Professional Documents
Culture Documents
Department of Mechanical, Aerospace and Nuclear Engineering, Jonsson Engineering Center, Rm. 2049, 110 8th Street, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA
Smart Engineering Asset Management Laboratory, City University of Hong Kong, Hong Kong
c
Faculty of Electrical and Computer Engineering, University of Tabriz, Iran
d
Faculty of Mechanical Engineering, University of Tabriz, Iran
b
a r t i c l e
i n f o
Keywords:
Condition monitoring
Signal processing
Vibration signal
Fault diagnosis
Neural network
Genetic algorithm
Mother wavelet
Gear
Piecewise cubic hermite interpolation
Daubechies
a b s t r a c t
This paper presents an optimized gear fault identication system using genetic algorithm (GA) to investigate the type of gear failures of a complex gearbox system using articial neural networks (ANNs) with
a well-designed structure suited for practical implementations due to its short training duration and high
accuracy. For this purpose, slight-worn, medium-worn, and broken-tooth of a spur gear of the gearbox
system were selected as the faults. In fault simulating, two very similar models of worn gear have been
considered with partial difference for evaluating the preciseness of the proposed algorithm. Moreover,
the processing of vibration signals has become much more difcult because a full-of-oil complex gearbox
system has been considered to record raw vibration signals. Raw vibration signals were segmented into
the signals recorded during one complete revolution of the input shaft using tachometer information and
then synchronized using piecewise cubic hermite interpolation to construct the sample signals with the
same length. Next, standard deviation of wavelet packet coefcients of the vibration signals considered as
the feature vector for training purposes of the ANN. To ameliorate the algorithm, GA was exploited to
optimize the algorithm so as to determine the best values for mother wavelet function, decomposition
level of the signals by means of wavelet analysis, and number of neurons in hidden layer resulted in a
high-speed, meticulous two-layer ANN with a small-sized structure. This technique has been eliminated
the drawbacks of the type of mother function for fault classication purpose not only in machine condition monitoring, but also in other related areas. The small-sized proposed network has improved the stability and reliability of the system for practical purposes.
2008 Published by Elsevier Ltd.
1. Introduction
The application of a condition-based maintenance strategy used
for detection and diagnosis of incipient faults can help us to minimize avoidable costs and impediments caused by the need to perform ad hoc repairs. Gearboxes consisting of three major
components (gears, bearings, and shafts) are considered to be
one of the essential systems on account of its inuential applications in a wide domain of industries (e.g. machine tools and vehicles). As more than half of the failures in gearboxes are due to gear
defects (Yesilyurt, 2004), the gear faults have been considered as a
case study for the proposed technique.
In general, acoustic emission (Wu, Chiang, Chang, & Shiao,
2008) and vibration analysis (Tse, Gontarz, & Wang, 2007) are
two main centers of fault detection and diagnosis systems in
machine condition monitoring. In this paper, vibration signals
* Corresponding author. Tel.: +1 (518) 276 6351; fax: +1 (518) 276 6025.
E-mail address: kraee81@gmail.com (J. Raee).
0957-4174/$ - see front matter 2008 Published by Elsevier Ltd.
doi:10.1016/j.eswa.2008.05.052
DB 4
DB 5
DB 6
DB 7
DB 8
DB 9
4863
Fig. 2. (A) Experimental set-up, (B) slight-worn gear, (C) broken-tooth gear, (D) accelerometer location, (E) slight-worn model, (F) schematic of the gearbox and (G) load
mechanism.
4864
Faultless
100
0
-100
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
Broken-teeth
400
200
0
-200
-400
0.005
0.01
0.015
0.02
0.025
Medium-worn
100
0
-100
0.005
0.01
0.015
0.02
0.025
Slight-worn
100
0
-100
0.005
0.01
0.015
0.02
0.025
Time (Sec)
Fig. 3. Raw vibration signals of four gearbox conditions recorded during one revolution of input shaft.
Faultless
100
0
-100
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
0.03
0.035
0.04
0.045
Broken-teeth
400
200
0
-200
-400
0.005
0.01
0.015
0.02
0.025
Medium-worn
100
0
-100
0.005
0.01
0.015
0.02
0.025
Slight-worn
100
0
-100
0.005
0.01
0.015
0.02
0.025
Time (Sec)
Fig. 4. Synchronized vibration signals of four gearbox conditions recorded during one revolution of input shaft.
