A Novel Approach for Detection of Shorted Turns Fault in Machine Using Combination of Flux and Instantaneous Power Signal

Transcription

1 A Novel Approach for Detection of Shorted Turns Fault in Machine Using Combination of Flux and Instantaneous Power Signal Intesar Ahmed *a, Manzar Ahmed b, M. Shuja han**a and ashif Imran a Abstract In this paper, non-invasive multiple sensors for the detection of shorted turns fault have been used. The fault has been detected using fault frequency components from flux and instantaneous power and a single signal (combination of flux and instantaneous power signals) called power-flux over a wide range of loads. To achieve this task, stator winding of the machine was re-wounded with taps to introduce shorted turns fault. An extensive series of laboratory tests were conducted to examine the fault frequency amplitudes for the healthy motor and faulty motors with different shorted turns. Results based on the variations in the amplitude of sideband components between healthy and faulty motors versus different frequencies over a wide range of loading conditions from different signals have been analyzed experimentally in detail. Overall, this paper provides comprehensive detail to detect the shorted turn faults in machines using a novel combination of axial flux and instantaneous power signals into a single signal called powerflux. Index Terms Condition Monitoring, Non-invasive techniques, Shorted turn fault, Power-Flux signal. I. INTRODUCTION It has been reported that 4% of machines failures are due to shorted turn faults [1, 2]. It is well understood that the impact of shorted turns in stator winding at early stages on the induction machines is small. Shorted turn faults result from insulation failure in a part of the winding. This insulation failure can be caused by thermal damage due to excessive currents, insulation breakdown as the result of voltage spikes/transients and mechanical damage. Shorted turn faults can produce either large fault currents which will rapidly trip the circuit breaker or else produce little external symptoms. In the latter case, however, the localized heating caused by the fault will gradually result in further insulation damage until the motor fails. It is useful to be able to detect such faults at an early stage so that a pre-planned shutdown can be arranged for the motor to be replaced by a healthy motor. Various causes of stator and rotor failures have been presented and discussed in detail. A specific methodology is proposed to facilitate an accurate analysis of these failures [3]. An approach, based on the computer-aided monitoring of the motor current Park s Vector representation is presented in [4]. Stator faults estimation Figure 1: View of shorted turns in stator winding of faulty motor. in induction machines have been detected by using Particle Swarm Optimization and Residual Saturation Harmonics in [5, 6]. Another method of detecting shorted turn stator fault is based on a fuzzy model [7]. A combination of wavelet and power-spectral-density techniques for the detection of shorted turn faults in induction motors has been reported in [8]. Although, condition monitoring is normally done for induction motors [9], and it can also be carried out for induction generators as well [1]. To overcome any shortcoming concerning the detection of broken rotor bar fault in the machines using the fault frequency components from the current, flux and instantaneous power signals under different levels of loading is presented in [11]. A comprehensive literature review of more than 2 existing methods, including the most common methods to assess the phase-to-ground, phase-to-phase, and turn-to-turn insulation conditions is presented in [12, 13]. Investigation of Multiple Faults Detection in Electric Machine Using Broken Rotor Bar and Eccentricity Fault Frequencies Techniques has also been described in [14]. This paper explores non-invasive multiple-sensor types for the detection of shorted turns fault using a novel approach of combination of fault frequency components of a single signal called power-flux over a wide range of loads. To achieve this task, stator winding of the machine was rewounded