You are on page 1of 15

Journal of Sound and Vibration (1996) 190(1), 105–119

COLLOCATED STRUCTUR AL CONTROL FOR
R EDUCTION OF AIR CR AFT CABIN NOISE
Doici\s G. M\cM\r1iN
United Technologies Research Center, East Hartford, Connecticut, 06108, U.S.A.
(Received 28 November 1994 and in final form 11 May 1995)
Various approaches to cabin noise reduction within turbo-prop aircraft have been
suggested. Since the primary source of internal noise is due to propeller-induced fuselage
vibration, the noise can be reduced by minimizing the noise transmission through the
fuselage using active structural control. The feasibility of a structural control approach
using piezoelectric actuators and structural sensors is investigated. Feedback of collocated
strain yields an approach which is simple, modular, and extremely robust. Actuator and
sensor selection is based on wave-number arguments. Simulations with a simple
vibro-acoustic aircraft model indicate that the desired goal of a 10 dBA reduction in peak
internal noise is achievable with relatively few actuators. Various actuator/sensor placement
and control architecture configurations are investigated with the eventual goal of
minimizing the total weight penalty.
7 1996 Academic Press Limited
1. INTRODUCTION
There is a desire to reduce the cabin noise within commercial turbo-prop aircraft, in order
to improve passenger comfort [1]. The cabin noise is generated by various sources,
including boundary layer flow noise, acoustic excitation of the fuselage from the propeller,
and structure-borne noise due to engine vibration and flow distortion over the wing. For
typical turbo-prop aircraft, the proximity of the propeller disc to the fuselage results in
the narrow-band acoustic excitation of the fuselage being the dominant source of noise.
The objective is to reduce the peak interior noise level by approximately 10 dBA.
Noise reduction options include both passive methods, such as structural modifications
[2] or damping augmentation [3, 4], and active methods, including synchrophasing, and
control of either the acoustic field [5, 6] or the structural vibration [7–10]. Stiffening
the structure tends to incur a substantial weight penalty. Broadband passive damping is
not useful because the response is primarily forced, rather than resonant. Narrowband
damping using tuned vibration absorbers has yielded 10 dB reductions [3], but the
improvement is limited by the difficulty of keeping them tuned in a varying environment.
An active approach, however, can be tuned to allow for varying engine speed or
environment, and can handle several frequencies simultaneously. There have already been
flight tests using active control of the internal sound field which have obtained 10 dB
reductions [5, 6]. However, many microphones and speakers are required to obtain this
performance, yielding a moderate weight penalty, and some difficulty in locating bulky
speakers within the passenger compartment.
An alternative active approach is to eliminate the problem before the sound gets into
the aircraft cabin by using active structural control to modify the acoustic transmission
properties of the fuselage. This approach has an advantage over directly controlling the
105
0022–460X/96/060105+15 $12.00/0 7 1996 Academic Press Limited
b. c. x\cx\r1iN 106
interior noise, in that fewer actuators and sensors are required, since they are only needed
near the propeller plane [11, 12]. The concept of controlling structural vibration to reduce
transmitted or radiated noise is not new [13], and has been investigated by a variety of
researchers for the aircraft noise control problem [7–12]. However, most of the literature
on this subject uses sound pressure information from within the cabin. By making the
control objective the reduction of noise transmission, rather than of internal noise, an
approach can be developed that uses only feedback of local structural information. High
gain feedback then reduces fuselage vibration in the propeller footprint area, which reduces
the power input from the propeller disturbance, and sound radiation into the aircraft
cabin.
Piezoelectric actuators are well suited to this problem due to their low weight, high
bandwidth, and because their area-averaging effect reduces their influence on structural
modes that do not couple with the acoustic response. In the context of the wave-number
based radiative efficiency arguments of reference [14], the distributed piezoelectric actuator
and collocated strain sensor act as spatial filters. Fuller et al. [10] have used piezoelectric
actuators bonded to an aircraft fuselage and obtained acceptable authority over internal
noise using microphone feedback, indicating that the required actuator strength is feasible.
To evaluate the capability of a collocated vibration control approach for reducing
internal noise, a vibro-acoustic model is developed. Various researchers have investigated
active structural and acoustic control for aircraft using simple cylinder models of the
structural vibration [15–17]. In addition, there are numerous more detailed modelling
techniques available, including finite element and boundary element methods [18]. Since
accurate predictions of absolute noise levels are not required, a simple model is sufficient.
The model developed herein treats the fuselage as an orthotropic finite cylinder with
equivalent bending stiffnesses to account for the frames and stringers, similar to the model
in reference [19]. The stiffening effect of the floor is partially included. The closed-loop
structural response is obtained for a computer prediction of the propeller sound pressure
field. The interior acoustic response is then computed from the structural vibration. The
primary objective is to demonstrate that the desired performance goal of a 10 dBA
reduction in peak internal noise can be obtained. In addition, different actuator and sensor
configurations and control architectures are evaluated, with the ultimate goal of
minimizing the weight penalty.
2. CONCEPTUAL DESIGN
One of the objectives that must be considered in developing an approach intended for
eventual production is the ease of implementation of the ultimate design. While difficult
to quantify, this includes factors such as ease of production and assembly, the impact on
other aspects of the overall system, the handling of inevitable sensor/actuator failures, and
substantial robustness to the wide range of operating conditions likely to be encountered
on numerous different aircraft. Finally, the total system cost and weight must be
minimized. With these design objectives in mind, the overall simplicity and modularity of
the approach can be as important as the performance it achieves.
One drawback to the use of speakers in the aircraft cabin for noise control is locating
them so as to preserve their effect, while protecting them from damage and minimizing
any loss in useful space. Prior work on active structural-acoustic control has indicated that
structural actuators can be used, rather than speakers. A purely structural control
approach that eliminates the needs for microphones can have considerable advantages with
regards to the overall design objectives, as described below. The key requirement is to
identify an appropriate structural sensor.
coiioc\1rb s1ric1ir\i coN1roi 107
Substantial simplifications result by defining the objective to be reducing the noise
transmission through the fuselage, rather than reducing the internal noise. These objectives
are equivalent in the sense that reducing the former will reduce the latter. However, the
internal noise is a global performance objective, while noise transmission depends only
on the local structural information at an actuator location. Thus it is sufficient to use
collocated feedback, and minimize the sensor output, as indicated schematically in
Figure 1. This yields a simple control design problem, and gives substantial robustness.
While the transfer function from a piezoelectric actuator to a microphone varies
considerably with changes in passenger load and operating conditions, the fundamental
characteristics of the local structural transfer functions remain unchanged. This is
particularly true if collocated strain sensors are used, since the resulting transfer functions
will be positive real. Positive real feedback can then be used to guarantee stability.
This approach is similar to the concept of disturbance isolation, wherein the disturbance
is prevented from entering the structure by high gain collocated feedback. This class of
solution techniques has typically achieved excellent performance, due in part to the fact
that the uncertain dynamics of the structure do not need to be accurately modelled. The
control system alters the mode shapes of the system so that the impedance seen by the
disturbance is increased. Performance can also be achieved by altering mode shapes to
reduce structural-acoustic coupling; this is analogous to performance isolation.
The limitation observed in the literature on purely structural control approaches to noise
reduction is that only low wave-number structural motion couples well with the acoustic
field [14, 9, 16]. Thus, minimizing structural motion may increase the acoustic response
unless the sensors and actuators do not respond to or excite that motion that does not
couple well. Choosing the sensor and actuator dimensions on the order of one half of
the acoustic wavelength at the frequency of interest results in spatial wave-number
filtering. Minimizing the sensor is therefore an approximation to minimizing the acoustic
radiation. Precisely sensing the structural motion that radiates efficiently would improve
performance, but is neither essential, nor simple.
The piezoelectric actuators will be bonded to the inner surface of the fuselage frames,
to give the largest possible moment arm to excite fuselage ring bending. It is the frames
which dominate the structural dynamics at the frequencies of interest. If the frames are
controlled, then the fuselage skin can still vibrate between the frames. However, on the
deHavilland Dash-8, the distance between frames is such that this remaining skin vibration
can only couple well with the acoustic field above approximately 300 Hz.
Because the disturbance source is not at a single point, actuators will be required at
several circumferential locations on several frames. This leads to several options for the
Figure 1. Piezoelectric actuator placement on fuselage frame.
b. c. x\cx\r1iN 108
control architecture. A purely decentralized control scheme wherein each actuator uses
only single-input/single-output (SISO) feedback of collocated information has numerous
advantages. This system would be totally modular, resulting in low computing
requirements and associated weight penalty, no redesigns necessary to add additional
sensor/actuator pairs, and a very graceful performance degradation to single point failures.
These factors can potentially lead to reduced product cost, and reduced installation and
maintenance costs, over other approaches.
Each SISO control loop can be tuned to give excellent performance using tachometer
information from the engines. Each sensor/actuator pair will effectively behave as a
distributed tuned vibration absorber, that is always perfectly tuned with the disturbance
frequency, cancels vibrations at the fundamental frequency and desired harmonics,
and concentrates its effort on the structural modes that couple efficiently with the
acoustics.
Replacing the structural information with microphone feedback would be guaranteed
to give better nominal performance, since the true quantity of interest would then be
minimized. However, the resulting compensator would have to be both MIMO, and
adaptive, which gives increased computational cost, and introduces robustness and fault
tolerance issues. Since the noise is caused by structural vibration, clever vibration control
can replace microphone feedback. The advantages are simplicity, and potentially lower
life-cycle cost.
The distributed collocated structural control architecture is validated analytically herein.
The internal noise must be reduced at the propeller blade passage frequency (BPF) and
the next few harmonics. The narrowband control for each of these frequencies is
completely independent. The control of the fundamental will require the most actuation,
since at higher harmonics, the propeller sound pressure distribution on the aircraft is more
compact. As a result, the predictions will concentrate on control at the blade passage
frequency. The properties of a deHavilland Dash-8 Series 100 fuselage are used for the
model, and a predicted propeller sound pressure field corresponding to the Dash-8 is used
as the forcing function.
3. VIBRO-ACOUSTIC MODEL DEVELOPMENT
The model must include those features of the real problem that will substantially affect
the noise reductions that will be achievable by structural control, however, it need not
accurately model the absolute noise levels inside the aircraft. A relatively simple model is
therefore sufficient.
The fuselage is modelled as a cylinder with uniform mass and stiffness properties. The
bending stiffnesses are calculated from those of the frames and stringers, with the effect
of the skin included using shear lag arguments. The primary limitation of this model is
that the dynamics are assumed to be dominated by the bending vibrations of the fuselage,
and the in-plane extensional vibration is assumed to be much stiffer. As a result of this
assumption, ring ‘‘breathing’’ modes are not included, and the lowest bending modes are
insufficiently stiff. The effective mass per unit surface area in the model was decreased to
match modal frequencies with data from deHavilland. Other simplifying assumptions
include:
(1) Ring and axial bending coupling is not included.
(2) Deviations of the real fuselage from a cylinder, such as the presence of wings, cockpit,
tail, variation in radius of curvature, doors, etc. are not modelled.
coiioc\1rb s1ric1ir\i coN1roi 109
(3) Since the floor adds significant stiffnesses to the structure in a region that would
otherwise be strongly affected by the propeller field, this symmetric stiffening effect
must be included. All other effects on the floor are ignored.
(4) All structural and acoustic modes are assumed to have 10% modal damping.
(5) The passive stiffness of the piezoelectric actuators is not included.
(6) The acoustic space is modelled as a pure closed cylinder (with no floor).
(7) The acoustic field does not affect the structural modes.
Polar co-ordinate axes will be used, with r being radial, y longitudinal (axial) and u
circumferential, as shown in Figure 2.
3.1. s1ric1ir\i xobri
The cylinder modes have independent variation in the y and u directions, indexed
separately by i and j. The symmetric (·)
s
and anti-symmetric (·)
a
mode shapes are [20]
0

