Download Intelligent fault classification of a tractor starter motor using vibration

CONDITION MONITORING
DOI: 10.1784/insi.2010.52.10.561
Intelligent fault classification of a tractor starter motor
using vibration monitoring and adaptive neuro-fuzzy
inference system
Submitted 13.05.10
Accepted 24.08.10
E Ebrahimi and K Mollazade
This paper presents an intelligent method for fault diagnosis
of the starter motor of an agricultural tractor, based on
vibration signals and an Adaptive Neuro-Fuzzy Inference
System (ANFIS). The starter motor conditions to be
considered were healthy, crack in rotor body, unbalancing
in driven shaft and wear in bearing. Thirty-three statistical
parameters of vibration signals in the time and frequency
domains were selected as a feature source for fault diagnosis.
A data mining filtering method was performed in order to
extract the superior features among the primary thirtythree features for the classification process and to reduce
the dimension of features. In this study, six superior features
were fed into an adaptive neuro-fuzzy inference system as
input vectors. Performance of the system was validated by
applying the testing data set to the trained ANFIS model.
According to the result, total classification accuracy was
86.67%. This shows that the system has great potential
to serve as an intelligent fault diagnosis system in real
applications.
Keywords:ANFIS, fault detection, statistical features, tractor,
vibration signal.
Ebrahim Ebrahimi* was born in 1978 in
Kermanshah, Iran, and received BSc, MSc
and PhD degrees in Mechanical Engineering
of Agricultural Machinery from the Urmia
University, Tehran University and Islamic Azad
University (Science and Research Branch),
Iran, in 2001, 2003 and 2007, respectively.
He is currently Assistant Professor and Head
of the Department of Mechanical Engineering
of Agricultural Machinery, Faculty of
Engineering, Islamic Azad University,
Kermanshah Branch, Kermanshah, Iran. His current research interests are
machine condition monitoring, modelling and simulation.
Kaveh Mollazade was born in 1984 in Kurdistan,
Iran, and received BSc and MSc degrees
in Mechanical Engineering of Agricultural
Machinery from the Urmia and Tehran
University, Iran, in 2007 and 2009, respectively.
He is now a PhD student in the Department of
Agricultural Machinery Engineering, Faculty
of Agricultural Engineering and Technology,
University of Tehran, PO Box 4111, Karaj
31587-77871, Iran. His research fields include
artificial intelligence, machine vision, condition
monitoring, postharvest engineering and NDT.
*Corresponding author. Email: [email protected]
Insight Vol 52 No 10 October 2010
1. Introduction
Because of the increasing demand for higher performance as
well as for increased safety and reliability of dynamic systems,
fault diagnosis has been becoming more important for machine
monitoring. Early diagnosis of machine faults while the machine is
still operating in a controllable region can help to avoid abnormal
event progression, which in turn can help to avoid major system
breakdowns and catastrophes. Hence, fault diagnosis is a major
research topic attracting considerable interest from industrial
practitioners as well as academic researchers(1).
One of the most common applications of condition monitoring
is fault diagnosis of electrical machines(2-5). Even though motor
current analysis has been widely utilised for electric machines,
vibration monitoring is also accepted for diagnosis of faults in
these machines(6). Vibration monitoring of electrical machines has
become an attractive field for many researchers and has also gained
industrial acceptance, since it is related to almost all machinery
failures and it does not require modification of the machine or access
to the supply lines(7-9). There are several fault types, mechanical and
electrical, which can induce undesired vibration levels in electrical
motors, such as misalignment, broken rotor bar, short circuits,
imbalance, stator winding faults and bearing failures(6).
During the last decade, a number of attempts have been made
to diagnose machine faults using artificial intelligence techniques
such as: Fuzzy Inference Systems (FISs) for external gear pumps(10),
railway wheels(11) and DC motors(12); Artificial Neural Networks
(ANN) for automotive generators(13), internal combustion engines(14)
and gearboxes(15); and Genetic Algorithms (GA) for rolling element
bearings(16) and so on. Other than these techniques, adaptive systems
have been used for intelligent fault classification. Nowadays, adaptive
neuro-fuzzy inference systems have found a wide gamut of industrial
and commercial applications that require analysis of uncertain
and imprecise information. ANNs and FISs are complementary
technologies in the design of adaptive intelligent systems. The
integrated neuro-fuzzy system combines the advantages of ANN and
FIS. While the learning capability is an advantage from the viewpoint
of FIS, the formation of a linguistic rule base is an advantage from
the viewpoint of ANN. An integrated neuro-fuzzy system shares data
structures and knowledge representations(17).