with taps to introduce shorted turns fault as shown in Fig. 1. An extensive series of laboratory tests were conducted to examine fault frequency amplitudes for the healthy and faulty motors with different shorted turns and at different loading conditions ranging from 3 to 12 percent load as given in Table 1. TABLE 1: NUMBER OF TURNS SHORT-CIRCUITING TO INVESTIGATE THE SEVERITY OF SHORTED TURNS FAULT VERSUS PERCENT OF RATED LOADS. Manuscript received September 6, 21.Department of Electrical Engineering, a COMSATS Institute of Information Technology, Lahore, Pakistan, b University of South Asia, Lahore, Pakistan,( *drintesarahmad@ciitlahore.edu.pk, **shuja@ciitlahore.edu.pk). 233 % of Rated Number of shorted turns Load

2 II. EXPERIMENTAL SETUP The experimental work was conducted using the test rig and data acquisition system as shown in Fig. 2. The tests were conducted on a set of healthy and faulty three-phase induction motors (T-DF1LA, manufactured by Brook Crompton), which were loaded by a dynamometer consisting of a 5 kw separately excited DC machine, using a variable-resistance bank. The detailed technical specifications of the induction motors under test are shown in Table 2. The induction motor under test was mechanically coupled to a separately excited DC machine, which was loaded using a variable-resistance bank. The wiring diagram of the AC and DC systems is shown in Fig. 3. The specifications of the DC machine used as a load in the testing are shown in Table 3. During the tests, the two current sensors for line currents, two voltage sensors for line voltages, one axial leakage flux sensor, two vibration sensors: drive-end horizontal (DEH) and drive-end vertical (DEV), and one speed sensor directly connected to the shaft of motor, were used. The position of the sensors in the test arrangement is shown in Fig. 3. The axial leakage flux measurement was taken with a circular search coil of comparable diameter to the motor, which is mounted concentrically with the shaft on the rear of motor. For reliability and consistency of the flux measurements, a fixed position for the flux coil is defined. The vibration sensors were screw-mounted to the motor housing to achieve the highest measurement bandwidth. Measurements of the input stator currents and voltages were taken using a custom-built measurement box that was located between the auto-transformer and the test motor. The speed sensor was directly attached to the motor shaft to obtain the running speed of motor as shown in Fig. 3. The analog signals from the sensors are passed through low-pass filters to remove any high frequency components that may cause aliasing. This is performed by an 8 channel, 8th order Butterworth analog low-pass filter unit with selectable cut-off frequencies of 1 Hz for the 4 Hz sampling frequency and 2 khz for the 8 khz sampling frequency. Table 4 illustrates the specifications of the sensors used for measurements in this study. TABLE 2 :TECHNICAL INFORMATION FOR THE INDUCTION MOTOR USED IN THE TESTING. Performance Data Value Rated voltage (V) 415 Rated frequency (Hz) 5 Rated current (A) 4.8 Power (kw) 2.2 Number of Poles 4 Rated speed (rpm) 1415 Rated torque (Nm) 14.8 Number of rotor slots 32 Power factor.81 TABLE 3 :TECHNICAL INFORMATION ABOUT THE DC MACHINE USED IN THE TESTING. Performance Data Value Rated shaft power (kw) 5.5 Rated speed (rpm) 15 Rated armature voltage (V) 22 Rated armature current (A) 28 Rated field current (A).65 There are number of factors which can affect the accuracy of the measurements. Some of these factors are the rigidity of the foundation of the test motor and the accuracy of the shaft alignment. The motors must have a firm and rigid foundation to eliminate soft foot problems and to reduce vibration [1]. During the measurements the sensors were sampled simultaneously and two different sampling rates used in this study are as follows: Low-frequency measurement with a 4 Hz sampling frequency that gives a Nyquist frequency of 2 Hz, and 1 second sampling time, which allow very highresolution frequency analysis (4, data points,.1 Hz resolution). Mains supply V A V B V C Current Voltage DEV Vibration Signal processing Low-pass filter Data acquisition system PC Axial Flux DEH Vibration Induction Motor 5 kw DC generator Speed Figure 3: The block diagram of the test set-up including the positions of the sensors [9]. Figure 2: Data-acquisition hardware (left) and motor/load test set-up (right). [1] High-frequency measurement at 8 Hz sampling frequency with a sampling time of 5 seconds (4, data points,.2 Hz resolution). 234