u
1
s
=
0
i cos iu cos jpy/l
sin iu cos jpy/l
1
,
0

u
1
a
=
0
i sin iu cos jpy/l
−cos iu cos jpy/l
1
, (1)
where r˜ and u are the deflections in the radial and circumferential directions, respectively.
The mode i=1 involves no frame deformation. The pure extensional mode of each frame,
i=0, is not included. The modal mass is m
ij
=
1
2
r
s
lRp(i
2
+1), the modal stiffness is
k
ij
=i
2
p
l
2
0
(EI)
y
d
f
R
3
(i
2
−1)
2
+
R(EI)
z
d
s 0
jp
l
1
4
1
(2)
and the frequencies are v
ij
=zk
ij
/m
ij
. The parameters (EI)
y
/d
f
and (EI)
z
/d
s
are the average
bending stiffnesses per unit length. For small j, the frequencies are dominated by the ring
bending stiffness.
The most important difference between the actual fuselage and a pure cylinder model
is the stiffening effect of the floor on the symmetric modes. All other effects of the floor
are ignored. The additional stiffness between symmetric modes with circumferential mode
numbers m and n is (DK)
mn
=k
f
f
f
m
f
f
n
, where f
f
m
is the relative horizontal displacement
between the floor attachment points for mode m, given by
f
f
m
=2(m cos mu
f
sin u
f
−sin mu
f
cos u
f
). (3)
The effective floor stiffness k
f
is chosen to be large with respect to the ring bending stiffness.
The magnitude of the computer predicted propeller sound pressure field at BPF on the
port side is shown in Figure 3. Note that, while only the magnitude is shown, the phase
Figure 2. Coordinate axes for model; r is radial, y is longitudinal, and u is circumferential.
b. c. x\cx\r1iN 110
Figure 3. Port side propeller sound pressure magnitude distribution, dB with respect to maximum value. The
fore and aft limits of the fuselage and the propeller plane are indicated.
variation is also important when performing the summation over modes. The effect of the
disturbance sound pressure is computed as a double integral over u and y:
b
wij
=
g
l
0
g
2p
0
P
ext
(u, y)f
ij
(u, y)R du dy, (4)
where P
ext
is the external sound pressure distribution, and f
ij
the radial displacement of
the ijth mode shape.
The free response of a piezoelectric to an applied voltage V is given by the induced strain
L=(d
31
/d
p
)V. The effect of a piezoelectric actuator bonded to a beam can be approximated
by assuming uniform shear within the beam, and constant strain in the piezoelectric [21].
This gives a resulting strain in the beam of
e=
1
1+c
b
L, where c
b
=
(EI)
s
t
2
(EA)
p
. (5)
A finite length piezoelectric is typically modelled as applying pin forces F at the ends
of the actuator, where
F=(EA)
p
c
b
1+c
b
L=mV. (6)
For actuators bonded to the inner surface of the fuselage frames, this force affects each
mode through the circumferential displacement u, as well as applying a moment. For the
fuselage parameters considered here, the force effect of the piezo is larger than the moment
effect for iQ4.
The pin force model is reasonable when the actuator length is small compared to the
wavelengths of the modes of interest. However, this is a poor assumption to use for this
model, since one of the advantages in using piezoelectric actuators is that their
area-averaging effect reduces their authority over high wavenumber modes that do not
couple effectively to the interior acoustic field. Instead, the piezo effectiveness on each
mode is computed by integrating over the linearly varying shear force distribution
generated by the actuator. For a piezoelectric actuator that runs from a circumferential
coiioc\1rb s1ric1ir\i coN1roi 111
angle of u to u , define u
c
=(u +u )/2 and Du=(u −u )/2. Then the shear stress as a function
of u and the equivalent pin force F is
S=
2F
RDu
(u−u
c
)
Du
. (7)
Evaluating the resulting integral yields the piezoelectric effectiveness per volt as
b
s
p
i
=
2
m
0
1+
t
R
i
2
10
(sin iu −sin iu )
(iDu)
2

