NAG/DAGA 2009 - Rotterdam

Experimental transfer path analysis on helicopters

E. Mucchi1, A. Vecchio2
1
Engineering Department, University of Ferrara, Ferrara, Italy, Email: [email protected]
2
LMS International, Leuven, Belgium, Email: [email protected]

terms of operational forces, path contributions and acoustic

Introduction pressure will be compared with those obtained with the
In complicated structures involving many sub-assemblies measured TFs. Eventually, some concluding remarks are
(such as helicopters) the vibro-acoustic sensations that are given.
experienced by an observer at any location (e.g. on the
passenger cabin) may easily have been caused by vibration The vibro-acoustic TPA on the AW-109
sources some way off. For example the energy from a source helicopter
in a helicopter is transmitted into the passenger cabin by a
The vibro-acoustic TPA is an analysis fully described by
number of different routes: from the gearbox connection
means of two sets of equations (1-2):
points, rotor bearings, engine mounts or even via the tail
1
rotors or main blade shafts. Airborne contributions from the ª X1 X1 X1 º
main and tail blades, for instance, may be important as well. « ! » ­  ½
­ F1 ½ « F1 F2 Fn » X1
Some of these paths are important, some are negligible. ° °
®# ¾ « # » ®° # ¾° [N] (1)
Some transfer paths may cause interference at certain °F ° « » °  °
frequency such that the observer does not notice anything ¯ n¿ « Xm » ¯Xm ¿
«F »
significant- until he moves position. ¬ 1 ¼

Transfer Path Analysis (TPA) is used to assess the structure

ª p1 p1 p1 º
and air-borne energy paths among excitation source(s) and «F "
receiver location(s) [1]. ­ p1 ½
« 1 F2 Fn »» ­ F1 ½
° ° « # » °®# °¾
®# ¾ [Pa] (2)
Hereafter a vibro-acoustic TPA is performed on a Agusta °p ° « »° °
Westland AW-109 helicopter. Usually vibro-acoustic TPA ¯ q¿ « pq » ¯ Fn ¿
involves measurements of accelerations in operational «¬ F1 »¼
conditions and of vibration and/or vibro-acoustic transfer
^ X ˜˜˜ X ` , ^F ˜˜˜ F ` ^ p ˜˜˜ p ` are the
t t t
functions (TFs). The transfer functions can be measured where and
1 m 1 n 1 q
using the most practical approaches – either using hammer
or shaker or volume acceleration source excitation acceleration, the operating force and acoustic response
techniques. Due to the complex testing conditions that vectors, respectively.
always happen in transfer function measurements of Equation (1) brings into play the relation between the
cumbersome systems [5], it is frequent to find corrupted data operating forces transmitted along the paths and the
in a few sets of measurements. structural accelerations caused by these forces, while the
In this scenario an attempt to perform the TPA using second set of equations relates the acoustic responses, e.g.
synthesized transfer functions, computed after a modal the noise inside a cabin, and the operating forces. Hence, it is
analysis, has been brought to completion with the aim of pretty clear that by exploiting the information that the first
verifying at first the reliability of such an approach and then set of equations carries it is possible to compute the acoustic
to assess whether this approach could be employed to responses from the second one. Indeed, it is much easier to
substitute corrupted data with synthesized ones when the measure the accelerations of a structure rather than the
database occurred to be unreliable. It will be proved that forces; these accelerations can then be employed in order to
using synthesized transfer functions yield significantly compute the operational forces which substituted in the
similar results to those obtained with the not-corrupted second set of equations will lead to the final result. On the
measured data. That being so, it could be thought that such left hand side of equation (1), there are the operating forces.
an approach could solve poor quality transfer function issues As it can be seen, it is necessary to invert the matrix linking
that otherwise would cause to be compelled to repeat the the accelerations and the forces. This constitutes the biggest
measurements, which is costly and time consuming. computational effort of the TPA. On the other hand, the
matrix in equation (2) will be much smaller, and since it has
The structure of this paper is as follows. In the first Section a not to be inverted, the computational cost of this second step
brief overview of the vibro-acoustic TPA performed on the is negligible compared to the first one. In the helicopter
AW-109 helicopter is presented giving a few details on the under study the vibro-acoustic TPA is performed
corrupted transfer functions acquired during tests. In the considering the gearbox as the source and the cabin cavity as
second Section the synthesized transfer functions that will the receiver location.
substitute the measured transfer functions in the TPA model
will be calculated by means of a modal analysis. The third The gearbox is connected to the cabin roof by means of two
Section is devoted to the verification of the proposed front struts, two rear struts and the anti-torque plate through
method: the results obtained with the synthesized TFs in four bolts leading to have eight structural paths. 24

