the limiting value on the right is the allowable


maximum control spillover. The robustness degree r can
Because the plant was symmetric, only symmetric modes were
also be interpreted as the generalized gain margin: GM =
excited and sensed. The frequency range from 10 Hz to
−20 log10 (1 − r ). The control spillover is particularly
1 kHz (within the vertical dotted lines in figure 2(a)) was
important when it occurs in the non-minimum phase frequency
targeted for control, and this is discussed in the following
bandwidth. The task here is to achieve the best performance
sections. It is clear from figure 2(b) that the plant behaves
as a minimum phase system below 4 kHz and then loses given by equation (2) within the control bandwidth of interest
this characteristic above this frequency. Thus, the plant is without violating the robustness constraint in equation (3) at all
not strictly positive real (non-SPR). Here, the term ‘minimum frequencies. The control bandwidth for the system in figure 1
phase’ refers to a particular bandwidth while SPR is defined was set as broad as two decades, from 10 Hz to 1 kHz. The
for all frequencies. Furthermore, because of the characteristics degree of robustness chosen was 0.5, corresponding to the
of the piezoelectric patch actuator, the high frequency non- generalized gain margin of 6 dB.
minimum phase response does not roll off naturally, but is From figure 2, it can be assumed that the generalized plant
very large, as can be seen in figure 2(a). In addition, the mobility is minimum phase (i.e. −90◦   [(jω)−1 P(jω)] 
response below 10 Hz is not reliable, having poor coherence. 90◦ in which  denotes the phase angle) within the frequency
These are typical characteristics of the plants encountered in range of interest (i.e. in the low frequency region) such that
the active control of flexible structures. Control of such plants 
M
is nontrivial and a classical control approach is presented in (jω)−1 P(jω) = cs−,m1 (jam ωs,m ω)s−,m1 , (4)
this paper, using analog compensators. m=1

3
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

Figure 4. Mechanically analogous models: (a) the proportional gain filter, (b) the first-order lowpass filter, (c) the second-order bandpass
filter, and (d) the coupling between a structural mode and the second-order bandpass filter.