3 The parameters of the two sampling rates, achieved frequency resolution, frequency range, setting of antialiasing filter and usable frequency ranges of each sampling rate used for experiments are given in [15]. Faulty motor (n=1 and k= -1) at 3% load III. SHORTED TURNS FAULT DETECTION USING FLUX SPECTRUM Measured fault frequencies and their respective magnitudes for the healthy and faulty motor (2 shorted turns) for n=1 and k= ±1, ±3, ±5 at 3% is shown in Fig. 4 and Fig. 5 respectively. The spectra showed significant decrease for both negative and positive sideband in the amplitudes of fault frequency components regarding the healthy and faulty motors. The amplitudes of the sidebands are clearly visible in the flux spectrum for different values of n and k, even in the presence of other peaks and high noise level when compared to the current spectrum. The noise level is about 65 db. The results show that it is useful to detect the shorted turn faults at light loads using the flux signal. However, the features of flux signal cannot provide useful information to detect the shorted turn faults at high loads. Since the flux signal showed could detect shorted turn faults at light loads therefore, the next section will examine the variability in healthy motors. This is made possible by test repeatability, difference between identical motors, and difference between the phases of the same motors tests. This will confirm the accuracy and reliability of the test results in finding shorted turn faults using shorted turn fault frequencies components. Healthy motor (n=1 and k= -1) at 3% load Figure 5: Under 3%-load test conditions, variations in the amplitudes of the healthy and faulty motors were found to be less than 4 db for all values of k except for k=1, which is clearly visible in the spectra noise level = - 65 db Faulty motor (n=1 and k=1) at 3% load Healthy Faulty(2 ST) at 3% load -1 () Healthy motor (n=1 and k=+1) at 3% load noise level = - 65dB Healthy Faulty(2 ST) -8-1 () Figure 6: Comparison of variation in shorted turn sideband amplitude versus percent of rated load of healthy motor at 3% load and. Figure 4: Highlights variations in the amplitudes of sideband components for n=1 and k= ±1 is more than 14 db for both the negative and positive sideband components at 3% load. A. Flux Spectrum Variations for Healthy Machines To examine the accuracy and reliability of the test results to detect shorted turn faults using fault frequency components from flux signal, a set of three tests on the same 235

4 healthy motor was conducted. Each test was repeated after the motor had been removed from the test rig and returned. Fig. 7 below summarizes the test results of three consecutive tests on the same healthy motor at 3% load for the different values of k. The results show the maximum variations in the sideband amplitudes for the healthy motor is less than 4 db at 3% load, which is not significant and does not affect the accuracy and reliability of the test results obtained by using the shorted turn fault frequency components. The faulty motor result is shown for reference. Fig. 8 shows a comparison of the variations in the amplitudes of the shorted turn sideband components versus % of rated load for two similar rating (2.2 kw) of motors and the variations for any value of k is less than 4 db, which are not significant. The finding demonstrates the reliability of test results. Fig. 9 shows variations in the sideband amplitudes between phases A and B for a healthy motor and faulty motor with 2 shorted turns for different values of k at 3% load. The variations in the amplitudes of fault frequency components for the two phases of the healthy and faulty motors showed slight variations of less than 3 db for any value of k. In summary all three tests showed that the test results