4865
Signal
A1
D1
Level 1
AA2
AD2
DD2
DA2
Level 2
Level n
Fig. 5. Approximations and details in wavelet packet analysis.
raw vibration signals which usually dont have the same length
in a complete revolution of the shaft. To overtake the impediment,
piecewise cubic hermite interpolation (Raee et al., 2007) was suggested to synchronize the signals. Afterwards, wavelet packet analysis (Liu, 2005) has been considered to preprocess the
synchronized signals for presenting a distinguished feature vector.
Then, selecting of the mother wavelet function and decomposition
level has been reformed by means of genetic algorithm as an obstacle in this regard. To present a more accurate fault classication
system, number of neurons in hidden layer has been also chosen
to be optimized by GA.
2.1. Experimental set-up
A proper data acquisition process can extensively improve the
sensibility for the detection of failures. Therefore, selecting an
appropriate method is important for acquiring vibration signal
data. The experimental set-up (Raee et al., 2007) to collect dataset
in this research consists of a four-speed motorcycle gearbox containing the oil during acquiring dataset, an electrical motor with
a constant nominal rotation speed of 1420 RPM, a load mechanism,
multi-channel pulse analyzer system, a triaxial accelerometer,
tachometer and four shock absorbers under the bases of test-bed.
The vibration signals were recorded by mounting the accelerometer on the outer surface of the gearboxs case near input shaft of the
gearbox. Ten different congurations were synthesized and tested.
For each conguration, three different fault conditions were tested
that were slight-worn, medium-worn, broken-tooth of gear. For
evaluating the meticulousness of the technique, two very similar
models of worn gear have been taken into account with partial difference. The rotational speed of the system was measured by
tachometer used as a measure to compensate the uctuations owing to uncertainties of the load mechanism which was a constant
static load. Moreover, the signals were also sampled at 16384 Hz
lasting eight seconds. Fig. 2 clearly shows all parts of fault simulator and related components in detail.
2.2. Preprocessing of raw vibration signals
4866
400 (0,0)
200
200
200 (1,0)
(1,1)
-200
-200
-200
200
400
600
100
400
50
200
100
300
-50
300
(2,2)
(2,1)
50
200
0
200
-200
(2,0)
-50
-400
50
100
150
50
100
50
150
100
40 (3,0)
400 (2,3)
100
150
(3,1)
20
200
-20
-100
50
100
20
150
(3,2)
200
40
60
80
100
20
40
60
80
100
20
(3,3)
200
0
-200
-20
0
-200
20
40
60
80
100
20
40
60
80
100
200
50
(3,5)
20
40
60
80
20
0
-100
20
40
60
80
100
60
20
150
40
10
20
-10
-20 (4,0)
100
40
60
80
100
40
60
80
100
(4,2)
50
0
(4,1)
-20
20
60
(3,7)
100
-100
100
40
200
(3,6)
100
-50
(3,4)
20
-50
40
60
20
40
60
Fig. 6a. Wavelet packet coefcients of a sample synchronized broken-tooth signal in four levels.
hk xk1 xk
dk yk1 yk =hk
Assume that dk designate the slope of the interpolant at xk
4867
50
(4,3)
100
-50
-100
20
40
(4,4)
200
0
-200
60
20
50
40
60
40
200
0
-50
20
-200
(4,7)
-400
(4,6)
-100
20
40
60
40
100
(4,10)
(4,9)
-50
20
40
60
20
40
60
20
40
60
20
40
60
(4,11)
-20
40
(4,8)
60
20
20
-20
20
50
(4,5)
-100
60
20
40
60
200
200
(4,12)
(4,13)
100
100
(4,14)
0
0
0
-100
- 200
20
40
-100
60
200
20
40
60
20
40
60
(4,15)
100
0
-100
Fig. 6b. Wavelet packet coefcients of a sample synchronized broken-tooth signal in four levels.
DB
order
Slight-worn
Standard
Deviation
of Wavelet
Packet
Coefficients
Medium-worn
0 1 1 0 1 0 1 1 10 11 0 1
Broken-tooth
Normal Gearbox
Output layer
Input layer
Hidden layer
dk P0 xk
Px
3hs 2s3
3
h
ss h2
h
Number of Decomposition
Neurons
Level
yk1
dk
h 3hs 2s3
h
yk
s2 s h
2
dk1
4
4868
Gearbox
Accelerometer
Tachometer
Number of generation
200
Population
Chromosome length
Selection operator
Fitness normalization
Elitism
Crossover
Mutation
30
12
Roulette
Rank
1
Pc = 0.8, two-point, uniform
Pm = 0.1, Gaussian, mean = 0.0, std = 1.0
Decomposition
Level
Yes
Stopping
Number of
Neurons
Criterion
Mother
Wavelet
Variable name
Range
Optimized value
Daubechies order
Decomposition level
Number of the hidden-layer neurons
DB2DB20
36
1035
DB11
4
14
No
components because it is able to simultaneously impart time and
frequency structures.