where s,m = [ωs2,m − ω2 + jam ωs,m ω]. The subscript m gain, figure 4(b) shows a serially connected skyhook spring–
denotes the structural mode number from 1 to M in which M is damper which is described by equation (6), and figure 4(c)
the total number of modeled modes, ω s,m is the m th structural
shows an SDOF dynamic absorber (i.e. EDA) consisting of
natural frequency in which ωs,m = ks,m /m s,m , the lumped a serially connected mass–damper–spring system which is
parameters m s,m, cs,m, and ks,m are respectively the m th modal described by equation (7). Note that − denotes a serial
mass, damper, and spring, and am = 2ζs,m is the half-power connection. The lumped parameters of the EDA can be
bandwidth of the m th mode normalized by ωs,m with ζs,m obtained from the relationships: ka = m a ωa2 = bωa ca ,
being the modal damping ratio. The relationships between the where m a , ca, and ka are respectively the mass, damper,
parameters are given by 1/m s,m = ωs2,m /ks,m = am ωs,m /cs,m . and spring of the absorber, ωa is the natural frequency,
To be compatible with the expression for the plant and b = 2ζa with ζa being the absorber damping ratio.
mobility in equation (4), the generalized impedance form of Figure 4(d) represents the coupling between a single structural
the controller jωC(jω) is analyzed instead, which means that mode given in equation (4) and the dynamic absorber in
there is a controller voltage output Vf in response to a velocity- equation (7) [11]. The coupling with the other analogous
compatible voltage input Va /jω. Thus, the product of the plant elements can be similarly represented. The structural and
mobility and the controller impedance yields the same open- absorber modal damping ratios are more specifically
√ defined
loop FRF: L(jω) = P(jω)C(jω). Three simple compensators as ζs,m = cs,m /(2 m s,m ks,m ) and ζa = m a ka /(2ca ),
could be considered for the broadband vibration control of respectively. The cutoff frequency in equation (6) can also be
the flexible structure: a proportional gain [4, 5]; a first-order described in terms of the mechanical parameters as ωo = ka /ca .
lowpass filter; and a second-order bandpass filter. These are It should be noted that the analogous models in figure 4 are
respectively written as exact provided the plant is SPR. If the plant is minimum phase
only in a limited bandwidth, as in figure 2, the complete system
jωC(jω) = ca , (5) is no longer guaranteed to be perfectly passive or absolutely
ωo stable. If the system is conditionally stable with the filters in
jωC(jω) = ca , (6) equations (5)–(7), however, these analogous models are still
ωo + jω
useful and can represent the control mechanism within the
jbωa ω control bandwidth.
jωC(jω) = ca , (7)
ωa2 − ω2 + jbωa ω The pass bandwidths of the filters in equations (5)–(7)
where ca is the common gain of the three filters, ωo is the are related to the control performance given in equation (2),
cutoff frequency of the first-order lowpass filter, and ωa and b while the stop bandwidths of the filters are related to the
are the center frequency and normalized half-power bandwidth control robustness given in equation (3). It is interesting to
(or the inverse of quality factor) of the second-order bandpass note that all three filters in equations (5)–(7) roughly have
filter, respectively. In the case when the plant is minimum ca in their pass bandwidths, revealing the control mechanism
phase for all frequencies (i.e. SPR), the control systems with that they all behave like a skyhook damper in the bandwidth
filters described by equations (5)–(7) are unconditionally stable of control. This indicates that the EDA in figure 4(c),
and are passive. The mechanically analogous models for the as well as the skyhook spring–damper in figure 4(b), acts
controllers are shown in figure 4 [11]. Figure 4(a) depicts a simply as a band-limited skyhook damper. The band-stop
skyhook damper in which equation (5) gives the proportional characteristics become particularly important when feedback

4
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

Figure 5. Active control filter consisting of a second-order lowpass filter realized in biquad circuit form and a first-order lowpass filter (inside
the dashed box).

control is implemented on a plant having uncertain and non- where ω2 and ω1 are the upper and lower frequency of interest,
minimum phase characteristics such as that shown in figure 2. respectively. The gain m a of the controller and the cutoff
Comparing the three filters in equations (5)–(7), it is found frequency ωc of the shaping filter can be determined from the
that the bandpass filter in equation (7) has the advantage of Nyquist robustness criterion with the generalized gain margin
robustness as it suppresses the unreliable plant response at very of 6 dB. The final control filter parameters determined are
low frequencies as well as the non-minimum phase response at given by
very high frequencies.
If the roll-off rate of the filter (−20 dB/decade) in m a = 204, b = 9. 9, ωa = 2π × 100 Hz,
(10)
equation (7) is insufficient and the control system is bound ωc = 2π × 4 kHz.
to instability at very high frequencies, there are two simple
options to choose from. One can either lower the upper cutoff The control filter in equation (9) with parameters given
frequency of the bandpass filter in equation (7) or introduce an in equations (10) was manufactured using four operational
additional lowpass filter to shape the high frequency loop gain amplifiers (National Semiconductor LF412CN, supply voltage
as given by of ±16 V) as shown in figure 5, where the values of the
ωc electrical elements used are also tabulated. The circuit
F(jω) = , (8)
ωc + jω consisting of the first three operational amplifiers (from the
where ωc is the cutoff frequency of the first-order lowpass left) is the second-order lowpass filter realized in biquad
filter. Care should be taken as the additional loop shaping circuit form for ease of tuning [19]. Simpler forms such as
filter can also destabilize the system due to the additional the Sallen–Key and multiple-feedback architectures are also
phase delay in the transition region of this filter. Combining available [20]. The circuit inside the dashed box in figure 5
equations (7) and (8), the complete control filter used is a serial is the first-order shaping lowpass filter. The shaping filter can
combination of the second-order and the first-order low pass be included (excluded) by adding (removing) capacitor Cb .
filters as given by The relationships between the filter and circuit parameters are
given by
 
