Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
J. Acoust. Soc. Am. 123 共3兲, March 2008 0001-4966/2008/123共3兲/1421/6/$23.00 © 2008 Acoustical Society of America 1421
(a) (b)
h h h h
h
h h h h FIG. 1. Smart panels with 16 decentralized velocity
feedback control units using 共a兲 square piezocer-
h
h h h h amic patch actuators and 共b兲 triangularly shaped pi-
ezoceramic patch actuators. The position of the pri-
h
h h h h
mary force excitation is shown as a solid circle.
y h h h h
h
x h h h h
Each control unit is composed of an isosceles triangular transfer functions between the sensor signals y共j兲兩 fp=0,
piezoceramic patch actuator and a high sensitivity induc- whereas no primary excitation is applied to the panel, and the
tively coupled plasma accelerometer error sensor 共PCB Pi- actuator control signals u共j兲 = 关u1u2 ¯ u16兴T. G p共j兲 is a
ezotronics, model 352C67兲. All 16 piezoceramic patches vector of the transfer functions between the sensor outputs
were made of Pz27 material, produced by E.P. Electronic y共j兲兩H=0, whereas no control excitation is applied to the
Component LTD. The patch actuators are bonded to the inner panel, and the primary disturbance input signal f p共j兲, which
side of the panel via a thin layer of nonconductive glue to is either the driving voltage of the loudspeaker Vin or the
ensure that the piezoceramic transducers and the plate are not applied force to the panel generated by the shaker Fin.
connected electrically. The error sensors are located on the For decentralized control, H共j兲 is a diagonal matrix of
opposite side of the panel at the top vertices of the triangular fixed control gains for each control unit, including the trans-
actuators. As illustrated in Fig. 1共b兲, these 16 piezoceramic fer function of the analog controller. In this study the decen-
patches are evenly distributed along the perimeter of the tralized controller has the same gain h0 in each control loop.
panel, with the base edges aligned along the border of the Thus, the feedback transfer matrix H共j兲 is given by
panel. The physical properties and geometry of the test rig H共j兲 = h0HT共j兲I, where HT共j兲 is the transfer function of
and the piezoceramic actuators are summarized in Tables I the analog controller. The identical control gain h0 is chosen
and II, respectively. to be the highest possible value for which the system remains
When the feedback control loops are closed, the error stable.
sensor output signals are fed to analog integrators with an
identical amplification gain in order to provide the feedback
controllers with velocity signals. These integrated signals are B. MIMO stability
then again amplified by charge amplifiers with an identical Velocity feedback control using collocated and dual
amplification gain in order to drive the piezoceramic actua- sensor–actuator is unconditionally stable, even for multiple
tors with the magnified velocity signals. As shown in Fig. 3, channels.3,7 In this case, the fixed feedback gains can, in
this decentralized multiple-input–multiple-output 共MIMO兲 principle, be increased without limit, so that the signals from
control system can be formulated in terms of a classic dis- the error sensors can be driven to zero. However, as dis-
turbance rejection block diagram. Provided that the control cussed by Gardonio et al.,5 the triangularly shaped piezoce-
system is stable, the vector of the spectra for the residual ramic actuator and the velocity sensor pairs considered in
signals at the sensors output y共j兲 = 关y 1y 2 ¯ y 16兴T is given by this study are strictly not dual or collocated, such that the
y共j兲 = 关I + G共j兲H共j兲兴−1G p共j兲f p共j兲, 共1兲 MIMO decentralized control system is only conditionally
stable. Therefore, the control performance is limited by the
where is the circular frequency, j = 冑−1, and I is the 16 by maximum stable control gain that can be implemented in the
16 identity matrix. G共j兲 is a fully populated matrix with the feedback loops.
TABLE I. Geometry and physical properies of the smart panel and the
Perspex box.
Parameter Values
FIG. 2. 共Color online兲 Smart panel mounted on a Perspex box. Reference 12.
1422 J. Acoust. Soc. Am., Vol. 123, No. 3, March 2008 Aoki et al.: Triangularly shaped piezoceramic actuators
The stability of MIMO decentralized control system can fp(j)
be assessed using the generalized Nyquist criteria.8 Assum-
ing that the open loop system is stable, this criteria states that Gp(j)
the closed loop system depicted in Fig. 3 is stable if and only
u(jw)
if the locus of the determinant of the measured return matrix, G(j) y(j)
det关I + G共j 兲H共j兲兴, does not encircle the origin, nor passes
through the origin, as the angular frequency varies be-
tween −⬁ and +⬁. -H(j)
Figure 4 shows the locus of det关I + G共j兲H共j兲兴 be-
tween 25 Hz and 42 kHz. The locus between 25 and 700 Hz FIG. 3. Disturbance rejection block diagram of the velocity feedback con-
is plotted by a faint line, and the locus for frequencies above trol system.
