Stator Winding Fault in Induction Motor

Transcription

1 Chapter 7 Stator Winding Fault in Induction Motor Chapter Outline Stator is one of the major fault areas in an induction motor. Stator fault initiates as a turn to turn short fault of its winding which gradually progresses to become severe. This chapter starts with a brief introduction along with thorough discussions on the origin and causes of the stator winding faults and their effects. The innovative and developed approaches which have been made for diagnosis of this fault using motor current signatures (both transient and steady state) through application of different signal processing and computational techniques have been presented as below: 1. Analysis of motor steady state current at no-load through DWT. Analysis of motor steady state current at no-load and load through DWT using reconstruction 3. Envelope analysis of motor steady state current at no-load 4. Envelope analysis of motor startup current at no- load and load 5. Analysis through extraction of harmonics using FFT Finally, the usefulness, advantages and difficulties using these methods have been discussed 7.1 Introduction Stator windings breakdown is one of the major causes of motor failure [1]. Failure of induction motors may cause plant shut down, reduced production, accidents etc. in the production line. Early monitoring and diagnosis improves the safety of the operator, reduces production time loss and minimize the expansive maintenance and repair cost []. Prevention of induction motor failure is thus a major issue in industries [3]. Most of the stator winding failures start as undetected turn to turn faults in a

2 138 Fault Diagnosis of Three Phase Induction Motor single phase due to destruction of insulation which finally grow and terminate in major faults [4]. Different types of stator winding faults that may develop are short circuit faults - (i) between turn to turn within a coil, (ii) between coil to coil of the same phase, (iii) between phase to phase, (iv) phase to earth and (v) open circuit fault [5]. Consequently, techniques able to detect these kinds of faults at an early stage of development are particularly welcome in order to prevent the catastrophic failure of the machine and also to allow for carefully planned repair actions [6]. Different types of stator winding faults are shown in Fig. 7.1 and Fig. 7.. Fig. 7.1 Different types of stator winding Fault (a) Healthy Stator Winding (b) Open circuit in one of the supply phases

3 Chapter 7 Stator Winding Fault in Induction Motor 139 (c) Turn to turn short (d) coil to coil short (e) Phase to ground short in the slot (f) Phase to phase short Fig. 7. Different types of stator winding fault in actual motor (Source: In Three Phase Stator Windings ) 7. Origin and Causes It is observed from the literature survey that 35-4 % of induction motor breakdowns are attributed to the stator winding insulation [7]. Normally, it is assumed that a large portion of stator winding related failures are instigated by insulation failures between turns of a stator coil within one phase. This type of fault is referred as a stator turn fault [8]. Various stresses are responsible for stator winding failures [9], [1]. These stresses may be classified into thermal, electrical, mechanical, and environment, among them thermal stress plays a pivotal role for the deterioration of the insulation in most of the cases. A stator turn fault in a symmetrical three-phase AC machine causes a large circulating current to flow and subsequently generates excessive heat in the s. If the heat which is proportional to the square of the circulating current exceeds the limiting value, the complete motor failure may occur [11]. (i) Thermal stress : These stresses might be due to thermal aging and thermal overloading. As a thumb rule, for every 1 C increase in temperature, the insulation life gets halved due to thermal aging. Due to thermal ageing, insulation system loses its physical integrity

4 14 Fault Diagnosis of Three Phase Induction Motor for which it fails to resist the other dielectric, mechanical, and environmental stresses. The effect of temperature on thermal ageing can be minimized either by lowering the operating temperature or by choice of high grade of insulation material. Voltage fluctuation, unbalance of phase voltage, repeated start of motor within very short interval of time may also cause thermal aging. Obstructed ventilation restricts the airflow as a result winding temperature rises due to heat generated in the stator and the rotor of the motor. Therefore, the motor should be kept clean inside and outside to ensure free flow of air for heat dissipation. If this is not practically possible, when the motor is working in hostile atmosphere, then these factors are to be taken into consideration during design stages by lowering the winding temperature and upgrading the insulation systems. (ii) Electrical stress Electrical stress leading to stator winding failures may be classified into dielectric, tracking, corona, and transient voltage conditions. The tracking occurs in the windings of the motors which operate over 6 volts and whose insulation system is not completely protected from environment. Corona discharge starts due to gaseous ionization in the insulation system where the voltage stress exceeds a critical value on the winding operating above 5 kv range. Finally, deterioration of winding insulation results due to heating, eroding, or a chemical reaction. Transient voltage conditions reduce winding life or premature failure. (iii) Mechanical stress This stress develops due to coil movement, vibration resulting from rotor unbalance, rotor striking the stator for which insulation system may be severely damaged. The force on the coils due to the stator winding current is maximum during starting or during transient loading, causing the coils to vibrate at twice the synchronous frequency with movement in both the radial and the tangential direction. This coil movement can cause damage to the coil insulation, loosen the top sticks, and may cause damage to the copper conductors. The rotor can strike the stator due to no. of reasons like broken rotor bar, worn bearings, and air-gap eccentricity. If the strike occurs during start up period, due to high striking force, the rotor may cause the stator laminations to puncture the coil insulation, resulting in premature grounding of the coil in the stator slot.