ωa2 ωc
C(jω) = m a 2 , (9) ωa2 = (R2 R4 C1 C2 )−1 , b = (R1 C1 ωa )−1 ,
ωa − ω2 + jbωa ω ωc + jω (11)
ma = K1 K2, ωc = (Rb Cb )−1 ,
where the absorber mass m a is the filter gain. Note that, rather
than the impedance form in equation (7), the controller C(jω) where K 1 = −R2 /R3 and K 2 = −Rb /Ra are the gains of
itself has been presented in equation (9) for the experimental the biquad and shaping filters, respectively. From the values
implementation that is described in section 3. The bandpass given in figure 5, it is seen that K 1 = −10 and K 2 =
filter in equation (7) is used only for analysis purposes. −20.4, and the large gain of m a = 204 [V /V ] was achieved
by the two-stage amplification. The reason for this was to
3. Experimental work avoid using resistors that were either too small or too large.
The resistor R3 is particularly important as it determines the
3.1. Control filter construction and loop gain measurements electrical input impedance of the control filter. To avoid the
mutual loading with the neighboring conditioning amplifier
From the control bandwidth of interest between 10 Hz and and voltage amplifier shown in figure 1, the input impedance of
1 kHz prescribed in section 2.2, it is straightforward to the control filter must be large and its output impedance must
determine ωa and b of the absorber from b = (ω2 − ω1 )/ωa , be small.

5
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

Figure 6. Measured controller impedance [V /V ] without (dashed)


and with (solid) the shaping filter: the response Vf to the time integral
of Va in figure 1. Predicted responses (dotted) are overlaid.
(a) Amplitude, (b) phase.

Figure 6 compares the amplitudes (a) and phases (b) of


the measured (dashed and solid lines) and predicted (dotted)
controller impedances jωC(jω) [V /V ] without (dashed) and
with (solid) the shaping filter. The measured and predicted
responses are nearly identical. The open-loop FRF was then
measured by disconnecting the Va point in figure 1, and the
results are shown in figure 7: (a) amplitudes of the measured
(dashed and solid lines) and simulated (dotted) open-loop
FRFs without (dashed) and with (solid) the shaping filter; ((b),
(c)) Nyquist plots of the measured and simulated (dotted) open-
loop FRFs without (b) and with (c) the shaping filter. Three
Figure 7. (a) Amplitudes of the measured (dashed and solid lines)
line types in (b) and (c) display the responses in different
and simulated (dotted) open-loop FRFs without (dashed) and with
frequency bandwidths, as detailed in the caption of the figure. (solid) the shaping filter. ((b), (c)) Nyquist plots of the measured and
The measured (solid and dashed) and simulated (dotted) results simulated (dotted) open-loop FRFs without (b) and with (c) the
in figure 7(a) again agree very well. This confirms that shaping filter. Three line types display the measured response: below
the mutual coupling between the controller and the voltage 4 kHz (solid line); between 4 and 12 kHz (dashed); and above
12 kHz (dash–dot).
amplifier was negligible. The robustness boundary circle of the
6 dB generalized gain margin is also overlaid in figures 7(b)
and (c). It is seen in figure 7(c) that the shaping filter has 3.2. Control experiments
improved stability at very high frequencies so as to satisfy the Impacts were applied as indicated in figure 1 with the feedback
generalized gain margin of at least 6 dB at all frequencies. loop first disconnected and then connected. The responses

