You are on page 1of 9

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.

10, October 2007

285

Diagnosis of Gear Systems by Spectral


Analysis of Vibration Signals
Noureddine Haloui1, Djamel Chikouche2, Messaoud Benidir1
1

Laboratory of the Signals and Systems, College of Electricity,


University of Paris-South, 3, Street Joliot-Curie, Postal Code 91192,
Gif-Sur-Yvette, Cedex, France
2

LIS Laboratory, Electronics Department, Engineering Faculty


University of Setif, 19000, Setif, Algeria.

Summary
The diagnosis of gearbox faults based on the Fourier analysis of
the vibration signal produced from a gear reductor system has
proved its limitations in terms of spectral resolution. In this
paper, we propose a method for the fault diagnosis of a gear
reductor made of two toothed wheels operating at constant
conditions. This technique is based on Prony method. The good
performances of the proposed method relatively to the
synchronous cepstral technique are demonstrated on the basis
of experimental data of the vibration signal delivered from the
gear reductor system under study. We show that the fault
diagnosis of the considered system can be performed by
observing the evolution of the power spectrum of the vibration
signal during the observation days of the reductor.

Key words:
Diagnosis, Power Spectrum, Prony, Gears, Vibration Signal.

1. Introduction
The gear reductors are present in all mechanical
machines. We find them in most industrial sectors such
as the speedbox in automobile industries. Researchers are
still very interested in the study of gear reductors because
of their relative weakness [1 - 7].
The analysis of vibration signals was often based on the
fast Fourier transform [3, 8, 9]. This approach suffers
from some limitations. Among these limitations, the FFT
is not

efficient to describe the non stationarities introduced by


faults in the vibration signal. The second limitation and
the most important
one is the frequency resolution, which is the ability to
distinguish the spectral responses to two or many
harmonics. Another limitation is due to the windowing of
data which appears during the FFT processing [9]. In
order to overcome these performance limitations inherent
to the FFT approach, many modern spectral estimation
techniques have been proposed during the last two
decades [10 - 23].
In this paper, we propose and analyse a spectral approach,
based on the Prony method, to the diagnosis of faults in
gear reductors. A comparison between the proposed
method and the synchonous cepstrum technique will be
performed by using the vibration signal delivered by gear
reductors.

2. Description of the Gear Reductor System


Under Study
The vibration signals of the gear reductor under study
have been provided from CETIM(1). They are delivered
from a reductor operating 24 hours over 24 hours. The
dimensions of gear wheels together with the operating
conditions (speed, couple) are adjusted so that we obtain
a spalling on all the width of a tooth [2, 3, 7]. During
experimentation, the system has been stopped every day
to observe the state of the wheel teeth. An expertise
report has thus been written (Table 1).

CETIM: Centre dEtudes Techniques des Industries Mcaniques, 52 av. Felix Louat,
60300 Senlis.

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

Measuring
point

Motor

Looping reductor
Ratio 40/42

Gear

Height of the tooth

286

Reductor under
test ratio
20/21

Fig. 1 Measuring system

The system characteristics are:


*The number of teeth is respectively 20 teeth on the first
wheel and 21 teeth on the second wheel.
*The rotation speed of the input tree of the reductor is
Rpm= 1000 turns/min, thus a rotation frequency of
fr=16.67 Hz and 15.87 Hz on the output tree.
*The meshing frequency is fe=333.33 Hz.
A measure has been recorded every day during a period
of 12 days. The machine has turned permanently during
the 12 days. Each measure is composed of approximately
60000 samples, thats about 50 rotation periods. The
sampling frequency is 20 kHz. During experimentation,
the testing gear goes from the well operating state to the
deteriorated gear state.

Fig. 2 Tooth 2 in the


10th day.

Fig. 3 Tooth 16 in
the 11th day.

3. The Meshing Signal


The meshing signal is principally caused by shocks
between teeth of wheels that compose the reductor (fig.
4). The importance of the shock depends on the teeth
shape during the well operation and the fault nature
during degraded operation [2, 3, 24].
The meshing signal Se(t) is amplitude and frequency
modulated by the signals Sr1(t) and Sr2(t) emitted from
the pinion and the wheel whose frequencies are
respectively fr1 and fr2. These modulations, being
multiplicative and convolved with meshing harmonics,
are defined by [3, 25]:

Table 1. Expertise Report [3, 24, 25]


Days

Observations

First day of acquisition, no irregularity

No irregularity

//

//

//

//

//

//

Tooth spalling

No evolution

Tooth 1/2 no evolution, tooth 15/16

a : Meshing beginning.

b: Beginning of the contact. unique

spalling beginning
9

Spalling evolution of tooth 15/16

10

//

//

11

//

//

12

Spalling on all the width of tooth


15/16

c: End of the contact.

d: End of the meshing.


