Академический Документы
Профессиональный Документы
Культура Документы
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
*
Professor, I&C Deptt., D.D.I.T., D.D.U, India
www.ijert.org 1
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
are suitable for on-line monitoring and fault implement the incipient fault detection scheme
detection purposes [5]. no modelling of the machine is necessary. As in
[6] With proper monitoring and fault detection
For the investigation in the area of fault schemes, the incipient faults can be detected;
diagnosis and protection some important issues thus, maintenance and down time expenses can
are be reduced while also improving safety. Many
of the conventional methods used to determine
1) A comprehensive noninvasive protection these faults are either very expensive to
scheme which covers all external and internal implement, performed off-line, require the need
faults and also provide operator heuristic of an expert, or impractical for small machines,
knowledge incorporation depend upon the ANN have been proposed and have
condition or trending. demonstrated the capability of solving the motor
2) Applicability of artificial intelligence to solve monitoring and fault detection problem using an
the problem like stalling, excessive start time, inexpensive, reliable, and noninvasive procedure
frequent start etc; An overview of feedforward nets and the
backpropagation training algorithm, along with
3) Quantitative fault detection in order to state their respective pseudocodes, and a general
an absolute threshold, independent of operating methodology for the design of feedforward
conditions [45, 27] artificial neural networks to perform motor fault
detection is discussed in [5]. Neural network
4) Fault detection in inverter-driven motors [45, design considerations such as network
27] performance, network implementation, size of
training data set, assignment of training
Noninvasive Methods which can be utilized for parameter values, and stopping criteria
the fault diagnosis in induction machine discussed and a fuzzy logic approach to
discussed here are classified in three categories, configure the network structure has been
Soft computing methods which include mainly presented to automate the network design in [8].
ANN, Fuzzy logic and expert systems based A Neural Network fault detector composed of
diagnosis, Signal processing based Approaches two parts one for filtering noise and disturbance
and Hybrid approaches. The use of combination while retaining steady state measurements and
of induction motor model and ANN, Model and other to detect fault developed for single phase
fuzzy logic, Neurofuzzzy approach, Model and motor. Neural network trained using
any signal processing method etc; are controllable data sources developed using a
considered here as hybrid approaches. There are computer programme for initial design and
some others approaches also like thermal training of ANN for bearing and stator turn fault
measurements, chemical analysis, axial flux, detection[9]. Feedfoward layered network
noise and vibration monitoring, infrared structure and backpropogation algorithm used in
recognition etc; for the problem solution worked ANN technique applied for identification of
upon. external faults of induction motor. Three phase
current and voltages from the induction motor
2. AI based approaches
are used in approach. RMS value calculated
The Noninvasive Parameter estimation using DFT [10]. Radial basis function utilized to
technique, require mathematical model and train ANN for the detection of bearing and stator
elaborate understanding of system dynamics interturn fault. Instantaneous current and angular
based on system parameters. The parameters are velocity depending upon rotor speed are utilized
usually chosen to reflect the motor conditions in the approach [12]. PC based monitoring and
and usually difficult to obtain. In addition, as in fault detection scheme for 3 phase I.M using
[5] the interpretation of the fault conditions- ANN developed in [11]. A scheme for detection
which is a fuzzy concept using rigorous of stator turn consists of a FF neural network
mathematical formulations is generally combined with a SOFM is shown in [14] to
impractical and inaccurate. ANN is proposed for display the operating condition of the machine
fault identification and other power system on two dimensional grid. The operating point
applications [10, 31] & is emerging technologies moves as faults start developing and can be used
promising for future widespread industrial usage to alert motor protection system to incipient
[5, 8]. Artificial Neural Network are used to fault but SOFM not able to reliably detect a fault
www.ijert.org 2
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
when autotransformer added to create unbalance and training procedure for Neurofuzzy fault
supply in different phase. A diagnosis of three detection system is discussed for single phase
phase induction motor stator fault carried out induction motor bearing and stator turn fault
with Hebbian based unsupervised NN and using detection in [6]. This procedure will be used to
Principal component analysis based algorithm determine the correct solutions by constructing
by deriving α and β component of line current. the fault detector using fuzzy membership
The direction of NN eigenvector obtained from function module and fuzzy rule module. In [19]
this component used to discern motor healthy or trained Neurofuzzy fault detector provide
not in [7]. accurate fault detector performance; also
An Artificial Neural Network used to learn provide the heuristic reasoning behind the fault
characteristic of good operating motor may detection process and the actual motor fault
contain harmonics due to load in [30] and a conditions through the fuzzy rules. A knowledge
frequency filter is employed to passes only base is built in [36] to support the fuzzy
harmonics importance to fault detection. After inference for the detection of external fault like
sufficient training period, neural network signals voltage unbalance and open phase wherein
a potential failure condition when a new cluster fuzzy subset and corresponding fuzzy
formed if persists for some time for online fault membership function describe stator current
prediction possible. Fault detection comparison amplitudes.