700 Hz is plotted by a dotted line. The locus is characterized
by a series of large circles at low frequency, which are de- with transducers but no feedback control 共thick solid lines兲
termined by the resonant response of the plate. At low fre- and the panel with transducers and decentralized velocity
quencies, the circles start close to the point 共1, 0j兲. As the feedback control 共dotted lines兲.
frequency rises, the circles tend to drift away from 共1, 0j兲,
and around 700 Hz the locus moves into the left-hand side of A. Structural vibration
the vertical axis passing through the point 共1, 0j兲 due to
propagation delays between the sensor and the actuator pairs. Figure 5 shows the measured narrow band spatially av-
This plot indicates that the decentralized feedback control eraged spectra of the out-of-plane vibration levels of the
system under study is only conditionally stable, and gener- panel per unit acoustic primary excitation input Vin 关Fig.
ates positive feedback effects at high frequencies. In general 5共a兲兴 and structural primary excitation input Fin 关Fig. 5共b兲兴.
the positive feedback are generated by the noncollocation The spectra have been derived by remotely measuring the
and nonduality properties of each sensor–actuator pair, velocity of the panel at 165 points evenly distributed over the
which originates both self-control spillover effects on each panel surface using a scanning laser vibrometer.
feedback loop and cross control spillover effects between Comparing the two plots in Fig. 5, it can be observed
neighbor feedback loops. that, when the panel is excited by a point force generated by
a shaker, the response of the panel is characterized by a
rather large number of resonant peaks. In contrast, when the
III. CONTROL IMPLEMENTATION panel is excited by acoustic field, the spectra show fewer
sharp resonant peaks, mainly at the first, third, and fifth reso-
This section presents the experimental results for the nance frequencies, which correspond to the odd modes,
control performance of the new smart panel. In Sec. III A, 共1,1兲, 共3,1兲, and 共1,3兲 of the panel, respectively. The peaks of
the control performance is discussed with reference to the the second and fourth resonances, which are controlled by
vibration level of the panel in the frequency range 0 – 1 kHz. 共2,1兲 and 共1,2兲 vibration modes, are relatively low. This is
Sections III B and III C present the reduction in terms of the due to the fact that the shaker excites nearly all structural
total radiated sound power in the same frequency range.
For both cases, the control performance is discussed in
two stages: the passive effect produced by the sensor– 1
actuator transducers, and the active control effect of the 16
0.8
decentralized control loops. The passive control performance
is assessed by comparing the responses and sound radiation 0.6
produced by the panel without transducers 共faint lines兲, and
the panel with transducers but no feedback control 共thick 0.4
TABLE II. Geometry and physical properties of the triangularly shaped −0.2
piezoceramic actuators.
−0.4
Parameter Values
−0.6
Base and height 共mm兲 40⫻ 40
Thickness 共mm兲 1.0 −0.8
Density 共kg/ m3兲 7700
Poisson’s ratio 0.39 −1
Elastic compliances sE11 17 0 0.5 1 1.5 2
共10−12 m2 / N兲 sE33 23 real
Piezoelectric charge d31 170
coefficient 共10−12 m / V兲 d33 425 FIG. 4. 共Color online兲 Locus of det关I + G共j兲H共j兲兴 between 25 and
700 Hz 共faint line兲 and between 700 Hz and 42 kHz 共dotted line兲.
J. Acoust. Soc. Am., Vol. 123, No. 3, March 2008 Aoki et al.: Triangularly shaped piezoceramic actuators 1423
(a) −20
−40
−50
−60
−70
−80
0 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
0
(b)
Spacially averaged velocity [dB rel. 1m s −1 / N ]
−10
−20
−30
−40
−50
−60
0 100 200 300 400 500 600 700 800 900 1000 FIG. 6. 共Color online兲 Vibratory fields of the panel excited by the loud-
Frequency [Hz]
speaker at the 258 Hz resonance frequency, where the response is controlled
by 共3,1兲 natural mode of the smart panel; without the control 共left column兲
FIG. 5. 共Color online兲 Narrow band spectra of 共a兲 the spatially averaged and with control 共right column兲 at phases = 0° 共top兲, = 120° 共middle兲, and
velocity of the panel, measured between 0 and 1 kHz, per unit voltage = 240° 共bottom兲.
driving the loudspeaker and 共b兲 per unit force generated by the shaker;
measured using the panel 共1兲 without transducers 共faint line兲, 共2兲 with trans-
ducers but no feedback control 共thick solid line兲, and 共3兲 with transducers
and feedback control 共dotted line兲.