In modern agriculture, it is important that farming works
be carried out in an appropriate time. Time is a vital factor
in agricultural systems and has a great influence on the total
performance of the farm. It is thus necessary that machines be in
service at, and for, the required time. Now the starter motor is one
of the critical components of a tractor. If the starter motor does not
work properly, the tractor engine cannot operate. Therefore, this
leads to a downtime in the farming schedule. Hence, the present
study tries to introduce a technique for intelligent fault diagnosis of
a tractor starter motor using acquired vibration signals and ANFIS.
Results of this study help agricultural workshop technicians to find
the faults of the tractor starter motor without dismantling it.
561
2. Proposed system for fault diagnosis
The classical way for detecting faults consists of checking the
measurable variables of a system in regard to a certain tolerance of
the normal values and triggering alarm messages if the tolerances
are exceeded, or taking appropriate action when they exceed a limit
value which signifies a dangerous process. In this research, we
want to present an intelligent fault diagnosis system so that it helps
us to give a rapid decision on machine structural health, without the
need for expert analysis.
In the present study, vibration signals are utilised for detecting
the faults of a tractor starter motor. The proposed system consists
of four procedures, as shown in Figure 1: data acquisition, signal
processing, feature extraction and fault classification. These are
specifically explained in the next sections. In this section, the
summary role of each procedure is described as follows:
q Data acquisition: this procedure is used to attain the vibration
signals.
q Signal processing: this includes transfer of data from time
domain into frequency domain.
q Feature extraction: the most significant features are calculated
using some feature parameters from both time and frequency
domains.
q Fault classification: the data obtained from feature extraction
section is fed into ANFIS. The results obtained from the data
test set indicate the total classification accuracy of ANFIS.
3. Data acquisition
Experiments were carried out on the starter motor of a Massey
Ferguson 285 tractor. This starter is an electric motor needed to
turn over the tractor engine to start it.
The starter consists of a very powerful DC electric motor and
a starter solenoid that is attached to the motor. When current from
the starting battery is applied to the solenoid, it pushes out the drive
pinion on the starter driveshaft and meshes the pinion with the ring
gear on the flywheel of the engine(18).
The experimental set-up is shown in Figure 2. A piezoelectric
accelerometer, type VMI 102 (VMI Ltd, Sweden), was mounted
on the starter motor body in the horizontal direction. Through
signal conditioners, the vibration data was acquired by an APC
40 Spectrum Analyser (A/D converter, APC Ltd, Korea) and Dell
Vostro 1320 laptop (data acquisition unit). The rotational speed
of the central shaft of the motor was evaluated using a contact
tachometer (DT-2235B model, Lotron Ltd, Taiwan). Vibration data
was acquired when the motor reached its maximum speed.
Vibration data of the motor in the good condition (healthy)
was used for comparison between healthy and faulty conditions
of motor. Considered faults were healthy motor, with crack in
rotor body (CRB), unbalancing in driven shaft (UDS) and wear
in bearing (WB), as shown in Figure 3. An unbalancing effect was
created by glueing three nuts to the outer body of the driven shaft
ring. Table 1 shows the description of fault conditions.
Table 1. The description of faulty starter motor
Fault condition
Fault description
Crack in rotor body (CRB)
Number of broken bars: 27
Unbalancing in driven shaft (UDS)
Unbalancing mass: 3 nuts × 2.6 gr
Wear in bearing (WB)
Increase of internal diameter: 1.6%
4. Signal processing
One of the common procedures to generate useful features is signal
transition from the time domain (for example, peak values) into
the frequency domain. A 1024-point Fast Fourier Transform (FFT)
is computed from each discrete time signal. Also, Power Spectral
Density (PSD) and FFT phase angle of vibration signals were
calculated using Matlab R2009a. Figure 4 shows an example of
the time domain signal, computed FFT amplitude, PSD and phase
angle, respectively, for different faults of the starter motor. The
FFT analysis produced 1024 sample data for each fault. Due to
even (odd) symmetry in PSD (phase), these features are halved.