5 Chapter 7 Stator Winding Fault in Induction Motor 141 (iv) Environmental stress Contamination due to unknown materials can result in undesirable outcomes on the stator winding s insulation. The presence of foreign contaminants can cause a reduction in heat dissipation [1] due to obstruction of ventilation duct. It is therefore of utmost importance to keep the motors dirt free and dry, in particular when the motors run in a hostile environment. Other problem arises due to condensation development in the winding for which winding is grounded. By using space heater or trickle heating during off period, this can be avoided. 7.3 Effects As revealed from the literature, it has been observed that most of the stator winding failure is supposed to be the consequence of turn short within a coil in a phase. Due to turn short in the coil, high current will flow in the winding, which will produce excessive heat. As a result, the temperature in the shorted area will start rising. If the temperature exceeds the limiting value, complete damage of the winding insulation may occur, even in the worst case, serious accident may happen, involving loss of precious human life. Motor operators should keep in mind that even the best insulation may fail quickly if motor is operated above its temperature limit. As a thumb rule, for every 1 C increase in temperature, the insulation life gets halved due to thermal aging [13]. Hence care must be taken so that electrical machine will not operate beyond its thermal capacity Due to high circulating current in the, the coil will vibrate at twice the line frequency with movement both in the radial and tangential directions. As a consequence, the rotor may rub the stator. If the rotor strikes the stator when the motor is running at full load, then the result is premature grounding of the coil in the stator slot caused by excessive heat generated at the point of contact 7.4. Diagnosis of Stator Winding Fault Due to stator inter- turn short in induction motor, various harmonics or spectral components are induced in low voltage stator windings of induction motor. These harmonics can be detected by using motor current signature analysis (MCSA) through FFT as reported in the articles [14], [15]. These induced harmonics in the stator current are expressed in (3.5) of chapter 3. Traditional difficulties of FFT techniques are mainly due to spectral leakage along with others like picket fence effect, smearing, aliasing and error introduced during averaging of spectrum etc.

6 14 Fault Diagnosis of Three Phase Induction Motor which make the use of FFT restricted, specially for small or medium size motor when operating at light load or no-load. Different methods have been used for diagnosis of stator inter-turn short in induction motor through analysis of motor current during steady state or starting period at load and no-load using FFT, DWT and HT. HT has employed for finding out the analyzing envelopes of the signals Experimentation Laboratory experiment was performed using the same experimental setup and block Diagram given in A and B of subsection of chapter 5. The same motor can be used as healthy and faulty with s (5% and 1%) in stator winding. The induction motor used has provisions for shorting the turns from outside tappings which are vividly shown in Fig.7.3. The electrical specification of the motor is given in table 7.1. The motor under healthy and faulty conditions with s were run at no-load and load from direct online supply at 33V, 5Hz. Motor current signatures of R-phase have been captured using Hall Probe (LEM PR3 ACV 6V CATIII 3Ampac / 3Vac) at a sampling frequency of 48 samples/sec.

7 Chapter 7 Stator Winding Fault in Induction Motor 143 Table 7.1 Induction motor s electrical specification Make Voltage Power Frequency Speed Local (Kolkata) 44V 1 HP 5 Hz 146 rpm 7.4. Diagnosis through Discrete Wavelet Transform of Motor Steady State Current Signature (MCS) at No-Load Theoretical overview A. Generation of harmonics Due to turn short, a circulating current flows through the s. This produces a negative MMF (magneto-motive-force). This backward MMF weakens the net MMF of the motor phase and as a result, the waveform of air gap flux, which is changed by the distortion of the net MMF, induces harmonic frequencies in a stator-winding current as given in (3.5) of chapter 3 which have been used for diagnosis of inter-turn short of induction motor B. Discrete wavelet transform (DWT) DWT decomposes a given signal into its constituent wavelet subbands or levels (scales) with different time and frequency resolution. In this work, DWT has been used for decomposition of the windowed steady state no-load currents of 5 secs. duration (using Hamming window) and reconstruction of the signals. The sampling frequency for capturing the signals was 48 samples /sec. Different spectral bands corresponding to this sampling frequency are given in Table 4., section 4.3 of Chapter 4. Details of DWT are elaborately described in section 4.3 of chapter 4 and section C. of Chapter Proposed technique This study presents a novel technique for identification of induction motor stator inter-turn short circuit fault at no-load. The algorithm steps are summarised below: 1 S.K. Ahamed, M. Mitra, Arghya Sarkar and S. Sengupta, Induction Machine Stator Inter-Turn Short Circuit Fault Detection Using Wavelet Transform, Proceedings of National Conference on Recent Developments in Electrical, Electronics & Engineering Physics, RDE3P-13, MCKV Institute of Engineering, Liluah, Howrah, pp.9-93, 6-7 October 13, ISBN:

8 144 Step 1: Step : Step 3: Step 4: Fault Diagnosis of Three Phase Induction Motor Use suitable window function to obtain specific size data vector from steady state portion of stator current of the induction motor. Window function is utilized to reduce the effect of transients. Perform discrete wavelet transform in such a fashion that the detailed coefficients at higher levels correspond to narrow band frequencies below 5 Hz. Compute statistical parameters RMS values of detailed coefficients and power detailed energy (PDE) at higher levels. PDE is defined (4.3) in section 4.5 of chapter 4. Compare the statistical parameters with set values to obtain the information about the health of the Induction Motor. Step 5: Go to Step 1. Present work selects Hamming window of length 14 (5 secs. duration) samples for transient suppression. Then DWT was performed using db4 as mother wavelet Result analysis From the experiment performed as given in section 7.4.1, steady state motor current signals of 5 sampling periods at no-load for the motor under healthy and faulty conditions were separated using window technique from the motor currents captured at a sampling frequency of 48 samples/sec. and DWT was performed on these signals using db4, to extract detail coefficients at wavelet levels corresponding to spectral bands below 5 Hz.. The decomposed details were then reconstructed, shown in Fig. 7.4, Fig. 7.5 and Fig Then the RMS values of the reconstructed details were estimated and they were further processed to get PDE (Power Detail Energy). RMS values and PDE given in Table 7. are analysed for diagnosis of inter-turn short fault. The percentage variations of the RMS values and PDE for the faulty motor with respect to the healthy one are given in Table 7.3. The analysis was performed using Matlab software. The following observations are made from the Table 7. and Table 7.3, (i) 6 th, 7 th and 8 th levels indicate higher RMS and PDE values of detail coefficients for the faulty motor with respect to the healthy motor. (ii) The percentage increases of the values of RMS for the motor with 5% turn short at 6 th, 7 th and 8 th levels are 15.8, and 8. whereas the percentage increases for the motor with 1% short are 318.6, 79. and (iii) In the energy level analysis, the percentage increases of the values of PDE for the motor with 5% short at 6 th, 7 th and 8 th levels are ,

9 Chapter 7 Stator Winding Fault in Induction Motor and whereas the percentage increases for the motor with 1% short are , and Therefore, it can be inferred from the above experimental data analysis as all the three levels below 5 Hz indicate higher percentage increases in the values of RMS and PDE for the faulty motor with s than those for the healthy one, the method can be successfully used for diagnosis of inter-turn short of induction motor. Advantage Measurement at higher wavelet levels, reduces spectral leakage as they are away from 5 Hz and overlapping zone from neighboring lower side band also disappears which improves the detectability and due to application of higher order wavelet, the method works with higher resolution. Disadvantage The main drawback is for a selected sampling frequency, the spectral bands become fixed which means some ranges of frequencies (presently, 3 Hz - 5 Hz) remain unexplored. Fig. 7.4 Detail coefficients at 6 th, 7 th, 8 th levels (healthy condition) Fig. 7.5 Detail coefficients at 6 th, 7 th, 8 th levels (5% turn short)

10 146 Fault Diagnosis of Three Phase Induction Motor Fig. 7.6 Detail coefficients at 6 th, 7 th, 8 th levels (1% turn short) Table 7. Table 7.3 The RMS values and PDE of the reconstructed details for the motor with healthy and faulty conditions Wavelet Level Motor Condition Conclusion Parameters RMS PDE 6 Healthy Faulty 5% Faulty 1% Healthy Faulty 5% Faulty 1% Healthy Faulty5% Faulty 1% Wavelet Level In this investigation, a novel monitoring system has been developed to detect the stator inter-turn short circuit fault of the induction motor. The investigation reveals that the proposed method can successfully be employed for monitoring of stator winding inter-turn short fault at no-load. The developed scheme is fast, highly reliable and low cost, ideally suited for industries. Percentage variation of the RMS values and PDE of the reconstructed details for the motor with faulty conditions with respect to the motor with healthy condition Motor Condition 5% shorted turn 1% shorted turn Faulty 5% Faulty 1% Faulty5% Faulty 1% Percentage variations of Parameters RMS PDE

11 Chapter 7 Stator Winding Fault in Induction Motor Diagnosis through Analysis of Reconstructed Details using Discrete Wavelet Transform of Motor Steady State Current Theoretical background A. Theoretical concept for generation of harmonics Due to turn short, the harmonics or spectral components induced in the stator current are utilized for diagnosis of this fault which have already been discussed in section A of this chapter. B. Discrete wavelet transform (DWT) In the present study, DWT has been used for decomposition of the windowed steady state motor current signatures at no-load and load using mother wavelet - db4 into details and approximate coefficients and reconstruction of decomposed details corresponding to spectral bands below 5 Hz. The spectral frequency bands at different decomposition levels are presented in Table 4. in section 4.3 of chapter 4 for the sampling frequency used for the work 48 samples/sec. Details of DWT are discussed in subsection B of section 7.4. of this Chapter Proposed scheme This work presents a novel technique for identification of induction motor stator inter-turn short circuit fault at no-load and at load conditions. The algorithm steps are summarized below: Step 1: Step : Step 3: Use suitable window function to obtain specific size data vector from the steady state portion of stator current of the induction motor. Window function is utilized to reduce the effect of transients. Perform discrete wavelet transform in such a fashion that the detailed coefficients at higher levels correspond to narrow band frequencies below 5 Hz. Compute statistical parameters RMS values and power detailed energy (PDE) at higher level wavelet coefficients. PDE is defined in (4.3) in section 4.5 of Chapter 4. Syed Kamruddin Ahamed, Arghya Sarkar, Madhuchhanda Mitra and Samarjit Sengupta, Induction Machine Stator Inter-Turn Short Circuit Fault Detection using Discrete Wavelet Transform, Journal Innovative Systems Design and Engineering, ISSN -177(Paper) ISSN -871(Online), vol. 5, no. 1, pp

