8th International Conference on Modelling, Identification and Control (ICMIC-2016)

Algiers, Algeria- November 15-17, 2016

A Summary of Vibration Analysis Techniques for

Fault Detection and Diagnosis in Bearing
Adel Boudiaf1, Abederrazek Djebala2, Hocine Bendjma1, Adel Balaska1, Amine Dahane3
1
Research Center in Industrial Technologies CRTI P.O.Box 64, cheraga 16014 Algiers, Algeria
2
Laboratory of Electrical Engineering of Guelma (LGEG), University of Guelma BP401, Guelma, 24000 Algeria
3
Intelligent Systems Research Laboratory, University of Sciences and Technology, Oran, Algeria
Emails: a.boudiaf@csc.dz, adjebala_abderrazek@yahoo.fr, h.bendjama@csc.dz, amineusto.laresi@gmail.com

Abstract Bearing is one of the most critical components of Mode Decomposition (EMD).The purpose of this study is to
rotating machinery .They are employed to support and rotate the review these techniques and explore their capabilities,
shafts in rotating machinery. Therefore; any fault in the bearings advantages and disadvantage in monitoring rolling element
can lead to losses on the level of production and equipments as bearings.
well as the creation of an unsafe working environment for The paper is organized as follows: Section II presents
humans. For that reason, the bearing fault diagnosis has received system and bearing faults descriptions. More details on the
considerable attention from the research and engineering fault diagnosis techniques and monitoring results are provided
communities in recent years. The purpose of this study is to in section III. In Section IV, concludes our contributions are
review the Vibration analysis techniques and explore their discussed, and we conclude with some remarks and future
capabilities, advantages and disadvantage in monitoring rolling work.
element bearings.
A. System Description
Keywords Vibration Analysis; bearing Fault diagnosis; Fast
The experiments presented in this paper used the vibration
Fourier Transform (FFT); Envelope Analysis (EA); Short Time
data obtained from the Case Western Reserve University
Fourier Transform (STFT); Empirical Mode Decomposition
(EMD).
Bearing Data Centre [9]. These signals were taken from an
accelerometer mounted on the bearing housing at the drive end
I. INTRODUCTION (HEADING 1) of the induction motor which is connected to a torque
transducer, coupled to a dynamometer, which plays the role of
The necessity of the use of the large and the small rotating
a load as shown in Fig .1.
machines in the present life and especially in industrial
systems since the beginning of last century imposes the
monitoring, maintenance and reparation. The main function of
condition monitoring of rotating machines is to provide
knowledge about the condition of machines at each moment
without stopping the line of production [1-3]. The vibration
monitoring is one of the most common technique of condition
monitoring and this is for its ability to detect, locate and
distinguish different types of faults since its inception before
they become critical and dangerous, these faults which may be
distributed or localized [3-4]. The bearings are the most
essential mechanical elements of rotating machinery. They are
employed to support and rotate the shafts in rotating
machinery. On the other hand, several studies showed that the
major source of most mechanical faults in rotating machinery
is the bearing fault [1-3].Therefore; any fault in the bearings
can lead to losses on the level of production and equipments as
well as the creation of an unsafe working environment for
humans. For these reasons, Condition monitoring and fault
diagnosis of these bearings has become a fundamental axis of
development and industrial research. [1-3].
There are several techniques that can be employed to
detect and diagnosis the bearing faults like Fast Fourier Fig.1. The bearing test stand.
Transform (FFT) [4], Short Time Fourier Transform (STFT)
[4-5],Wigner-Ville Distribution (WVD) [6], Envelope analysis The bearing fault signals were obtained from different states
(EA) [7-8], Wavelet Transform (WT) [7-8] and Empirical
of the bearing: (1) normal state; (2) inner race fault; (3) ball