Also, since PSD has FFT amplitude information in itself, it was not
considered further(19).
Figure 1. Proposed system for fault diagnosis
Figure 2. The experimental set-up
562
Figure 3. Starter motor faults. Top left: crack in rotor body; Top
right: unbalancing in driven shaft; Bottom: wear in bearing
Insight Vol 52 No 10 October 2010
Figure 4. Typical vibration signals of starter motor faults. From top to bottom: time domain, FFT magnitude, PSD and FFT phase
angle
5. Feature extraction
A fault in rotating machinery leads to a change of the time domain
signal. Both its amplitude and distribution may be different from
those of a time domain signal in a healthy condition. Also, the
frequency spectrum and its distribution may change, which signifies
that new frequency components may appear. In the present study,
the time domain data set is divided into some signals of 1024 data
points. On the other hand, data points of signals in the frequency
domain, ie PSD and FFT phase angle, amounted to 512. These
signals were processed to extract thirty-three feature parameters.
The eleven parameters (T1–T11) are time domain statistical
characteristics and the remaining parameters (P1-P11 and A1-A11)
are frequency domain statistical characteristics(10,20). These features
are shown in Table 2. Since the number of features (33) is large as
input for ANFIS, feature extraction was carried out using a data
mining technique in order to select the most significant features. To
this end, an attribute selection filter of Weka software was used(21).
After filtering, a huge reduction in features was observed. According
to the results of this data mining filtering, only six features were
most significant for fault classification. These features are T1, T4,
T10, P2, A5 and A7.
6. ANFIS structure
An architecture of a fuzzy system with the aid of neural networks
was used to make an intelligent decision for starter motor faults.
The neuro-fuzzy system combines the learning capabilities of neural
networks with the linguistic rule interpretation of a fuzzy inference
system. Fuzzy systems are suitable for uncertain knowledge
representation, while neural networks are efficient structures capable
of learning from examples. The hybrid technique brings the learning
capability of neural networks to the fuzzy inference system.
Insight Vol 52 No 10 October 2010
The parameters associated with the membership functions of
a Sugeno-type FIS will change through the learning algorithm
of the neural network. The computation and adjustment of these
parameters are facilitated by a gradient vector, which provides a
measure of how well the FIS is modelling the input/output data for
a given set of parameters. From the topology point of view, ANFIS
is an implementation of a representative fuzzy inference system
using a back propagation (BP) neural network-like structure. Figure
5 shows the topology of ANFIS with q node for each input, which
consists of five layers. A description of each layer follows(22):
q Layer 1 – In the first layer each node corresponds to one linguistic
term. The number of linguistic terms is determined by the expert
of problem domain. In this layer for i = 1, 2, 3,…, P; xi denotes
the ith input of ANFIS and Oi1 is the output of node i. Here
there is a node function where its rule is equal with that of fuzzy
membership functions. ANFIS uses either back propagation or
a combination of least squares estimation and back propagation
for membership function parameter estimation:
Oi1 = M i (xi )
q Layer 2 – The output of every node in this layer, which is the
product of all incoming signals, represents the firing strength of the
reasoning rule. Each rule represents one fuzzy logic rule. Here, to
calculate the output of the layer, AND (min) operation is used:
Oi2 = M i (xi )ANDM j (x j )
q Layer 3 – Comparison between firing strength of the rules and
the sum of all firing strength is done in this layer. The output of
this layer is the normalised firing strength:
Oi3 =
Oi2
∑ Oi2
i
563
Table 2. Time and frequency domain features
Time domain features
Frequency domain features
PSD
N
K
∑ x(n)
T1 =
n==1
2
n=1
T4 = max x(n)
P5 =
T4
P6 =
N
1
∑ x(n)
N n=1
T4
T3
T8 =
P8 =
n(n − 1)
1
∑ s(k)
K k =1
P4
P3
N
k =1
P10 =
N
n
x(n) − T1 3
(
)
∑
(n −1)(n − 2) n=1
T8
where x(n) is a signal series for
n = 1, 2, …, N
N is the number of data points
P11 =
j =1
Sample variance
∑ (i( j) − A )
Kurtosis
A10 =
K × (P9 )2
K
k
s(k) − P1 3
(
)
∑
(k − 1)(k − 2) k =1
P8
i=1
q Layer 5 – Defuzzification process is occurred in this layer and
the outputs of layer 4 are aggregated:
4
1
k =1
p
Sample standard deviation
j( j − 1)
J
Oi4 = Oi3 ∑ Pj x j + c j
i
J
j =1
4
where s(k) is a signal series for
k = 1, 2, …, K
K is the number of data points
Oi5 = ∑ Oi4
Crest factor
A4
A3
A9 = (A8 )2
q Layer 4 – This layer is a consequent layer and implements
the Sugeno-type inference system, ie a linear combination
of the ANFIS input variables (x1, x2,…, xp) plus a constant
term (c1, c2,…, cp) from the output of each fuzzy rule.