12 148 Step 4: Fault Diagnosis of Three Phase Induction Motor Compare the statistical parameters with set value to obtain the information about the health of the induction motor. Step 5: Go to Step 1. This study selects Hamming window of length 14 (5 secs. duration) samples. Then discrete wavelet transform was performed using db4 mother wavelet Result analysis Experiment has been carried out as in described in of this Chapter. Then the windowed (Hamming window) steady state portions of the captured signals of 5 secs. duration were separated for the motor under healthy and faulty conditions. Wavelet transform using db4 have been performed on the windowed (Hamming window) steady state signals (sample size of duration of 5secs.) at no-load and load for decomposition and reconstruction. Reconstructed details for the motor with the healthy and faulty conditions corresponding to 6 th, 7 th and 8 th levels i.e. spectral bands below 5 Hz are shown in Fig. 7.7, Fig. 7.8, Fig. 7.9 at no-load and Fig. 7.1, Fig. 7.11, Fig. 7.1 at load. Then the RMS values of reconstructed details and PDE were estimated and the details were processed to obtain PDE (power detailed energy) given in Table 7.4, Table 7.5. The variations of the values of RMS and PDE of the motor under faulty condition with respect to the motor under healthy condition are given in Table 7.6 and 7.7. The following observations are made from Table 7.4, Table 7.5, Table 7.6, Table 7.7 at (a) no-load and (b) at load (a) At no-load (i) (b) At load 6 th, 7 th and 8 th levels indicate higher RMS and PDE values of detail coefficients of the faulty motor with respect to the healthy motor at noload (ii) The percentage increases of the values of RMS for the motor with 5% short at 6 th, 7 th and 8 th levels are 11.34, 1. and whereas the percentage increases for the motor with 1% short are 65.5, 43.9 and respectively. (iii) In the energy level analysis, these percentage variations are 337.1, 387. and for 5% stator turn short and for 1% stator turn short, the percentage variations are 11.78, and respectively at 6 th, 7 th and 8 th levels. (i) 6 th, 7 th and 8 th levels indicate higher RMS and PDE values of detail coefficients of the faulty motor with respect to the healthy motor

13 Chapter 7 Stator Winding Fault in Induction Motor 149 (ii) The percentage increases of the values of RMS for the motor with 5% short at 6 th, 7 th and 8 th levels are 4.79, 39.8 and 3.49 whereas the percentage increases for the motor with 1% short are 43.4, 84.8, and (iii) In the energy level analysis, the percentage increases of the values of PDE for the motor with 5% short at 6 th, 7 th and 8 th levels are 95.4, 39.6 and whereas the percentage increases for the motor with 1% short are 58.86, and respectively. From the above discussion, it can be inferred that these variations in the values of RMS and PDE of the reconstructed detailed coefficients can be efficiently used for detection of stator turn short both at no-load and load condition as they indicate higher values at all the three levels. Fig. 7.7 Detail coefficients at 6 th, 7 th, 8 th levels at no-load (healthy Condition) Fig. 7.8 Detail coefficients at 6 th, 7 th, 8 th levels at no-load (5% turn short)

14 15 Fault Diagnosis of Three Phase Induction Motor Fig. 7.9 Detail coefficients at 6 th, 7 th, 8 th levels at no-load (1% turn short) Fig. 7.1 Detail coefficients at 6 th, 7 th, 8 th levels at load (healthy condition) Fig Detail coefficients at 6 th, 7 th, 8 th levels at load (5% turn short)

15 Chapter 7 Stator Winding Fault in Induction Motor 151 Fig. 7.1 Detail coefficients at 6 th, 7 th, 8 th levels at load (1% turn short) Table 7.4 RMS value and PDE of detailed coefficients for the motor with faulty conditions and the motor with healthy condition at no-load Table 7.5 RMS value and PDE of detailed coefficients for the motor with faulty conditions and the motor with healthy condition at load Wavelet Level Motor Parameters Condition RMS PDE Healthy Faulty 5% Faulty 1% Healthy Faulty 5% Faulty 1% Healthy Faulty5% Faulty 1% Wavelet Level Motor Condition Parameters RMS PDE Healthy Faulty 5% shorted turn Faulty 1% Healthy Faulty 5% shorted turn Faulty 1% Healthy.5.8 Faulty5% shorted turn Faulty 1% Table 7.6 Table 7.7 Variation of parameters for the motor with faulty conditions with respect to the motor with healthy condition at no-load Wavelet Level Motor Condition 5% shorted turn 1%shorted turn 5% shorted turn 1%shorted turn 5% shorted turn 1%shorted turn Percentage variations of Parameters RMS PDE 11, Variation of parameters for the motor with faulty conditions with respect to the motor with healthy condition at load Wavelet Level Motor Condition 5% shorted turn 1% 5% shorted turn 1% 5% shorted turn 1% Percentage variations of Parameters RMS PDE