(cos iu +cos iu )
iDu
1
cos
jpy
l
(8)
for the symmetric modes, with a corresponding expression for the anti-symmetric modes.
For Du:0, equation (8) is equivalent to assuming pin forces applied at the weighted centre
of the shear force distribution. The modal response of the collocated strain sensors is the
same as the modal influence of the piezoelectric actuators.
The structural model ultimately yields the amplitudes q
s
and q
a
of the assumed symmetric
and anti-symmetric mode shapes, from which the surface velocity of the fuselage can be
predicted by summation. In addition to computing the resulting acoustic response, it is
useful to compute the power input to the structure by the propeller disturbance, and the
total vibrational energy in the structure, given by
P
in
=v
0
s
i
s
j
i(b
s
w
ij
q
s
ij
+b
a
w
ij
q
a
ij
), (9)
E
s
=s
i
s
j
(M
ij
v
2
0
+K
ij
)((q
s
ij
)
2
+(q
a
ij
)
2
). (10)
The power input at a point, P
in
(u, y), can be computed from the surface velocity and the
propeller disturbance, and is useful for determining potential actuator locations. The
open-loop contribution of different modes to the total energy, shown in Figure 4, indicates
that circumferential modes iq5 and axial modes jq13 do not contribute substantially, and
can be eliminated from the model. Modes i=3, jE9 dominate, due to a good match with
the disturbance in both frequency and mode shape. At higher harmonics, many more
Figure 4. Contribution to total energy by mode number at the blade passage frequency. Contributions from
symmetric and anti-symmetric modes are grouped together.
b. c. x\cx\r1iN 112
modes contribute, due to both the higher frequency of the disturbance, and to the higher
spatial wavenumber content of the disturbance. As a result, it is more difficult to get an
accurate characterization of the response.
3.2. \cois1ic rrsioNsr
The mode shapes for the sound pressure in a cylinder are indexed by i, j, and k;
corresponding to variations in the circumferential, axial, and radial directions, respectively.
The symmetric mode shapes are [20]:
c
s
ijk
(u, y, r)=J
i
(l
ik
r/R) cos ( jpy/l) cos iu, i, j, k=0, 1, . . . , (11)
with frequency v
ijk
=czl
2
ik
/R
2
+j
2
p
2
/l
2
, where J
i
is a Bessel function, and l
ik
satisfies
J'
i
(l
ik
)=0. The anti-symmetric modes have identical frequencies, and circumferential
variation sin iu rather than cos iu.
From Nelson and Elliot [22] the total sound pressure is given by
P(u, y, r)=−v
0
rc
2
s
i,j,k
0
f
ijk
c
ijk
(u, y, r)
V
ijk
(−v
2
0
+2z
a
v
ijk
v
0
+v
2
ijk
)
1
. (12)
The summation includes both the symmetric and anti-symmetric modes. The modal
volume V
ijk
is the integral of c
2
ijk
over the internal volume, and z
a
is the assumed damping
ratio of the acoustic modes. The quantity f
ijk
is the generalized force on mode (i, j, k), given
by the integral over the fuselage surface area of the mode shape c
ijk
(u, y, 1) multiplied by
the surface velocity predicted by the structural model. By orthogonality, the acoustic mode
c
s
ijk
couples only with the structural mode f
s
ij
, and c
a
ijk
couples only with f
a
ij
. The generalized
forces in equation (12) are given by
f
ijk
=q
ij
iRplv
0
2
J
i
(l
ik
), (13)
with q
s
ij
being used for symmetric acoustic modes, and q
a
ij
for antisymmetric ones. Note that
the i=0 acoustic modes do not couple with any of the assumed structural modes; this
corresponds to a pure breathing mode of the ring which does not involve any bending
motion.
The goal is to minimize the peak internal sound pressure. An alternative performance
measure is the total acoustic potential energy, given by
E
ac
=s
i,j,k
V
ijk
4rc
2
((q
s
ijk
)
2
+(q
a
ijk
)
2
), (14)
where q
ijk
are the amplitudes of the acoustic modes, given by the terms in brackets in
equation (12).
One of the key assumptions made in the conceptual design of the control architecture
was that only the low wave-number structural modes couple well with the acoustic
response. While this is true for an infinite flat plate radiating into a half-space [14], the
coupling is more complicated for an enclosed cylinder, and there is no cut-off phenomenon
wherein certain modes do not couple at all. The structural-acoustic coupling predicted by
this model is plotted in Figure 5. The coupling is computed per structural mode, and
plotted as a function of wave number. Also shown is the cut-off wave-number predicted
by infinite flat plate theory, and the wave-number filtering provided by the distributed
actuator and sensor.
coiioc\1rb s1ric1ir\i coN1roi 113
Figure 5. Structural-acoustic coupling as a function of wave-number: × predicted by model; · · · cut-off wave
number predicted by infinite flat plate theory; ---- wave-number filtering provided by the distributed sensor and
actuator.
3.3. xobri v\iib\1ioN
A deHavilland Dash-8 Series 100 fuselage was available for testing. This fuselage was
used in previous studies of propeller noise transmission [23]. Although a complete modal
analysis of the fuselage was not performed, the predicted circumferential bending
frequencies are within about 10% of the measured values for the modes near the blade
passage frequency. Because of the high acoustic modal density on both the model and the
aircraft in the frequency range of interest, it is difficult to compare acoustic mode
frequencies and shapes.
The turbo-prop noise control application demands that the piezoelectrics be capable of
generating substantial strains in the fuselage frames. The maximum strain that can be
induced in the structure is related to the relative stiffness of the piezoelectric actuator and
the structure, given by the parameter c
b
in equation (5). Therefore, a relatively thick crystal
that is approximately impedance matched to the structure is desired. For implementation,
a layered actuator can be used to reduce voltage requirements.
In order to verify the strain that could be induced, several actuators were bonded to
the fuselage, and the resulting strain measured. The predicted levels were achieved,
indicating that equation (5) is approximately valid, even though flat actuators were bonded
to a curved surface. Estimates of the in-flight dynamic strain that the piezoelectric
actuators will enounter are lower than the maximum capability indicated by the
preliminary experimentation.
4. CONTROL DESIGN
There are several approaches that could be used to design control systems for this
problem. Much of the work on noise control has used an LMS adaptive disturbance
feedforward algorithm, while much of the research in the related area of vibration control
has used standard feedback design techniques. It can be shown, however, that for harmonic
disturbances, the two approaches yield identical controllers [24], and therefore for analysis
purposes, linear feedback was used. Controllers were designed using low order, classical
b. c. x\cx\r1iN 114
frequency domain techniques, and applied to the state space model of the fuselage to
obtain the closed loop modal amplitudes.
The simplest control architecture is to design independent controllers for each
sensor/actuator pair. High gain is desired at the disturbance frequency, v
0
, in order to
minimize the average strain over the region covered by each actuator. The transfer function
for each compensator is of the form
K
i
(s)=
2av
0
=G( jv
0
)=
0
s
s
2
+2z
c
v
0
s+v
2
01
, (15)
where G( jv
0
) is the transfer function from an actuator to the collocated sensor, evaluated
at the excitation frequency v
0
, and a is a gain parameter. This lightly damped second order
system is a standard compensator for tonal control, and can be easily implemented with
a slowly time-varying excitation frequency v
0
[25]. Stability-robustness is guaranteed
because both the compensator and the plant are positive real. Performance is obtained with
high gain at the disturbance frequency; the sensor output is reduced by a factor of a/z
c
.
Increasing a increases the gain at frequencies away from the disturbance, which may
increase broadband noise problems. Decreasing z
c
makes the frequency range of
disturbance attenuation narrower, which is acceptable if the disturbance frequency is
constant. For the simulations, the compensator design parameters were selected to reduce
the sensor output to 1% of its former value. Further reductions will not yield substantial
further improvements in the acoustic response, since this will now be dominated by the
fuselage response elsewhere.
5. PREDICTIONS FROM MODEL
The structural modelling and sound pressure distribution information can be used to
predict the surface vibrational velocity distribution, and the corresponding distribution of
input power at the blade passage frequency. The predicted input power is shown in
Figure 6, with the starboard side of the aircraft on the left, and the front of the cabin at
the bottom. The bottom of the fuselage is in the centre of the plot, and the top at the edges.
Figure 6. Predicted open-loop input power distribution, in dB with respect to the peak input power. A potential
set of actuator locations is shown with dashed lines. The axial location of the propeller plane, and circumferential
angle of the floor locations are indicated with dotted lines.
coiioc\1rb s1ric1ir\i coN1roi 115
Figure 7. Predicted open-loop internal acoustic response at passenger head height, dB with respect to peak.
The axial location of the propeller plane, and the horizontal location of passenger seats are shown with dotted
lines.
The purpose in plotting the input power distribution is to determine good actuator
locations, and a potential set is shown in this figure.
The predicted open-loop internal acoustic field is shown in Figure 7. The peak response
is on the port side of the aircraft, near the propeller plane. The actuator locations indicated
in Figure 6 were then used to obtain the closed-loop acoustic field in Figure 8. This
configuration meets the performance goal of a 10 dB reduction in peak internal noise,
indicating that this goal is feasible. Independent control laws were used for all
actuator/sensor pairs, and all actuators covered a 50° arc of a single frame. The total
weight of piezoelectrics in this configuration is approximately 16 kg. The weight penalty
could be reduced with additional optimization of the control architecture, actuator
locations, and the actuators themselves.
Neither the total input power, nor the total vibrational energy within the structure are
representative of the peak internal noise, as was noted by Thomas et al. [16]. This is due
to the fact that not all structural modes couple equally well with the acoustic field, and
thus the energy in some modes is not as important as the energy in others. This result is
evident in Figure 9, which gives both the open-loop and closed-loop energy in the structure
Figure 8. Predicted closed-loop internal acoustic response at passenger head height, dB with respect to
open-loop peak, using 18 actuators. The axial location of the propeller plane, and the horizontal location of
passenger seats are shown with dotted lines.
b. c. x\cx\r1iN 116
Figure 9. Structural energy as a function of circumferential wave-number; - - - - open-loop, —— closed-loop.
as a function of circumferential wave-number. While the total structural energy has been
decreased only slightly, some energy has been shifted from low wave-number modes which
couple well with the acoustic field, to higher wave-number modes where the energy has
little effect (see Figure 5). This result is particularly important, as it indicates how the
control approach results in the achieved noise reduction. Also note that while the total
structural vibrational energy is a poor measure of the acoustic performance, structural
vibration is also a problem, and therefore it is necessary to monitor the structural energy
to ensure that it does not increase substantially.
Closed-loop results for various actuator configurations are summarized in Table 1,
which lists the power input to the structure, the structural energy, the acoustic potential
energy, and the peak internal noise. The cases with 1 and 3 actuators have all of the
actuators near the peak of the sound pressure disturbance on the port side, the next case
has actuators on both sides of the aircraft. These show the gradual improvement with
additional actuators, however, none of them has sufficient actuators to prevent sound from
entering the cabin. An inadvertant variation in the modal damping between symmetric and
asymmetric modes is responsible for the occasional slight increase in total structural energy
observed, even though the total input power is reduced, and the compensator is purely
dissipative.
T\nir 1
Input power, structural vibrational energy, acoustic potential energy, and reduction in peak
internal cabin noise for various actuator and sensor locations, and configurations; all dB with
respect to open-loop
Case (Actuators Pin (dB) Es (dB) Eac (dB) Pmax (dB)
— 0 0 0 0 0
1 1 −0·38 0·56 0·10 −0·55
2 3 −1·33 −0·22 −0·29 −2·09
3 9 −2·92 2·09 −2·32 −4·95
4 18 −2·90 −0·81 −3·56 −11·21
5 18/4 −0·71 0·21 −3·28 −9·88
6 12/2 −1·05 0·22 −3·07 −8·78
coiioc\1rb s1ric1ir\i coN1roi 117
Case 4 corresponds to the actuator locations in Figure 8. This is the minimum coverage
required for this model to obtain sufficient noise reductions. Optimizing the placement is
important, as different configurations with the same number of actuators can yield
significantly different performance. While a complete numerical optimization was not
performed, the locations were chosen to cover the peak areas of the input power
distribution, so that the specified external disturbance could be cancelled.
In addition to optimizing actuator locations, a comparison of alternate control
architectures with a given selection can be made. Two additional cases are shown in
Table 1. For case 5, the same actuators were grouped together so that only four separate
control loops were used. Although some modularity is sacrificed by requiring connections
between actuators on adjacent frames, the control hardware is simplified substantially. The
additional simplicity is gained with only a 1·4 dB increase in internal noise. The final case
indicates that it may not be necessary to actuate in all of the locations indicated in Figure 8.
It may be sufficient only to sense the strain in the frames furthest from the propeller plane,
and allow actuators to use sensor information from neighbouring frames. In case 6, all
the actuators on one side of the aircraft were grouped together. Similar information to
the previous two cases was available, but only 12 of the previous 18 actuators were used.
The reduction obtained in this case is 1·6 dB better than if only collocated information
were available.
Finally, the 18 actuator locations from Figure 6 were used with a different external
sound pressure distribution corresponding to a 15% change in engine rpm. In
addition to the change in the applied pressure distribution, the structural and acoustic
dynamics evaluated at the disturbance frequency also changed, which resulted in
different modes being excited. At this second frequency, the results were superior. The
control achieved a 13 dB reduction in peak sound pressure, accompanied by
significant decrease in input power, structural energy, and acoustic energy. This verifies
that the control architecture and actuator locations are robust to variations in both the
details of the propeller sound pressure footprint, and in the structural and acoustic
dynamics.
6. CONCLUSIONS
There is a desire to reduce the level of cabin noise in turbo-prop aircraft. A promising
approach is to use active structural control to minimize the fuselage noise transmission
in the vicinity of the propeller noise ‘‘footprint’’. A simple, robust, and modular system
can be obtained by placing piezoelectric actuators on the fuselage frames near the propeller
plane, and using collocated strain feedback to reduce the noise transmission. Simulations
with a simple vibro-acoustic model indicate that the desired goal of a 10 dBA reduction
in peak internal noise is feasible with fewer actuators and a lower weight penalty
than existing speaker-based active noise control systems. A proper comparison of
structural actuators versus speakers, and collocated feedback of structural sensors versus
microphone feedback, must be based on the total life-cycle cost of any option for a given
performance.
The weight penalty required to achieve the specified performance objective can be
minimized by studying different actuator/sensor placements and control architectures.
Using feedback of strain information from neighbouring frames in addition to the
collocated information was shown to improve performance. The computational burden
can be reduced by grouping actuators on different frames together. The simulations
indicate that only a few distinct control loops are required.
b. c. x\cx\r1iN 118
ACKNOWLEDGMENTS
This work was conducted as part of a collaborative research and development project
between deHavilland Inc. and the National Research Council of Canada, while the author
was with the Institute for Aerospace Research at NRC. Propeller sound pressure field and
structural information were obtained from deHavilland Inc. The assistance of deHavilland,
and of the rest of the staff at the Aeroacoustics Facility of NRC is greatly appreciated.
Thanks also to Dr. Andreas von Flotow for his comments.
REFERENCES
1. J. S. MixsoN and C. A. Povrii 1984 AIAA/NASA 9th Aeroacoustics Conference, (Williamsburg,
VA). Review of Recent Research on Interior Noise of Propeller Aircraft. (AIAA Paper 84-2349).
2. R. B. Bn\1, J. Sonirszcz\Nsii and J. S. MixsoN 1977 Noisexpo, 40–45. Reduction of aircraft
cabin noise by fuselage structural optimization.
3. E. H. W\1rrx\N, D. K\i1riN and S. L. S\riN 1983 SAE Paper 830736. Fokker’s activities in
cabin noise control for propellor aircraft.
4. W. G. H\ivorsrN and U. Exnorc 1989 SAE Paper 891080, General aviation aircraft meeting
and exposition. Interior noise control of the Saab 340 Aircraft.
5. S. J. Eiiio1, P. A. NrisoN, I. M. S1o1nrrs and C. C. Boicnrr 1990 Jounal of Sound and
Vibration 140, 219–238. In-flight experiments on the active control of propeller-induced cabin
noise.
6. I. U. Borcnrrs, K. RrNcrr, J. P\iii\rb, G. Biiioib, P. Korrs and E. DoiirNnrrc 1993
Inter-Noise, (Leuven, Belgium), 59–64. Selected flight test data and control system results of the
CEC BRITE/EURAM ASANCA study.
7. M. A. SixisoN, T. M. LioNc, C. R. Fiiirr and J. D. JoNrs 1991 AIAA J. Aircraft 28, 208–215.
Full-scale demonstration tests of cabin noise reduction using active vibration control.
8. C. R. Fiiirr, S. D. SN.brr, C. H. H\NsrN and R. J. Siicox 1992 AIAA Journal 30, 2613–2617.
Active control of interior noise in model aircraft fuselages using piezoceramic actuators.
9. R. J. Siicox, S. Lrrrnvrr, V. L. Mr1c\ir, T. B. Br.rr and C. R. Fiiirr 1992 Proc.,
DGLR/AIAA 14th Aeroacoustics Conference, 542–551. Evaluation of piezoceramic actuators for
control of aircraft interior noise.
10. C. R. Fiiirr and G. P. Ginns 1994 Noise-Con 94, (Ft. Lauderdale, Florida). Active control of
interior noise in a business jet using piezoceramic actuators.
11. R. J. Siicox, C. R. Fiiirr and H. C. Lrs1rr 1990 AIAA Journal 28, 1397–1404. Mechanisms
of active control in cylindrical fuselage structures.
12. D. J. Rossr11i, M. A. Norris, S. C. Soi1nv\rb and J. Q. SiN 1993 Recent Advances in Active
Control of Sound and Vibration, (Blacksburg, Virginia), VPI and SU, Technomic. A comparison
of speakers and structural-based actuators for aircraft cabin noise control.
13. A. I. V.\i.snrv and B. D. T\r1\iovsiii 1976 Soviet Phys. Acoustics 22, 382–385. Radiation
compensation of a flexurally vibrating plate at subcritical frequencies.
14. C. R. Fiiirr and R. A. Birbisso 1991 Journal of Sound and Vibration 148, 355–360. A
wavenumber domain approach to the active control of structure-borne sound.
15. A. J. Biiixorr, P. A. NrisoN and S. J. Eiiio1 1990 Journal of Sound and Vibration 140,
191–217. Theoretical studies of the active control of propeller-induced cabin noise.
16. D. R. Tnox\s, P. A. NrisoN and S. J. Eiiio1 1993 Journal of Sound and Vibration 167, 91–111.
Active control of the transmission of sound through a thin cylindrical shell, Part I: The
minimization of vibrational energy.
17. H. C. Lrs1rr and S. Lrrrnvrr 1993 J. of Intelligent Materials Systems and Structures 4, 295–306.
Piezoelectric actuator models for active sound and vibration control of cylinders.
18. R. S. L\Ncir. 1993 Journal of Sound and Vibration 163, 207–230. A dynamic stiffness/boundary
element method for the prediction of interior noise levels.
19. H. C. NrisoN, B. Z\io1ovsii and M. BrrNs1riN 1958 Proceedings, National Specialists Meeting
on Dynamics and Aeroelasticity (Fort Worth, Texas) 77–87. Vibration analysis of an
orthogonally stiffened circular fuselage and comparison with experiment.
20. R. D. BirviNs 1979 Formulas for Frequency and Mode Shape. Malabar, Florida: Robert Krieger.
21. E. H. ANbrrsoN 1989 Master’s thesis, Massachusetts Institute of Technology—Space Systems
Laboratory Report 5–89. Induced strain actuation of one- and two-dimensional structures.
coiioc\1rb s1ric1ir\i coN1roi 119
22. P. A. NrisoN and S. J. Eiiio1 1992 Active Control of Sound. London: Academic Press.
23. D. G. M\cM\r1iN, G. L. B\sso and F. W. SiiNcrri\Nb 1995 Journal of Sound and Vibration
187, 467–483. Aircraft fuselage noise transmission measurements using a reciprocity technique.
24. D. G. M\cM\r1iN 1994 American Control Conference (Baltimore, MD), 1632–1636. A feedback
perspective on the LSM disturbance feedforward algorithm.
25. K. B. ScrinNrr, L. A. Sirvrrs and A. H. voNFio1ov 1993 Journal of Sound and Vibration 167,
17–40. Active narrow-band vibration isolation of machinery noise from resonant substructures.
APPENDIX: NOTATION
(EI)z axial bending stiffness per stringer v0 disturbance frequency
R radius of fuselage (EA)p extensional stiffness of piezo
l length of aircraft cabin dp thickness of piezoelectric
df frame spacing t distance, piezo to frame neutral axis
ds stringer spacing d31 piezoelectric strain constant
uf floor angle, from top of aircraft r density of air
rs effective mass per unit surface area c speed of sound in air
(EI)y ring bending stiffness per frame

