Current signature analysis of induction motor mechanical ... · stationary nature of stator current can be accurately considered. ... and classification. The goal of feature extraction

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 50, NO. 6, DECEMBER 2003 1217

Current Signature Analysis of Induction MotorMechanical Faults by Wavelet Packet Decomposition

Zhongming Ye, Member, IEEE, Bin Wu, Senior Member, IEEE, and Alireza Sadeghian, Member, IEEE

Abstract—This paper presents a novel approach to inductionmotor current signature analysis based on wavelet packet decom-position (WPD) of the stator current. The novelty of the proposedmethod lies in the fact that by using WPD method the inherent non-stationary nature of stator current can be accurately considered.The key characteristics of the proposed method are its ability toprovide feature representations of multiple frequency resolutionsfor faulty modes, ability to clearly differentiate between healthyand faulty conditions, and its applicability to nonstationary signals.Successful implementation of the system for two types of faults,i.e., rotor bar breakage and air-gap eccentricity is demonstratedhere. The results are validated based on both simulation and ex-periments of a 5-hp induction motor.

Index Terms—Air-gap eccentricity, induction motor, mechan-ical fault, rotor bar, stator current signature analysis, waveletpacket decomposition (WPD).

I. INTRODUCTION

A S the backbone of modern industry, induction motors arevirtually used in every industry. Online fault diagnostics

of induction motors are very important to ensure safe operation,timely maintenance, increased operation reliability, and preven-tive rescue especially in high power applications. The inductionmotor faults are generally classified as either mechanical or in-sulation system faults. Common mechanical faults include rotorbar breakage, rotor end ring cracking, static and/or dynamicair-gap irregularities, stator winding faults, bent shaft, misalign-ment, and bearing gearbox failures. Statistical data show that themechanical faults are responsible for more than 95% of all fail-ures [1]–[7].

There are different methods for the detection of mechanicalfaults. Typical examples as related to this may include detectionof air-gap eccentricity, shaft and bearing faults by monitoringthe vibration signal [6] or by analyzing the lubricating oildebris in the case of latter fault; and detection of stator windingfaults means of searching coils that measure the leakageflux along the shaft [3]. Other diagnostic methods includeacoustic noise analysis, temperature measurement, infraredmeasurement, radio frequency emission monitoring, partialdischarge measurement and motor current signature analysis(MCSA) [1], [2], [7], [8]. Among all these methods, MCSA is

Manuscript received January 7, 2002; revised December 11, 2002. Abstractpublished on the Internet September 17, 2003.

Z. Ye was with the Department of Electrical and Computer Engineering, Ry-erson University, Toronto, ON M5B 2K3, Canada. He is now with the Depart-ment of Electrical and Computer Engineering, Queen’s University, Kingston,ON K7L 3N6, Canada (e-mail: [email protected]).

B. Wu and A. Sadeghian are with Ryerson University, Toronto, ON M5B2K3, Canada (e-mail: [email protected]).

Digital Object Identifier 10.1109/TIE.2003.819682

one of the most popular used methods because of the followingreasons. Firstly, it is noninvasive. The stator current can bedetected from the terminals. Secondly, it can be measuredonline therefore makes online detection possible. Thirdly, mostof the mechanical and electrical faults can be detected by thismethod. The mechanism of MCSA can be explained as below.

The occurrence of motor mechanical faults usually resultsin an asymmetry in the windings and eccentricity of air gap,which lead to a change in the air-gap space harmonics distri-bution. This abnormality exhibits itself in the spectrum of thestator current as unusual harmonics. The current spectrum canbe obtained through fast Fourier transform (FFT) of the statorcurrent under steady-state conditions. Fig. 1(a) and (b) depict,respectively, the stator current waveform and spectrum aroundthe fundamental component for a three-phase induction motorwith two broken rotor bars. The low sideband harmonic com-ponent around 60 Hz is caused by the rotor bar breakage and itdoes not appear in a healthy motor [Fig. 1(c) and (d)]. The fre-quency of this component is given by

(1)

where is the slip frequency of the induction motor andisthe fundmanetal frequency.