Fig. 4 Evolution of the teeth meshing.

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

287

se (t) = ( xc (t nTe ))(1+ sr1(t vTr1) + se (t iTr2 ))) (1)

4.2 The cepstre technique

with:

The meshing signal being frequency and amplitude


modulated by signals s r1 (t ) and s r 2 (t ) emitted from
the pinion and the wheel, the cepstre constitutes one of
the means to detect the periodicities in these signals [2, 3,
25].
The cepstre of a signal x(t) is defined as:
2
(4)
X e (t ) = TF 1 [ln X e ( f ) ]

n=

v=

i=

xc (t ) the signal produced by the shock between two


gear teeth.
Te = 1 / fe With: T the meshing period and f the
e
e
meshing frequency.
Tr 1 = 1 / f r 1 = N 1 .Te With:

Tr1 the rotation period of

wheel 1 (pinion of 20 teeth), f r1 : the rotation frequency


of wheel 1 and N 1 : the number of teeth of wheel 1.

Tr 2 = 1 / f 2 = N 2 .Te With: Tr 2 : the rotation period of


wheel 2, f r 2 : the rotation frequency of wheel 2 and
N 2 : the teeth number of wheel 2.
The Fourier transform (TF) of the signal se (t ) is:
+

Se( f ) = Xc ( f ).( f nfe) + Xc ( f )( f nfe) *Sr1( f ).


n=

( f f

r1

i =

) + Xc ( f ) ( f nfe ) * Sr 2 ( f ). ( f ifr 2 )

(2)

Where TF stands for the inverse Fourier transform


and X e ( f ) the Fourier transform of xe(t).

4.3 Prony method


The model proposed in the Prony method is a sum of
exponentials whose amplitude, phase, frequency and
damping factor are arbitrary [9, 26- 29].
The Prony model to approximate a sequence x0, x1, x2,
x3,..,xn-1, is given by:

x en =

Where: is the Dirac impulse.


S e ( f ) = TF [ s e ( t )]

i =1

b i Z in for n=0,1,.,N-1

The coefficients bi and Zi are generally complex:


(6)
bi = Ai exp( ji )

X c ( f ) = TF[xc (t)]
S r 1 ( f ) = TF [ s r 1 ( t )]

Zi = [exp(i + j 2fi)t ]

damping factor,

of

(7)

with Ai the amplitude, i the phase in radiants,

S r 2 ( f ) = TF[s r 2 (t )]

4. Presentation
Techniques

(5)

Vibratory

Analysis

4.1 First order cyclostationarity: synchronous mean


The mean of a random function of time x(t) is defined as
the expected value of this function [5, 8, 13]:

i the

fi the oscillation frequency in Hertz,

and t the sampling period in seconds.


To estimate the polynomial coefficients and the model
order, we use one of the estimation algorithms for the
AR model parameters. In our work, we have used the
RLS (recursive least square) algorithm.
To determine the frequencies and the damping
coefficients, we use the following equations:

x e (t ) = E[ xe (t )]

Generally, this mean is a function of time. It is constant


for a stationary signal; it is periodic for a cyclostationary
signal [8]. The first order cyclostationarity is one means
for the diagnosis of gear reductor faults. For the kind of
reductor under study including a pinion of period T1 and
a wheel of period T2 [2, 8], it is shown that the signal is
cyclostationary at the periods T1 and T2 [3, 24]. It is
possible to estimate, in our application, the synchronous
mean relatively to one of these periods as follows:
p ( ) = 1 / N

N 1

(3)

s ( + k . motif )

k =0

The signal is divided into N blocks of

motif

samples.