Continuous wavelet transform (CWT), a function of two parameters entitled translation and scale, bears a remarkable data
redundancy for the signal or function to be analyzed. By skipping
several steps rather than continuously adjusting two parameters,
in other words, by reducing the volume of calculations, discrete
wavelet transform (DWT) has been dened to analyze the signals
with a smaller set of scales and specic number of translations at
each scale (Soman & Ramachandran, 2004), as efcient as CWT
with identical accuracy. In DWT, the signals are decomposed into
a hierarchical structure of detail and approximations at limited levels as follows:
Selection of parameters by GA
StD of WPC
Training ANN
Testing ANN
Training Time
ANN Performance
Fitness
Function
f t
ij
X
Di t Aj t
i1
where Di(t) denotes the wavelet detail and Aj(t) stands for the wavelet approximation at the jth level.
Wavelet packet analysis (Avci, Turkoglu, & Poyraz, 2005), a
generalization of wavelet decomposition offering a richer range
of possibilities for signal analysis, is based upon DWT. Every wavelet detail component is further decomposed to get their own
domain into timefrequency domain. Wavelet transform (WT) offers a satisfactory time and poor frequency resolution at high frequencies and a satisfactory frequency and poor time resolution at
low frequencies (Peng et al., 2005). It is used to detect transient
0,0
Synchronized Signal
Level 1
1,0
Level 2
2,0
Level 3
Level 4
1,1
2,1
3,0
4,0
4,1
3,1
4,2
4,3
3,2
4,4
2,2
3,3
4,5
4,6
4,7
3,4
4,8
2,3
3,5
4,9
3,6
3,7
4869
(4,0), StD=14.58
(4,1), StD=9.99
20
20
-20
-20
(4,4), StD=50.04
(4,2), StD=10.72
20
20
-20
-20
(4,5), StD=73.18
100
200
-100
- 200
(4,6), StD=13.98
(4,7), StD=58.92
50
100
0
- 100
-50
(4,9), StD=4.36
(4,8), StD=3.22
(4,3), StD=14.26
50
10
10
-10
-10
20
0
0
-20
-50
100
100
-100
50
0
100
-50
-50
Fig. 11. Wavelet packet coefcients in one sample broken-tooth signal in level 4 (decomposition tree and standard deviation of the coefcients in title).
wij;k t 2j=2 wj 2j t k; i 1; 2; 3; . . .
The wavelet functions wj are determined from the following recursive equations:
w2j t
1
p X
2
hkwi 2t k
w2j1 t
1
1
p X
gkwi 2t k
7
8
1
f t
2j
X
fji t
i1
while the wavelet packet component signal fji t are stated by a linear combination of wavelet packet functions wij;k t as follows:
1
X
fji t
cij;k twij;k t
10
k1
cij;k t
1
f twij;k tdt
providing that
orthogonality:
the
n
wm
j;k twj;k t 0 if mn
11
wavelet
packet
functions
satisfy
the
12
4870
(4,0), StD=19.94
(4,1), StD=15.49
(4,2), SD=11.51t
100
(4,3), StD=23.52
100
50
20
50
0
-20
-50
-100
-50
(4,4), StD=67.50
(4,5), StD=157.57
(4,6), StD=41.60
200
500
200
-200
-500
500
200
(4,9), StD=8.54
(4,8), StD=5.58
-500
20
20
(4,7), StD=116.60
100
0
0
-20
-20
-100
(4, 12), StD=54.96
0
-200
200
200
-100
200
100
-200
-100
200
Fig. 12. Wavelet packet coefcients in one sample broken-tooth signal in level 4 (decomposition tree and standard deviation of the coefcients in title).
of training data available and can manage the imprecision of condition monitoring process thanks to adaptability dynamism. Unlike
the classic digital-processing techniques, ANNs possess the ability
to perform parallel processing of data, cope with noisy data and
adapt to different circumstances.