In the context of the wave-number based radiative efficiency arguments of reference [14]. the impact on other aspects of the overall system. and has been investigated by a variety of researchers for the aircraft noise control problem [7–12]. To evaluate the capability of a collocated vibration control approach for reducing internal noise. since they are only needed near the propeller plane [11. The stiffening effect of the floor is partially included. high bandwidth. which reduces the power input from the propeller disturbance. CONCEPTUAL DESIGN One of the objectives that must be considered in developing an approach intended for eventual production is the ease of implementation of the ultimate design. Fuller et al. and because their area-averaging effect reduces their influence on structural modes that do not couple with the acoustic response. In addition. 12]. in that fewer actuators and sensors are required.  interior noise. with the ultimate goal of minimizing the weight penalty. By making the control objective the reduction of noise transmission. including finite element and boundary element methods [18]. High gain feedback then reduces fuselage vibration in the propeller footprint area. and sound radiation into the aircraft cabin. rather than of internal noise. The closed-loop structural response is obtained for a computer prediction of the propeller sound pressure field. as described below. a vibro-acoustic model is developed. The primary objective is to demonstrate that the desired performance goal of a 10 dBA reduction in peak internal noise can be obtained. most of the literature on this subject uses sound pressure information from within the cabin. Since accurate predictions of absolute noise levels are not required. rather than speakers. While difficult to quantify. One drawback to the use of speakers in the aircraft cabin for noise control is locating them so as to preserve their effect. [10] have used piezoelectric actuators bonded to an aircraft fuselage and obtained acceptable authority over internal noise using microphone feedback. A purely structural control approach that eliminates the needs for microphones can have considerable advantages with regards to the overall design objectives. Various researchers have investigated active structural and acoustic control for aircraft using simple cylinder models of the structural vibration [15–17]. similar to the model in reference [19]. the total system cost and weight must be minimized. Prior work on active structural-acoustic control has indicated that structural actuators can be used. With these design objectives in mind. the distributed piezoelectric actuator and collocated strain sensor act as spatial filters. The concept of controlling structural vibration to reduce transmitted or radiated noise is not new [13]. . there are numerous more detailed modelling techniques available. the handling of inevitable sensor/actuator failures. Finally. a simple model is sufficient. indicating that the required actuator strength is feasible. The model developed herein treats the fuselage as an orthotropic finite cylinder with equivalent bending stiffnesses to account for the frames and stringers. In addition. The key requirement is to identify an appropriate structural sensor. and substantial robustness to the wide range of operating conditions likely to be encountered on numerous different aircraft. . different actuator and sensor configurations and control architectures are evaluated. while protecting them from damage and minimizing any loss in useful space. the overall simplicity and modularity of the approach can be as important as the performance it achieves. an approach can be developed that uses only feedback of local structural information. this includes factors such as ease of production and assembly. Piezoelectric actuators are well suited to this problem due to their low weight.106 . However. 2. The interior acoustic response is then computed from the structural vibration.