F (Z ) = (Z Z k ) =
k =1

i = ln Z i / t

i=0

aiZ

P i

,a0=1

(8)
(9)

(10)
fi =tg 1[Im( Z i ) / Re( Z i )]/ 2 t
The determination of amplitudes and phases is based on
equations (9) et (10).
The approximation function becomes then:

X e (n) = Ai exp( i it ) exp[ j (2f i it + i )] (11)


i =1

Finally, we obtain the power spectral density of Prony as


follows:

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

X e ( f )=

Day (08)

0 .3

S e Pr ony ( f ) = X e ( f )

(12)

Ai exp( j i ) 2 + [2 ( mf f
i =1

) ]2

The temporal and spectral processing of data, using the


investigated techniques, will be presented in the
following section.

Amplitude relative

288

0 .2

0 .1

- 0 .1

- 0 .2

- 0 .3

- 0 .4
0

of

Data

and

Day (06)

Amplitude relative

0 .4
0 .3
0 .2
0 .1
0
- 0 .1
- 0 .2
- 0 .3
- 0 .4
- 0 .5
0 .0 5

0 .1

0 .1 5

0 .1 5

0 .2

0 .2 5

0 .2

0 .2 5

Day (12)

The synchronous cepstre method and the Prony based


technique proposed in this work have been applied to the
vibration signal emitted by the gear system under study.
The order of Prony model is estimated with the optimal
criterion of Akaike (AIC) [9, 27].
The spectral analysis methods proposed in this paper
have been simulated using an amplitude modulated
signal composed of a white gaussian noise and three
sinusoids of respective normalised frequencies 0.1, 0.2
and 0.21 and amplitudes 0.1, 1, and 1 (already available
mechanical model [9]).
The technique based on Prony algorithm and
synchronous cepstre method are then applied to the
recorded vibration signal emitted by the gear system
under study.

0 .1

Time (s)

Amplitude relative

5. Temporal Processing
Interpretation

0 .0 5

0 .8
0 .6
0 .4
0 .2
0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
-1
0

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5

Time (s)
Fig. 5 Vibrations recorded during days 6, 7,8 and 12.
Displaying over 4 periods of rotation relative to the
pinion.

We notice, in figure 5, that the temporal representation of


signals observed each day presents oscillations caused by
teeth meshing and a modulation of long duration
corresponding to the period of the two wheels (pinion of
20 teeth and wheel of 21 teeth). The vibration signal
keeps this shape until the 12th day during
which the fault is supposed to appear. We observe a very
high increase of signal amplitude around modulations
relative to oscillations between these last ones. These
observations allow the diagnosis of a fault in the 12th and
13th days.

Time (s)

6. Synchonous Mean and Spectrum

Day (07)

Amplitude relative

0 .4

the meshing signal is modulated by the signal of the


pinion and the wheel. In order to separate the different
components of the recorded signal, we compute the
synchronous mean which is one of the means of
separation and elimination of noise components that
affect the vibration signal.

0 .3
0 .2
0 .1
0
- 0 .1
- 0 .2
- 0 .3
- 0 .4
0

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5

Time (s)

7. Synchronous Cepstre Technique


This technique is applied to the signal recorded each day
of the experimentation on 8 blocks of 6 periods relative
to the pinion (figure 6). Figure 7 presents the evolution
of peaks corresponding to the pinion (with 20 teeth) and

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

Day 12

Amplitude relative

the wheel (with 21 teeth) during the 12 days of


experimentation.
We observe, according to figure 5, from the 2nd to the 7th
day the presence of a relatively weak change in
amplitude of peaks of the wheel with 21 teeth and of the
pinion with 20 teeth. This change is due to many
phenomena such as the level and quality of the lubricant.
At the 8th day, we observe the appearance of a fault on
the pinion in an obvious manner by an increase of the
peak characterising it. These observations can be
valuable for the diagnosis of the fault and its localisation
through the synchronous cepstre.

289

Day 06

Time (s)

Amplitude relative

Fig. 6 The cepstre of the signal recorded during days

6,7,8 and 12.


A: peak of wheel with 21 teeth.
B: peak of wheel with 20 teeth.