Fault detection and the diagnostic system of inductionmotors include at least two important parts: feature extractionand classification. The goal of feature extraction is to extractfeatures which are related to specific fault modes. Usuallythe features are obtained by processing the spectrum of statorcurrent of the induction motors using FFT. Other possiblemethods include wavelet transform and short-time Fouriertransform (STFT). The goal of classification is to classify faultymode from normal mode and different fault modes. Artificialintelligence (AI), for instance, expert system and artificialneural network, are often used. Much work has been reportedin the literature on AI-based fault detection and diagnosticsystems [3], [4], [6]–[8]. Most of the published methods arebased on the features obtained by FFT spectrum. Unfortunately,not much work is reported on the feature extraction methodswhich can be very important for the performance of the wholediagnostic system.

The task of distinguishing faulty conditions from normal con-ditions based on the resultant FFT spectrum can be done accu-rately as long as the signals are stationary, the induction mo-tors are run around full-load condition, and the terminal volt-ages are sinusoidal. The stator current, however, is a nonsta-tionary signal whose properties vary with the time-variant oper-

0278-0046/03$17.00 © 2003 IEEE

1218 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 50, NO. 6, DECEMBER 2003

(a) (b)

(c) (d)

Fig. 1. Stator current waveforms and spectrum around 60 Hz, for healthy and faulty induction motor at speed of 1749 r/min. (a) Stator current waveform of ahealthy induction motor. (b) Corresponding spectrum around the fundamental frequency. (c) Stator current waveform for faulty induction motor withtwo brokenrotor bars. (d) Corresponding spectrum around fundamental frequency.

ating conditions of the motors such as fluctuations in load torqueand power supply. To remedy the problems associated with timevariations, STFT, which is a very useful time-frequency local-ization tool, is usually used. However, STFT is only applicablewhere the low-pass window function can be suitably chosen andwell localized. To this end, a fixed-width window for all fre-quency components is usually used by this approach and, there-fore, STFT cannot provide either multiple frequency resolutionor temporal resolution [9], [10].

The importance of a study on new feature extraction methodsalso arises from the two requirements from the current industryapplications. Firstly, most of the induction motors are poweredfrom power electronics equipment, which generate a lot of har-monics. These harmonics make the FFT-based feature extrac-tion difficult for online detection. Secondly, for offline detectionand the diagnostic system of large induction motors, it is expen-

sive to run the motor at full-load conditions. While the featuresobtained from light-load conditions normally cannot ensure anaccurate classification. One method is to do the detection and di-agnosis during startup. However, the FFT method is incapablefor such transient signals. Only the wavelet-based method canbe used in these cases.

Wavelets are mathematical tools that have recently emergedfor applications such as waveform representations and segmen-tations, time-frequency analysis, detection of irregularities, fea-ture extractions, and compression of digital data. The popu-larity of wavelets is due to properties such as the dilation prop-erty that can be used to adjust the width of the frequency bandalong with the location of its center frequency, and the transla-tion property that can be used to automatically zoom in and outin order to locate the positions of high-frequency and low-fre-quency changes. However, the data obtained from wavelet trans-

YE et al.: CURRENT SIGNATURE ANALYSIS OF INDUCTION MOTOR MECHANICAL FAULTS BY WPD 1219

form cannot be used for feature extraction unless by further FFTanalysis, which increases the complexity of the algorithm.

This paper proposes a new method for mechanical fault fea-ture extraction of induction motors based on wavelet packet de-composition (WPD) of the stator current. Two important charac-teristics, time localization ability and multiresolution analysis,make WPD very attractive for the purposes of fault detectionand diagnosis. The underlying idea of the proposed method isto use WPD to decompose the stator current into the time–fre-quency spectrum, and then use the results to calculate and tochoose proper feature coefficients which best represent the me-chanical faults of the induction motor. The feature signatures forair-gap eccentricity fault as well as rotor bar breakage are inves-tigated in this paper.