16 15 Fault Diagnosis of Three Phase Induction Motor Conclusion In this work, a novel monitoring system has been described to detect the stator inter-turn short circuit fault of the induction motor. The spectral leakage is minimized and the effects of transients are suppressed due to application of window function. The investigation reveals that the proposed method can successfully be employed for on-line monitoring of stator winding turn short both at no-load and load. The developed scheme is fast, highly reliable and low cost, ideally suited for industries Diagnosis through Envelope Analysis of Motor Steady State Current at No- Load Theoretical overview A. Stator Current envelope of an induction motor at the presence of interturn short of stator winding In the investigation of inter-turn short, the stator current has long been used. The faults affect the current according to motor slip as given in (3.5) of chapter 3. The effects cause fluctuations on the current which is modulated. The modulation of the current is the so-called envelope which is cyclically repeated. B. Hilbert transform and envelope detection Hilbert Transform is used for finding out the envelopes of the signals which work on instantaneous frequency. Envelope is the modulus of complex analytic signal. It is a new dimension in the area of fault detection from the spectrum analysis of envelope which is almost free from 5 Hz frequency. Details of Hilbert transform, instantaneous frequency are discussed in section 4.4 C. Discrete wavelet transform The decomposition and reconstruction properties have been utilised for analysis motor steady state current. Spectral bands corresponds to different wavelets for the present sampling frequency 48 samples/sec. are given in Table 4. in section 4.3. Details of DWT are elaborately discussed in section Syed Kamruddin Ahamed, Arghya Sarkar, Madhuchhanda Mitra and Samarjit Sengupta, Novel Approach for Detection of Inter-Turn Short Circuit of Induction Motor s Stator Winding through Envelope Analysis, 8th International Conference on Electrical and Computer Engineering, ICECE 14, Dhaka, Bangladesh, pp , - December 14, IEEE, Available in IEEE Xplore.

17 Chapter 7 Stator Winding Fault in Induction Motor Proposed technique This work presents a novel approach for detection of induction motor stator inter-turn short circuit fault at no-load condition. The algorithm steps are summarised below: Step 1: Step : Step 3: Step 4: Use suitable window function to obtain specific size data vector from the steady state portions of stator current of the induction motor. Window function is utilized to reduce the effect of transients. Perform discrete wavelet transform in such a fashion that the reconstructed detailed coefficients at higher levels correspond to narrow band frequencies below 5 Hz. Compute fault parameters RMS values, mean values of the detailed coefficients and power detailed energy (PDE) at higher levels. PDE is defined in (4.3) in section 4.5 of Chapter 4. Compare the statistical parameters of the faulty motor with respect to the healthy one to obtain the information about the health of the induction motor. Step 5: Go to Step 1. This work selects Hamming window of length 48 (one second duration) samples. Then discrete wavelet transform was performed using db4 mother wavelet which has the most compact support for a given number of vanishing moments, thus allowing the lowest computational loads Result analysis Discrete Wavelet Transform using db4 were performed to decompose the envelopes of windowed steady state signals (sample size of duration of one sec. i.e. 48) at no-load. The signals and the envelopes of the steady state signals at no load for the motor under healthy, 5% and 1% inter-turn short conditions are shown in Fig Then the decomposed details below 5 Hz were reconstructed, shown in Fig.7.14, Fig.7.15 and Fig The reconstructed details for the motor under healthy and faulty conditions corresponding to 6 th, 7 th and 8 th level below 5 Hz are the analyzing signal. Mean, RMS values of the reconstructed details and PDE at all the three levels were then estimated shown in Table 7.8 and graphically represented in Fig The following observations are made from Table 7.8 and Fig (i) All the three levels indicate considerable higher values of RMS, mean and PDE for the motor under faulty conditions than those for the motor under healthy condition.

18 154 Fault Diagnosis of Three Phase Induction Motor (ii) The most significant increases of these values are noticed at 8 th level At 8 th level, the RMS values are 1.74 times and 6.61 times for 5% and 1% s with respect to the values for healthy condition whereas the rises in the mean values are 1.57 and 6. times for 5% and 1% s. (iii) In the energy level analysis at 8 th level, the values of PDE are 3.1 and 4.84 times for 5% and 1% s with respect to the values for healthy condition. Hence 8th level is considered the most suitable for detection of stator inter-turn short. The main advantages in this method are - the envelope analysis which makes diagnosis easier and cleaner as the power frequency is eliminated and the spectral leakage is minimized and the application of DWT on the steady state current which overcomes the difficulty of FFT analysis. The main drawback is that once sampling frequency is selected, some ranges of frequencies remain unexplored. In the present analysis, it is (3 Hz - 5 Hz). (a) (b) (c) Fig Signals and their envelopes (red) under (a) healthy, condition, (b) faulty condition with 5% turn short and (c) faulty condition with 1% turn short at no-load

19 Chapter 7 Stator Winding Fault in Induction Motor 155 Fig The detailed coefficients at 6 th, 7 th and 8 th levels of envelopes for motor under healthy condition at no load Fig The detailed coefficients at 8 th, 9 th and 1 th levels of envelopes for motor with 1% inter-turn short at no load Fig The detailed coefficients at 6 th, 7 th and 8 th levels of envelopes for motor with 5% inter-turn short at no load

20 156 Fault Diagnosis of Three Phase Induction Motor Table 7.8 Fault parameters of the detailed coefficients of the envelopes Motor Condition Healthy Wavelet Level RMS value Mean value Power Detail Energy in joules % short % short Healthy % short % short Healthy % short % short (a) (b) (c) Fig (a) Mean Curve, (b) RMS Curve, (c) PDE Curve