The output of the node is a weighted sum of these
intermediate outputs. In the following output, parameters
p1, p2,…, Pp and c1, c2,…, cp are referred to as the consequent
parameters:
564
A8 =
1
N × (T9 )2
1
∑ s( j)
J j =1
j ∑ i 2 ( j) − (∑ i( j))2
K
n=1
Impulse factor
A4
J
J
∑ (s(k) − P )
4
1
1
∑ s( j)
J j =1
A7 =
K
Shape factor
J
A6 =
P9 = (P8 )2
∑ (x(n) − T )
T11 =
P4
K
k(k − 1)
T9 = (T8 )2
T10 =
1
∑ s(k)
K k =1
k =1
Peak value
A3
A5 =
k ∑ s 2 (k) − (∑ s(k))2
n=1
n=1
P3
K
N
N
n∑ x 2 (n) − (∑ x(n))2
J
A4 = max i( j)
K
P7 =
2
j =1
A3 =
K
P4 = max s(k)
T3
1 N
∑ x(n)
N n=1
T7 =
k =1
P3 =
N
Root mean square (RMS)
J
∑ (i( j))
∑ (s(k))
2
T6 =
J −1
K
N
Standard deviation
j =1
A2 =
K −1
∑ (x(n))
T3 =
∑ (i( j) − A1 )2
k =1
P2 =
N −1
J
J
∑ (s(k) − P1 )2
n=1
j ==1
A1 =
K
K
∑ (x(n) − T1 )2
T5 =
∑ i( j)
k ==1
P1 =
N
Mean value
J
∑ s(k)
N
T2 =
Feature description
FFT phase angle
A11 =
j =1
J × (A9 )2
J
j
i( j) − A1 3
(
)
∑
A8
( j − 1)( j − 2) j =1
Skewness
where i(j) is a signal series for
j = 1, 2, …, J
J is the number of data points
7. Fault diagnosis
In order to evaluate the proposed approach, it was applied to the
fault diagnosis of a starter motor. The data set was collected under
different fault categories. The data sets were divided into two
separate data sets – the training data set and the testing data set.
Table 3 shows the detailed description of the data set. According to
the feature extraction results mentioned in Section 5, six superior
features out of 33 features have been selected by the data mining
technique and then presented to the ANFIS classifier. These
six features, consisting of three time domain features and three
frequency domain features, are superior to the others in classifying
the four classes of the starter motor.
The ANFIS classifier was implemented by using the Matlab
software package (Matlab version R2009a with fuzzy logic toolbox).
The training data set was used to train the ANFIS model, whereas the
Insight Vol 52 No 10 October 2010
testing data set was used to verify the accuracy and the effectiveness
of the trained ANFIS model for classification of the four classes of
starter motor fault. ANFIS used six input data sets, including a total
of 2880 training data in 1000 training epochs and the step size for
parameter adaptation had an initial value of 0.01. Figure 6 shows the
topology of ANFIS designed for fault diagnosis. Two Gaussian type
functions were used as a membership function of the input variables.