II. WPD AND FEATURE COEFFICIENTS

A. WPD

For any given signal , the discrete wavelettransform is defined as the inner product of the wavelet func-tion and the signal, that is,

(2)

where is the signal to be analyzed and is the dis-crete wavelet function. The original signal can be approximatedwith the wavelet functions and the wavelet coefficients

(3)

where is the scale factor and is the displacement. Thewavelets are derived from a so-called mother wavelet by thedilation and translation factors. The mother wavelet is normal-ized with zero average and meets the following admissibilitycondition

(4)

Applying the wavelet transform to the original signal dividesthe signal into two parts, the high-frequency part and the low-frequency part. The low-frequency part is called an approxima-tion of the original signal. A series of approximations can beobtained by reiterating such decompositions. The difference ofthe approximations between two successive decompositions iscalled the details. The multiresolution analysis (MRA) is an al-gorithm based on the reiterative decomposition of the low-fre-quency parts only. The peeling-off process in MRA can also bedefined as decomposing the approximation spaceinto a sub-sequent approximation subspace and the correspondingdetail subspace . The detail space related to the ap-proxiamtion space , however, remains undecomposed.

WPD is an extension of wavelet transformation achieved bymeans of generalizing the link between multiresolution approx-imation and wavelets. In WPD, both the approximation space

and the detail space are decomposed further. The trans-

(a)

(b)

Fig. 2. Wavelet packet filter bank decomposition and corresponding binarytree.h : low-pass filter;h : high-pass filter. Downsampling# operator.

formation of the input sequence at scalecan be described by[9], [10]

(5)

where and represent low pass and high pass having afinite-impulse response of size.

In Fig. 2(a), an example is shown for such a division using theconjugation mirror filter banks. The original space is dividedinto detail and approximation spaces by the low-pass filterand high-pass filter , respectively. The resultant detail spaceis further divided. The reiterative splitting of vector spaces isrepresented in a binary tree in Fig. 2(b), where the binary treenodes are labeled by their Depth(a dilation factor) and nodenumber (frequency factor), and the corresponding space isdenoted as . It has been proven that there are more than

different wavelet packet orthonormal bases included ina full wavelet packet binary tree of Depth[9]. Each of thesepacket has a limitted time support as well as frequency support.

B. Feature Coefficients

WPD adds redundancy to the transformation by expandingeach packet repeatedly. The obtained time–frequency represen-tation is a matrix containing the wavelet packet coefficients forall Depths and Nodes. For a signal of length, the total numberof Depths is . The maximum number of coefficientsfor one Depth is . At Depth 0, the coefficients are exactlythe original signal. Most of the coefficinets are irrelevant to themechanical faults to be detected. Further processing is requiredto make these data useful for the intelligent fault detection anddiagnostics using AI techniques such as multilayer perceptronnetworks and adaptive neural fuzzy systems.


Fig. 3. Approximation of the fourth-order Coiflet function.

(a)

(b)

Fig. 4. Feature coefficients for a small three-phase induction motor with twobroken rotor bars at different load conditions (torques), 16 sampling points percycle for 1024 points. (a) At Depth 8. (b) At Depth 9.

Owing to the frequency localization, the WPD coefficients ofdifferent harmonic components distribute at different Nodes of

(a)

(b)

Fig. 5. Difference of the feature coefficients between faulty and normalconditions for different torques. (a) At Depth 8. (b) At Depth 9.

the spectrum of the WPD. The Feature Coefficient forNode at Depth is defined in terms of WPD coefficients as

(6)

where is the number of WPD coefficients used for the calcu-lation of the features at Depthand Node .

For a given frequency component, the energy of the compo-nent is localized at a certain number of Nodes at a given Depth,and the strength of the energy depends on the amplitude of thefrequency component [9]. In other words, this frequency com-ponent can be represented with the feature coefficients at theseNodes. A set of Nodes from different Depths can be selectedto calculate the feature coefficients that best represent the fre-quency components caused by the faults of the induction mo-tors. In the following section, the current signature extraction ofrotor bar breakage and air-gap eccentricity is discussed.


TABLE IMEANS OF THECOEFFICIENTS FORROTOR BAR BREAKAGE FORDIFFERENTLOAD CONDITIONS

(a) (b)

(c) (d)

Fig. 6. Feature coefficients for air-gap eccentricity at various loading conditions. Circles: normal conditions; squares: faulty conditions withair-gap eccentricity(simulation results). (a) Node 14 of Depth 8. (b) Node 16 of Depth 8. (c) Node 32 of Depth 9. (d) Node 8 of Depth 7.