between support vector machine and back
propagation algorithm using experiment data in
[28] and the training patterns are obtained using 3. Signal processing based approaches
MCSA and spectral Park‟s vector. A experiment
test were conducted in [31] by varying thermal One of the most frequently used fault detection
induced stress between ambient and maximum methods is the MCSA. This technique depends
rated temperature of winding and online upon locating faults according to the position of
sequence components were calculated, specific harmonic components in stator current
monitored by previously offline trained neural spectrum like broken rotor bars, air-gap
network act as negative sequence model of eccentricity, bearing fault, faults in stator
induction motor to detect the turn fault. windings, etc. As advantage of the method only
Simulated RMS values of parameters are taken a single current transducer is required for its
for ANN training for external faults classical method and can also be applied on-line.
classification and Applicability of NN for fault Signal processing techniques are applied to the
detection and protection discussed in [38]. Using measured sensor signals in order to generate
simulated faults and radial basis function Neural features or parameters (e.g. amplitudes of
network train for external fault detection. frequency components associated with faults)
Instantaneous values of faults are taken and which are sensitive to the presence or absence of
shown network classified the faults correctly in specific faults. Calculation of simple statistical
[39]. ]. A filter bank of Neural network time parameters such as the overall RMS value of a
series models are created for normal and faulty signal can give useful information. For instance,
motor, Bayesian classifier is used for correct the RMS value of the vibration velocity is a
classification of fault and tested with different convenient measure of the overall vibration
types of FEM simulated fault data in [46]. severity [4, 27]. In the same way, the RMS
value of the stator current provides a rough
Conventional ANN can correctly detect the indication of the motor loading. Some of the
faults but cannot provide heuristic interpretation main stator current signature based techniques
of the solution due to its numerically oriented discussed in [41,27] are frequency analysis
structures. Some Engineers prefer the accurate based Classical FFT & Instantaneous FFT,
fault detection as well as the heuristic Bispectrum periodic with a period of 2 and
knowledge behind the fault detection process. preserves both magnitude and phase
Fuzzy logic is a technology that can easily information, High Resolution spectral analysis
provide heuristic reasoning while being difficult and Wavelet analysis. The other frequency
to provide exact solutions. Methodology behind analysis technique can also be well associated
a novel hybrid Neurofuzzy system which for MCSA are Park‟s vector approach. Fourier
merges the neural network and fuzzy logic analysis is useful for stationary signals while in
technologies to solve fault detection problems transitory signal Short time Fourier Transform
www.ijert.org 3
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
and Wavelet analysis. STFT represents a sort of experientially during healthy condition of motor
compromise between time and frequency based and DWT coefficients of fault currents using
view of signal but used with limited precession selected mother wavelet „db3‟ at the sixth level
as fixed size of window. Wavelet transform of resolution utilized and second based on
overcome this limitation. A windowing comparison of modulus maxima of the DWT
technique with variable-size region is then used coeffiecients. Wavelet packet transform based
to perform the signal analysis, which can be the protection system developed in [40] coefficients
stator current. Wavelet analysis allows the use of the Wavelet Packet Transform line currents
of long time intervals where we want more compared experimentally decided threshold for
precise low-frequency information, and shorter detecting and diagnosing various disturbance
regions where we want high-frequency occurring in induction motor. Single phasing,
information [41]. Park vector approach is based phase to earth and short circuit faults.
on the locus of the instantaneous spatial vector
sum of three stator currents. This locus is 4. Hybrid Approaches
affected by stator winding faults and air gap
eccentricity. Park‟s vector monitored in [48] for Combination like I.M Model and any signal
identifying stator interturn fault whose ellipticity processing method, Model and soft computing
increased with the severity of the fault and its method etc; are considered here as a hybrid
major axis orientation is associated to faulty approach for the fault detection of induction
phase. motor. Induction machine model have been
[16] measures current spectra to detect faults in simulated in [17, 18, 26] for its behaviour
three phase induction motors but does not analysis in Matlab/Simulink using symmetrical
distinguish supply voltage unbalance condition. I.M d-q model. As in [17] dynamic simulation
[20] analyzed motor current spectrum under of small induction motor based on mathematical
faulty conditions (inter-turn short circuits) and modelling is one of the key steps in the
illustrated that RSH in the motor current are validation of the design process of motor drive
amplified under stator winding fault conditions. system and [34] discussed induction machine
In [22] RSH in the motor spectrum are sensitive model wherein all parameter are accessible for
to supply voltage unbalance. They have control and verification purposes. d-q model
presented a method using the RSH in the provides guide lines for dynamic simulation of
terminal voltage spectrum just after switch off induction motor, which can also applied for
so it does not have aforementioned problem but some faults data generation. [26] Shows the
is included in off line methods. A method results of studies under acceleration, variable
proposed in [15] based on motor current load and open phase using this model.