nor simple. Choosing the sensor and actuator dimensions on the order of one half of the acoustic wavelength at the frequency of interest results in spatial wave-number filtering. the distance between frames is such that this remaining skin vibration can only couple well with the acoustic field above approximately 300 Hz. The limitation observed in the literature on purely structural control approaches to noise reduction is that only low wave-number structural motion couples well with the acoustic field [14. Precisely sensing the structural motion that radiates efficiently would improve performance. and gives substantial robustness. It is the frames which dominate the structural dynamics at the frequencies of interest. 9. due in part to the fact that the uncertain dynamics of the structure do not need to be accurately modelled. then the fuselage skin can still vibrate between the frames. This is particularly true if collocated strain sensors are used. These objectives are equivalent in the sense that reducing the former will reduce the latter. while noise transmission depends only on the local structural information at an actuator location. and minimize the sensor output. Piezoelectric actuator placement on fuselage frame. . While the transfer function from a piezoelectric actuator to a microphone varies considerably with changes in passenger load and operating conditions. as indicated schematically in Figure 1. Positive real feedback can then be used to guarantee stability. wherein the disturbance is prevented from entering the structure by high gain collocated feedback. This leads to several options for the Figure 1. This yields a simple control design problem. but is neither essential. If the frames are controlled. Minimizing the sensor is therefore an approximation to minimizing the acoustic radiation. Because the disturbance source is not at a single point. Thus it is sufficient to use collocated feedback. Thus. to give the largest possible moment arm to excite fuselage ring bending. actuators will be required at several circumferential locations on several frames. rather than reducing the internal noise. on the deHavilland Dash-8. The control system alters the mode shapes of the system so that the impedance seen by the disturbance is increased. Performance can also be achieved by altering mode shapes to reduce structural-acoustic coupling. the internal noise is a global performance objective. The piezoelectric actuators will be bonded to the inner surface of the fuselage frames. since the resulting transfer functions will be positive real. However.   107 Substantial simplifications result by defining the objective to be reducing the noise transmission through the fuselage. minimizing structural motion may increase the acoustic response unless the sensors and actuators do not respond to or excite that motion that does not couple well. This approach is similar to the concept of disturbance isolation. This class of solution techniques has typically achieved excellent performance. the fundamental characteristics of the local structural transfer functions remain unchanged. However. 16]. this is analogous to performance isolation.