III. FEATURE COEFFICIENTS FORDETECTION OFBROKEN

ROTOR BARS AND AIR-GAP ECCENTRICITY

Dynamic simulations for a 7.5-hp induction motor (param-eters are given in Appendix A) with and without mechanicalfaults can be accomplished using the Winding Function Method(WFM) [11]–[13] that is a convenient and a fast method for sim-

ulation of the mechanical faults, such as rotor bar breakage, endring crack, shaft alignment, etc. The parameters of the equiva-lent circuits representing the rotor and stator can be deteriminedfrom the dimensions of the rotor and stator using the WFM.Any mechanical fault will lead to the change of the parame-ters of the equivalent circuits. Two mechanical fault modes, i.e.,


TABLE IIMEAN AND STANDARD DEVIATION OF THE FEATURE COEFFICIENTS FOR

AIR-GAP ECCENTRICITY FORDIFFERENTLOAD CONDITIONS

rotor bar breakage and air-gap eccentricity are simulated withthis method.

Although detection and diagnostics of mechanical and elec-trical faults of induction motors can be done with full load ap-plied. However, the following practical conditions have to beunder consideration: For an online fault detection system, it isexpected the system can also work when the motor is not runin rated torque. For offline diagnostics system, it is expensiveto build a test bench for full-load operation for large power mo-tors. Therefore, a system which can perform diagnostics underreduced load torque is better. Therefore, the conditions for thefull range of the torque will be investigated, that is, from 0% to110% of rated torque.

Using WPD, the WPD coefficients of the stator current atsteady state can be obtained. The feature coefficients can becalculated for all Depths and Node numbers according to (5). Inthis paper, fourth-order biorthogonal wavelet Coiflets [11] areused in the WPD. The approximation of the wavelet is plottedin Fig. 3. The Coiflets are one family of compactly supportedwavelets with highest number of vanishing moments for a givensupport width [11]. For an th order Coiflet, the support widthis and the vanishing moment for the wavelet function is

and the scaling function is , respectively.

A. Rotor Bar Breakage

Rotor bar breakage is one of the most common motor faults.Several factors may contribute to this fault, such as, hot spots,sparking and thermal unbalance, chemical contamination,and moisture abrasion of the rotor materials. As a rotor baris broken, a sideband component next to the fundamentalfrequency increases in the spectrum of the stator current. Usingthe WPD analysis and (6), the feature coefficients for differentload torques are calculated. The mesh plots of the featurecoefficients of the stator current at Depths 8 and 9 aregiven in Fig. 4. This presents the simulation case where two(out of the 44) rotor bars of a three-phase induction motor arebroken. The stator current waveform is sampled with 16 pointsper cycle for a window of 64 cycles. The large amplitude of thecoefficients around Node 32, Depth 8, and Node 64, Depth 9,are corresponding to the fundamental component of the statorcurrent. As the torque increases, the fundamental component ofthe stator current increases which, in turn, leads to the increaseof the corresponding feature coefficients.

At Depth 8, the feature coefficients plot in Fig. 5(a) exhibitslocal peaks at the proximity of Node 32. This implies a low side-band frequency component in the stator current which corre-

Fig. 7. Experiment setup. Induction motor and generator: 5 hp, 60 Hz, 208 V,1750 r/min.

(a)

(b)

Fig. 8. Feature coefficients for three-phase induction motor with two brokenrotor bars at different load conditions (torques), 16 sampling points per cycle for1024 points. They axis is exactly the experiment sequence number; the largerthis number, the larger the torque. (a) At Depth 8. (b) At Depth 9.

sponds to the broken rotor bar fault. As the torque increases, thenode number, which represents the local peak, decreases. Sincethe node number is propotional to the frequency, this indicatesthat the sideband component of the rotor bar breakage decreasesas the slip speed increases. For light-load conditions, the slipspeed is too small, and according to (1), the sideband compo-nent is very close to the fundamental component. Therefore, thefeature coefficients are too close to the fundamental to be dis-tinguished at this Depth. A similar observation can be made for


(a) (b)

(c) (d)

Fig. 9. Feature coefficients at Depth 10 for rotor bar breakage. Circles: healthy conditions; squares: faulty conditions. (a) Node 60. (b) Node 61. (c) Node 62.(d) Node 63.

the feature coefficients at Depth 9 as shown in Fig. 4(b). The dif-ference of the feature coefficients between the normal and faultconditions are denoted as

(7)

where and are the feature coefficients fornormal and faulty conditions, respectively.