are accurate and reliable noise level = -65 db test 1 test 2 test 3 faulty Load = 3% Figure 7: Comparison of variations in flux sideband amplitudes for three tests on the same healthy motor noise level = -65dB motor1 motor 2 faulty Load = 3% Figure 8: Variations in shorted turn sideband amplitudes for tests on two nominal identical healthy motors. The result of faulty motor is shown for reference noise level = -9 db phase A faulty A Load = 3% phase B faulty B Figure 9: Variations in shorted turn sideband amplitudes for tests on two nominal identical healthy motors. Faulty motor result is shown for reference. IV. SHORTED TURNS FAULT DETECTION USING INSTANTANEOUS POWER SPECTRUM The Frequency spectra of the power signal for a healthy and a faulty motor for n=1 and k=±1 are shown in Fig. 1, which are used to detect shorted turn faults in the machines. The spectrum showed the significant variations emerged in the amplitude of fault frequency sideband components in the faulty motor and are clearly visible in comparison to the healthy motor. The results shown in Fig.1 are much clearer separated in the amplitude of sideband component at 1% load for n=1 and k=±1. It can be concluded that the instantaneous power signal showed slightly more variations in the amplitude of negative sideband components at 1% load when compared to the result for the flux spectra. However, results obtained at 3 % load showed slight variations in the amplitudes of fault frequency components for the healthy and faulty motor. Therefore, on the basis of these results, instantaneous power is useful for detecting shorted turn faults at 1% load. Since the instantaneous power signal can detect shorted turn faults at higher loads, it is important to examine the variability of the instantaneous power signal in order to confirm the reliability of the test results to detect shorted turn fault using instantaneous power signal. A. Instantaneous Power Variations For Healthy Machines Fig.12, 13 and 14 depict the healthy motor results regarding: repeatability tests on the same motor; differences between motors; and differences between the motor s phases for confirming the reliability of the test results in detecting shorted turn faults from the instantaneous power signal. Fig.12 summarizes the variation in the sideband amplitudes versus different values of k at 1% load and a set of three tests. The variation was found to be less than 3 db, which is very small. Fig. 13 compares the instantaneous power sideband amplitudes versus different values of k at 1% load for the two nominally identical healthy motors and demonstrated a 236

5 small variation of 4 db, which is not significant. Fig. 14 highlights the variation in the sideband amplitudes between phases A and B for a healthy and faulty motor. The variation in the amplitudes for both phases is less than 3 db, which is also very small. It can be concluded from the three different tests on the healthy motor that no significant variations were found in the amplitudes of fault frequency components, and this confirms the accuracy and reliability of the test results noise level = - 95 db Instt. power Spectrum : f p[(n/p)(1-s)+/-k)] Healthy Faulty(2 ST) at 3% load Healthy motor (n=1 and k=-1) -1 () Healthy motor (n=1 and k=+1) Figure 11: Variations in the shorted turn sidebands amplitudes versus % of rated load of healthy motor for different values of k from instantaneous power spectrum. Intt.Power Spectrum : f p[(n/p)(1-s)+/-k)] -2 Faulty motor (n=1 and k=-1) at full- load test 1 test 2 faulty test 3 Load = 1% noise level = -95 db -1 Figure 12: Variations in shorted turn sideband amplitudes for three tests on the same healthy motor. Instt.Power Spectrum : f p[(n/p)(1-s)+/-k)] -2 Faulty motor (n=1 and k=+1) -4-6 motor 1 motor 2 faulty Load = 3% -8 noise level = -95 db Figure 1: indicates variations in the amplitudes of the healthy and faulty motor with 2 shorted turns at 1% load tests for n=1 and k=±1. -1 Figure 13: Variations in the shorted turn sideband amplitudes for tests on two identical healthy motors at 1% load. 237