788
 NAG/DAGA 2009 - Rotterdam

accelerometers, one along each of the three orthogonal

0.96
directions, have been placed on these eight paths. The
transfer function matrix (equation (1)) has been obtained by
exciting the structure with an impact hammer and measuring
0.70
the acceleration responses. In the present work, this matrix

Amplitude
has dimension of 48×24 and each TF is composed of 4096

/
spectral lines. Indeed the analysis has been pushed up to
4096 Hz (thus frequency resolution equal to 1 Hz) despite Coherence for TF 1
being hardly reliable at such a high frequency. But being the Coherence for TF 2
most important resonance peak situated far before, the
analysis retains its interest. In order to have an over-
determined matrix (hence 48×24) for matrix inversion 0.00

purpose [3] another point close to each of the path points has 0.00 Hz 4100.00
been placed. In these extra positions, 24 accelerometers were Figure 1: Coherence functions concerning two corrupted
placed thus building up the necessary database. The fact that transfer functions (TF 1 and TF 2) in the frequency range
the matrix is rectangular means that theoretically it is not till 4 kHz.
possible to invert it. Nevertheless it is still possible to apply
-10.00
the singular value decomposition (SVD) and then to pseudo
invert the diagonal matrix that this mathematical tool yields.
Finally the acceleration vector of equation (1)

^  ˜˜˜ X ` ) has been calculated in flight operational

t
(i.e. X 1 m (g/N)
dB
conditions in the 48 acceleration responses, so the
operational forces can be calculated.
Concerning equation (2), the vibro-acoustic transfer
TF 1
functions have been obtained taking advantage of the vibro- TF F2 FRF BODY:20:+Z/BODY:31:+X

 X j / qi , where qi is

F FRF BODY:20:+Z/BODY:31:+Y
acoustic reciprocity (i.e. pi / Fj -90.00
0.00 Hz 2000.00
the volume velocity at location j). Therefore the TFs have
been calculated exciting by means of two volume
Figure 2: Corrupted transfer functions (TF 1 and TF 2) due to
acceleration sources (LMS Q-sources) working at different
ground loops during measurements in the frequency range till
frequencies, 0 - 400 Hz and 400 - 4000 Hz, and measuring
2kHz.
the responses using the 24 accelerometers located in the
eight structural paths. The sources were placed in the
helicopter cabin, close to the pilot seat.
Synthesis of the transfer functions
The entire structural transfer function matrix of equation (1)
In the structural transfer functions matrix two full columns has been reconstructed by means of the modal parameter
of data concerning two structural paths were completely extraction algorithm PolyMAX [3]. The modal analysis of
unreliable. The coherence functions concerning these two the cabin roof has been performed exclusively with the aim
transfer functions are particularly poor (Figure 1). In fact it of obtaining the synthesized transfer functions. PolyMAX
has been discovered that in the transfer functions sharp works properly when the modal synthesis is applied
peaks located between 200 and 700 Hz due to some considering the full set of responses but only one reference
electrical ground loops are present (Figure 2). Their at a time, as if the system were single input. Then, in order to
magnitude could spoil the further analysis because it could have a good set of transfer functions the modal synthesis has
make appear unjustified peaks. been repeated for each of the eight paths and for all the
directions, thus building the full matrix constituted by 24
references and 48 responses. More details about this modal
analysis can be found in [2].
In Figure 3, an example of the application of the PolyMAX
algorithm is shown. As it can be seen, the number of
selected poles in the frequency range of analysis (0-4kHz) is
limited to 26. Notice no poles have been selected in the
range 0-100 Hz. In fact PolyMAX begins to find stable poles
only at high model rank and furthermore no stable poles
occur in the range 0-100Hz.

789
 NAG/DAGA 2009 - Rotterdam

of the main paths, it can be stated that the models are

providing the same result. Both in fact highlight the present
of the resonances at the frequencies of 1793 and 1825 Hz.
Finally, the comparison between the acoustical responses
calculated in the cabin cavity by using the measured TFs and
the synthesized ones is shown. Looking at the RMS values
(Table 1) the dB level is the same for both the TFs,
furthermore the maximum peak level in the spectrum has
similar value for both the models and occurs at the same
frequency.