and a predicted propeller sound pressure field corresponding to the Dash-8 is used as the forcing function. Each sensor/actuator pair will effectively behave as a distributed tuned vibration absorber. resulting in low computing requirements and associated weight penalty. and the in-plane extensional vibration is assumed to be much stiffer. ring ‘‘breathing’’ modes are not included. VIBRO-ACOUSTIC MODEL DEVELOPMENT The model must include those features of the real problem that will substantially affect the noise reductions that will be achievable by structural control. and introduces robustness and fault tolerance issues. the propeller sound pressure distribution on the aircraft is more compact. The properties of a deHavilland Dash-8 Series 100 fuselage are used for the model. clever vibration control can replace microphone feedback. are not modelled. The primary limitation of this model is that the dynamics are assumed to be dominated by the bending vibrations of the fuselage. Other simplifying assumptions include: (1) Ring and axial bending coupling is not included. the predictions will concentrate on control at the blade passage frequency. This system would be totally modular. The advantages are simplicity. over other approaches. Replacing the structural information with microphone feedback would be guaranteed to give better nominal performance. no redesigns necessary to add additional sensor/actuator pairs. The narrowband control for each of these frequencies is completely independent. Each SISO control loop can be tuned to give excellent performance using tachometer information from the engines. The distributed collocated structural control architecture is validated analytically herein. . tail. however. since the true quantity of interest would then be minimized. and adaptive. etc. and a very graceful performance degradation to single point failures. it need not accurately model the absolute noise levels inside the aircraft. that is always perfectly tuned with the disturbance frequency. and reduced installation and maintenance costs. doors. since at higher harmonics. The bending stiffnesses are calculated from those of the frames and stringers. The fuselage is modelled as a cylinder with uniform mass and stiffness properties. A relatively simple model is therefore sufficient. with the effect of the skin included using shear lag arguments.  control architecture. . As a result of this assumption. The effective mass per unit surface area in the model was decreased to match modal frequencies with data from deHavilland. However. such as the presence of wings. A purely decentralized control scheme wherein each actuator uses only single-input/single-output (SISO) feedback of collocated information has numerous advantages. Since the noise is caused by structural vibration. These factors can potentially lead to reduced product cost. and potentially lower life-cycle cost. cockpit. and concentrates its effort on the structural modes that couple efficiently with the acoustics. 3. cancels vibrations at the fundamental frequency and desired harmonics. the resulting compensator would have to be both MIMO. The control of the fundamental will require the most actuation. which gives increased computational cost. The internal noise must be reduced at the propeller blade passage frequency (BPF) and the next few harmonics. variation in radius of curvature. As a result.108 . (2) Deviations of the real fuselage from a cylinder. and the lowest bending modes are insufficiently stiff.

y longitudinal (axial) and u circumferential. with r being radial. u s sin iu cos jpy/l 1 01 0 r˜ i sin iu cos jpy/l = . the frequencies are dominated by the ring bending stiffness.1. given by f f m=2(m cos muf sin uf−sin muf cos uf ). All other effects of the floor are ignored. and u is circumferential. .   109 (3) Since the floor adds significant stiffnesses to the structure in a region that would otherwise be strongly affected by the propeller field. Note that.   The cylinder modes have independent variation in the y and u directions. The mode i=1 involves no frame deformation. The modal mass is mij=1 rs lRp(i 2+1). (7) The acoustic field does not affect the structural modes. Coordinate axes for model. indexed separately by i and j. 3. u a −cos iu cos jpy/l 1 (1) where r˜ and u are the deflections in the radial and circumferential directions. is not included. For small j. Polar co-ordinate axes will be used. y is longitudinal. The parameters (EI)y /df and (EI)z /ds are the average bending stiffnesses per unit length. where f m is the relative horizontal displacement between the floor attachment points for mode m. this symmetric stiffening effect must be included. The most important difference between the actual fuselage and a pure cylinder model is the stiffening effect of the floor on the symmetric modes. The symmetric (·)s and anti-symmetric (·)a mode shapes are [20] 01 0 r˜ i cos iu cos jpy/l = . the phase Figure 2. The pure extensional mode of each frame. r is radial. while only the magnitude is shown. i=0. the modal stiffness is 2 kij=i 2p l (EI)y 2 R(EI)z jp (i −1)2+ ds l 2 df R 3 0 0 11 4 (2) and the frequencies are vij=zkij /mij . The additional stiffness between symmetric modes with circumferential mode f f f numbers m and n is (DK)mn=kf f m f n . (4) All structural and acoustic modes are assumed to have 10% modal damping. (3) The effective floor stiffness kf is chosen to be large with respect to the ring bending stiffness. respectively. All other effects on the floor are ignored. (6) The acoustic space is modelled as a pure closed cylinder (with no floor). as shown in Figure 2. (5) The passive stiffness of the piezoelectric actuators is not included. The magnitude of the computer predicted propeller sound pressure field at BPF on the port side is shown in Figure 3.

this force affects each mode through the circumferential displacement u . variation is also important when performing the summation over modes. and constant strain in the piezoelectric [21]. For a piezoelectric actuator that runs from a circumferential . Instead. the piezo effectiveness on each mode is computed by integrating over the linearly varying shear force distribution generated by the actuator. where F=(EA)p cb L=mV. 1+cb t (EA)p (5) A finite length piezoelectric is typically modelled as applying pin forces F at the ends of the actuator. This gives a resulting strain in the beam of e= 1 (EI)s L. y)R du dy. The effect of a piezoelectric actuator bonded to a beam can be approximated by assuming uniform shear within the beam. The fore and aft limits of the fuselage and the propeller plane are indicated. where cb= 2 . However. Port side propeller sound pressure magnitude distribution. . The effect of the disturbance sound pressure is computed as a double integral over u and y: bwij= gg l 0 2p Pext (u. the force effect of the piezo is larger than the moment effect for iQ4. and fij the radial displacement of the ijth mode shape. dB with respect to maximum value. since one of the advantages in using piezoelectric actuators is that their area-averaging effect reduces their authority over high wavenumber modes that do not couple effectively to the interior acoustic field. (4) 0 where Pext is the external sound pressure distribution. The pin force model is reasonable when the actuator length is small compared to the wavelengths of the modes of interest.  Figure 3. this is a poor assumption to use for this model. For the fuselage parameters considered here.110 . The free response of a piezoelectric to an applied voltage V is given by the induced strain L=(d31 /dp )V. y)fij (u. 1+cb (6) For actuators bonded to the inner surface of the fuselage frames. as well as applying a moment.

and is useful for determining potential actuator locations. The modal response of the collocated strain sensors is the same as the modal influence of the piezoelectric actuators. RDu Du (7) Evaluating the resulting integral yields the piezoelectric effectiveness per volt as s bpi= 2 t 1+ i 2 m R 0 10 (sin iu−sin iu) (cos iu+cos iu)   jpy − cos (iDu)2 iDu l 1 (8) for the symmetric modes. jE9 dominate. Then the shear stress as a function    of u and the equivalent pin force F is S= 2F (u−uc ) . At higher harmonics. The open-loop contribution of different modes to the total energy. shown in Figure 4. and the total vibrational energy in the structure. For Du:0. given by s s a a Pin=v0 s s i(bwij qij+bwij qij ). indicates that circumferential modes iq5 and axial modes jq13 do not contribute substantially. In addition to computing the resulting acoustic response. Contribution to total energy by mode number at the blade passage frequency. Contributions from symmetric and anti-symmetric modes are grouped together. i j (10) The power input at a point. equation (8) is equivalent to assuming pin forces applied at the weighted centre of the shear force distribution. from which the surface velocity of the fuselage can be predicted by summation. can be computed from the surface velocity and the propeller disturbance. and can be eliminated from the model. due to a good match with the disturbance in both frequency and mode shape. it is useful to compute the power input to the structure by the propeller disturbance.   111 angle of u to u. with a corresponding expression for the anti-symmetric modes. The structural model ultimately yields the amplitudes q s and q a of the assumed symmetric and anti-symmetric mode shapes. many more Figure 4. . Pin (u. define uc=(u+u)/2 and Du=(u−u)/2. y). i j (9) 2 s a Es=s s (Mij v0 +Kij )((qij )2+(qij )2 ). Modes i=3.