6 Instt. Power Spectrum : f p [(n/p)(1-s)+/-k)] phase A phase B faulty B faulty A Load = 3% db db Instt.Power : f p [(n/p)(1-s)+/-k)] k=1 healthy motor faulty motor -1 noise level = -95 db Figure 14: Variations in shorted turn sideband amplitudes of the healthy motor between phases A and B. V. COMBINATION OF FLUX AND INSTANTANEOUS POWER SPECTRA Fig. 15 shows the measured shorted turn sideband amplitudes for healthy and faulty motor (2 shorted turns) over a wide range of loads from the flux, instantaneous power and single power-flux signal. The results obtained from the flux signal indicate that shorted turn causes sideband amplitudes to decrease as the load on the motor increases. However, flux signal proved more suitable and in fact preferable for detecting shorted turn faults at light loads, while the instantaneous power is able to detect shorted turn faults at heavy loads. Therefore, it is difficult to set a threshold between the healthy and faulty motors. To overcome this difficulty, a novel approach has been used as a combination of both flux and instantaneous power signals into a single signal called Power-Flux Signal. Flux signal and instantaneous power signal are used in a combination of Power-Flux signal. The best feature of the Power-Flux signal is that the sideband amplitudes between the shorted turn fault frequency components are generally higher as shown in Fig. 15 (bottom row). It is thus easier to set a threshold between the healthy and faulty motors to determine the shorted turn fault severity using combined power-flux signal Flux Spectrum : f 1 [(n/p)(1-s)+/-k)] -26 db -4 db k=1 healthy motor faulty motor % Rated Load % Rated Load dB db Combined Single Signal k=1 healthy motor faulty motor % Rated Load Figure 15: Variations in the sideband amplitudes for the healthy and faulty motor (with 2 shorted turns). Top row (left) and top row (right) show the flux and instantaneous power. The Bottom (row) represents a single signal. In summary, the single signal is preferable for detecting shorted turn faults over a wide range of loads for the different values of k when compared to the results obtained from flux and instantaneous power signals separately. VI. CONCLUSION Non-invasive multiple sensors for the detection of shorted turns fault using fault frequency components from flux signal, instantaneous power signal and a single signal (combination of flux and instantaneous power signals) called power-flux over a wide range of loads has been presented in this paper. Using fault frequency components to detect the faults and estimate fault severity in the machines relies heavily on variations in the amplitudes of sideband components. The results for the flux signal showed that it is not able to provide useful information to detect shorted turn faults at higher load and is only useful for finding shorted turn faults at light loads but only for the specific values of k i.e. -5 to +5. Instantaneous power signal exhibited that there is enough separation between the healthy and faulty motors in the frequency spectra at fullload for the different values of k. In contrast, results obtained at 3 % load showed small variations in the amplitudes of fault frequency components. Therefore, on the basis of the results obtained, instantaneous power is useful for detecting shorted turn faults only at heavy loads. The use of single power-flux signal demonstrated that it can 238

7 be used to detect shorted turns fault and showed much clearer variations between healthy and faulty motors over a whole range of loads (from no-load to full-load). Frequency Spectrum Technique, International Journal of Computer and Electrical Engineering, IJCEE (Accepted for Publication in Vol.3, Issue 1, Available online in February 211). ACNOWLEDGEMENTS The authors would like to acknowledge their respective universities for providing adequate resources and environment for research. REFERENCES [1] W. R. Finlay, Troubleshooting motor problems, IEEE Trans. Ind. Applicat., vol. 3, pp , Sept./Oct [2] F. Filippetti and M. Martelli, Development of Expert System nowledge Base to On-Line Diagnosis of Rotor Electrical Faults of Induction Motors, IEEE Ind. Applicat. Society Annual Meeting, vol. 1, 1992, pp [3] A. H. Bonnett and G. C. Soukup, "Cause and Analysis of stator and Rotor Failures in Three-phase Squirrel-Cage