ault diagnosis in rotating machines using

advertisement
AULT DIAGNOSIS IN ROTATING MACHINES
USING VIBRATION MONITORING AND
ARTIFICIAL NEURAL NETWORKS
By
K.S. SRINIVASAN
INDUSTRIAL TRIBOLOGY, MACHINE DYNAMICS AND
MAINTENANCE ENGINEERING CENTRE
Submitted
fulfilment of the requirements for the degree of Doctor of Philosophy
to the
INDIAN INSTITUTE OF TECHNOLOGY, DELHI
INDIA
AUGUST, 2002
CERTIFICATE
This is to certify that the thesis entitled "FAULT DIAGNOSIS IN
ROTATING MACHINES USING VIBRATION MONITORING AND
ARTIFICIAL NEURAL NETWORKS"
which is being submitted by
Mr.KS.Srinivasan to the Indian Institute of Technology, Delhi for the award of the
degree of Doctor of Philosophy in Industrial Tribology, Machine Dynamics and
Maintenance Engineering Centre, is a record of bonafide research work carried out
by him. He has worked under our guidance and supervision and has fulfilled the
requirements for the submission of this thesis, which is to our knowledge, has attained
the standard required for PhD. degree of this institute. The results presented in this
thesis have not been submitted elsewhere for the award of any other degree or
diploma.
P
piA/4/
( Dr.G.& Yadava )
( Dr.B.0 Nakra )
Chief Design Engineer (SG)
Professor
Dedicated to my beloved Parents
And
Respectable Teachers
Acknowledgements
I would like to sincerely express my deep sense of gratitude and indebtedness
to Prof.B.C.Nakra and Dr.G.S.Yadava for their valuable guidance and
encouragement throughout this work.
I immensely thank Prof. V.P.Agarwal, Head, ITMMEC for his encouragement
during the course of this work.
I express my sincere thanks to the Principal, PES College of Engineering,
Mandya for sponsoring me under Quality Improvement Programme.
I extend my gratitude to Mr. Ashok Kumar and Paramanand Jha of Machine
Dynamics Laboratory for all their help during experimentation. My thanks are also
due to Mr. J.C. Tuteja for his help in the preparation of the drawings.
I express my indebtedness to Mr. Saravanan and friends, for all their help. I
would also like to express my sincere thanks to my co-research scholars Mr.Rajesh
Dwivedi, Mr. B.C.Sharma, Mr. K.M. Ramakrishna Mr. John Rajesh and Mr.
A.P.Harsha for their courtesy and constant interaction.
I express my heartfelt thanks to my wife Mrs. Mala Srinivasan and all my
family members for all their co-operation, understanding and help during the course
of my work.
Abstract
Efficient maintenance is the key to optimum reliability and economy of
modern industrial process plants. Reduction of equipment down time is essential for
increasing the efficiency of the plants. Condition monitoring helps in reducing the
machine downtime. Vibration monitoring is a useful technique for application to
rotating machines and provides valuable information regarding symptoms of
machinery failures. This may avoid costly breakdowns. There is also a need for a
diagnostic system based on artificial neural networks to diagnose the rotating
machinery faults.
The present work involves the simulation of rotor faults and studying their
effects on the frequency components of the vibration signals. The faults like parallel
misalignment, angular misalignment, combined parallel and angular misalignment,
unbalance, mechanical looseness, combined unbalance and misalignment, rub,
bearing clearance, crack and combined crack and unbalance have been simulated. An
artificial neural network system has been applied for quantifying and classifying the
rotor faults using frequency domain data.
To simulate the faults two test rigs with two different lower critical speeds ale
used. In rotor rig 1, the faults like parallel misalignment, angular misalignment,
combined parallel and angular misalignment; unbalance, mechanical looseness,
combined unbalance and misalignment have been simulated. In rig 2, the other faults
like rub, bearing clearance, crack and combined crack and unbalance have been
simulated. The frequency analysis of the vibration signatures due to these faults has
been carried out. For the case of rub, the effect has also been studied on the shape of
orbits of shaft center motion. The experimental simulation studies of the rotor faults
have shown that the faults affect the magnitude of the various harmonics of vibration
ii
signatures significantly. Further, the shapes of orbits in the case of rub are also
affected.