0.35

Peak amplitudes

Time (s)

Amplitude relative

Day 07

0 .2

Wheel 20 teeth
Wheel 21 teeth

Detection of
pinion fault

0.3

0.25

0.2

0.15

0.1

0.05

0.15

10

11

12

13

Days

0 .1

0.05

Fig. 7 Peak evolution during 12 days


(Synchronous cepstre method)

-0.0 5
0

0.05

0.1

0.15

Time (s)

0.2

0 .25

8. Power Spectrum and Interpretation

Amplitude relative

Day 08

0 .2

0.15

0 .1

0.05

-0.0 5
0

0.05

0.1

Time (s)

0.15

0.2

0 .25

The power spectral density is estimated each day of the


experimentation by the proposed Prony based technique.
The vibration signal recorded each day is filtered
between the third and the fifth meshing harmonics,
where there is a higher energy with a bandwidth equals
to two times the meshing frequency. The filtered
passeband signal is modulated, and then it is lowpass
filtered to eliminate the meshing peak in the purpose of
observing the information contained in the lateral bands.
Figure 8 shows the spectrum of the vibration signal
recorded at the 4th day over 20 rotation periods of the
pinion before and after the demodulation and the lowpass
filtering.

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

290

In the next part, we will present the experimental results,


in the frequency domain, that are obtained from the
Prony based method after the synchronous mean, the
demodulation and the lowpass filtering.

(Day 07)

170

D S P (dB)

160

Spectrum after demodulation and filtering

Amplitude

150

140

130

120

110
20

40

60

80

100

120

140

160

1 80

200

120

1 40

160

180

200

120

1 40

Frequency (Hz)
2

Spectrum before demodulation

1 00 0

2000

3 00 0

4000

5 00 0

6000

7 00 0

8000

(Day 08)

9 00 0

170

Frequency (Hz)
Fig. 8. Spectrum before and after demodulation

160

D S P (dB)

and lowpass filtering

9. Prony Based Technique

150

140

130

120

110

20

40

60

80

1 00

Frequency (Hz)
(Day 12)

190

180

170

D S P (dB)

The experimental results of the power spectral density


for the vibration signal, estimated by Prony based
technique, are represented on figure 9. The examination
of these power spectra reveals that the spectrum keeps
almost the same shape until the 6th day during which the
fault manifests itself through the appearance of an
amplitude increase of the peaks corresponding to
multiples of the meshing frequency for the 20 teeth
pinion (figure 10). We notice the non complete
disappearance of peaks that characterise the 21 teeth
wheel.

160

150

140

130

120

(Day 06)

110

100

20

40

60

80

100

160

1 80

200

Frequency (Hz)

1 70

D S P (dB)

1 60

Fig. 9 Spectrum of the recorded signal estimated from


Prony based
technique at days 6, 7, 8 and 12.
A: peak of 21 teeth wheel
B: peak of 20 teeth wheel

1 50

1 40

1 30

1 20

1 10

1 00
20

40

60

80

1 00

12 0

1 40

1 60

1 80

Frequency (Hz)

This spectral analysis technique detects the fault, its


localisation and its evolution. We can conclude that
Prony based technique is an efficient means for the
diagnosis of gear reductor faults.

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

0.35

Wheel 20 teeth
Wheel 21 teeth

Detection of a
pinion fault

Peak Amplitudes

0.3

0.25

0.2

0.15

0.1

0.05
2

10

11

12

13

Days

Fig. 10 Peak evolution during 12 days


(Prony based technique)

10. Comparative Study of the Proposed


Method Relative to the Expertise Report [3]
Table 2 presents the day of fault detection using the
investigated techniques according to the evolution of
peaks corresponding to the rotation frequencies of the 21
teeth wheel and the 20 teeth pinion.

Table2. Day of fault detection.


Days

Synchronous

Prony based

Cepstre

technique

291

vibration signal emitted from the gear system under


study have revealed that the cepstral method shows
clearly the presence of two modulations, whereas in the
temporal signal, it seems that there exists only one
modulation. Moreover, this method permits to precisely
measure the peaks corresponding to the 21 teeth wheel
and the 20 teeth pinion. It detects well the fault at the 8th
day according to the change in the vibrations nature,
whereas the temporal signal does not present yet
characteristic particularities on the appearance of a fault
until the 12th day only.
Furthermore, the Prony based technique detects the fault
appearance at the 6th day in the gear reductor system
through a simple visual interpretation and consequently a
simple observation of the presence or the absence of gear
faults.
In summary, this study demonstrates, relatively to the
expertise report, that the proposed Prony based method is
more performant than the cepstre synchronous method. It
can be interesting in the supervision of complex gear
reductors. In this case, we must follow carefully the
evolution of particular peaks, both in amplitude and
frequency, in the purpose of detecting faults.