signature analysis for detection of inter-turn Conventional d-q model and development of its
short circuits in the stator winding of I.M. current, torque and power relationships are
utilizes three phase current spectra to overcome based on the assumptions that MMF produced
the problem of distinguishing stator interturn by stator winding excitation is sinusoidally
fault form voltage unbalance as in case of distributed in space and that the rotor MMF due
classical MCSA where only one transducer used to the slip frequency induced currents is
in one phase. Although the magnitude of these similarly distributed. Space harmonic are not
harmonics depends upon the level of voltage considered which may exist due to motor
unbalance, they have the same magnitude in asymmetry. The existence of space harmonics
three phase in these conditions, While in short creates detrimental effect on the steady state and
circuit the magnitudes differ from each other. transient characteristic of the machine.
However machine asymmetry is not considered
while supply unbalance voltage condition [1,34] developed a transient model using
occurs. Using high resolution spectral analysis winding function approach and magnetic
of stator current spectrum through experiment coupled circuit theory, model [34] can extended
the voltage unbalance and open phase external for the solution of rotor bar, end ring and stator
fault condition identified [32]. In [33] two faults. Rotor fault cause-effect relationships
approaches based on discrete wavelet transform studied in [1] using stator current and frequency
utilized for the induction motor fault detection signature. [2] suggested way for formulate the
wherein first fault detection criteria is the suitable model using the decomposition of
comparison between threshold determined mutual inductance matrix into Fourier series and
www.ijert.org 4
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
www.ijert.org 5
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
Squirrel Cage I.M. using ANN”, IEEE [21] Perdo Vicente J, Marian N, Antero A, “ A
Transactions on Industrial Electronics, Vol. 06, simplified scheme for I.M condition monitoring
No. 3 ,pp. 536-545, April 1991 “, Mechanical systems and signal processing,
[10] Sri kola, L. Varatharasa, “Identifying 3 Science direct, Elsevier, pp. 1216-1236, Nov
phase I.M faults using neural networks”, ISA 2007.
Transactions 39 (2000), Elsevier, Science Direct
[11] Sri kola, Shwan D, “ANN based fault [22] Nandi S, Toliyat HA. “Novel frequency-
identification scheme implementation for 3 domain-based technique to detect stator interturn
phase I.M.”, ISA Transactions 39, Elsevier, faults in induction machines using stator-
Science Direct, pp.433-439, 2007. induced voltages after switch-off.” IEEE
[12] Ahmed N, Abd Alla, “Three phase I.M Transactions on Industry Application
motor faults detection by using Radial Basis 2002;38(1):101–9.
function neural Network”, Journal of Applied [23] Mini V.P, Setty S, Dr. S.Ushakumari, “
Science 6(13), Asian Network for scientific Fault detection and Diagnosis of an I.M using
information, 2006. Fuzzy logic”, IEEE Region 8 SIBIRCON-2010,
[13] Khalaf Salloum Gaeid and Haider Irkutsk Listvyanka, Rusia, July 11-15, pp 459-
A.F.Mohamed, “Dignosis and Fault tolerant 464, 2010.
control of the induction motors techniques a
Review”, Australian Journal of Basic and [24] Hamid A. Toliyat, Thomas A. Lipo,, “
Applied Sciences, 4(2): 227-246, 2010. Transient Analysis of cage induction machines
under stator, Rotor bar and End Ring faults”,
[14] R M Tallam, David J G. Bruce H. “NN IEEE, pp. 241-247, 1994.
based on line stator winding turn fault detection
for induction motors”, IEEE, pp. 375-380,2000. [25] Samir Hamdani, Omar Touhami, Rachid
[15] Rasool Sharifi, Mohammad Ebrahimi, “ Ibtiouen, “ A Generalized Two axis model of a
Detection of stator winding faults in induction squirrel cage induction motor for Rotor fault
motors using three phase current monitoring”, diagnosis”, Serbian Journal of Electrical
ISA Transactions, Elsevier , pp. 1-7, 2010. Engineering, Vol 5., pp. 155-170, May 2008.
[16] Mihet-Popa L. “Current signature analysis [26] A.I.Orille, G.M.A.Sowilam, J.A Valencia,
as diagnosis media for incipient fault “A New simulation of symmetrical 3 phase I.M
detection.”, Advances in Electrical and under transformation of park” Comput. Ind.
Computer Engineering 7(2):11–15,2007. Eng., vol. 37, pp. 359–362, 1999.
[17] K. L. Shi, T.F.Chan, Y.K.Wong, S.L.Ho, [27] M.L Sin, W.L.Soong and N Ertugrul.