21 Chapter 7 Stator Winding Fault in Induction Motor Conclusion The present approach reveals that this method can successfully be employed for detection of stator winding inter-turn short. The method works with higher detectability and higher resolution due to DWT and Hilbert transform. The method is faster with less power consumption and hence less costly which makes it ideally suitable for industry Diagnosis through Tracking of Low Frequency Oscillation using Envelope Analysis of Motor Startup Current Theoretical background A. Stator current envelope of an induction motor at the presence of stator inter-turn short Inter-turn short-circuits in stator winding causes a profile modulation on the three phase stator current leading to envelope cyclically repeated at a rate equal to the power frequency (f). An inter-turn short-circuit principally affects only the stator current of the faulty phase in both profile and peak value. The other stator phase currents suffer smaller interferences. Thus, the stator current profile of each phase is not equally affected by the fault. The frequency of repetition of this envelope is the power frequency, f and not a function of the slip frequency, sf. The stator current envelope of the healthy phases is slightly affected by the faulty phase, while the envelope of the faulty phase is highly modulated. The characteristic(spatial) harmonics in the stator current due to change in the air gap flux caused by stator inter-turn short in stator winding as revealed from the literature review expressed in (3.5) of Chapter 3. B. Discrete wavelet transform (DWT) Wavelet transform provides flexibility in describing signals that include regions of different frequency contents. Wavelets are localized in both time and frequency domain for which wavelet signal processing technique is suitable for those signals whose frequency content changes with time, especially transient signal [16], [17]. In discrete wavelet transform, the mother wavelet is not scaled continuously, but scaled in the power of. Hence, it is easy to implement in digital computers and takes less execution time [18]. DWT decomposes a sampled signal s(t) by passing it 4 Syed Kamruddin Ahamed, Samarjit Sengupta, Madhuchhanda Mitra and Arghya Sarkar, Tracking of Low Frequency Oscillations for detection of Inter-Turn Short of Stator Winding of Induction Motor Through Envelope Analysis using Startup Current, International Journal of Computer and Electrical Engineering (IJCEE)

22 158 Fault Diagnosis of Three Phase Induction Motor through HPF (high pass filter) and LPF (low pass filter) into its approximate signal a n and several detail signals d n [19] as given in (7.1) n j j s() t n n t t i i i j 1 i i i an dn d 1 (7.1) where n i, j i are scaling and wavelet function, n () t is the scaling function at level n and j t is the wavelet function at level j; a n & decomposition level n and detail signal at level j []. d j is approximate signal at In the present work, DWT has been used for decomposition of the starting current signatures at no load and load. The decomposed details corresponding to wavelet levels below 5 Hz. were determined and analyzed. After decomposition of the signal, each detail at different levels was analyzed. The spectral frequency bands corresponding to sampling frequency - 48 samples/sec for present analysis are shown in Table 4. in section 4.3 of chapter 4. Details of DWT are given in section 4.3 of Chapter 4. C. Hilbert transform and envelope analysis Hilbert Transform is used for finding out the envelopes of the signals which work on instantaneous frequency. Envelope is the modulus of complex analytic signal Proposed methodology This investigation presents a new technique for detection of inter-turn short of stator of induction motor at no-load and load condition. The algorithm follows the following sequence of steps: Step 1: Capture motor current signatures at sampling frequency 48 samples/sec. under healthy and faulty conditions with 5% and 1% inter-turn short and separate the starting portions of the signatures for analysis. Step : Perform Discrete Wavelet Transform using db8 in such a fashion that the detailed coefficients at higher levels correspond to narrow band frequencies below 5 Hz. Step 3: Determine the envelopes of the decomposed details at 6 th, 7 th and 8th levels corresponding to spectral bands below 5 Hz. Step 4: Compute statistical parameters of these envelopes RMS values, standard deviation, variance of higher level details corresponding to spectral bands

23 Chapter 7 Stator Winding Fault in Induction Motor 159 below 5 Hz Step 5: Compare the statistical parameters with set values to obtain the information about the health of the Induction Motor Step 6: Go to Step 1. Step 7: Algorithm will continue till the estimation and comparison of all the statistical parameters are completed for the levels under considerations. Here the set values are the estimated parameters for the healthy motor under no-load and load condition Results and discussion The experiment has been carried out accordingly as described in section The signatures were captured at a sampling frequency of 48 samples/sec for the motor under healthy and faulty conditions with 5% and 1% s at no-load and load. The starting portions of the captured motor current signatures were separated and decomposed using db8 of Daubechies family shown in Fig. 7.18, Fig Then the envelopes of the decomposed details at 6 th, 7 th and 8th levels corresponding to spectral bands below 5 Hz were obtained. The decomposed details d 6, d 7, d 8 at 6 th, 7 th and 8th levels of the starting current transients and their envelopes for the motor under healthy and faulty conditions with shorted inter-turns (5% and 1%) at no- load and at load are shown in Fig. 7., Fig. 7.1, Fig. 7., Fig. 7.3, Fig. 7.4 and Fig. 7.5 respectively. Then the statistical coefficients RMS values, standard Deviations (STD) and variances of the envelopes of the details d 6, d 7, d 8 at 6 th, 7 th and 8th levels of the starting current transients for the motor under healthy and faulty conditions with shorted inter-turns (5% and 1%) at no- load and at load were estimated and are shown graphically in Fig. 7.6, Fig. 7.7, Fig. 7.8, Fig. 7.9, Fig and Fig These statistical coefficients are considered as fault parameters or features. The variations of the statistical parameters were calculated and are given in Table 7.9. The following observations are made (i) The parameter variations for 5% and 1% inter-turn short under faulty condition at no-load indicates very higher percentage increases at 6 th, 7 th and 8 th levels except at 8 th level for 5% inter-turn short, the variations in the values of RMS, standard deviation (STD) are 3.9% and.8%, not too small. But the variations in the values of variance at 8 th level are quite high, 49.49% for 5% and 13.6% for 1% inter-turn shorts.