r)=Ji (lik r/R) cos ( jpy/l) cos iu. 1) multiplied by the surface velocity predicted by the structural model. y. 2 2 Vijk (−v0 +2za vijk v0+vijk ) 1 (12) The summation includes both the symmetric and anti-symmetric modes. The anti-symmetric modes have identical frequencies. 2 (13) s a with qij being used for symmetric acoustic modes. y. and the wave-number filtering provided by the distributed actuator and sensor. From Nelson and Elliot [22] the total sound pressure is given by P(u. j.2. The structural-acoustic coupling predicted by this model is plotted in Figure 5. and radial directions. The goal is to minimize the peak internal sound pressure. (11) 2 with frequency vijk=czlik /R 2+j 2p 2/l2 .112 . . and to the higher spatial wavenumber content of the disturbance. the acoustic mode s s a a cijk couples only with the structural mode fij . it is more difficult to get an accurate characterization of the response.  modes contribute. and za is the assumed damping ratio of the acoustic modes. y. respectively. 1. The modal 2 volume Vijk is the integral of cijk over the internal volume. i. and lik satisfies J' (lik )=0. where Ji is a Bessel function. and cijk couples only with fij . 4rc 2 ijk (14) where qijk are the amplitudes of the acoustic modes. and plotted as a function of wave number.k 0 fijk cijk (u. due to both the higher frequency of the disturbance. The symmetric mode shapes are [20]: s cijk (u. y.j.j. k=0. An alternative performance measure is the total acoustic potential energy. . k). j. the coupling is more complicated for an enclosed cylinder. r)=−v0 rc 2 s i. Also shown is the cut-off wave-number predicted by infinite flat plate theory. . given by the integral over the fuselage surface area of the mode shape cijk (u. The quantity fijk is the generalized force on mode (i. One of the key assumptions made in the conceptual design of the control architecture was that only the low wave-number structural modes couple well with the acoustic response.   The mode shapes for the sound pressure in a cylinder are indexed by i. given by the terms in brackets in equation (12).k Vijk a ((q s )2+(qijk )2 ). As a result. and there is no cut-off phenomenon wherein certain modes do not couple at all. r) . . axial. By orthogonality. . The generalized forces in equation (12) are given by fijk=qij iRplv0 Ji (lik ). Note that the i=0 acoustic modes do not couple with any of the assumed structural modes. . 3. and qij for antisymmetric ones. and k. j. The coupling is computed per structural mode. given by Eac= s i. corresponding to variations in the circumferential. and circumferential i variation sin iu rather than cos iu. While this is true for an infinite flat plate radiating into a half-space [14]. this corresponds to a pure breathing mode of the ring which does not involve any bending motion.

The turbo-prop noise control application demands that the piezoelectrics be capable of generating substantial strains in the fuselage frames. linear feedback was used. It can be shown. Much of the work on noise control has used an LMS adaptive disturbance feedforward algorithm. ---. a layered actuator can be used to reduce voltage requirements. Because of the high acoustic modal density on both the model and the aircraft in the frequency range of interest. the two approaches yield identical controllers [24]. This fuselage was used in previous studies of propeller noise transmission [23]. classical . given by the parameter cb in equation (5). 4. In order to verify the strain that could be induced. even though flat actuators were bonded to a curved surface. Controllers were designed using low order. the predicted circumferential bending frequencies are within about 10% of the measured values for the modes near the blade passage frequency. Estimates of the in-flight dynamic strain that the piezoelectric actuators will enounter are lower than the maximum capability indicated by the preliminary experimentation. Structural-acoustic coupling as a function of wave-number: × predicted by model. Although a complete modal analysis of the fuselage was not performed. indicating that equation (5) is approximately valid. and therefore for analysis purposes. a relatively thick crystal that is approximately impedance matched to the structure is desired. however. For implementation. Therefore. that for harmonic disturbances. several actuators were bonded to the fuselage. while much of the research in the related area of vibration control has used standard feedback design techniques.   113 Figure 5.   A deHavilland Dash-8 Series 100 fuselage was available for testing. · · · cut-off wave number predicted by infinite flat plate theory. it is difficult to compare acoustic mode frequencies and shapes. and the resulting strain measured.wave-number filtering provided by the distributed sensor and actuator. 3. The predicted levels were achieved.3. The maximum strain that can be induced in the structure is related to the relative stiffness of the piezoelectric actuator and the structure. CONTROL DESIGN There are several approaches that could be used to design control systems for this problem.

Decreasing zc makes the frequency range of disturbance attenuation narrower. For the simulations. The predicted input power is shown in Figure 6. Figure 6. and can be easily implemented with a slowly time-varying excitation frequency v0 [25]. . and the corresponding distribution of input power at the blade passage frequency. and applied to the state space model of the fuselage to obtain the closed loop modal amplitudes. 5. with the starboard side of the aircraft on the left. The bottom of the fuselage is in the centre of the plot. Predicted open-loop input power distribution. which is acceptable if the disturbance frequency is constant. which may increase broadband noise problems. =G( jv0 )= s 2+2zc v0 s+v0 0 1 (15) where G( jv0 ) is the transfer function from an actuator to the collocated sensor. Performance is obtained with high gain at the disturbance frequency. The simplest control architecture is to design independent controllers for each sensor/actuator pair. The axial location of the propeller plane. Stability-robustness is guaranteed because both the compensator and the plant are positive real. . and a is a gain parameter.114 . Further reductions will not yield substantial further improvements in the acoustic response. v0 . and circumferential angle of the floor locations are indicated with dotted lines. since this will now be dominated by the fuselage response elsewhere. PREDICTIONS FROM MODEL The structural modelling and sound pressure distribution information can be used to predict the surface vibrational velocity distribution. and the front of the cabin at the bottom. and the top at the edges. evaluated at the excitation frequency v0 .  frequency domain techniques. the sensor output is reduced by a factor of a/zc . A potential set of actuator locations is shown with dashed lines. the compensator design parameters were selected to reduce the sensor output to 1% of its former value. in order to minimize the average strain over the region covered by each actuator. in dB with respect to the peak input power. Increasing a increases the gain at frequencies away from the disturbance. The transfer function for each compensator is of the form Ki (s)= 2av0 s 2 . This lightly damped second order system is a standard compensator for tonal control. High gain is desired at the disturbance frequency.

Neither the total input power.   115 Figure 7. The axial location of the propeller plane. dB with respect to open-loop peak. Independent control laws were used for all actuator/sensor pairs. Predicted closed-loop internal acoustic response at passenger head height. dB with respect to peak. using 18 actuators. actuator locations. and the actuators themselves. This configuration meets the performance goal of a 10 dB reduction in peak internal noise. and all actuators covered a 50° arc of a single frame. and the horizontal location of passenger seats are shown with dotted lines. [16]. The peak response is on the port side of the aircraft. The predicted open-loop internal acoustic field is shown in Figure 7. . near the propeller plane. This result is evident in Figure 9. nor the total vibrational energy within the structure are representative of the peak internal noise. Predicted open-loop internal acoustic response at passenger head height. indicating that this goal is feasible. which gives both the open-loop and closed-loop energy in the structure Figure 8. The weight penalty could be reduced with additional optimization of the control architecture. The actuator locations indicated in Figure 6 were then used to obtain the closed-loop acoustic field in Figure 8. The axial location of the propeller plane. and a potential set is shown in this figure. as was noted by Thomas et al. and thus the energy in some modes is not as important as the energy in others. The total weight of piezoelectrics in this configuration is approximately 16 kg. The purpose in plotting the input power distribution is to determine good actuator locations. This is due to the fact that not all structural modes couple equally well with the acoustic field. and the horizontal location of passenger seats are shown with dotted lines.

The cases with 1 and 3 actuators have all of the actuators near the peak of the sound pressure disturbance on the port side. and configurations.. none of them has sufficient actuators to prevent sound from entering the cabin.  Figure 9.. acoustic potential energy. While the total structural energy has been decreased only slightly. Closed-loop results for various actuator configurations are summarized in Table 1. even though the total input power is reduced. . Also note that while the total structural vibrational energy is a poor measure of the acoustic performance. These show the gradual improvement with additional actuators. .open-loop. the structural energy. An inadvertant variation in the modal damping between symmetric and asymmetric modes is responsible for the occasional slight increase in total structural energy observed.116 . and the peak internal noise. and reduction in peak internal cabin noise for various actuator and sensor locations. and therefore it is necessary to monitor the structural energy to ensure that it does not increase substantially. to higher wave-number modes where the energy has little effect (see Figure 5). all dB with respect to open-loop Case — 1 2 3 4 5 6 (Actuators 0 1 3 9 18 18/4 12/2 Pin (dB) 0 −0·38 −1·33 −2·92 −2·90 −0·71 −1·05 Es (dB) 0 0·56 −0·22 2·09 −0·81 0·21 0·22 Eac (dB) 0 0·10 −0·29 −2·32 −3·56 −3·28 −3·07 Pmax (dB) 0 −0·55 −2·09 −4·95 −11·21 −9·88 −8·78 . some energy has been shifted from low wave-number modes which couple well with the acoustic field. however. the next case has actuators on both sides of the aircraft. the acoustic potential energy. which lists the power input to the structure. This result is particularly important. and the compensator is purely dissipative. Structural energy as a function of circumferential wave-number.. T 1 Input power. as it indicates how the control approach results in the achieved noise reduction. structural vibrational energy. as a function of circumferential wave-number. —— closed-loop. structural vibration is also a problem.