6
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

Figure 8. Measured acceleration responses of the controlled and


uncontrolled plants. (a) Original plant, before control (dash–dot) and
after control (solid) in comparison with the simulated controlled
performance (dotted). (b) Both the original (solid) and mass-added
(thick dashed) plants after control and before control (dash–dot).
Figure 9. Nyquist plots of the measured open-loop FRFs for the
mass-added plant shown in figure 8(b) without (a) and with (b) the
shaping filter.
were measured with the loop shaping filter in place. Figure 8(a)
shows the measured acceleration responses of the plant before
(dash–dot) and after (solid) control, with the dB reference of To test the robustness of the controller, two lumped masses
1 m s−2 . The simulation result (dotted) is also overlaid for (in total, 16.5 g) were attached symmetrically, one at each end
comparison, which was obtained from equation (1) using the of the beam (originally, 166 g) as shown in figure 1 (photo).
uncontrolled plant response d and the open-loop FRF L(jω). The same impact tests were also conducted. The results before
Both measured and simulated results agree extremely well, and after control are shown in figure 8(b), where the original
which also confirms that the mutual coupling between the plant responses (solid and dash–dot) in figure 8(a) are also
control filter and the signal conditioner was negligible. Thus, overlaid for comparison. As expected, robust performance
the manufactured control filter using the electrical elements was achieved as long as the modes were located within the
listed in figure 5 was a functional design that was free from control bandwidth. To examine the robust stability, the open-
any mutual loading problems when connected to the particular loop FRFs of this mass-added plant were measured without
electrical devices K s and K a in figure 1. In figure 8(a), it and with the additional shaping filter and the results are
can be seen that the vibrations within the bandwidth between shown in figures 9(a) and (b), respectively. Regardless of the
10 Hz and 1 kHz are greatly suppressed. It is interesting that shaping filter, both systems are quite robust but figure 9(a)
reductions occur not only at the peaks but also at the troughs. is slightly more robust. Figure 9(b) is the case where the
In fact, they can occur wherever the open-loop FRF locus is on additionally introduced shaping filter acts adversely due to the
the right half of the Nyquist plot. additional phase delay in the transition region of the filter. Both

7
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

figures 9(a) and (b) suggest that the performance of this mass-
added plant can be further improved.
If the EDA method is applied to a general structure,
the control gain and bandwidth should be optimally retuned
depending on the structure under control. There are some
general guidelines as regards preparing the plant for the
EDA. First, it is important to maximize the minimum phase
bandwidth of the plant. The larger this bandwidth is, the
broader the control bandwidth is and the more robust the
control system is to changes of the plant. Positioning of
the sensor and the actuator is critical and a rule of thumb is
collocation. Second, it is important to maximize the plant
sensitivity (excluding the electrical devices used) within the
control bandwidth. This can relax the requirement for a high-
gain voltage amplifier for actuation. The piezoelectric actuator
should be made suitable for excitation by careful choices of
the material, dimensions, and installation location. The type
of sensor (e.g. strain gauge, PVDF, accelerometer) must also
be appropriately chosen. Third, it is important to minimize
the plant sensitivity outside the control bandwidth, particularly
the non-minimum phase and uncertain characteristics at very
high frequencies. To achieve this, passive constrained layer
damping treatment and other passive means such as adding
lumped masses as in this paper and in [9] could be used;
stiffeners could also be appropriately applied.