24 16 Fault Diagnosis of Three Phase Induction Motor (ii) At load, values of RMS, standard deviation (STD) and variance decreases at 6 th level for 5% and 1% inter-turn short from which no inference can be drawn. At 7 th level for 5% inter turn short, the percentage variations in the values of RMS, standard deviation and variances are 13.8, and 35. which are not too high but trend is upward which signifies the presence of inter-turn short. The 8th level indicates very small rise in the values for 5% inter-turn short which can never be considered for detection of this type of fault. In the case of 1% inter-turn short of the motor under load, the increases in the values of the parameters - RMS, standard deviation (STD) and variances are 13.65, and in percentage at 7 th level, very higher rise of the values whereas at 8 th level, the rise in the variations in these parameters in percentage are 38.5, and , more or less sufficient to detect inter turn short Therefore, for 1% inter-turn short at load, 7 th level is most sensitive to detect fault and 8th level may also be used for diagnosis whereas at no-load, all the three levels 6 th, 7 th and 8 th are sufficient to detect inter-turn short fault. The main advantage of this method is that the envelope is free from supply frequency and hence the spectral leakage is minimized which improves detectibility. The resolution is higher due to choice of higher order wavelet and its use at higher wavelet level to extract left side harmonics below 5 Hz is possible due to enhancement of amplitudes by HT. The main disadvantage is that for selected sampling frequency, the bands become fixed which means some ranges of frequencies are not covered. Here in the present case, the range (3 Hz - 5 Hz) is unexplored (a) (a)

25 Chapter 7 Stator Winding Fault in Induction Motor NO.OF SAMPLES (b) 4 (b) (c) Fig Starting current signatures with (a) healthy, (b) 5% inter-turn short and (c) 1% interturn short conditions at no-load (c) Fig.7.19 Starting current signatures with (a) healthy, (b) 5% inter-turn short and (c) 1% inter-turn short conditions at load (a) (a) (b) (b)

26 16 Fault Diagnosis of Three Phase Induction Motor (c) Fig. 7. Details and their envelopes (red) at (a) 6 th, (b) 7 th and (c) 8 th levels at no-load with healthy condition (a) (b) (c) Fig. 7.1 Details and their envelopes (red) at (a) 6 th, (b) 7 th and (c) 8 th levels at no-load with 5% inter-turn short condition (a) (b) (c) Fig. 7. Details and their envelopes ( red) at 6 th, (b) 7 th and (c) 8 th levels at no-load with 1% inter-turn short (c) Fig. 7.3 Details and their envelopes ( red) at 6 th, (b) 7 th and (c) 8 th levels at load with healthy condition

27 Chapter 7 Stator Winding Fault in Induction Motor (a) (b) (c) Fig. 7.4 Details and their envelopes (red) at 6 th, (b) 7 th and (c) 8 th levels at load with 5% inter-turn short (a) (b) (c) Fig.7.5 Details and their envelopes (red) at 6 th, (b) 7 th and (c) 8 th levels at load with 1% interturn short

28 164 Fault Diagnosis of Three Phase Induction Motor Fig. 7.6 RMS curve of the envelopes of the details of no-load startup current at 6 th, 7 th and 8 th levels Fig. 7.7 RMS curve of the envelopes of details of load startup current at 6 th, 7 th and 8 th levels Fig. 7.8 Standard deviation curve of the envelopes of the details of no-load startup current at 6 th,7 th and 8 th levels Fig. 7.9 Standard deviation curve of the envelopes of details of load startup current at 6 th, 7 th and 8 th levels Fig. 7.3 Variance curve of the envelopes of the details of no-load startup current at 6 th, 7 th and 8 th levels Fig Variance curve of the envelopes of the details of load startup current at 6 th,7 th and 8 th levels

29 Chapter 7 Stator Winding Fault in Induction Motor 165 Table 7.9 Variations of the parameters of the envelopes of the details for the motor with faulty conditions with respect to the motor with healthy condition at no-load and load Wavelet level 6 Motor Condition Percentage (%) variation of Parameters at no-load Percentage (%) variation of Parameters at load RMS STD Variance RMS STD Variance Faulty 5% Faulty 1% Faulty 5% Faulty 1% Faulty5% Faulty 1% Conclusion The main contribution in this study is the introduction of envelope which works on instantaneous frequency. It can handle short data effectively for which this method is suitable for online industrial application Diagnosis through Extraction of Harmonics using MCSA Technique Theoretical background A. Generation of spectral harmonics or spectral components in stator current signature Due to change and distortion in the air gap flux caused by stator winding interturn short, several harmonics or spectral components are induced in the motor current signatures given in (3.5) of Chapter 3. These harmonics were utilized for diagnosis of the present fault. B. Signal Decomposition and reconstruction using discrete wavelet transform B1. Decomposition of the signal using discrete wavelet transform (DWT) The discrete captured signature were decomposed at first level and then reconstructed to obtain the original signal for analysis using FFT technique. A first level decomposition and reconstruction are shown in Fig. 7.3 and Fig Syed Kamruddin Ahamed, Arghya Sarkar, Madhuchhanda Mitra and Samarjit Sengupta, Harmonic Extraction for Detection of Induction Motor Stator Inter-turn Short through MCSA, ESM-15, International Journal of Engineering Technology, Management and Applied Sciences, volume 3, Special Issue, ISSN , pp. 17 7, September 15.