CONCLUSIONS There is a desire to reduce the level of cabin noise in turbo-prop aircraft. A promising approach is to use active structural control to minimize the fuselage noise transmission in the vicinity of the propeller noise ‘‘footprint’’. the locations were chosen to cover the peak areas of the input power distribution. and collocated feedback of structural sensors versus microphone feedback. Although some modularity is sacrificed by requiring connections between actuators on adjacent frames. A simple. The final case indicates that it may not be necessary to actuate in all of the locations indicated in Figure 8. and using collocated strain feedback to reduce the noise transmission. While a complete numerical optimization was not performed. It may be sufficient only to sense the strain in the frames furthest from the propeller plane. Finally. Two additional cases are shown in Table 1.   117 Case 4 corresponds to the actuator locations in Figure 8. a comparison of alternate control architectures with a given selection can be made. For case 5. but only 12 of the previous 18 actuators were used. The reduction obtained in this case is 1·6 dB better than if only collocated information were available. as different configurations with the same number of actuators can yield significantly different performance. This is the minimum coverage required for this model to obtain sufficient noise reductions. robust. The computational burden can be reduced by grouping actuators on different frames together. In addition to the change in the applied pressure distribution. . and modular system can be obtained by placing piezoelectric actuators on the fuselage frames near the propeller plane. This verifies that the control architecture and actuator locations are robust to variations in both the details of the propeller sound pressure footprint. accompanied by significant decrease in input power. must be based on the total life-cycle cost of any option for a given performance. At this second frequency. the results were superior. Optimizing the placement is important. the 18 actuator locations from Figure 6 were used with a different external sound pressure distribution corresponding to a 15% change in engine rpm. Similar information to the previous two cases was available. so that the specified external disturbance could be cancelled. Using feedback of strain information from neighbouring frames in addition to the collocated information was shown to improve performance. In addition to optimizing actuator locations. structural energy. the structural and acoustic dynamics evaluated at the disturbance frequency also changed. The simulations indicate that only a few distinct control loops are required. 6. The weight penalty required to achieve the specified performance objective can be minimized by studying different actuator/sensor placements and control architectures. all the actuators on one side of the aircraft were grouped together. and acoustic energy. The additional simplicity is gained with only a 1·4 dB increase in internal noise. and allow actuators to use sensor information from neighbouring frames. and in the structural and acoustic dynamics. In case 6. The control achieved a 13 dB reduction in peak sound pressure. Simulations with a simple vibro-acoustic model indicate that the desired goal of a 10 dBA reduction in peak internal noise is feasible with fewer actuators and a lower weight penalty than existing speaker-based active noise control systems. the control hardware is simplified substantially. the same actuators were grouped together so that only four separate control loops were used. A proper comparison of structural actuators versus speakers. which resulted in different modes being excited.

S. 191–217. D. . (Leuven. Reduction of aircraft cabin noise by fuselage structural optimization. R. L. B. DGLR/AIAA 14th Aeroacoustics Conference. Virginia). L and S. J. F and R. S. and the National Research Council of Canada. 17. Part I: The minimization of vibrational energy. 5. J. B 1991 Journal of Sound and Vibration 148. 59–64. 219–238. Massachusetts Institute of Technology—Space Systems Laboratory Report 5–89. W. Thanks also to Dr. I. M. F 1992 Proc. while the author was with the Institute for Aerospace Research at NRC. E 1990 Journal of Sound and Vibration 140. 4. 355–360. The assistance of deHavilland. B. S. R. Q. A 1989 Master’s thesis. Technomic. and of the rest of the staff at the Aeroacoustics Facility of NRC is greatly appreciated. 15. S 1993 Recent Advances in Active Control of Sound and Vibration. C. R. Florida: Robert Krieger. A. H. Florida). Lauderdale. (Ft. 207–230. B. S. Piezoelectric actuator models for active sound and vibration control of cylinders.118 . 91–111. M. D. H. M. (Blacksburg. B. R. 2. D 1993 Inter-Noise. C. (AIAA Paper 84-2349). Acoustics 22. A. of Intelligent Materials Systems and Structures 4. W. Aircraft 28. Propeller sound pressure field and structural information were obtained from deHavilland Inc. N. E. 382–385. P. A. E 1989 SAE Paper 891080. A. Active control of interior noise in model aircraft fuselages using piezoceramic actuators. P. C. Induced strain actuation of one. 208–215. Radiation compensation of a flexurally vibrating plate at subcritical frequencies. L 1993 Journal of Sound and Vibration 163. J. J. VA). 10. Vibration analysis of an orthogonally stiffened circular fuselage and comparison with experiment. F and H. . VPI and SU. T. S. In-flight experiments on the active control of propeller-induced cabin noise. M. P. M 1977 Noisexpo. L 1993 J. 2613–2617. L 1990 AIAA Journal 28. H. 16. 295–306. H. A. G 1994 Noise-Con 94. I. C. R. A. REFERENCES 1. Interior noise control of the Saab 340 Aircraft. P 1984 AIAA/NASA 9th Aeroacoustics Conference. Review of Recent Research on Interior Noise of Propeller Aircraft. S. C. P. R. K. S and J. H and R. D. A. (Williamsburg. C. Active control of interior noise in a business jet using piezoceramic actuators.  ACKNOWLEDGMENTS This work was conducted as part of a collaborative research and development project between deHavilland Inc. S and J. C. Belgium). N. 18. J. L. G. Selected flight test data and control system results of the CEC BRITE/EURAM ASANCA study. R. I. P. B. K and S. C.. F. C. J. Andreas von Flotow for his comments. R. B. D. H. E. S. R. P. Malabar. 19. 7. S. National Specialists Meeting on Dynamics and Aeroelasticity (Fort Worth. Texas) 77–87. J 1991 AIAA J. R. A. U. S and C. 1397–1404. R. S 1992 AIAA Journal 30. S 1983 SAE Paper 830736. Active control of the transmission of sound through a thin cylindrical shell. Mechanisms of active control in cylindrical fuselage structures. Full-scale demonstration tests of cabin noise reduction using active vibration control. A comparison of speakers and structural-based actuators for aircraft cabin noise control. A. B 1979 Formulas for Frequency and Mode Shape. Fokker’s activities in cabin noise control for propellor aircraft. S. L. 542–551. T 1976 Soviet Phys. 9. Theoretical studies of the active control of propeller-induced cabin noise. D. M and C. B 1990 Jounal of Sound and Vibration 140. R. C. 40–45. D. M. 6. J. 13. B. E. 3. 12. A wavenumber domain approach to the active control of structure-borne sound. S. N and S. J. 11.and two-dimensional structures. S. G. N and S. D. E 1993 Journal of Sound and Vibration 167. F and G. Z and M. C. J. B and C. N. J. J. Evaluation of piezoceramic actuators for control of aircraft interior noise. H and U. F and J. T. 20. 21. V and B. R. R. K and E. 14. L. General aviation aircraft meeting and exposition. T. A dynamic stiffness/boundary element method for the prediction of interior noise levels. V. B 1958 Proceedings. 8.

B. Active narrow-band vibration isolation of machinery noise from resonant substructures. K. from top of aircraft effective mass per unit surface area ring bending stiffness per frame (EI)z (EA)p dp t d31 r c axial bending stiffness per stringer extensional stiffness of piezo thickness of piezoelectric distance. L. G. piezo to frame neutral axis piezoelectric strain constant density of air speed of sound in air . MM 1994 American Control Conference (Baltimore. 17–40. 25. G. B and F. MD).  F 1993 Journal of Sound and Vibration 167. S.   119 22. J. 467–483. S and A. D. D. APPENDIX: NOTATION v0 R l df ds uf rs (EI)y disturbance frequency radius of fuselage length of aircraft cabin frame spacing stringer spacing floor angle. E 1992 Active Control of Sound. 1632–1636. W. H. 23. London: Academic Press. N and S. L. A feedback perspective on the LSM disturbance feedforward algorithm. A. MM. S 1995 Journal of Sound and Vibration 187. G. A. Aircraft fuselage noise transmission measurements using a reciprocity technique. P. 24.