university of huddersfield repositoryeprints.hud.ac.uk/id/eprint/29476/1/ballmsb bearing - adb -...
TRANSCRIPT
University of Huddersfield Repository
Ball, Andrew, Wang, Tian T., Tian, X. and Gu, Fengshou
A robust detector for rolling element bearing condition monitoring based on the modulation signal bispectrum,
Original Citation
Ball, Andrew, Wang, Tian T., Tian, X. and Gu, Fengshou (2016) A robust detector for rolling element bearing condition monitoring based on the modulation signal bispectrum,. In: COMADEM 2016, the 29th International Congress on Condition Monitoring and Diagnostic Engineering Management, 20th22nd August 2016, Empark Grand Hotel in Xi’an, China.
This version is available at http://eprints.hud.ac.uk/id/eprint/29476/
The University Repository is a digital collection of the research output of theUniversity, available on Open Access. Copyright and Moral Rights for the itemson this site are retained by the individual author and/or other copyright owners.Users may access full items free of charge; copies of full text items generallycan be reproduced, displayed or performed and given to third parties in anyformat or medium for personal research or study, educational or notforprofitpurposes without prior permission or charge, provided:
• The authors, title and full bibliographic details is credited in any copy;• A hyperlink and/or URL is included for the original metadata page; and• The content is not changed in any way.
For more information, including our policy and submission procedure, pleasecontact the Repository Team at: [email protected].
http://eprints.hud.ac.uk/
Professor Andrew BallPro Vice-Chancellor for Research and Enterprise
&Director of The Centre for Efficiency and Performance Engineering
A robust detector for rolling element bearing condition monitoring, based on the modulation
signal bispectrum, and its performance evaluation against the kurtogram
Contents
1. Introduction
2. The modulation signal based detector
3. Simulation study
4. Application case studies
5. Conclusions
Introduction The MSB detector Application case studiesSimulation study Conclusions 2
1. Introduction
• Bearings are at the heart of almost every rotating machine• they have received a lot of attention in
the field of vibration analysis • because they are common sources of
machine faults• To keep machinery operating reliably many
methods for bearing fault detection and diagnosis have been developed• vibration measurement and associated
signal processing are the most widely used approach
3Introduction The MSB detector Application case studiesSimulation study Conclusions
1. Introduction
High frequency resonance technique
Cyclostationary spectral analysis
Cepstrum analysis
Bispectrum analysis
Time-frequency analysis
Self-adaptive noise cancellation
Minimum entropy deconvolution
Empirical mode decomposition
A review of signal processing methods
Most methods are based ontracking the amplitude variationsof characteristic fault frequencies
Only limited attention has beengiven to utilisation of modulationcharacteristics in extracting thediagnostic information
But the Modulation SignalBispectrum can be used toextract fault features fromenvelope signals giving reliablebearing fault detection, diagnosisand severity assessment
4Introduction The MSB detector Application case studiesSimulation study Conclusions
• To develop and evaluate a robust detector for bearing fault diagnosis based on modulation signal bispectrum (MSB) analysis
Purpose of the research
1. Introduction
5
The modulation signal bispectrum
• MSB analysis can be used to suppress random noise and todecompose nonlinear modulation components in a measuredsignal, eg vibration
• Advantages of the modulation signal bispectrum include:
1) Highly effective suppression of random noise
2) Revelation of the weak nonlinear characteristics of signals
Introduction The MSB detector Application case studiesSimulation study Conclusions
Outer race
Inner race
Roller
Cage
𝐷𝑟𝐷𝑐
2. The modulation signal based detector
Contact Angle φ
Ball Diameter
Pitch
Diameter
(1 cos )2
r rs
c
N DBPFO F
D
(1 cos )2
r rs
c
N DBPFI F
D
22
2(1 cos )
2
c s r
r c
D F DBSF
D D
Outer race fault frequency:
Inner race fault frequency:
Ball fault frequency:
Cage fault frequency (often called the fundamental train frequency):
1(1 cos )
2
rs
c
DFTF F
D
Bearing fault frequencies
6Introduction The MSB detector Application case studiesSimulation study Conclusions
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
2
4x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
2
4x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
2
4x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Time(s)
Acc
.(m
/s2)
0 100 200 300 400 500 6000
5x 10
-3
Frequency(Hz)
Small outer race fault
HealthyRaw data
Filtered
Demodulated
Spectrum of the envelope
7
2. The modulation signal based detector
Envelope analysis is the current state of the art in industry, and it can be exemplified as follows:
Introduction The MSB detector Application case studiesSimulation study Conclusions
Rectified & Smoothed
Envelope analysis is a time domain technique which involves filtering, de-modulating, rectifying and smoothing time data before transferring to the frequency domain.The challenge is selecting the most appropriate filter edges.