11. Conclusion
In this paper, a gear box diagnosis technique based on
Prony model has been proposed. The performances of
this technique in the gear system diagnosis have been
compared to those of the synchronous cepstre method
and the expertise report. The application of this
technique to the vibration signal emitted by the gear
reductor system permits to conclude that it can play an
important role in the study of gear vibrations. In fact, the
use state of a reductor is strongly related to modulation
phenomena that present the vibrations relative to the
meshing signal. We have shown that vibration analysis
through the Prony based technique permits to detect the
fault presence and determine the deteriorated wheel at
the 6th day, whereas the synchronous cepstre detects the
fault at the 8th day. Consequently, this technique is very
efficient for the diagnosis of faults in gears reductors.

No fault

No fault

//

//

//

//

//

//

//

Fault detection

//

No evolution

Fault detection

Fault evolution

Fault evolution

//

10

//

//

References

11

//

//

12

//

//

13

//

//

[1] Isa Yesilyurt Fault detection and location in gears


by the smoothed instantaneous power spectrum
distribution NDT&E International 36 (2003) 535
542
[2] M. El Badaoui, Cahouet V., Guillet F., Danire J.,
Velex P., "Modelling and detection of localized tooth
defects in geared systems", ASME Journal of
mechanical design, Vol. 123, sept. 2001, p 422-430.

The proposed Prony based technique and the


synchronous cepstre method that we have applied to the

292

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007

[3] Noureddine Haloui, Djamel Chikouche, Messaoud


Benidir, and Rais Elhadi Bekka , Diagnosis of Gear
Systems by Specral Analysis of Vibration Signals
Using Synchronous Cepstre Technique , GESTS
Internationl Transactions on Communication and
Signal Processing, August 2006, Vol 8, N 1, p 27
36.
[4] Mcfadden P. D, Detecting fatigue cracks en gears
by amplitude and phase demodulation of the meshing
vibration , Journal of vibration, Acoustics, stress,
and Reliability in Design, avril 1986, Vol. 108, p.
165 170.
[5] Antoniadis I. Glossiotis G, Cyclostationnary
analysis of rolling element bearing vibration signals
Journal Sound of Vibration, Dc 2001, Vol. 248, N
5, p 829 845.
[6] Shiroishi J., L I Y., Liang S., Kurfess T., Danyluk S.
Bearing condition diagnostics via vibration and
acoustic emission measurements Mechanical
Systems and Signal Processing, Sept 1997, Vol. 11,
N 5, p 693 705.
[7] W. J. Wang, P. D. Mcfadden, Application of
wavelets to gearbox vibration signals for fault
detection, Journal of sound and vibration (1996),
192(5), pp. 927- 939.
[8] H. Akaike, A new look at the statistical model
identification, IEEE. Transactions on automatic
control, vol. AC-19, No . 6, Dec. 1974.
[9] S. M. Kay, Modern spectral estimation, Printice
hall signal processing series, Englewood cliffs, New
Jersey 07632, 1988.
[10] J.A. Cadzow, Spectral estimation: An Overdetermined Rational Model Equation
Approach,
Proc. IEEE, vol.70, N 9, Sep. 1982, pp. 907-937.
[11] R. H. Jones, Identification and autoregressive
spectrum estimation,
IEEE. Transaction on
automatic contrl, vol. AC 131, No . 13, Dec. 1974.
[12] R. E. Bekka, D. Chikouche, Pouvoir de dtection
et de rsolution de la mthode AR : Application
aux signaux courts, Revue
Sciences
&c
Technologie, Univ. Constantine, N12, Dec. 1999,
pp. 49- 53.
[13] S . Kay,
S.
L.
Marpele, Spectrum
Analysis: A modern perspective , Proc. IEEE,
vol. 69, N 11, Nov 1981, pp.1380-1419.
[14] B. Samanta Gear fault detection using artificial
neural networks and support vector machines with
genetic algorithms Mechanical Systems and Signal
Processing 18 (2004) 625644.
[15] K. R. Al- Balushi, B. Samanta, Gear fault diagnosis
using energy- based features of acoustic emission
siganals, Proceedings of institution of Mechanical
Engineers, Part I: Journal of Systems and control
Engineeing 216 (2002) 249- 263.