Fig. 5 illustrates the feautre coefficient differences at Depths8 and 9 in a contour form. In the contour plot, the dark arearepresents a large difference value, while the light area repre-sents a small difference. As can be seen in Fig. 5, the nodes withlarge differences of feature coefficients appear in three lines:the fundamental components (Node 32, Depth 8 and Node 64,Depth 9), the left sideband components shifting toward the lownodes as the torque increases, and the right sideband compo-nents shifting toward high nodes as the torque increases. Thefeature coefficients on the left-half side correspond tocomponent in the frequency domain and the right-half-side fea-tures correspond to the component.

The obvious differences of the feature coefficients for the side-band components indicate that these feature coefficients can beused for the applications of detection and diagnosis of the rotorbar fault. However, it is noted that for light-load conditions, thedifferences of the feature coefficients between faulty and healthyare not very obvious and the nodes are too close to the funda-mental component. Therefore, it is difficult to perform a faultdiagnosis for these conditions using the stator current. Table Isurmmarizes the mean values of the feature coefficients for bothfaulty and healthy modes of operation of an induction motor witha synchronous frequency of 60 Hz. Three different load condi-tions, i.e., light, medium, and heavy are considered, where theyrefer the torque of 0%–30% of the full load, 30%–70% of the fullload, and 70%–110% of the full load, respectively. Each of thestatistics mean is estimated from about 20 samples. Table I showsthat normal and faulty conditions can be clearly distinguishedbased on the statistical properities of the feature coefficients.

Further investigations reveal that the feature coefficients fornormal conditions approximately follow a normal distribution,


(a) (b)

Fig. 10. Feature coefficients at Depth 9 for rotor bar breakage. Circles: healthy conditions; squares: faulty conditions. (a) Node 30. (b) Node 31.

while for faulty conditions, the statistic characteristics vary withthe load conditions. Table I shows that for different load con-ditions, the feature coefficients that best describe the differencebetween the faulty and normal conditions are located at differentNodes and Depths. For example, for light-load conditions, fea-ture coefficients at Depth 10 are better than those at Depth 8, andthe feature coefficients at Nodes 62 and 63 of Depth 10 are betterthan those at Nodes 60 and 61 of the same Depth. According tothe results obtained from Table I, small node numbers should bechosen to detect faults at heavy-load conditions, whereas largenode numbers should be used for light-load conditions.

B. Air-Gap Eccentricity

Air-gap eccentricity is another typical fault related to induc-tion motors. One of the traditional methods to detect the air-gapeccentricity is to monitor the behavior of the motor stator cur-rent at the sideband of the fundamental component, that is,

(8)

where is the rotation frequency,is the number of pole pairs,and is the slip frequency.

Feature coefficients related to air-gap eccentricity are also an-alyzed here in a similar fashion to those of rotor bar breakage.It is observed that the large difference between the feature coef-ficients of normal and faulty motors appear around: Nodes 31,32, and 33 at Depth 10; Nodes 15, 16, 17 at Depth 9; and Nodes7 and 8 at Depth 8. Fig. 6(a) and (b) depicts feature coefficientsof Nodes 14 and 16 (whose frequency is half the fundamentalfrequency) at Depth 9 for both normal and faulty conditions.The solid lines in these figures are obtained from the feature co-efficients using a linear polynormal fit. The feature coefficientsof Node 16 tend to decrease with the increase of the slip speedor load torque, while the feature coefficients of Node 14 haveanother tendency. According to (8), the frequency of the side-band component next to the fundamental component tends todecrease as the slip increases. Hence, the feature coefficients

TABLE IIIMEAN VALUES OF THEFEATURE COEFFICIENTS FORROTORBAR BREAKAGE

shift from Node 16 to 14 as load increases. Similar observationscan be made for the feature coefficients illustrated in Fig. 6(c)and (d).