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Time(s)
Acc
.(m
/s2)
0 100 200 300 400 500 6000
2
4
6x 10
-3
Frequency(Hz)
Small outer race fault
Healthy
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Time(s)
Acc
.(m
/s2)
0 100 200 300 400 500 6000
2
4
6x 10
-3
Frequency(Hz)
Small outer race fault
Healthy
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Time(s)
Acc
.(m
/s2)
0 100 200 300 400 500 6000
2
4
6x 10
-3
Frequency(Hz)
Small outer race fault
Healthy
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1A
cc.(
m/s
2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Acc
.(m
/s2)
0 2000 4000 6000 8000 100000
1
2
3x 10
-3
0 0.02 0.04 0.06 0.08 0.1-0.1
0
0.1
Time(s)
Acc
.(m
/s2)
0 100 200 300 400 500 6000
2
4
6x 10
-3
Frequency(Hz)
Small outer race fault
Healthy
BPFO=88.5Hz
8
2. The modulation signal based detector
The kurtogram is current state of the art in bearing monitoring research – and hence is our comparator
Introduction The MSB detector Application case studiesSimulation study Conclusions
Frequency(Hz)
lev
el k
fb-kurt.1 - Kmax
=0.7 @ level 4, Bw= 625Hz, fc=2187.5Hz
0 2000 4000 6000 8000 10000
0
1
1.6
2
2.6
3
3.6
4
4.6
5
5.6
6
6.6
70
0.1
0.2
0.3
0.4
0.5
0.6
The Kurtogram reveals a colour map of thresholded Kurtosis across frequency, and hence it allows resonance regions to be identified and envelope filter edges to be selected.
Kmax — Maximum kurtosis valueLevel 4 — Optimised filter band at level 4Bw — filter bandwidth of optimised filterfc — central frequency of optimised filter
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1-0.05
0
0.05Raw data
Time(s)
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
0.005
0.01Envelope of the filtered signal
Time(s)
0 50 100 150 200 250 300 350 400 450 5000
1
2x 10
-6 Fourier transform magnitude of the squared envelope
Frequency(Hz)
BPFO=88.5Hz
)()()()()()( tntxtxtxtxtx sbsqf
Impulses produced by bearing fault
AM due to the non-uniform load distribution
NoiseMachinery induced vibration
Bearing-induced vibration determined by the bearing
structural dynamics
The bearing vibration model
2. The modulation signal based detector
9Introduction The MSB detector Application case studiesSimulation study Conclusions
The vibration signature x(t) of a faulty rolling element bearing is comprised of several components, and these are represented in the model as follows:
The presence of modulation in the time data means that there will be sidebands in the frequency domain, and these are key to bearing fault detection/diagnosis
It was decided to use 2 ways of evaluating the new method: via simulated data and real data. To simulate bearing fault data, a bearing vibration model is needed.