[16] L. B. Jack, A.K. Nandi, Fault detection using


support vector machines and artificial neural network
augmented by genetic algorithms, Mechanical
Systems and Signal Processing 16 (2002) 373- 390.
[17] Tim Toutountzakis, Chee Keong Tan, David Mba
Application of acoustic emission to seeded gear
fault detectionT. Toutountzakis et al. / NDT&E
International 38 (2005) 2736
[18] S.J. Loutridis Instantaneous energy density as a
feature for gear fault detection, Mechanical Systems
and Signal Processing 20 (2006) 12391253.
[19] D. Chen, W. J. Wang Classification of Wavelet
Map Patternsusing Multi-Layerneural Networks For
Gear Fault Detection Mechanical Systems and
Signal Processing (2002) 16(4), 695704.
[20] B. Samanta Artificial neural networks and genetic
algorithms for gear fault detection Mechanical
Systems and Signal Processing 18 (2004) 12731282.
[21] B. Samanta, K. R. Al- Balushi, Artificial neural
network based fault diagnosis of rolling element
bearings using time- domain features, Mechanical
Systems and Signal Processing, 2002.
[22] P. D. McFadden, Detection of gear faults by
decomposition of matched differences of vibration
signals, Mechanical Systems and Signal Processing
14 (2002) 805- 817.
[23] J. Antoni, R. B. Randall, Differential diagnosis of
gear and bearing faults , Transactions of the ASME,
Journal of Vibration and Acoustics 124 (2002) 165171.
[24] J. Antoni, J. Danire, F. Guillet., Effective
Vibration
Analysis
of IC
Engines
using
Cyclostationarity Part 2 New Results on the
reconstruction of the cylinder Pressures., Journal of
Sound and Vibration, volume 257 issue 5. 7
novembre 2002, page 839-856.
[25] ElBadaoui M, Guillet F, Danire J. Contribution
de cepstre dnergie au diagnostic des de rducteur
complexe
engrenages .
3me Confrence
Internationale Mthodes de Surveillances et
Technique de Diagnostique Acoustique et Vibration,
Senlis, France, 13-15 Oct. 1998.
[26]
Haloui, N.
Chikouche, D.
Benidir,
M. Application of spectral parameters analysis
methods for the detection of gear faults in rotating
machines IEEE CCECE02 Proceedings; vol. 1,
pp. 190-194, 13-15 mai 2002.
[27] H. Akaike, A new look at the statistical model
identification, IEEE. Transactions on automatic
control, vol. AC-19, No . 6, Dec. 1974.
[28] A. Apostoliouk, Amlioration de la mthode
de Prony dans le problme
didentification
des systmes dynamiques, Proc, Frrst. Int. Conf
on Electronics end Automatic Control. Mai. 1994.

IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.10, October 2007
[29] W. M. Steedly, C. J. Ying, A Modified TLS-Prony
Method Using Data Decimation, IEEE, Trans. Sig.
Proc., vol. 42, N 9, Sept. 1994, pp. 2292- 2303.

Biography
Name: Chikouche Djamel
Address: Electronics Department, Engineering, Faculty,
University of Setif, Postal Code 19000, Setif, Algeria.
Education & Work experience: Professor at the
Electronics
Department,
Engineering Faculty,
University of Setif, Algeria. I got my MS
degree in Electrical Engineering from Ohio State
University, USA in 1984 and my PhD in Signal
Processing from the University of Setif, Algeria
in 2000. My research interests are: Signal
processing,
spectral analysis, fast algorithms,
diagnosis of gearbox faults, and parallel processors.

Name: Benidir Massaoud


Address: Laboratory of the Signals and Systems, College
of Electricity, University of Paris- South, 3, Street JoliotCurie, Postal Code 91192, Gif- Sur-Yvette, Cedex,
France.
Education & Work experience: Professor of signal
processing and Dean of the
Doctoral education,
University of Paris-South, France.
Name: Haloui Noureddine
Address: 2, Avenue of Savigny
Postal Code 93600, Aulnay Sous Bois, France.
Education & Work experience: I got my MS degree in
Signal Processing from the University of Setif, Algeria
in 2002. My research interests are: Signal processing,
spectral analysis and diagnosis of gearbox faults.

293

You might also like