The most popular neural network is the multi-layer perceptron,
which is a feed-forward network and frequently exploited in fault
detection and diagnosis systems, which has found an immense
popularity in machine condition monitoring applications as it constitutes more than 90% (Bartelmus, Zimroz, & Batra, 2003) of the
current ANNs in the eld, and therefore, in this study, has been
used as the base of failure classier system. The feed-forward network rst learns through already existing data by iteratively
changing its weights using the back propagation of error algorithm which is performed by iteratively seeking to minimize the
root-mean-square-error (RMSE) using the gradient search technique (Kwak & Ha, 2004).
An important network design issue is the selection and implementation of the network conguration. In the proposed system,
the number of nodes in hidden layer, usually determined by
trial-and-error methods in several papers, has been optimized
using GA to provide the maximum throughput. Hyperbolic tangent
sigmoid transfer function was applied to all net layers with Resilient Backpropagation training algorithm (Riedmiller & Braun,
1993). A matrix of a data set with 800 processed elements, comprising 5 4 40 sampled data for every four gear conditions
4871
(4,0)
40
30
(4,2)
(4,1)
25
40
20
30
MW
45
NG
MW
NG
15
20
10
10
SW
NG
MW
SW
SW
10
SW
BT
(4,4)
NG
MW
100
BT
NG
SW
40
BT
SW MW
NG
50
20
BT
BT
(4,10)
(4,11)
20
60
SW
NG
10
45
SW
MW
40
NG
SW
30
NG
MW
20
5
MW
(4,7)
NG
MW
50
15
10
BT
100
(4,9)
SW MW
20
SW
60
(4,8)
BT
15
(4,6)
150
SW
25
MW BT
BT
(4,5)
50
15
30
NG
20
75
(4,3)
MW
15
BT
BT
BT
(4,13)
(4,12)
NG
(4,14)
(4,15)
75
SW
90
60
SW
50
MW
60
20
BT
30
75
SW
40
NG
25
NG
NG
MW BT
NG
SW
MW
50
MW
BT
25
BT
Fig. 13. Standard deviation of wavelet packet coefcients for 90 segmented sample signals in each condition (decomposition level: 4, gearbox conditions: normal gearbox,
slight-worn, medium-worn, broken-tooth).
4872
120
BT
100
80
MW
60
SW
40
20
120
BT
100
80
MW
60
SW
40
20
NG
NG
2
10
12
14
16
10
12
14
16
BT
100
80
MW
60
SW
40
20
NG
2
10
12
14
16
120
BT
100
80
MW
60
SW
40
20
NG
2
10
12
14
16
Fig. 14. Standard deviation of WPC of 30 sample signals in each gearbox condition with different Daubechies order (gearbox conditions: normal gearbox, slight-worn,
medium-worn, broken-tooth).
F apbt
13
where a and b are the weights, p and t denote network performance and training time, respectively. In general, the network
performance which is directly related to the accuracy of the network outweighs the training time constraint due to the signicance of the system accuracy and off-line attribute of the
network in the fault detection and diagnosis systems. Therefore,
to optimize the system, network performance has more impact
on tness function than training time; the weights were applied
where a was assumed to be 0.7 vs. 0.3 which was assigned to b.
As the more training time has the more negative impact on the
whole fault identication system negative sign has been considered for b in tness function. In simple words, the network performance and training time have inverse proportion on the tness
function. Depending on different case studies, these values for
constant coefcients a and b may deliberately be selected. For
example, perhaps the training time may not be very important
for a fault identication system having ofine training process.
Accordingly, the weight of training time can verge to be zero in
that circumstance.
To sum up, GA was used to optimally search Daubechies order,
decomposition level of the signals, and the number of neurons in
hidden layer. Fig. 9 depicts the whole scheme of the fault identication system. Tables 1 and 2 explore, respectively, the parameters
and variables of the GA exploited in the research.