At the end of 1000 training epochs, the network error (root mean
square error) convergence curve of ANFIS was derived as shown in
Figure 7. From the curve, the final convergence value is 0.162. Also,
the 64 rules were obtained as follows:
Rule1: If (input1 is in1mf1) and (input2 is in2mf1) and (input3 is
in3mf1) and (input4 is in4mf1) and (input5 is in5mf1) and (input6
is in6mf1) then (output is out1mf1) (1)
Rule2: If (input1 is in1mf1) and (input2 is in2mf1) and (input3 is
in3mf1) and (input4 is in4mf1) and (input5 is in5mf1) and (input6
is in6mf2) then (output is out1mf2) (1)
……
Rule63: If (input1 is in1mf2) and (input2 is in2mf2) and (input3 is
in3mf2) and (input4 is in4mf2) and (input5 is in5mf2) and (input6
is in6mf1) then (output is out1mf63) (1)
Rule64: If (input1 is in1mf2) and (input2 is in2mf2) and (input3 is
in3mf2) and (input4 is in4mf2) and (input5 is in5mf2) and (input6
is in6mf2) then (output is out1mf64) (1)
After training, 120 testing data were used to validate the
accuracy of the ANFIS model for classification of the starter motor
faults. The confusion matrix showing the classification results of
the ANFIS model is given in Table 4. The diagonal elements in the
confusion matrix show the number of correctly classified instances.
In the first column, the first element shows the number of data points
belonging to the healthy class and classified by ANFIS as healthy.
The second element shows the number of data points belonging to
the healthy class but misclassified as WB. The third element shows
the number of data points misclassified as CRB and so on.
Figure 5. ANFIS structure(22)
Table 3. Dataset description
Number of
training samples
Number of
testing samples
Operating
condition
Label of
classification
120
30
Healthy
1
120
30
WB
2
120
30
CRB
3
120
30
UDS
4
Figure 6. Topology of ANFIS for fault diagnosis of tractor starter
motor
Table 4. Confusion matrix of testing data
Output/desired
Healthy
WB
CRB
UDS
Healthy
28
0
2
0
WB
0
26
1
2
CRB
2
0
25
3
UDS
0
4
2
25
Sensitivity, specificity and total classification accuracy are
three criteria to determine the test performance of classifiers. These
criteria are defined as:
q Sensitivity: number of true positive decisions/number of actually
positive cases.
q Specificity: number of true negative decisions/number of
actually negative cases.
q Total classification accuracy: number of correct decisions/ total
number of cases.
According to the values of statistical parameters (see Table
5), ANFIS classified sets healthy, WB, CRB and UDS as 93.33,
86.67, 83.33 and 83.33%, respectively. Also, the total classification
accuracy of ANFIS was obtained to be 86.67%.
Insight Vol 52 No 10 October 2010
Figure 7. ANFIS curve of network error convergence
Table 5. The values of classification accuracy criteria
Fault condition
Statistical parameter
Sensitivity
(%)
Specificity
(%)
Healthy
93.33
97.78
WB
86.67
95.56
CRB
83.33
94.45
UDS
83.33
94.45
Total classification
accuracy (%)
86.67
565
8. Conclusion
The aim of this paper is to introduce an intelligent method to
diagnose the fault type of the starter motor of agricultural tractors
accurately and quickly. The vibration data were collected from the
starter motor under different fault categories using a piezoelectric
acceleration sensor and data acquisition system. Statistical
features from the time and the frequency domains were extracted
to reflect different faults of the starter motor. Input vectors to the
ANFIS are six superior features, which were extracted using a
data mining technique. The final ANFIS model has 64 rules with
a network error convergence of 0.162. The trained ANFIS model
was evaluated using 120 testing data and it was observed that the
total classification of this technique is 86.67%. The results show
the applicability and effectiveness of this method to detect faults
in starter motors.
Acknowledgement
This research is supported by Islamic Azad University, Kermanshah
Branch. This assistance is gratefully appreciated.
References
1. Q Yang, ‘Model-based and data driven fault diagnosis methods
with applications to process monitoring’, PhD Thesis, Case
Western Reserve University, 2004.
2. H Mohamadi Monavar, H Ahmadi, S S Mohtasebi
and S Hasani, ‘Vibration condition monitoring techniques
for fault diagnosis of electromotor with 1.5 kW power’,
Journal of Applied Sciences, Vol 8, No 7, pp 1268-1273,
2008.
3. H Ahmadi and K Mollazade, ‘Fault diagnosis of an electropump in a marine ship using vibration condition monitoring’,
Insight, Vol 51, No 8, pp 431-438, 2009.