“Modelling and simulation of 3 phase induction “Induction machine on line condition
motor using simulink”, International journal of monitoring and fault diagnosis- A Survey”
Electrical Engg. Education, Vol. 36, [28] Pablo S., Antonio Z, Arturo H, Albeto O,
Manchester, pp. 163-172, 1999. “ANN for Diagnosing Stator I.M. faults”,
[18]A.A.Ansari,D.M.Deshpande,“Mathematical Neural Networks, Vol..3, issue 1
Model of Asynchronous Machine in Matlab/ [29] Jorge Nei B, Robson P, “A Hybrid Neural /
simulink”, International journal of Engineering Expert system to diagnosis problems in I.M”,
science and Technology, pp. 1260-1267, 2010. 1999.
[19] P.V.Goode, M.Y.Chow, “Using a [30] R.R.Schoen, B.K.lin, T.G.Habetler,
Neural/Fuzzy system to extract Heuristic J.H.Schlag & S.Farag, “An unsupervised, on
knowledge of incipient faults in I.M: part II- line system for I.M fault detection using stator
Application”, IEEE transactions on industrial
current monitoring”, IEEE Trans. Ind Appl., vol.
Electronics, Vol. 42, No.2, 139-146, April 1995
31, no. 6, pp., 1280–1286, Nov./Dec. 1995.
[20] Joksimovic GM, Penman J., “The detection [31] R. M. Tallam, T. G. Habetler, and R. G.
of inter-turn short circuits in the stator windings Harley, "Experimental testing of a neural-
of operating motors.” IEEE Transactions on network-based turn-fault detection scheme for
Industry Applications 2000; 47 (5):1078–84. induction machines under accelerated insulation
failure conditions," in Diagnostics for Electric
www.ijert.org 6
International Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181
Vol. 1 Issue 5, July - 2012
Machines, PowerElectronics and Drives, 2003. [43] Lane M. R.B, Benjamim R.M., Walmir
SDEMPED 2003. 4th IEEE International M.C., “fault induction dynamic model, suitable
Symposium on, pp. 58-62, 2003. for computer simulation: simulation results and
experimental validation, Mechanical systems
[32] Mohamed El Hachemi, Benbouzid, and Signal Processing 24 (2010) 300-311,
Michelle Vieira, and C´eline Theys , “Induction ELSEVIER
Motors‟ faults detection and localization using [44] Thomas G.H, Ronal G.H., R Tallam, Sang-
stator Current advanced signal Processing Bin Lee, R. Obaid and Jason, “Complete current
Techniques”, IEEE Transactions on power based induction motor condition moitoring:
electronics, Vol. 14, No. 1, January 1999. stator , Rotor , Bearings and Load, CIEP,IEEE ,
pp. 3-8, 2002.
[33] M.A.S.K.Khan and M.A.Rahman, “Discrte [45] A. Bellini, F. Filippetti, C. Tasoni, and
Wavelet transform based detection of Geard Capolino, “Advances in diagnostic
disturbances in induction motors”, ICECE 2006, techniques for induction machines”, IEEE
19-21 Dec.,2006, Bangladesh Transactions on industrial Elecgtrconics”, Vol.
[34] Xiaogang Luo,Yuefeng Liao.Hamid 55, No. 12, Dec. 2008.
Toliyat, Ahmed El-Antably and Thomas Lipo, [46] Jarmo Lehtonen , Heikki N. Koivo , Fault
“Multiple coupled circuit modelling of induction Diagnosis of Induction Motors with Dynamical
machines”, 1993, pp. 203-210, IEEE. Neural Networks”, IEEE SMC 2005, pp2979-
2984,2005.
[35] Burak Ozpinee, Leon M.T, “Simulink [47] Fatiha Z., M. H. Benbouzid, Demba
implementation of induction Machine Model – Diallo,and M. S. NaïSaïd, “,Induction Motor
A modular Approach, IEEE, pp. 728-7342003. Stator Faults Diagnosis by a Current Concordia
Pattern-Based Fuzzy Decision System”, IEEE
[36] M.E.H. Benbouzid and H. Nejjari, “A Trans. On Energy. Conv., Vol. 18, NO. 4, Dec.
Simple Fuzzy Logic Approach for induction 2003, p.p. 469-475.
Motor Stator Condition Monitoring”, pp. 634- [48] A.J.m. Cardoso, S.M.A.Cruz and
639, 2001, IEEE. D.S.B.Fonseca, “ Inter-Turn Stator Winding
Fault Diagnosis in Three-phase Induction
[37] M Sudha, P Anbalagan, “A protection Motors, by Park's Vector Approach”, IEEE
Scheme for three phase I.M form Incipient faults Trans. on Energy Conv., Vol. 14, No. 3,
using Embedded Controller”, Asian Journal of September 1999. P.p.595-598
Scientific Research 2(1),2009.
www.ijert.org 7