3. Results and discussion
In this paper, by running GA, DB11, level 4 and 14 neurons
have been selected as the best values for Daubechies order,
decomposition level, and the number of nodes in hidden layer,
respectively. Decomposition tree of wavelet packet has been depicted in Fig. 10 and related wavelet coefcients of two sample
signals for normal gearbox and broken-tooth gear have been
illustrated in Figs. 11 and 12. With close scrutiny to the amplitude of the wavelet packet coefcients, it has been shown that
there is a considerable correlation between DB11 and the failure
condition (broken-tooth gear). The values of standard deviation
of wavelet packet coefcients in 16 sub-signals are mostly higher than those of the normal one. On the other hand, as shown in
Fig. 13, the uctuations of the standard deviations in different
sample segmented signals are considerable and this is the most
appropriate factor for training the neural networks. Stability and
convergence of the network are directly dependent on the distinguished feature vector with enough samples in ample ranges. As
a matter of fact, the range of the samples in feature vector is
more important than the number of samples in training that
4873
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Wavelet function
order
Decomposition
level
No. of hidden-layer
neurons
Samples for
testing
Slightworn
Mediumworn
Brokentooth
Faultless
Best net
structure
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
3
3
3
3
3
3
4
4
4
4
4
4
5
5
5
5
5
5
13
15
13
13
15
15
17
19
17
14
15
14
24
25
25
23
23
23
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
94
93.33
95.66
98.53
97.33
93.33
94.33
92
100
100
100
97.86
97.33
90.66
96
100
100
90.66
90
92.66
93.33
97.66
96.06
95.33
100
98.46
98.66
100
100
100
96.86
92
100
100
96
100
98.06
97.13
97.33
99.33
98.2
98.06
99.66
99.13
100
100
100
99
100
99.4
100
100
100
100
96.66
92.33
95.66
97.66
96.73
99
96
94.66
93.33
100
100
100
98.66
97.53
98.66
100
100
94.66
8:10:4
8:12:4
8:10:4
8:10:4
8:12:4
8:12:4
16:14:4
16:16:4
16:14:4
16:11:4
16:13:4
16:14:4
32:21:4
32:22:4
32:22:4
32:21:4
32:23:4
32:23:4
8
9
10
11
12
13
8
9
10
11
12
14
8
9
10
11
12
13
Table 4
Various arrangements of the three optimized parameters with a constant number of neurons and the related results
1
2
3
4
5
6
7
8
9
10
11
12
Wavelet function
order
Decomposition
level
No. of hidden-layer
neurons
Samples for
testing
Slightworn
Mediumworn
Brokentooth
Faultless
Net
structure
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
DB
4
4
4
4
4
4
5
5
5
5
5
5
14
14
14
14
14
14
14
14
14
14
14
14
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
5 4 75
97
96.33
98.66
100
100
96
89.66
94.33
95.06
95
95.66
93.66
93.33
97.33
100
100
100
100
91
89.66
97
98.66
94
91.66
100
100
100
100
100
100
96.66
95.33
98.66
96.66
96.33
96.66
98.66
100
100
100
100
100
95
96.66
100
98
96.66
94
16:11:4
16:11:4
16:11:4
16:11:4
16:11:4
16:11:4
32:11:4
32:11:4
32:11:4
32:11:4
32:11:4
32:11:4
8
9
10
11
12
13
8
9
10
11
12
13
4874
Segmentation
Raw signal
Synchronization by P.C.H.I.
Selection
Crossover
Generation
Mutation
0
1
1
0
1
0
1
1
0
1
0
1
0
1
1
0
1
0
1
1
0
1
0
1
DB order
No. of
Neurons
F =. p+ . t
Decomposition
Level
Genetic Algorithm
Fitness Function
No
Stopping
Criteria Fulfilled
Yes
END
Fig. 15. Schematic of the entire proposed algorithm.
4. Conclusion
An intelligent gear fault identication system was developed
and implemented to pinpoint the gearbox faults; Three parameters
were recognized to be of paramount signicance to be optimized
using GA i.e. Daubechies wavelet function order, decomposition level, and number of neurons in hidden layer of ANN, which play a
key role in size and performance of the network and the soundness
of the feature vector. The Feature vector was also obtained from
optimized standard deviation of wavelet packet coefcients acquired from DB11 at fourth level of decomposition following a
piecewise cubic hermite interpolation for synchronizing. Eventually, a MLP network with well-formed and optimized structure
(16:14:4) and remarkable accuracy was presented providing the
capability to identify slight-worn, medium-worn and broken-tooth
of gears faults perfectly in 100% of the ve tested congurations of
gearbox. By the proposed system which has been elaborately depicted in Fig. 15, the drawback of implementing mother wavelet
function through trial-and-error based methods has been also improved for fault classication purposes.
Acknowledgements
This work described in this paper was partially supported by
Vibration and Modal Analysis Lab at University of Tabriz and partially supported by a grant from the Research Grants Council of
Hong Kong Special Administrative Region, China (Project no. CityU
120506). The authors would like to write in memoriam of two dedicated mentors, Professor Vahhab Pirouzpanah and Professor Samad Nami Notash who devoted more than four decades to
enlighten a myriad of students. They would also like to thank the
4875