978-0-9567157-6-0 IEEE 2016

37
8th International Conference on Modelling, Identification and Control (ICMIC-2016)
Algiers, Algeria- November 15-17, 2016

fault; (4) outer race fault, and at different speeds of the motor: Motor Speed 1797 (rpm)
1730, 1750, 1772, and 1797 rpm.
The diameters of the faults created on the inner race, outer 0.2

Amplitude (g)
race and the balls are 0.007", 0.014" and 0.021 inch 0.1
(1inch=25.4 mm), all faults have a depth of 0.011 inch. All 0
experiments were repeated for four different load conditions: -0.1
0, 1, 2 and 3 horse power (hp). The data is sampled at a
-0.2
frequency of 12 kHz and the duration of each vibration signal
of vibration is 10 seconds. 0 0.1 0.2 0.3 0.4
Time (s)
B. Fault Bearing Characteristic Frequencies .
Defective bearings generate vibration equal to the Motor Speed 1730 (rpm)
rotational speed of each element bearing frequencies. They
relate notably to the rotation of the balls, the cage and the 0.2

Amplitude (g)
passage of the balls on the inner and outer rings. 0.1
Frequencies associated with defective inner and outer races 0
are as follows: -0.1
- Inner-race defect is given by equation (1):
-0.2

n d (1) 0 0.1 0.2 0.3 0.4

BPFI f r 1 cos
2 D Time (s)

Fig.3. Vibration signals of normal state.

Where, fr is the rotational frequency, d the ball diameter, D the
pitch diameter, n the number of balls and the contact angle. Motor Speed 1797 (rpm)
The components of bearing as shown is Fig .2. 2

Amplitude (g) 1

-1

-2
0 0.1 0.2 0.3 0.4
Time (s)

Motor Speed 1730 (rpm)

2

1
Amplitude (g)

Fig.2. The components of bearing 0

Table 1. Calculated Frequencies of bearing faults -1

Location of the data The Drive-End bearing (DE) -2

0 0.1 0.2 0.3 0.4
collection Time (s)
Fault Diameter 0.007 inches
Faults Types Inner race Fig.4. Vibration signals of inner races fault.
Motor Load (HP) 0 3
Recently, there are many signal processing methods which
Motor Speed (rpm) 1797 1730
Frequencies Speed of Motor 29.94 28.83 can be used for fault detection and diagnosis in bearing such
fr (Hz) as Frequency Analysis, Envelope Analysis and Empirical
Frequencies of bearing 162.2 156.13 Mode Decomposition (EMD).
faults in Hz calculated by
equations (1) and (2) A. Frequency analysis
Frequency analysis is the most traditional method which can
Figures (3) and (4) represent respectively the vibration be employed for analyzing the vibration signals [10-13].
signals of normal state, inner races fault.

978-0-9567157-6-0 IEEE 2016

38
8th International Conference on Modelling, Identification and Control (ICMIC-2016)
Algiers, Algeria- November 15-17, 2016

The spectrum X (f) of a given signal x (t) is defined by

t x t 2 xt 2 (5)

i 2ft
X( f ) x(t )e dt (2)

Motor Speed 1797 (rpm)
Where: is the signal frequency

The spectrums of inner races fault at 1797 rpm and 1730

Amplitude (g)
3580Hz
0.1
rpm are shown in Fig. 6. Obviously, it is difficult to identify
the bearing fault, because the spectral analysis presents some 0.05
limitations in the analysis of non-stationary signals.
0
Motor Speed 1797 (rpm) 0 2000 4000 6000
Frequency (Hz)

Motor Speed 1730 (rpm)

Amplitude (g)

3580Hz
0.1
3600Hz

Amplitude (g)
0.05
0.1

0
0 2000 4000 6000 0.05
Frequency (Hz)

Motor Speed 1730 (rpm) 0

0 2000 4000 6000
3600Hz Frequency (Hz)
Amplitude (g)

0.1 Fig.6. Spectrums of signals with inner races fault.

0.05

Laod Vibration Signal

0
0 2000 4000 6000
Frequency (Hz)