The feature coefficients of Nodes 7, 8, and 9 at Depth 8,Nodes 14, 15, 16, 17, and 18 at Depth 9, and Nodes 30, 31,32, 33, and 34 at Depth 10 can be used for detection of air-gapeccentricity. Table II shows the mean and standard deviation forthe feature coefficients for normal and faulty conditions, respec-tively. It can be observed that the mean values for normal con-ditions are much smaller than those of the faulty conditions.

IV. EXPERIMENT VERIFICATION

A. Experiment Layout

Experiments on a three-phase 5-hp 208-V 60-Hz inductionmotor (parameters are given in Appendix B) are carried out fornormal and faulty conditions (one broken rotor bar, two brokenrotor bars, and air-gap eccentricity). The rated speed of the in-duction motor is 1750 r/min. The induction motors used have analuminum rotor with 28 rotor bars. The layout of the circuit forthe experiment is shown in Fig. 7. Two identical induction mo-tors are installed on the same bench where one is used as a motorand the other as a generator. The shafts of the two machines arein the same axial direction. A delta-connected capacitor bank isconnected to the output terminals of the generator to provide aself-excitation current.


(a) (b)

(c) (d)

Fig. 11. Feature coefficients at Depth 10 for rotor bar breakage. Solid lines: normal conditions; dashed lines: conditions of one broken rotor bars; dotted lines:two broken rotor bar conditions. (a) Node 60. (b) Node 61. (c) Node 62. (d) Node 63.

B. Rotor Bar Breakage

In order to emulate a real rotor bar breakage, first, a hole in themiddle of a rotor bar is drilled, so that it is electrically broken.Since in real situations nearby rotor bars are more likely to bebroken than remote ones, a second bar which is physically nextto the first broken bar is damaged in a similar fashion to createthe two broken rotor bar condition.

Using the WPD analysis and (6), the feature coefficients fordifferent load torques are caluclated. The mesh plots of the fea-ture coefficients, , of the stator current at Depths 8 and 9are given in Fig. 8. This experiment presents the case where two(out of the 28) rotor bars are broken, and the stator current wave-form is sampled with 16 points per cycle for a window of 64cycles. The large amplitude of the coefficients around Node 32,Depth 8, and Node 64, Depth 9, corresponds to the fundamentalcomponent of the stator current. As the torque increases, thefundamental component of the stator current increases which,in turn, leads to the increase of the corresponding feature coef-ficients.

In Fig. 9, feature coefficients obtained by analyzing the statorcurrent of the induction motor at Depth 10 are plotted for healthyconditions and faulty conditions with two broken rotor bars.As can be observed, distinguishable differences exist betweenhealthy and faulty conditions. For healthy motors, the featurecoefficients are constantly small for all load conditions com-pared to those of the faulty motors, while for faulty conditions,the feature coefficients are changing with the load torque. InFig. 9, the solid lines are linearly fitted to the feature coef-ficients. The notable difference in shapes of the lines at dif-ferent nodes corresponds to the variation tendency of the fea-ture coefficients at a specific node when load changes. For in-stance, at Node 60, which corresponds to large slip frequency,the feature coefficients increase as the load or slip increases.The feature coefficients are, therefore, fit for detecting the faultat heavy-load conditions. On the contrary, the features at Node63 are fit for light-load conditions, and for medium-load condi-tions, the proper Nodes for the fault detection are 62 and 61. Thefeature coefficients at Nodes 31 and 32 of Depth 9 are given in


(a) (b)

Fig. 12. Feature coefficients at Depth 9 for rotor bar breakage. Solid lines: normal conditions; dashed lines: conditions of one broken rotor bar; dotted line: twobroken rotor bar conditions. (a) Node 30. (b) Node 31.

(a) (b)

(c) (d)

Fig. 13. Feature coefficients for air-gap eccentricity. Circles: healthy conditions; squares: faulty conditions. (a) Node 15 of Depth 9. (b) Node 16of Depth 9.(c) Node 32 of Depth 10. (d) Node 8 of Depth 8.