The MATLAB neural network toolbox with multi-layer feed forward back
propagation network has been applied to diagnose the rotating machinery faults using
frequency domain data as input to train the network. The faults like parallel
misalignment, angular misalignment, combined parallel and angular misalignment,
unbalance, crack, and combined crack and unbalance have been quantified using a
three-layer neural network. Further, the faults like parallel misalignment and
mechanical looseness have been classified. The networks have been trained and
tested with combinations of different error goals and different number of neurons,
using the frequency domain data generated from the experiments conducted for
various rotor faults.
The quantification of faults carried out using artificial neural network has
revealed that faults like parallel misalignment, angular misalignment, combined
parallel and angular misalignment, unbalance, crack, combined crack and unbalance,
can be effectively carried out by the combinations of different error goals and
different number of neurons. The classification of parallel misalignment and
mechanical looseness has also been carried out . In cases of both quantification and
classification of faults, the training and testing of the neural networks, have been
demonstrated with reasonable success.
iii
CONTENTS
Acknowledgement
Abstract
ii
Contents
iv
List of Figures
vii
List of Tables
xv
Chapter —1 Introduction and Literature Review
1.1
Introduction
1
1.2
Literature Review
3
1.3
Present Work
22
Chapter — 2
Experimental Set-up and Measurements Methodology Used
2.1
Introduction
25
2.2
Instrumentation
25
2.2.1
2.3
2.4
2.4
Experimental set up and Measurement Methodology Used 25
a)
Test Rig - 1
25
b)
Test Rig - 2 (Bently - Nevada Rotor kit)
27
29
Studies on Rig 1
a)
Parallel Misalignment
29
b)
Angular Misalignment
30
c)
Combined Parallel and Angular Misalignment
30
d)
Unbalance
30
e)
Combined Unbalance and Misalignment
30
f)
Mechanical Looseness
30
Studies on Rig 2
31
a)
Rotor Rub
31
b)
Bearing Clearance
31
c)
Crack in Rotor
31
d)
Combined Crack and Unbalance
32
32
Conclusions
iv
Chapter — 3 Simulation of Faults and Vibration Measurements
3.1Introduction
33
3.2Simulation of Faults on Test Rig 1
33
3.2.1 Parallel Misalignment
33
3.2.2 Angular Misalignment
37
3.2.3 Combined Parallel and Angular Misalignment 42
3.2.4 Unbalance
46
3.2.5 Combined Unbalance and Misalignment
46
3.2.6 Mechanical Looseness
52
3.3Simulation of Faults on Bently Nevada Rotor Kit
55
3.3.1 Rotor Rub
55
3.3.2 Bearing Clearance
66
3.3.3 Crack in Rotor
75
3.3.4 Combined Crack and Unbalance
76
3.4 Conclusions
98
Chapter — 4 Diagnostics using Neural Networks
4.1Introduction
101
4.2Neural Networks
101
4.2.1 Techniques used in Neural Network Applications103
4.3Application of Neural Networks for Diagnostics of Rotor Faults 105
4.3.1 Parallel Misalignment
106
4.3.2 Angular Misalignment
109
4.3.3 Combined Parallel and Angular Misalignment 109
4.3.4 Unbalance
113
4.3.5 Crack in Rotor
120
4.3.6 Combined Crack and Unbalance
120
4.3.7 Classification of Parallel Misalignment and Mechanical -looseness
4.4 Conclusions
128
132
Chapter — 5 Discussion of Results and Conclusions
5.1Discussion of Results
134
5.1.1 Experimental Results
134
5.1.2 Neural networks applications for diagnosis
137
5.2 Conclusions
139
5.3Future scope of work
141
References
142
Bio-data
149
vi
Download