Fig.5. Spectrums of signals with inner races fault. Filtered Signal

B. Envelope Analysis (EA)

The envelope Analysis is an important signal processing
technique which used to extract the fault features from
modulation signals.
The envelope Analysis can be decomposed into three
steps:: signal filtering, envelope extraction of the filtered
signal by the application of Hilbert transform (HT), and the
determination of the spectrum of the envelope by the applying
the Fast Fourier transform (FFT) [14-18] as shown in Fig.7.
The Hilbert Transform in time domain, for a given signal
x(t), is given by [14-18].
1 x
x t d (3)
t studied. The envelopes of inner races fault at 1797 rpm and
It is defined as the convolution of the signal x(t) with
function 1/t, which is the impulse response function of the
Hilbert Transformer. The phase shifted and original signals are
summed up to obtain an analytic signal x+(t) defined as
follows [14-18].
x t x t j x t
Envelope Extraction of
the Filtered signal
Spectrum of Bearing
Fault
Fig.7. Procedure for Envelope Analysis (EA).
The center frequency of band pass filter should be selected
to coincide with the center frequency of the resonance to be
1730 rpm are shown in Fig 8. The envelopes spectrums of
inner race fault collected at 1797 rpm and 1730 rpm are shown
in Fig.9. It can be clearly seen from Fig.9 the frequency peaks
at 162 Hz, 156 Hz and its multiples. Also, the frequency peaks
at the rotation frequency of the motor (29 and 30Hz). The
frequency peaks at 162 Hz and 156 Hz are very close to
calculated frequencies of inner race fault at 162.2 Hz and
156.13Hz as listed in table 1. This clearly indicates an inner
(4) race fault.

978-0-9567157-6-0 IEEE 2016

39
8th International Conference on Modelling, Identification and Control (ICMIC-2016)
Algiers, Algeria- November 15-17, 2016

The identification of the bearing faults is possible by using Hence, the signal will be represented by two elements of:
envelope analysis. However, a critical drawback of this time and frequency shown in Fig .10 [19-22].
technique is that it requires a pre-knowledge of the resonance
frequency and filtering band.

Motor Speed 1797 (rpm)

0.8

0.6
Amplitude (g)

0.4
Fig.10. Short-Time Fourier Transform.
0.2
For a signal x(t), its short time Fourier transform is defined
0
0 0.1 0.2 0.3 0.4 as follows [19-22]:
Time (s)
(6)
STFTxw ( , f ) x(t ) w (t ) e j 2ft dt
*
Motor Speed 1730 (rpm)
0.8 t

0.6
Where:
Amplitude (g)

W (t): is the window of width T which centered at

0.4 : is the parameter of temporal location of the window g.
w *(t): is the conjugate of the complex function w (t)
0.2 The short time Fourier transform spectrum of inner race
0
fault collected at 1797 rpm and 1730 rpm are shown in
0 0.1 0.2 0.3 0.4 Fig.11.We can be seen clearly the different colors in red in
Time (s) Fig.11. Which correspond to calculated frequencies of inner
Fig.8. Envelopes of signals with inner races fault. race fault at 162.2 Hz and 156.13Hz as listed in table.1, the
inner race fault is detected successfully with STFT.
Motor Speed 1797 (rpm)
0.2 Short-time Fourier Transform spectrum for Motor Speed 1797 (rpm)
162Hz 250
0.15
Amplitude (g)

200
0.1
324Hz
Freq (Hz)

30Hz 150
0.05
100
0
0 100 200 300 400 500
Frequency (Hz) 50
Motor Speed 1730 (rpm)
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Time (secs)
156Hz
Amplitude (g)

0.1 Short-time Fourier Transform spectrum for Motor Speed 1730 (rpm)
312Hz 250

0.05 200
29Hz
Freq (Hz)