TABLE IVMEAN VALUES OF THEFEATURE COEFFICIENTS FORAIR-GAP ECCENTRICITY

Fig. 10. The statistics of the feature coefficients at some of theDepths and Nodes are given in Table III. The differences of thefeature coefficients between normal and faulty conditions areclearly demonstrated.

It is further found that the feature coefficients are related tothe number of broken rotor bars. The feature coefficients forserious faulty conditions are larger than those of slight faultyconditions or normal conditions. For example, the feature coef-ficients for two broken bars are larger than for one broken rotorbar, and the feature coefficients for one broken rotor bar arelarger than for normal conditions. In Figs. 11 and 12, the fea-ture coefficients with zero, one, and two broken rotor bars areplotted versus slip frequency for normal and faulty conditions.

C. Air-Gap Eccentricity

The same motor–generator set is used for the air-gap eccen-tricity experiment. The induction motor is intentionally placedout of axis with the generator to make a slight air-gap eccen-tricity. The displacement is about 1 mm. The feature coefficientsat Nodes 15 and 16 of Depth 9, Node 32 of Depth 10, and Node8 of Depth 8 are plotted in Fig. 13. The statistic characteris-tics of the feature coefficients at some of the Depths and Nodesare given in Table IV for two eccentricity conditions: one is forabout 1-mm displacement and the other is for 1.5-mm displace-ment. The differences of the feature coefficients between normaland air-gap eccentricity are clearly demonstrated.

As can be observed, compared to the traditional FFT-basedmethod, the WPD-based feature extraction method providesmultiresolution presentation with which the features for light-load conditions can be easier to identify. Although the FFTanalysis for steady-state current is accurate, the multiresolutionpresentation facilitates the training of the artificial-neural-network-based classification network and, therefore, improvesthe performance of the fault detection and diagnostics. Forthe transient signal, the WPD is inherently superior to theFFT-based method because of its time localization ability.

V. CONCLUSION

New feature coefficients for induction motor mechanicalfaults are obtained by WPD of the stator current. The featurecoefficients differentiate the healthy and faulty conditions withan obvious difference. They can be used for the purpose ofonline noninvasive detection and diagnosis of such mechanicalfaults. One of the major advantages of this method is that it canbe used for nonstationary signal analysis. As is well known,induction motors are commonly used together with the powerelectronics drives. The traditional FFT-based MCSA method is

TABLE V7.5-hp 220-V 60-Hz INDUCTION MOTOR

TABLE VI5-hp 208-V 60-Hz INDUCTION MOTOR

subject to not only the harmonics disturbance, but also frequentdynamics of the drives. This method can be used for suchapplications. Another advantage is that the feature coefficientsobtained using the proposed method are of multiple frequencyresolutions. The same frequency component can be representedwith different frequency resolution. It is advantageous for faultdetection and diagnosis. This method can also be extended toother MCSA-based fault detection applications.

APPENDIX A

See Table V.

APPENDIX B

See Table VI.

REFERENCES

[1] H. Toliyat and T. Lipo, “Transient analysis of cage induction machinesunder stator, rotor bar and end ring faults,”IEEE Trans. Energy Conver-sion, vol. 10, pp. 241–247, June 1995.

[2] M. Elkasabgy, A. Eastam, and G. Dawson, “Detection of broken barsin the cage rotor on an induction machine,”IEEE Trans. Ind. Applicat.,vol. 28, pp. 165–171, Jan./Feb. 1992.

[3] G. Kliman et al., “Noninvasive detection of broken bars in operatinginduction motors,”IEEE Trans. Energy Conversion, vol. 3, pp. 874–879,Dec. 1988.

[4] A. H. Bonnett and G. C. Soukup, “Rotor failures in squirrel cage in-duction motors,”IEEE Trans. Ind. Applicat., vol. 22, pp. 1165–1173,Nov./Dec. 1986.

[5] A. H. Bonnettet al., “Cause and analysis of stator and rotor failures inthree-phase squirrel cage induction motors,”IEEE Trans. Ind. Applicat.,vol. 28, pp. 921–937, July/Aug. 1992.