3.3. Comparison with other methods

Two interesting comparisons are made in this section using


simulations only. The EDA is compared first with an electrical
damper (ED), which corresponds to CVF, and then with Figure 10. (a) Simulated performances of the EDA (solid line)
a mechanical dynamic absorber (MDA). When the EDA is shown in figure 8(a) and a comparable ED (thick dashed) for the
used with m a = 204 as given by equations (10), the original plant (dash–dot), (b) simulated performances of the same
EDA in (a) and several MDAs (five dashed lines) with different
corresponding electrical damping coefficient in equation (7) is absorber masses of 0.3 × m s , 0.6 × m s , 0.9 × m s , or 1.2 × m s , where
1.3 × 104 , because ca = m a ωa /b . For an ED, the maximum the total mass of the original plant m s = 166 g.
allowable gain determined from the Nyquist robustness
criterion described in section 2.2 with the generalized gain
margin of 6 dB is only about 17% of that of the EDA. Thus,
having different absorber masses of 0.3×m s , 0.6×m s , 0.9×m s ,
the EDA can outperform the ED by a factor of about 5.8
or 1.2 × m s , where the total mass of the original plant m s =
within the control bandwidth for the plant considered. This
166 g. Due to the high damping used for the MDA, no new
is precisely shown in figure 10(a) by comparing the simulated
dominant peak is observable. Due to the MDA mass, there
performances of the EDA (solid line) previously shown in
are reductions in the low frequency responses (consequently
figure 8(a) and the ED (dashed) for the original plant (dash–
dot). in the rigid body modal responses of the beam). The high
Now suppose that the MDA in figure 4(c), which is frequency responses are little reduced as the mobility YS is
assumed to have the same natural frequency of 100 Hz and the small in the region. The behavior with the MDA considered is
same normalized bandwidth of 9.9, is connected to the center very different from that with the EDA used. Note, in particular,
of the beam. The plant velocity responses v before and after that the low frequency response for the EDA (solid line) is
attachment of the MDA are respectively given by [11] nearly unchanged, which suggests that the EDA used reduces
flexible modes without mass-loading the beam (i.e., without
v = YS f, v = YS (1 + YS Z A )−1 f, increasing the total weight of the system). It is thus clear that
the MDA, which is directly connected to the beam, does not
where f is the impact force applied to the center of the represent the mechanism of the EDA used. This is because the
beam. The input mobility of the beam YS can be obtained force-excited input mobility YS of the beam is very different
by dividing the uncontrolled plant response d (dash–dot line from the piezo-excited input mobility (jω)−1 P(jω) of the beam
in figure 10(a)) by jω, and the absorber impedance Z A is shown in figure 2. The mechanism of the EDA used can rather
given by equation (7). Figure 10(b) compares the simulated be explained in the modal domain by using a number of such
performances of the EDA and several MDAs (dashed lines) analogous two-DOF coupled models, as shown in figure 4(d).