150
0 100
0 100 200 300 400 500
Frequency (Hz) 50
Fig.9. Spectrums of inner race fault using envelope analysis.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
C. Short-Time Fourier Transform (STFT) Time (secs)
Fig.11. Spectrums of inner race fault using STFT.
Short-Time Fourier Transform (STFT) or windowed
Fourier transform (WFT) is the most widely used tool for time The identification of the bearing faults is possible by
frequency (TF) analysis of non-stationary signals. The idea of using STFT. However, the problem with STFT is that it
STFT is to analyze the signal into segment by segment (or provides constant resolution for all frequencies since it uses
window by window). It uses a window function to slide on the the same window for the entire signal. Therefore, once the
window function is chosen, the time and frequency
signal studied and then divide it into several equal length resolution are fixed. So there is a tradeoff to choose a
segments (or window). The inside signal of the segments is proper window function between the time resolution and the
supposed to be stationary. After that Fourier transform is frequency resolution: a longer window will lead to a higher
applied in each segment to find out the frequencies contained frequency resolution with a lower time frequency and vice
in that segment. versa.

978-0-9567157-6-0 IEEE 2016

40
8th International Conference on Modelling, Identification and Control (ICMIC-2016)
Algiers, Algeria- November 15-17, 2016

imf3

Amplitude (g)
0.5
D. Empirical Mode Decomposition (EMD)
0

Empirical Mode Decomposition (EMD) is an adaptive -0.5

0 0.1 0.2 0.3 0.4
time-frequency technique for analyzing nonlinear and non- Time (s)
stationary signals, which can decompose the time domain imf4

Amplitude (g)
0.5
signal into a set of oscillatory functions in the time-domain,
called intrinsic mode functions (IMF). The nonlinear signal 0

(i) is then decomposed into M intrinsic modes and a residue -0.5

0 0.1 0.2 0.3 0.4
RM [23-25]. Time (s)
imf5

Amplitude (g)
M 0.5
i (n) imf m (n) RM (n) (7)
0
m 1
-0.5
0 0.1 0.2 0.3 0.4
The first five IMFs after EMD of inner race fault collected Time (s)
at 1797 rpm and 1730 rpm are shown in Figs.12 and 13.
Fig.13. The first five IMFs after EMD of inner race fault
imf1
collected at 1730 rpm.
Amplitude (g)

0
The envelope spectrum of IMF2 of inner race fault
-1
0 0.1 0.2 0.3 0.4 collected at 1797 rpm is shown in Fig.14.
Time (s)
imf2 The envelope spectrum of IMF3 of inner race fault
Amplitude (g)

0.5
collected at 1730 rpm is shown in Fig.15.
0

-0.5
It can be clearly seen from Figs.14 and 15, the frequency
0 0.1 0.2 0.3 0.4 peaks at 162 Hz, 156 Hz and its multiples. The frequency
Time (s)
imf3
peaks at 162 Hz and 156 Hz are very close to calculated
frequencies of inner race fault at 162.2 Hz and 156.13Hz as
Amplitude (g)

0.5

0
listed in table 1. This clearly indicates an inner race fault.
-0.5
0 0.1 0.2 0.3 0.4 Empirical Mode Decomposition (EMD) is an excellent
Time (s)
imf4
tool for bearing fault diagnosis. Unfortunately, there are two
problems in EMD, which are the selection of the suitable
Amplitude (g)

0.2

0 decomposition level and its intrinsic mode functions (IMF)

which contains the necessary information for faults diagnosis.
-0.2
0 0.1 0.2 0.3 0.4
Time (s)
Spectrum Envelope of imf 2
imf5
0.1
Amplitude (g)

0.2
162Hz 1797 rpm
Amplitude (g)

-0.2 0.05
0 0.1 0.2 0.3 0.4
Time (s)

0
Fig.12. The first five IMFs after EMD of inner race fault 0 100 200 300 400 500
collected at 1797 rpm. Frequency (Hz)

imf1 Fig.14. The envelope spectrum of IMF2 of inner race fault

Amplitude (g)

0 Spectrum Envelope of imf 3

-2
0.1
0 0.1 0.2 0.3 0.4 156Hz 1730 rpm
Amplitude (g)