30 166 Fault Diagnosis of Three Phase Induction Motor B. Reconstruction of the signal The captured signal in the form of discrete dataset is required to be reconstructed perfectly so that no important spectral information to diagnose the fault is lost before applying FFT. This is successfully done by first level decomposition and reconstruction.first level decomposition produces approximate and detailed coefficient a 1 and d 1, which have been reconstructed to get the original signal expressed as in (7.). x( t) a1 d1 x / t (7.) For perfect reconstruction, x(t)=x / (t) where x(t) and x / (t) are original and reconstructed signal Fig. 7.3 A single level decomposition Fig A single level reconstruction C. Fast Fourier transform Before discussion on FFT, Fourier Transform and discrete Fourier Transform are required to be to be understood first. C1. Fourier transform It is a signal processing technique which converts a time domain signal into frequency domain signal. This is mathematically defined in (4.1) in section 4. of chapter 4 [1] C. Discrete Fourier transform It is the discrete version of Fourier transform. In the field of digital signal processing, the discrete Fourier transform (DFT) is used. It is employed to analyze the frequencies contained in sampled signals and is defined in (4.) in section 4. of chapter 4 []. As the number of point increases in DFT, the number of arithmetic operation will be much larger, the analysis will be very difficult. The Fast Fourier Transform does not refer to a new or different type of Fourier transform. It refers to a very efficient algorithm for computing the DFT. FFT is a

31 Chapter 7 Stator Winding Fault in Induction Motor 167 faster version of discrete Fourier transform (DFT) [3]. FFT breaks the set of data to be transformed into smaller data sets. At each stage of processing, the results of the previous stage are combined in special way. Finally, the DFT of each small data set may be obtained. The Fast Fourier Transform is the most conventional approach for implementing the DFT in real time. Power spectrum obtained through FFT using MCSA is widely used technique for detecting induction motor faults, it is also cheaper and simpler. The amplitude of the power spectrum from FFT [4] can be determined using (4.3), given in section 4. of Chapter Proposed technique This work presents a new technique for detection of inter-turn short fault of induction motor stator winding. The algorithm follows the following sequence of steps: Step 1: Step : Step 3: Capture motor current signatures at sampling frequency 48 samples/sec under healthy and faulty conditions with 5% and 1% inter-turn short and separate steady state portions of the sampled dataset of window size of 819 samples. Multiply the selected portions of the datasets by Hanning window. Perform Discrete Wavelet Transform using db8 for first level decomposition to obtain the approximate coefficient a 1 and detailed coefficient d 1. Step 4: Step 5: Step 6: Step 7: Reconstruct these first level decomposed coefficients to obtain the original signal Eliminate 5Hz- the supply frequency from the resultant signatures by using notch Band filter (band used 49Hz-51Hz) Perform FFT on these filtered signals to get power spectrum and to determine the spectral components. Algorithm will continue till the estimation and comparison of spectral components for the motor under healthy and faulty conditions with 5% and 1% s are completed This work selects Hanning window to minimize leakage and to suppress transients. Then discrete wavelet transform was performed using db4 mother wavelet for perfect reconstruction of the signal. The spectral frequency bands at first level decomposition for the sampling frequency used in this work are presented in Table 7.1.

32 168 Fault Diagnosis of Three Phase Induction Motor Table 7.1 Spectral bands at first level decomposition Decomposition Details Frequency Bands(Hz) Results and discussion The experiment was carried out accordingly as given in Then the captured signatures at no-load and load for the motor under healthy and faulty conditions with 5% and 1% turn short have been windowed (Hanning window, sample size 819). The windowed signatures were then decomposed and reconstructed using db4. From these reconstructed signatures, 5 Hz was eliminated using band notch filter (49Hz-51 Hz). Finally, FFT have been performed on these signatures for spectrum analysis. The spectra at no load are shown at Fig. 7.34, Fig and Fig. 7.36, whereas the spectra at load are given in Fig. 7.37, Fig and Fig Then the percentage increases in the amplitudes of the harmonics at no-load and load for the motor healthy and faulty conditions were determined and presented in Table 7.11 and Table 7.1. The following observations are made (i) (ii) At no-load, percentage increases in the amplitudes of 5th and 9 th harmonics are higher for both 5% and 1% turn short whereas the amplitudes of 3 rd and 7th harmonics indicate lower values for 1% turn short. Again the side band harmonics close to fundamental show considerable higher increase in the amplitudes. These side band harmonics though readable, are not sharply distinguished. Hence they are not suggested for detection where 5 th and 9 th harmonics may be used for diagnosis of turn short. At load, with respect to the motor under healthy condition, the amplitudes of 7 th and 9 th harmonics of the motor under faulty conditions with 5% and 1% turn short show very sharper increase whereas for 1% turn short, the amplitudes of 3 rd and 5 th harmonics decrease considerably. Moreover the side band harmonics close to fundamental indicate higher increase in the values of their amplitudes and can be used for diagnosis of turn short. Therefore, under load condition, 7 th and 9 th harmonics including sidebands component can be used for detection of turn short. Hence finally, it can be inferred that 9th harmonic is most sensitive to detect inter-turn short fault both at load and no load. Approximate Level 1-51 Detail Level