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude 1/f
o
(a)
0 50 100 150 2000
0.005
0.01
0.015
Am
pli
tude f o
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude
(b)
1/fi
1/fr
0 50 100 150 2000
0.005
0.01
Am
pli
tude f i
fi-f r
f i+fr
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude
(c)
1/fb
1/fc
0 50 100 150 2000
2
4x 10
-3
Am
pli
tude f b
fb-f c
f b+f c
0 0.05 0.1 0.15 0.2-5
0
5
Time(s)
Am
pli
tude
(d)
1/fc
0 50 100 150 2000
1
2
3x 10
-3
Frequency(Hz)
Am
pli
tude
f c
Simulated fault time series and spectra of a rolling element bearing with a localised defect on the (a) outer race, (b) inner race (c) rolling element, and (d) cage
2. The modulation signal based detector
Simulated fault data
10Introduction The MSB detector Application case studiesSimulation study Conclusions
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e 1/fo
(a)
0 50 100 150 2000
0.005
0.01
0.015
Am
pli
tud
e f o
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e
(b)
1/fi
1/fr
0 50 100 150 2000
0.005
0.01
Am
pli
tud
e f i
fi-f r
f i+fr
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e
(c)
1/fb
1/fc
0 50 100 150 2000
2
4x 10
-3
Am
pli
tud
e f b
fb-f c
f b+f c
0 0.05 0.1 0.15 0.2-5
0
5
Time(s)
Am
pli
tud
e
(d)
1/fc
0 50 100 150 2000
1
2
3x 10
-3
Frequency(Hz)
Am
pli
tud
e
f c
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e 1/fo
(a)
0 50 100 150 2000
0.005
0.01
0.015
Am
pli
tud
e f o
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e
(b)
1/fi
1/fr
0 50 100 150 2000
0.005
0.01
Am
pli
tud
e f i
fi-f r
f i+fr
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tud
e
(c)
1/fb
1/fc
0 50 100 150 2000
2
4x 10
-3
Am
pli
tud
e f b
fb-f c
f b+f c
0 0.05 0.1 0.15 0.2-5
0
5
Time(s)
Am
pli
tud
e
(d)
1/fc
0 50 100 150 2000
1
2
3x 10
-3
Frequency(Hz)
Am
pli
tud
e
f c
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude 1/f
o
(a)
0 50 100 150 2000
0.005
0.01
0.015
Am
pli
tude f o
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude
(b)
1/fi
1/fr
0 50 100 150 2000
0.005
0.01
Am
pli
tude f i
fi-f r
f i+fr
0 0.05 0.1 0.15 0.2-5
0
5
Am
pli
tude
(c)
1/fb
1/fc
0 50 100 150 2000
2
4x 10
-3
Am
pli
tude f b
fb-f c
f b+f c
0 0.05 0.1 0.15 0.2-5
0
5
Time(s)
Am
pli
tude
(d)
1/fc
0 50 100 150 2000
1
2
3x 10
-3
Frequency(Hz)
Am
pli
tude
f c
Conventional bispectrum (CB)
𝑓1 𝑓3 = 𝑓2 + 𝑓1𝑓2
• Conventional bispectrum offers:– Nonlinear identification capability
– Retention of phase information
– Noise suppression capability
)()()( * fXfXEfPS
2. The modulation signal based detector
11Introduction The MSB detector Application case studiesSimulation study Conclusions
But it is limited to f1 + f2 = f3 ……so it only shows the higher sideband
1 2 2 1 2 1( , ) ( ) ( ) ( )CB f f f f f f
An average must be performed to suppress random noise
E
1 2 1 2 1 2( , ) ( ) ( ) ( )B f f E X f X f X f f
Definition of CB:
Phase of CB:
Coherence of CB:
The modulation signal bispectrum (MSB)
* *( , ) ( ) ( ) ( ) ( )MS c x c x c x c cB f f E X f f X f f X f X f
2. The modulation signal based detector
,,
,0
MS c xSEMS c x
MS c
B f fB f f
B f
12Introduction The MSB detector Application case studiesSimulation study Conclusions
The MSB is based on the CB, but is more attractive in this work because it contains both upper and lower sideband components
The MSB can itself be modified to enable precise quantification of sideband amplitudes, by removing the influence of the carrier frequency fc. We call this the MSB sideband estimator (MSB-SE), defined as:
The MSB-SE only includes the information of sidebands (fc + fx) & (fc - fx)
2. The modulation signal based detector
of
2 of
Typical results of the MSB detector - B(fx) - formed from slices shown along B(fc), show that the optimal frequency band for detecting a bearing fault is at a specific value of fc. Symptomatic features are labelled *.