[6] D. Dorrell, W. Thomson, and S. Roach, “Analysis of air gap flux, cur-rent and vibration signals as function of the combination of static anddynamic air gap eccentricity in 3-phase induction motors,”IEEE Trans.Ind. Applicat., vol. 33, pp. 24–34, Jan./Feb. 1996.

[7] M. E. Benbouzidet al., “Induction motor asymmetrical faults detectionusing advanced signal processing techniques,”IEEE Trans. Energy Con-version, vol. 14, pp. 147–152, June 1999.

[8] T. Breenet al., “New developments in noninvasive online motor diag-nostics,” inProc. IEEE PCIC, 1996, pp. 231–236.

[9] S. Mallat,A Wavelet Tour of Signal Processing. San Diego, CA: Aca-demic, 1998.

[10] I. Daubechies,Ten Lectures on Wavelets. Philadelphia, PA: SIAM,1992.


[11] L. Xiaogang and H. Toliyat, “Multiple coupling circuit modeling of in-duction machines,”IEEE Trans. Ind. Applicat., vol. 31, pp. 311–318,Mar./Apr. 1995.

[12] H. Toliyat, “A method for dynamic simulation of air gap eccentricity ininduction machines,”IEEE Trans. Ind. Applicat., vol. 32, pp. 910–918,July 1996.

[13] , “Simulation and detection of dynamic air gap eccentricity insalient pole synchronous machines,”IEEE Trans. Ind. Applicat., vol.35, pp. 86–93, Jan./Feb. 1999.

Zhongming Ye (M’01) was born in Jiangsu, China.He received the B.S. degree from Xi’an Jiaotong Uni-versity, Xi’an, China, in 1992, the M.S. degree fromShanghai Jiao Tong University, Shanghai, China, in1995, and the Ph.D. degree from Zhejiang Univer-sity, Hangzhou, China, in 1998, all in electrical engi-neering.

From 1998 to 1999, he was with the Department ofElectrical and Electronics Engineering, University ofHong Kong, Hong Kong. From 1999 to 2001, he waswith the Department of Electrical and Computer En-

gineering, Ryerson University, Toronto, ON, Canada. Since 2001, he has beenwith the Department of Electrical and Computer Engineering, Queen’s Uni-versity, Kingston, ON, Canada. His research interests include high-frequencypower conversion, high-frequency ac power distribution systems, electrical ma-chine fault diagnostics, power quality, artificial intelligence, neural networks,and fuzzy logic.

Bin Wu (S’89–M’92–SM’99) received the M.A.Sc.and Ph.D. degrees in electrical and computer engi-neering from the University of Toronto, Toronto, ON,Canada, in 1989 and 1993, respectively.

After being with Rockwell Automation Canadaas a Senior Engineer, he joined Ryerson University,Toronto, ON, Canada, where he is currently aProfessor in the Department of Electrical and Com-puter Engineering. His research interests includehigh-power converter topologies, motor drives, andapplication of advanced control in power electronic

systems.Dr. Wu is the recipient of the Gold Medal of the Governor General of Canada,

the Premier’s Research Excellence Award, and the NSERC Synergy Award forInnovation. He is a Registered Professional Engineer in the Province of Ontario,Canada.

Alireza Sadeghian(S’94–M’00) was born in Tehran,Iran. He received the B.A.Sc. (Hons.) degree in elec-trical engineering from Tehran Polytechnic Univer-sity, Tehran, Iran, in 1989, and the M.A.Sc. and Ph.D.degrees in electrical and computer engineering fromthe University of Toronto, Toronto, ON, Canada, in1994 and 1999, respectively.

In l999, he joined the Department of Mathematics,Physics and Computer Science, Ryerson University,Toronto, ON, Canada, as an Assistant Professor. Hisresearch interests include knowledge-based expert

systems, artificial neural networks, fuzzy logic systems, adaptive neuro-fuzzynetworks, and nonlinear modeling.

Documents

Current signature analysis of induction motor mechanical ... · stationary nature of stator current can be accurately considered. ... and classification. The goal of feature extraction