ity beneath the panel. In fact, the first three predicted natural
frequencies of the rigid walled cavity underneath the panel
are calculated as 414.3, 476.4, and 546.2 Hz, which corre-
modes of the panel, whereas the acoustic field generated by spond to the 共1,0,0兲, 共0,0,1兲, and 共0,1,0兲 acoustic modes,
the loudspeaker efficiently excites only specific structural respectively.10 The measured resonant frequencies are
modes of the panel that are well coupled to the acoustic slightly lower than the predicted ones, because the cavity
modes of the cavity underneath the panel.1 does not have perfectly rigid walls, especially due to the
The faint and thick solid lines in Fig. 5 show that, when smart panel on the top side. This cavity–plate coupling11
a set of 16 control units is bonded on the panel, the reso- problem is particularly strong with this test rig setup. There-
nance frequencies are shifted toward higher frequencies, and fore, it is expected that in applications where the panel has
the amplitudes of the peaks are reduced by approximately no backing cavity, and is excited by an acoustic disturbance
3 dB with acoustic excitation, and by a maximum 10 dB that comes from a free field or a very large enclosure with
with structural excitation. A significant increase in the natural diffuse acoustic fields, the control performances would be
frequencies is generated by piezoceramic actuators, which higher than those shown in Fig. 5. At frequencies higher than
generate a local stiffness effect on the panel especially at low 700 Hz, the response of the panel slightly increases when
frequencies.9 The response at resonant frequencies is re- control loops are closed, because of the positive feedback
duced, because the bonding layer between the actuator and control effect mentioned in Sec. II B.
the panel adds damping to the panel. The sensor–actuator When active control is implemented, the spectrum of the
control units also produce mass loading effects on the panel vibration level has fewer peaks, as active damping can effec-
in the mass controlled frequency zone, i.e., high frequencies, tively reduce the response at resonances of the panel. This is
which are above the frequency range considered in this confirmed by comparing the vibratory field of the panel at
study.9 resonant frequency with and without control. Figure 6 shows
The thick solid and dotted lines in Fig. 5 show that the the vibratory field of the smart panel excited by the acoustic
active control system smooths the resonant peaks between field produced by the loudspeaker in the cavity at 258 Hz,
100 and 700 Hz by approximately 6 – 18 dB, except for the which is the resonant frequency corresponding to the 共3,1兲
resonance peaks at around 390, 445, and 500 Hz. These non- natural mode of the panel. The pictures on the left-hand side
controlled peaks are dominated by the resonances of the cav- show the vibratory fields of the panel without control at
1424 J. Acoust. Soc. Am., Vol. 123, No. 3, March 2008 Aoki et al.: Triangularly shaped piezoceramic actuators
(a) 60
(a) 60
W/V]
50 50
30 30
20 20
10 10
0 0
0 100 200 300 400 500 600 700 800 900 1000 50 63 80 100 125 160 200 250 315 400 500 630 800
Frequency [Hz] Center frequency [Hz]
(b) 70
(b)
60
W/N]
60
50
−12
50
40
40
30
Π
30
20
20
10
10
0 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz] 0
50 63 80 100 125 160 200 250 315 400 500 630 800
Center frequency [Hz]
FIG. 7. 共Color online兲 Narrow band spectra of the total radiated sound
power, measured between 0 and 1 kHz, 共a兲 per unit voltage driving the FIG. 8. 共Color online兲 Total radiated sound power in one-third octave bands
loudspeaker and 共b兲 per unit force generated by the shaker; measured using between 50 and 800 Hz center bands 共a兲 per unit voltage driving the loud-
the panel 共1兲 without transducers 共faint line兲, 共2兲 with transducers but no speaker and 共b兲 per unit force generated by the shaker; Measured using the
feedback control 共thick solid line兲, and 共3兲 with transducers and feedback panel without transducers 共first column兲, with transducers but no feedback
control 共dotted line兲. control 共middle column兲, and with transducers and feedback control 共last
column兲.
J. Acoust. Soc. Am., Vol. 123, No. 3, March 2008 Aoki et al.: Triangularly shaped piezoceramic actuators 1425
is responsive to frequency bands rather than individual fre- controlled by resonances of the cavity under the panel. At
quencies. Therefore, in this section, one-third octave frequencies higher than around 700 Hz, the vibration level
spectra10 is used to evaluate the control performance in terms and the radiated sound power slightly increases.