13Introduction The MSB detector Application case studiesSimulation study Conclusions
Vibration signal
Calculate the MSB using
Calculate the MSB-sideband estimator using
Calculate the compound MSB slice using
Calculate the robust MSB detector using
* *( , ) ( ) ( ) ( ) ( )MS c x c x c x c cB f f E X f f X f f X f X f
,,
,0
MS c xSEMS c x
MS c
B f fB f f
B f
2
1( ) ,
1
SENic MS cB f B f i f
N
1
1( ) , 0SE kK
kx MS c x xB f B f f fK
2. The modulation signal based detector
Flow chart of the robust MSB detector calculation
14Introduction The MSB detector Application case studiesSimulation study Conclusions
Vibration signal
Calculate the MSB using
* *( , ) ( ) ( ) ( ) ( )MS c x c x c x c cB f f E X f f X f f X f X f
Calculate the MSB-sideband estimator using
,,
,0
MS c xSEMS c x
MS c
B f fB f f
B f
Calculate the compound MSB slice using
2
1( ) ,
1
SENic MS cB f B f i f
N
Calculate the robust MSB detector using
1
1( ) , 0SE kK
kx MS c x xB f B f f fK
3. Simulation study
1020log /s nType 1 SNR P P 1020log /s nType 2 SNR A A
Scenario White noiseAperiodic impact
interference
Type 1
SNR value
Type 2
SNR value
Low noise signal without impact interferences
Level 1 None -15dB n/a
High noise signal without impact interferences
Level 2 None -30dB n/a
Low noise signal with low level impact interferences
Level 1 Level 1 -15dB -22dB
High noise signal with high level impact interferences
Level 1 Level 2 -22dB -48dB
Four different simulation scenarios were developed
15Introduction The MSB detector Application case studiesSimulation study Conclusions
They use different levels of random noise and different amounts of aperiodic interference, to represent the noisy in-field measurements typically encountered in the vibration-based condition monitoring of rolling element bearings
3. Simulation study
The extent of white noise contamination
16Introduction The MSB detector Application case studiesSimulation study Conclusions
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tude
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tude
3
471H
z
712
0Hz
11750Hz
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tude
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
7
120H
z
11750HzA
mpli
tude
0 0.02 0.04 0.06 0.08 0.1-200
0
200
Time(s)
Am
pli
tude
(c)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tude 11750Hz
SNR=-15dB
SNR=-30dB
R2 R3
Low noise, no impacts
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tud
e
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tud
e
3
471H
z
712
0Hz
11750Hz
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tud
e
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
7
120H
z
11750Hz
Am
pli
tud
e
0 0.02 0.04 0.06 0.08 0.1-200
0
200
Time(s)
Am
pli
tud
e
(c)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tud
e 11750Hz
SNR=-15dB
SNR=-30dB
R3
High noise, no impacts
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tud
e
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tud
e
3
471H
z
712
0Hz
11750Hz
0 0.02 0.04 0.06 0.08 0.1-20
0
20
Am
pli
tud
e
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
7
120H
z
11750HzA
mp
litu
de
0 0.02 0.04 0.06 0.08 0.1-200
0
200
Time(s)
Am
pli
tud
e
(c)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tud
e 11750Hz
SNR=-15dB
SNR=-30dB
1/ of
R2 R3R1
Pulse train of impacts from bearing fault
BPFO=88.5Hz Simulated structural resonance regions which naturally amplify bearing frequency harmonics
6000 7000 8000 9000 10000 11000 12000 130000
0.02
0.04
0.06(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-30dB
R3
Kurtogram based Detector
MSB Robust Detector
High noise, no impulses
6000 7000 8000 9000 10000 11000 12000 130000
0.005
0.01
(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-15dB
R2
R3
Kurtogram based Detector
MSB Robust Detector
Low noise, no impulses
3. Simulation study
The MSB-SE results compared to the Kurtogram results
17Introduction The MSB detector Application case studiesSimulation study Conclusions
3. Simulation study
The addition of aperiodic impact interference, along with white noise
18Introduction The MSB detector Application case studiesSimulation study Conclusions
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.1 0.2 0.3 0.4 0.5-100
-50
0
50
100
Time(s)
Am
pli
tud
e
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tud
e
7
120H
z
11750Hz
0 0.1 0.2 0.3 0.4 0.5-1000