4. H Ahmadi and K Mollazade, ‘A practical approach to
electromotor fault diagnosis of Imam Khomaynei silo by
vibration condition monitoring’, African Journal of Agricultural
Research, Vol 4, No 4, pp 383-388, 2009.
Ultrasonic Flaw Detection for
Technicians, 3rd Edition
by J C Drury
In the twenty-five or so years since
the first edition of ‘Ultrasonic Flaw
Detection for Technicians’ was
published, there have been a number
of advances in transducer technology
and flaw detection instruments. The
gradual acceptance by industry that
the sizing of weld defects by intensity
drop was not as accurate as had been
claimed led to the development of the
TOFD technique. Modern digital flaw
detectors and computer technology allow far more information
to be stored by the operator. The author thus felt that it was time
to give the book a thorough review and to try to address some
of the advances. The result is this new edition.
Available price £25.00 (Non-Members); £22.50 (BINDT
Members) from The British Institute of Non-Destructive
Testing, Newton Building, St George’s Avenue, Northampton
NN2 6JB, UK. Tel: +44 (0)1604 89 3811; Fax: +44 (0)1604 89
3861; Email: [email protected]
Order online at www.bindt.org
566
5. K S Gaeid and H A F Mohamed, ‘Diagnosis and fault tolerant
control of the induction motors techniques – a review’,
Australian Journal of Basic and Applied Sciences, Vol 4, No 2,
pp 227-246, 2010.
6. H D Bloch and F K Geither, ‘Machinery failure analysis and
troubleshooting’, Gulf Publishing Company, Houston, Texas,
1990.
7. J Undacheck and V R Dodd, ‘Progress and way out of a
machinery surveillance and diagnostic system’, In Proceedings
of the ASME Conference on Pressure Vessel and Piping,
Mexico City, Mexico, September 1976.
8. G K Sing and S A K S Ahmed, ‘Vibration signal analysis using
wavelet transform for isolation and identification of electrical
faults in induction machine’, Electrical Power Systems
Research, Vol 68, No 2, pp 119-136, 2004.
9. I Chindurza, D G Dorrell and C Cossar, ‘Vibration analysis
of a switched reluctance machine with eccentric rotor’, In
Proceedings of the Second International Conference on Power
Electronics, Machines and Drives (PEMD), Vol 2, No 3, pp
481-486, March-April 2004.
10. K Mollazade, H Ahmadi, M Omid and R Alimardani, ‘An
intelligent model based on data mining and fuzzy logic for fault
diagnosis of external gear hydraulic pumps’, Insight, Vol 51,
No 11, pp 594-600, 2009.
11. D Skarlatos, K Karakasis and A Trochidis, ‘Railway wheel
fault diagnosis using a fuzzy logic method’, Applied Acoustics,
Vol 65, No 10, pp 951-966, 2004.
12. L J D Miguel and L F Blázquez, ‘Fuzzy logic-based decision
making for fault diagnosis in a DC motor’, Engineering
Applications of Artificial Intelligence, Vol 18, No 4, pp 423450, 2004.
13. J D Wu and J M Kuo, ‘An automotive generator fault diagnosis
system using discrete wavelet transform and artificial neural
network’, Expert Systems with Applications, Vol 36, pp 97769783, 2009.
14. J D Wu, C K Huang, Y W Chang and Y J Shiao, ‘Fault diagnosis
for internal combustion engines using intake manifold
pressure and artificial neural network’, Expert Systems with
Applications, Vol 37, No 2, pp 949-958, 2010.
15. N Saravanan and K I Ramachandran, ‘Incipient gear box fault
diagnosis using discrete wavelet transform (DWT) for feature
extraction and classification using artificial neural network
(ANN)’, Expert Systems with Applications, Vol 37, No 6, pp
4168-4181, 2010.
16. Y Zhang and R B Randall, ‘Rolling element bearing fault
diagnosis based on the combination of genetic algorithms and
fast kurtogram’, Mechanical Systems and Signal Processing,
Vol 23, No 5, pp 1509-1517, 2009.
17. M A Akcayol, ‘Application of adaptive neuro-fuzzy controller
for SRM’, Advances in Engineering Software, Vol 35, No 3-4,
pp 129-137, 2004.
18. C E Goering and A C Hansen, ‘Engine and tractor power’,
American Society of Agricultural and Biological Engineering,
4th edition, USA, 2004.