8
Smart Mater. Struct. 20 (2011) 075002 S-M Kim et al

4. Conclusions [2] Park C H and Baz A 1999 Vibration control of bending modes
of plates using active constrained layer damping J. Sound
In this paper, broadband control of a free–free beam has been Vib. 227 711–34
considered by using an EDA. The transducers used in the [3] Pietrzko S J and Mao Q 2008 New results in active and passive
control of sound transmission through double wall structures
control system were an accelerometer and a PZT ceramic
Aerosp. Sci. Technol. 12 42–53
patch actuator. A single EDA was employed to simultaneously [4] Karnopp D, Crosby M J and Harwood R A 1974 Vibration
reduce the vibration of multiple vibration modes in a frequency control using semi-active force generators ASME Trans. J.
bandwidth between 10 Hz and 1 kHz. The EDA had the Eng. Ind. 96 619–26
natural frequency of 100 Hz and the normalized bandwidth [5] Balas M J 1979 Direct velocity feedback control of large space
of 9.9. The EDA was realized by feeding back the structural structures J. Guid. Control 2 252–3
[6] Lee Y S, Elliott S J and Gardonio P 2003 Matched piezoelectric
acceleration at one position to a collocated piezoceramic patch
double sensor/actuator pairs for beam motion control Smart
actuator via an analog controller consisting of a second-order Mater. Struct. 12 541–8
lowpass filter. This EDA is equivalent to an SDOF mechanical [7] Gardonio P, Lee Y S, Elliott S J and Debost S 2001 Analysis
oscillator consisting of a serially connected mass–spring– and measurement of a matched volume velocity sensor and
damper system. A first-order lowpass filter was additionally uniform force actuator for active structural acoustic control
used to improve the stability at high frequencies. The J. Acoust. Soc. Am. 110 3025–31
[8] Yang S Y and Huang W H 1998 Is collocated piezoelectric
experiments conducted showed that the single EDA could
sensor/actuator pair feasible for an intelligent beam?
simultaneously suppress multiple vibration modes within the J. Sound Vib. 216 529–39
control bandwidth. Moreover, the control system was robust to [9] Gatti G, Brennan M J and Gardonio P 2007 Active damping of
slight changes in the plant, such as natural frequencies. The a beam using a physically collocated accelerometer and
use of the single dynamic absorber for multiple modes was piezoelectric patch sensor J. Sound Vib. 303 798–813
feasible because the design parameters are easily controllable [10] Kim S M, Pietrzko S and Brennan M J 2008 Active vibration
isolation using an electrical damper or an electrical dynamic
and a heavy absorber mass can be imposed without increasing
absorber IEEE Trans. Control Syst. Technol. 16 245–54
the total weight of the system. [11] Kim S M, Wang S and Brennan M J 2011 Dynamic analysis
The major advantages of the EDA presented in this paper and optimal design of a passive and an active piezo-electrical
are that it is easy to design, is effective, and is very stable. This dynamic vibration absorber J. Sound Vib. 330 603–14
is because the EDA is, in principle, an electrical realization [12] Kim S M, Wang S and Brennan M J 2011 Comparison of
of a mechanical dynamic vibration absorber. It should be negative and positive position feedback control of a flexible
structure Smart Mater. Struct. 20 015011
emphasized that the method can also be applied to other
[13] Kim S M, Wang S and Brennan M J 2011 Optimal and robust
one-dimensional structures embedded with different types of modal control of a flexible structure using an active dynamic
sensors (strain, velocity, or acceleration). A low pass filter can vibration absorber Smart Mater. Struct. 20 045003
be employed when an accelerometer (acceleration–position [14] Song G, Sethi V and Li H N 2006 Vibration control of civil
feedback: APF) is available as in this paper, a band pass filter structures using piezoceramic smart materials: a review Eng.
is used with a velocity sensor (velocity–velocity feedback: Struct. 28 1513–24
VVF), and finally a high pass filter is used with a strain sensor [15] Rao M D 2003 Recent applications of viscoelastic damping for
noise control in automobiles and commercial airplanes
(position–acceleration feedback: PAF). These are the three J. Sound Vib. 262 457–74
different ways to realize an EDA [12]. If a cantilever beam [16] Jiang J P and Li D X 2010 Optimal placement and
is fitted with a pair of collocated piezoelectric sensor and decentralized robust vibration control for spacecraft smart
actuator at its root [12], for example, a high pass filter can solar panel structures Smart Mater. Struct. 19 085020
thus be employed. The method can also be applied to other [17] Kar I N, Miyakura T and Seto K 2000 Bending and torsional
structures such as plates and three-dimensional structures. As vibration control of a flexible plate structure using H∞ -based
robust control law IEEE Trans. Control Syst. Technol.
the EDA is a broadband control device, the method is suitable 8 545–53
for complex structures which have a high modal density at low [18] Aphale S S, Fleming A J and Moheimani S O R 2007 Integral
frequencies. It is also possible to extend the approach to a resonance control of collocated smart structures Smart
decentralized multiple-input–multiple-output (MIMO) control Mater. Struct. 16 439–46
system [16, 21]. [19] van Valkenburg M E 1982 Analog Filter Design (Oxford:
Oxford University Press)
[20] Karki J 2000 Active low-pass filter design Application Report
References SLOA049A, Texas Instruments
[21] Bianchi E, Gardonio P and Elliott S J 2004 Smart panel with
[1] Luo Z H 1993 Direct strain feedback control of flexible robot multiple decentralized units for the control of sound
arms: new theoretical and experimental results IEEE Trans. transmission. Part III: control system implementation
Autom. Control 38 1610–22 J. Sound Vib. 274 215–32

You might also like