-500
0
500
1000
Time(s)
Am
pli
tud
e
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tud
e
11750Hz
SNR=-22dB
SNR=-49dB
R2
R3
Low noise, low level impacts
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-1000
-500
0
500
1000
0 0.1 0.2 0.3 0.4 0.5-100
-50
0
50
100
Am
pli
tude
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tude
7
120H
z
11750Hz
0 0.1 0.2 0.3 0.4 0.5-1000
-500
0
500
1000
Time(s)
Am
pli
tude
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tude
11750Hz
SNR=-21dB
SNR=-50dB
R3
Low noise, high level impacts
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.1 0.2 0.3 0.4 0.5-100
-50
0
50
100
Time(s)
Am
pli
tud
e
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tud
e
7
120Hz
11750Hz
0 0.1 0.2 0.3 0.4 0.5-1000
-500
0
500
1000
Time(s)
Am
pli
tud
e
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tud
e
11750Hz
SNR=-22dB
SNR=-49dB
++Pulse train of impacts from bearing fault Random noise Impacts
BPFO=88.5Hz
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-100
-50
0
50
100
0 0.2 0.4-1000
-500
0
500
1000
0 0.1 0.2 0.3 0.4 0.5-100
-50
0
50
100
Am
pli
tude
(a)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Am
pli
tude
7
120H
z
11750Hz
0 0.1 0.2 0.3 0.4 0.5-1000
-500
0
500
1000
Time(s)
Am
pli
tude
(b)
0 0.5 1 1.5 2
x 104
0
0.2
0.4
0.6
0.8
Frequency (Hz)
Am
pli
tude
11750Hz
SNR=-21dB
SNR=-50dB
+ +
3. Simulation study
Performance evaluation of the MSB-SE against the Kurtogram
6000 7000 8000 9000 10000 11000 12000 130000
0.005
0.01
(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-22dB
R2
R3
Kurtogram based Detector
MSB Robust Detector
6000 7000 8000 9000 10000 11000 12000 130000
0.01
0.02
0.03(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-50dB
R2 R3
Kurtogram based Detector
MSB Robust Detector
Low noise, low level
impact interference
Low noise, high level
impact interference
19Introduction The MSB detector Application case studiesSimulation study Conclusions
6000 7000 8000 9000 10000 11000 12000 130000
0.005
0.01
(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-22dB
R2
R3
Kurtogram based Detector
MSB Robust Detector
6000 7000 8000 9000 10000 11000 12000 130000
0.01
0.02
0.03(a)
fc(Hz)
B(
f c )
88.5 177 265.5 3540
0.5
1(b)
fx(Hz)
B(
f x )
fo
2x
3x
4x
Kurtogram based Detector
MSB based Detector
SNR=-50dB
R2 R3
Kurtogram based Detector
MSB Robust Detector
4. Application case studies
Dynamic
brake
Supporting
bearing
Flexible
coupling
Vibration
sensor
Supporting
bearingAC motor Shaft encoder
The first real application: motor bearing fault detection
20Introduction The MSB detector Application case studiesSimulation study Conclusions
4. Application case studies
Electric motor bearing with a small seeded outer race fatigue defect (defect simulated by EDM)
Parameter Measurement
Pitch Diameter 46.4mm
Ball Diameter 9.53mm
Ball Number 9
Contact Angle 0˚
Specification of NSK Type 6206ZZ deep groove ball bearing
21Introduction The MSB detector Application case studiesSimulation study Conclusions
4. Application case studies
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-0.05
0
0.05
(a)
Time (s)
Am
pli
tud
e(m
s-2)
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 100000
0.5
1
1.5
x 10-3
(b)
Frequency (Hz)
Am
pli
tud
e(m
s-2)
R2
R1
22Introduction The MSB detector Application case studiesSimulation study Conclusions
The motor data reveals 2 clear resonant regions
4. Application case studies
23Introduction The MSB detector Application case studiesSimulation study Conclusions
1000 2000 3000 4000 5000 6000 7000 8000 90000
0.5
1x 10
-7
(a)
fc(Hz)
B(
f c )
89.33 178.7 2680
0.5
1
fx(Hz)
B(
f x )
(b)
1f c
age
3f c
age
5f c
age
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
MSB Robust Detector
89.33 178.7 2680
0.5
1
Frequency(Hz)
Norm
ali
sed
(c)
1f c
age
3f c
age
5f c
age
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
Bw= 625Hz, fc=2187.5Hz Kurtogram based Detector
R2
R1
Motor bearing fault detection capabilityThe slices of the first three highest peaks are selected for calculating the MSB detector
Clear 1x fo showing outer race fault
Possible cage damage also apparent
of
2 of
Clear 1x fo showing outer race fault
1000 2000 3000 4000 5000 6000 7000 8000 90000
0.5
1x 10
-7
(a)
fc(Hz)
B(
f c )
89.33 178.7 2680
0.5
1
fx(Hz)
B(
f x )
(b)
1f c
age
3f c
age
5f c
age
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