Time (s)
imf2
Amplitude (g)

0.5 0.05
0

-0.5
0 0.1 0.2 0.3 0.4 0
Time (s) 0 100 200 300 400 500
Frequency (Hz)
Fig.15. The envelope spectrum of IMF2 of inner race fault

978-0-9567157-6-0 IEEE 2016

41
8th International Conference on Modelling, Identification and Control (ICMIC-2016)
Algiers, Algeria- November 15-17, 2016
III. CONCLUSION
In this paper, A Comparative Study of FFT, STFT, EA and
EMD for bearing fault diagnosis were presented in order to
review the Vibration analysis techniques and explore their
capabilities, advantages and disadvantage in monitoring
rolling element bearings. After this study, the following points
are concluded.
1. The identification of the bearing faults by using
frequency analysis is difficult because, it is not
suitable for non-stationary signal analysis.
2. The identification of the bearing faults is possible by
using envelope analysis. However; the envelope
analysis has a major drawback consisting of the
requirement of a preliminary research of the
resonance frequencies.
3. The identification of the bearing faults is possible by
using STFT. However, the problem with STFT is that
it provides constant resolution for all frequencies
since it uses the same window for the entire signal.
Therefore, once the window function is chosen, the
time and frequency resolution are fixed. So there is a
tradeoff to choose a proper window function
between the time resolution and the frequency
resolution: a longer window will lead to a higher
frequency resolution with a lower time frequency and
vice versa.
4. The identification of the bearing faults is possible by
using EMD. Unfortunately, there are two problems in
EMD, which are the selection of the suitable
decomposition level and its intrinsic mode functions
(IMF) which contains the necessary information for
faults diagnosis
REFERENCES
[1] N. Tandon, A. Parey , Condition Monitoring of Rotary Machines,
Springer Series in Advanced Manufacturing , Vol.5, pp. 109-136,2006.
[2] A . Heng, S. Zhang, A. Tan, J. Mathew, Rotating machinery
prognostics: State of the art, challenges and opportunities, Journal of
Mechanical Systems and Signal Processing, Vol.23, no. 3, pp. 285
293, 1997.
[3] A. Boudiaf, S. Bouhouche ,A.K,Moussaoui,S.Taleb, An effective
method for bearing faults diagnosis," Proceedings of the 3nd 3rd
International Conference on Control, Engineering & Information
Technology (CEIT2015), Tlemcen (Algeria), pp. 1-6, 2015.
[4] K. Shibata, A. Takahashi, T. Shirai, Fault diagnosis of rotating
machinery through visualisation of sound signal, Journal of Mechanical
Systems and Signal Processing, Vol. 14, pp. 229241 , 2000.
[5] S. Seker, E. Ayaz, A study on condition monitoring for induction
motors under the accelerated aging processes, IEEE Power
Engineering, Vol.22, no. 7, pp. 3537, 2002.
[6] N. Baydar, A. Ball, A comparative study of acoustic and vibration
signals in detection of gear failures using wigner ville distribution,
Mechanical Systems and Signal Processing, Vol.15, no. 6, pp. 1091
1107, 2001.
[7] H. Bendjama, S. Bouhouche, M.S. Boucherit, Application of Wavelet
Transform for Fault Diagnosis in Rotating Machinery, International
Journal of Machine Learning and Computing,Vol.2, no. 1, pp. 8287,
2012.
[8] H. Bendjama , S. Bouhouche , A.k. Moussaoui, Wavelet Transform for
Bearing Faults Diagnosis, proceedings of the 2013, Conference of
Advances in Control Engineering, , no.14, pp. 8588, 2013.