0.10

Figure 3: Stabilization diagram used for the modal analysis and

(g/N)
Log
for the further synthesis of transfer functions in the frequency
range till 4 kHz. The PolyMAX algorithm has been used to
estimate the system poles. The red curve is the FRF-sum, i.e. the
complex sum of the TFs of all the measured structural points.
Measured
F TF
Synthesized FRF BODY:31:+Z/BODY:41:+X
F
SynthesizedFRF
TF BODY:31:+Z/BODY:41:+X
Method assessment 100e-6

In order to verify the proposed method, the two corrupted 1000.00 Hz 2500.00

columns of transfer functions has been measured again

obtaining good quality data (called hereafter measured TFs). Figure 4: Synthesized (blue) and measured (black) TF in the
Four different comparisons will be presented in order to frequency range 1000-2500Hz.
assess the method.
40.00
First the measured TFs are compared with the synthesized
one. In particular, Figure 4 depicts a measured TF and the
relative synthesized one in the frequency range 1000-
2500Hz. As it can be seen the general trend is quite
respected.
dB
N

Secondly, the operating forces which are calculated once the

matrix inversion is computed are compared in Figure 5 and
Figure 6. In particular the spectra in Figure 6 are referred to
Measured TF
the range 800 - 2200 Hz in which most of the resonances Synthesized TF
takes place. The match between the two curves is quite
-60.00 -
impressive and gives a strong confirmation about the quality
0.00 Hz 800.00
of the analysis that can be obtained using synthesized
transfer functions. The fact that this bandwidth is the one
which contains the most of the resonance peaks means also Figure 5: Operational forces calculated by the TPA model by
that this is where PolyMAX was able to find the largest using the synthesized and measured TFs in the frequency range
number of stable poles. Hence it was expected to find a till 800Hz.
certain match between the two TPA models (i.e. the one
obtained by using synthesized transfer functions and the one
by measured TFs). It is at low frequency (Figure 5) where
the models differ the most: this is due to the fact, already
underlined, that at low frequency only a limited number of
poles were selectable, therefore the description of the system
is not complete and the synthesizes TFs can not properly
represent the real behavior.
The third comparison is made considering the path
contributions in terms of acoustical pressure at the receiver
locations (cabin cavity), see Figure 7. The absolute values
have been neglected for confidential reasons, nevertheless
the color-scale is the same for both the maps. Considering
both the amplitude of the path contributions and the location

790
 NAG/DAGA 2009 - Rotterdam

30.00 Acoustic response RMS (in 1500 – Maximum peak [dB @

Measured TF curve 2200 Hz) Hz]
Synthesized TF
Calculated with 111.36 dB 104.45 dB @ 1824 Hz
measured TF

Calculated with 111.81 dB 105.32 dB @ 1824 Hz

dB
N

synthesized TF

Table 1: Acoustic response comparisons in terms of RMS

values (second column) and maximum peak amplitude
1825.00
-60.00 (third column).
800.00 Hz 2200.00

Figure 6: Operational forces calculated by the TPA model by Concluding remarks

using the synthesized and measured TFs in the frequency range
A vibro-acoustic TPA performed on the AW-109 helicopter
800-2200Hz.
is presented. Since corrupted structural TFs were present in
the database an attempt to perform the TPA using
synthesized transfer functions, computed after a modal
analysis, has been done. The proposed method has been
verifying comparing the results of the TPA using both the
measured TFs and the synthesized ones. The comparisons
has been made in terms of TFs, operating forces, path
contribution and acoustic response at the receiver locations.
It has been proved that using synthesized transfer functions
yields significantly similar results to those obtained with the
not-corrupted measured data. That being so, it could be
thought that such an approach could solve poor quality
transfer function issues that otherwise would cause to be
compelled to repeat the measurements, which is costly and
time consuming.
(a)
References
[1] J.W. Verheij, Multi-path sound transfer from resiliently
mounted shipboard machinery, PhD thesis, 1982.
[2] L. Testa, A. Vecchio, L. Bregant, Transfer path analysis
of the Agusta Westland AW-109 performed by means of
in-flight data and PolyMAX synthesized FRFs,
Proceedings of ISMA2008, Leuven, Belgium, 2008,
September 15-17.
[3] P. Mas, P. Sas, K. Wyckaert, Indirect force
identification based upon impendance matrix inversion:
a study on statistical and deterministical accuracy,
Proceedings of ISMA19, Leuven, Belgium, 1994.
[4] B. Peeters, H. Van der Auweraer, P. Guillaume, J.
(b) Leuridan, The PolyMAX frequency-domain method: a
new standard for modal parametrs estimation?, Shock
Figure 7: Path contribution comparisons in the frequency range and Vibration 11 (2004) 395-410.
1700-1900Hz, (a) obtained with the measured TFs and (b) with
the synthesized ones.
[5] E. Pierro, E. Mucchi, L. Soria, A. Vecchio, On the
vibro-acoustic operational modal analysis of a
helicopter cabin, Mechanical Systems and Signal
Processing 23 (2009) 1205-1217.

791