MSB Robust Detector
89.33 178.7 2680
0.5
1
Frequency(Hz)
No
rma
lise
d
(c)
1f c
age
3f c
age
5f c
age
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
Bw= 625Hz, fc=2187.5Hz Kurtogram based Detector
R2
R1
4. Application case studies
24Introduction The MSB detector Application case studiesSimulation study Conclusions
The second real application - planetary gearbox bearing fault detection
DC
Generator
Planetary
gearbox
Helical
gearboxMotor
Vibration
sensor
Specifications of deep groove ball bearing (SKF 6008)
Parameter Specification
Ring gear teeth number 62
Planet gear (3) teeth number 26
Sun gear teeth number 10
Transmission ratio 7.2
Maximum torque 670 Nm
Maximum input speed 2800 rpm
Maximum output speed 388 rpm
Planetary gearbox specifications (David Brown)
Parameter Diameter (mm)
Pitch circle 54
Ball 7.9
Ball number 12
Contact angle 0
4. Application case studies
Planetary gearbox specification
Bearing
positionSensor
position
Input Output
25Introduction The MSB detector Application case studiesSimulation study Conclusions
Inner race fault
Bearing
positionSensor
position
Input Output
4. Application case studies
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-100
0
100
Time (s)
Am
pli
tude(
ms-2
)
(a)
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 100000
0.2
0.4
Frequency (Hz)
Am
pli
tude(
ms-2
)
(b)R4
R1 R2 R3
26Introduction The MSB detector Application case studiesSimulation study Conclusions
Planetary gearbox time data and spectrum Note presence of high levels of random noise, in addition to impacts
In this case, there are 4 resonance regions
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-100
0
100
Time (s)
Am
pli
tud
e(m
s-2)
(a)
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 100000
0.2
0.4
Frequency (Hz)
Am
pli
tud
e(m
s-2)
(b)R4
R1 R2 R3
But when choosing the resonant region(s) to use in the calculation of the MSB-SE, it is always wise to check the coherence…
4. Application case studies
27Introduction The MSB detector Application case studiesSimulation study Conclusions
MSB-SE coherence for the planetary gearbox
R1 (1kHz-2kHz)
R2 (around 4kHz)
R3 (around 6kHz)
R4 (around 9kHz)
Although resonance R4 has high MSB amplitude, it has low coherence and so it is excluded from the calculations of the MSB-SE detector
1000 2000 3000 4000 5000 60000
1
2
3x 10
-3
(a)
fc(Hz)
B(
f c )
65.17 130.3 195.50
0.5
1(b)
fx(Hz)
B(
f x )
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
MSB Robust Detector
65.17 130.3 195.50
0.5
1(c)
Frequency(Hz)
No
rma
lise
d
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
Bw= 156.3Hz, fc=5703.1Hz Kurtogram based Detector
R1
R2 R3
4. Application case studies
Choice of slices and performance evaluation
28Introduction The MSB detector Application case studiesSimulation study Conclusions
Clear 1x fi showing inner race fault
The slices of the first three highest peaks are selected for calculating the MSB detector.
No clear inner race fault feature
1000 2000 3000 4000 5000 60000
1
2
3x 10
-3
(a)
fc(Hz)
B(
f c )
65.17 130.3 195.50
0.5
1(b)
fx(Hz)
B(
f x )
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
MSB Robust Detector
65.17 130.3 195.50
0.5
1(c)
Frequency(Hz)
Norm
ali
sed
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
Bw= 156.3Hz, fc=5703.1Hz Kurtogram based Detector
R1
R2 R3
1000 2000 3000 4000 5000 60000
1
2
3x 10
-3
(a)
fc(Hz)
B(
f c )
65.17 130.3 195.50
0.5
1(b)
fx(Hz)
B(
f x )
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
MSB Robust Detector
65.17 130.3 195.50
0.5
1(c)
Frequency(Hz)
Norm
ali
sed
1frs
1fsf
2fsf
3fsf
4fsf
5fsf
6fsf
1f o
1f i
1f b
2f o
2f i
2f b
3f o
3f i
3f b
1f c
age
3f c
age
5f c
age
Bw= 156.3Hz, fc=5703.1Hz Kurtogram based Detector
R1
R2 R3
Conclusions
29Introduction The MSB detector Application case studiesSimulation study Conclusions
• The MSB is demonstrably effective in suppressing noise and decomposing the nonlinear modulation components
– The MSB-SE is effective in the suppression of both stationary white noise and aperiodic impact impulses.
• Simulated signal and real data studies shows that the capability of the MSB-SE exceeds that of a kurtogram-based detector.
• The application to signals from a planetary gearbox shows that the new approach can successfully detect bearing faults in circumstances where no other method is able to do so.