[9] K.A. Loparo, Bearings vibration data set, Case Western Reserve
University, (http://www.eecs.cwru.edu),2003.
978-0-9567157-6-0 IEEE 2016
[10] V .Saxena, N Chowdhury,S. Devendiran, Assessment of Gearbox Fault
Detection Using Vibration Signal Analysis and Acoustic Emission
Technique, Journal of Mechanical and Civil Engineering, Vol.7, no. 4,
pp. 52 60, 2013.
[11] A. Aherwar, M. Saifullah Khalid , Vibration analysis techniques for
gearbox Diagnostic: a review, International Journal of Advanced
Engineering Technology, Vol.3, no. 2, pp. 0412, 2012.
[12] T.Karacay, N.Akturk, Experimental diagnostics of ball bearings using
statistical and spectral methods, Tribology International,Vol.42, pp.
836843, 2009.
[13] A.Bhende, G.Awari ,S.Untawale , Assessment of Bearing Fault
Detection Using Vibration Signal Analysis, Journal of Technical and
Non-Technical , Vol.2, no.5, pp. 249-261, 2011.
[14] A. Djebala ,N. Ouelaa ,C. Benchaabane, D.F. Laefer, Application of the
Wavelet Multi-resolution Analysis and Hilbert transform for the
prediction of gear tooth defects, Journal of Meccanica, Vol.47, no.7, pp.
16011612, 2012.
[15] M. Feldman , Hilbert transform in vibration analysis, Mechanical
Systems and Signal Processing , Vol.25, pp. 735802, 2011.
[16] R. B.Randall , J. Antoni , Rolling element bearing diagnostics,
Mechanical Systems and Signal Processing, Vol.25, pp. 485520, 2011.
[17] J.Slavic, A. Brkovic, M. Boltezar, Typical Bearing-Fault Rating Using
Force Measurements-Application to Real Data, Journal of Vibration and
Control,Vol.17, no.14, pp. 2164-2174, 2012.
[18] A. Boudiaf,A.K,Moussaoui,A. Dahane,I.Atoui A Comparative Study of
Various Methods of Bearing Faults Diagnosis Using the Case Western
Reserve University Data," Journal of Failure Analysis and Prevention,
Vol.16, no.02, pp. 271-284, 2016.
[19] S. Dass , M. S. Holi and K. Soundara Rajan, A Comparative Study on
FFT,STFT and WT for the Analysis of Auditory Evoked Potentials,
International Journal of Engineering Research & Technology (IJERT),
Vol.02, pp. 636640, 2013.
[20] N. Mehala, R. Dahiya, A Comparative Study on FFT,STFT and WT for
Induction Machine Fault Diagnostic Analysis, Proceedings of the
7th WSEAS international conference on Computational intelligence,
man-machine systems and cybernetics, pp. 203208, 2008.
[21] M. Cocconcelli, R. Zimroz, R. Rubini, W. Bartelmus, STFT Based
Approach for Ball Bearing Fault Detection in a Varying Speed Motor,
Proceedings of the Second International Conference "Condition
Monitoring of Machinery in Non-Stationnary Operations" CMMNO, pp.
4150, 2012.
[22] F. Al-Badour, M. Sunar, L. Cheded, Vibration Analysis ofRotating
Machinery Using Time-Frequency Analysis and Wavelet Techniques,
Mechanical Systems and Signal Processing ,Vol.25, no.06, pp. 2083-
2101, 2011.
[23] M.Pan , W.Tsao , Using appropriate IMFs for envelope analysis in
multiple fault diagnosis of ball bearings, International Journal of
Mechanical Sciences, Vol.69, pp. 114124, 2013.
[24] J.A. Rosero, L. Romeral, J.A. Ortega and E. Rosero, Short-circuit
detection by means of empirical mode decomposition and WignerVille
distribution for PMSM running under dynamic condition, IEEE Trans.
Industrial Electronics, vol. 56, no.11, pp. 4534-4547, November 2009.
[25] Q. Gao, C. Duan, H. Fan and Q. Meng, Rotating machine fault
diagnosis using empirical mode decomposition, Mechanical Systems &
Signal Processing, vol. 22, no.5, pp. 1072-1081, July 2008.
42