of the total radiated sound power per unit acoustic primary These results have been obtained from a prototype smart
excitation input Vin 关Fig. 8共a兲兴 and structural primary excita- panel. It is believed that the control performance can be fur-
tion input Fin 关Fig. 8共b兲兴. The primary excitation sources are ther enhanced by properly designing the size and shape of
driven by a pink noise signal between 0 and 1 kHz in order the triangular piezoceramic actuators in such a way as to
to give equal energy in all octave bands. improve the limit of the stable control gain and extend the
The two plots in Fig. 8 highlight that when the 16 trans- frequency range where the response of the panel can be re-
ducers are attached to the panel the radiated sound power is duced by the decentralized velocity feedback control.
significantly reduced, by 5 – 15 dB at the first three one-third
ACKNOWLEDGMENTS
octave bands for both acoustic and structural primary excita-
tion. These reductions are generated by the shift of the reso- This project has been carried out with joint support from
nant frequencies due to the increased stiffness of the panel. InMAR European Commission project and RENAULT. One
When active control is implemented, the radiated sound of the authors 共Y.A.兲 was also supported by Nakajima Foun-
power is reduced by 1 – 7 dB for the fourth to the eleventh dation, Japan.
one-third octave band in both plots. No control effect is 1
F. J. Fahy and P. Gardonio, Sound and Structural Vibration, 2nd ed. 共Aca-
found in the first three one-third octave bands, where the demic, London, 2006兲.
response of the panel is not characterized by resonances, and 2
C. R. Fuller, S. J. Elliott, and P. A. Nelson, The Active Control of Vibration
thus cannot be controlled by implementing active damping. 共Academic, New York, 1997兲.
3
At higher frequencies, very little or no control effect is ob- S. J. Elliott, P. Gardonio, T. C. Sors, and M. J. Brennan, “Active vibroa-
coustic control with multiple local feedback loops,” J. Acoust. Soc. Am.
served. 111, 908–915 共2002兲.
4
P. Gardonio, E. Bianchi, and S. J. Elliott, “Smart panel with multiple
IV. CONCLUSIONS decentralized units for the control of sound transmission, Part I: Theoret-
ical predictions,” J. Sound Vib. 274, 163–192 共2004兲; “Smart panel with
This paper summarizes the experimental results of a multiple decentralized units for the control of sound transmission, Part II:
Design of the decentralised control units,” J. Sound Vib. 274, 193–213
study to assess the control effectiveness of a smart panel 共2004兲; “Smart panel with multiple decentralized units for the control of
using triangularly shaped piezoceramic actuator and acceler- sound transmission, Part III: Control system implementation,” J. Sound
ometer sensor pairs. Sixteen decentralized velocity feedback Vib. 274, 215–232 共2004兲.
5
control loops have been implemented in order to generate P. Gardonio and S. J. Elliott, “Smart panels with velocity feedback control
systems using triangularly shaped strain actuators,” J. Acoust. Soc. Am.
active damping and to reduce both the response and the 117, 2046–2064 共2005兲.
sound radiation via the panel at resonant frequencies. 6
J. M. Sullivan, J. E. Hubbard, Jr., and S. E. Burke, “Modeling approach for
The experimental results have highlighted that, when the two-dimensional distributed transducers of arbitrary spatial distribution,”
control transducers are attached to the thin aluminum panel, J. Acoust. Soc. Am. 99, 2965–2974 共1996兲.
7
M. J. Balas, “Feedback-control of flexible system,” IEEE Trans. Autom.
its natural frequencies are shifted up due to the passive stiff- Control 23, 673–679 共1978兲.
ening effect of the piezoceramic patch actuators, and the am- 8
S. Skogestad and I. Postlethwaite, Multivariable Feedback Control;
plitudes of the resonant peaks are slightly reduced due to Analysis and Design, 2nd ed. 共Wiley, New York 2005兲.
9
passive damping effects generated by the bonding layer and Y. Aoki, P. Gardonio, and S. J. Elliott, “Modeling of piezoceramic patch
actuator for velocity feedback control,” Smart Mater. Struct. 17, 015052
the dissipative effects induced by the short circuited piezo- 共2008兲.
10
ceramic actuators. The closed loop experimental results have D. A. Bies and C. H. Hansen, Engineering Noise Control; Theory and
highlighted that significant control effects are obtained in the Practice, 3rd ed. 共E & FN Spon, London, 2003兲.
11
I. L. Ver, Noise and Vibration Control Engineering: Principles and Appli-
frequency band between 100 and 700 Hz with reductions of
cations 共Wiley, New York 2005兲.
up to 15 dB at the resonant peaks in both structural vibration 12
S. S. Rao, Mechanical Vibrations, 3rd ed. 共Addison-Wesley, Reading, MA,
and sound radiation, except for some resonance frequencies 1995兲.
1426 J. Acoust. Soc. Am., Vol. 123, No. 3, March 2008 Aoki et al.: Triangularly shaped piezoceramic actuators