19. M Omid, A Mahmoudi and M H Omid, ‘An intelligent system
for sorting pistachio nut varieties’, Expert Systems with
Applications, Vol 36, pp 11528-11535, 2009.
20. Y A Lei, Z He and Y Zi, ‘A new approach to intelligent
fault diagnosis of rotating machinery’, Expert Systems with
Applications, Vol 35, pp 1593-1600, 2008.
21. I H Witten and E Frank, Data Mining: Practical Machine
Learning Tools and Techniques, Morgan Kaufmann Press, 2nd
edition, 560 pages, 2005.
22. S Alavandar and M J Nigam, ‘Adaptive neuro-fuzzy inference
system based control of six DOF robot manipulator’, Journal
of Engineering Science and Technology Review, Vol 1, pp 106111, 2008.
Insight Vol 52 No 10 October 2010
Vibration Monitoring
& Analysis Handbook
This new book is compiled by Simon Mills, a chartered marine
SER IES
BIN DT CM
VI BR AT IO N
MON
S
BINDT CM SERIE
engineer with over 30 years’ experience in practical engineering,
AN ALYS IS
ITOR IN G &
ONITORING
VIBRATION M
and has been developed from his BINDT PCN accredited series of
HANDBOOK
& ANALYSIS
n
training modules for Vibration Analyst Training and Qualification
ring Cer tificatio
dition Monito
BINDT PCN Con
Reference for
An Approved
in accordance with ISO 18436-2. It approaches the subject in
an informal and informative manner, with a selection of clear
diagrams, pictures and tabulated methodologies.
It addresses all levels of the syllabus given in ISO 18436-2 for
OK
LYSIS HANDBO
ITORING & ANA
VIBRATION MON
OOK
The British Institute of Non-Destructive Testing
Condition Monitoring Series
Vibration Condition Monitoring and Diagnostics and the BINDT
specification – General requirements for qualification and
assessment of condition monitoring and diagnostic personnel
(BINDT CM Gen Appendix D for Vibration Analysis), giving
practical advice, examples and case histories.
BINDT has sponsored the publication of this handbook as part
of its portfolio of CM handbooks, to fill in and reference required
areas of knowledge. In conjunction with required ISO, BS and
MILLS
SIMON R W
other textbook references, it provides a basis for effective training
and accountable qualification and certification as required by
ISO 18436.
66).
(Reg No 2606
.
The Handbook (A4 size) is printed on high-quality paper with a durable matt finish.
ISBN 978-0-903132-39-7.
Produced and published by The British Institute of Non-Destructive Testing on behalf of its Condition Monitoring Group (COMADIT).
#
ORDER FORM
Vibration Monitoring & Analysis Handbook
Name:............................................................................................................................................................
Please supply ........................... copies @
Organisation: ...........................................................................................................................................
q
£85.00 each plus postage and packing* (BINDT Member)
q
£90.00 each plus postage and packing* (BINDT Non-Member)
............................................................................................................................................................................
METHOD OF PAYMENT
I enclose cheque/International Money Order for £......................................................
made payable to The British Institute of NDT.
Please charge £................................................ to my Visa/Mastercard/Amex/etc
Card no:
__/__/__/__
__/__/__/__
Address: ......................................................................................................................................................
............................................................................................................................................................................
__/__/__/__
__/__/__/__
Expiry date: __/__ Start date (Maestro): __/__ Security No: __/__/__
Signature......................................................................................................................................................
If the name and address on your credit card differ from the delivery
address, please state separately.
Postcode: ....................................................................................................................................................
Country: ......................................................................................................................................................
Tel: ..................................................................................................................................................................
Fax: .................................................................................................................................................................
E-Mail: .........................................................................................................................................................
If you are registered for VAT or a local sales tax, please provide your number to
comply with EC regulations.
*Orders with cheque and credit orders are sent carriage free within the UK.
Postage and packing for other orders will be invoiced according to weight.
Reply to: The British Institute of Non-Destructive Testing, Newton Building, St George’s Avenue, Northampton NN2 6JB, UK.
Tel: +44 (0)1604 89 3811 Fax: +44 (0)1604 89 3861 E-mail: [email protected]
VAT Reg. No. 581 2105 68
Enquiry No 010-10