Induction Motors," IEEE Trans. on Industry Applications, vol. 28, no 4, pp , July/August [4] A.J.M. Cardoso, S.M.A. Cruz, D.S.B. Fonseca, Inter-turn stator winding fault diagnosis in three-phase induction motors, by Park's Vector approach IEEE International Electric Machines and Drives Conference Record, Digital Object Identifier: 1.119/IEMDC Publication Year: 1997, pp MB1/5.1 - MB1/5.3 [5] H.M. Emara, M.E. Ammar, A. Bahgat, H.T. Dorrah, Stator fault estimation in induction motors using particle swarm optimization IEEE International Electric Machines and Drives Conference, 23. Volume: 3 Publication Year: 23, pp [6] S. Nandi, Detection of Stator Faults in Induction Machines Using Residual Saturation Harmonics IEEE Transactions on Industry Applications, Volume: 42, Issue: 5 Digital Object Identifier: 1.119/TIA Publication Year: 26, pp [7] Wang Xu-hong; He Yi-gang, Fuzzy Model based On-line Stator Winding Turn Fault Detection for Induction Motors Sixth International Conference on Intelligent Systems Design and Applications, 26. Volume: 1 Digital Object Identifier: 1.119/ISDA , 26, pp [8] J. Cusido, L. Romeral, J.A. Ortega, J.A. Rosero, A.G. Espinosa, Fault Detection in Induction Machines Using Power Spectral Density in Wavelet Decomposition IEEE Transactions on Industrial Electronics, Volume: 55, Issue: 2, Digital Object Identifier: 1.119/TIE , Publication Year: 28, pp [9] Ahmed, I., Supangat, R., Grieger, J., Ertugrul, N., and Soong, W.L. "A Baseline Study for On-Line Condition Monitoring of Induction Machines", Australian University Power Engineering Conference, Brisbane, Australia, September 24, ISBN , pp [1] I. Albizu, A. Tapia, J.R. Saenz, A.J. Mazon, I. Zamora, Stator winding fault diagnosis in induction generators for renewable generation, Proceedings of the 12th IEEE Mediterranean Electrotechnical Conference, 24. Publication Year: 24, pp Vol.3. [11] Ahmed, I., Ahmed, M., "Comparison of Stator Current, Axial Flux and Instantaneous Power Used to Detect the Broken Rotor Bar Fault in Machines". Australian University Power Engineering Conference, Sydney, Australia, 14-17th December 28, ISBN , pp [12] A. Siddique, G.S. Yadava, B. Singh, A review of stator fault monitoring techniques of induction motors IEEE Transactions on Energy Conversion, Volume: 2, Issue: 1, Digital Object Identifier: 1.119/TEC , 25, pp [13] S. Grubic, J.M Aller, B. Lu, T.G. Habetler, A survey of testing and monitoring methods for stator insulation systems in induction machines International Conference on Condition Monitoring and Diagnosis, 28. Digital Object Identifier: 1.119/CMD Publication Year: 28, pp [14] Intesar Ahmed, Manzar Ahmed, M. Shuja han, ashif Imran, Investigation of Multiple Faults Detection in Electric Machine Using Broken Rotor Bar and Eccentricity Fault Frequencies Techniques, International Journal of Electrical and Computer Sciences, IJECS/IJENS. Vol. 1, Issue 5, pp 24 31, (21). [15] Intesar Ahmed, Manzar Ahmed, ashif Imran, M. Shuja han, S. Junaid Akhtar, Detection of Eccentricity Faults in Machine Using Intesar Ahmed received Phd Electrical Power Engineering from University of Adelaide, South Australia and M.Sc. in Electrical Power Engineering from University of New South Wales, Sydney, Australia in 27 and 1994 respectively. Dr. Intesar has been involving in various research projects relating to online condition monitoring of electrical machines, power electronics, and study of multiple faults in electric machines via signal processing techniques. Manzar Ahmed received MS in Telecommunication from Asian Institute of Technology, Thailand and B.Sc. in Electrical Engineering from University of Engineering & Technology, Lahore Pakistan in 21 and 1991 respectively. His current research interests include electric machines and drives, electromagnetic computational and electromechanical actuation as well as techniques for energy savings. Muhammad Shuja han received MS degree in Electronic Engineering from Ghulam Ishaq han Institute (G.I..I.) of Engineering. Science & Technology, Swabi Pakistan and B.Sc. Electrical (POWER) Engineering from University of Engineering & Technology, Lahore Pakistan in June 29 and July 27 respectively. He joined COMSATS Institute, Lahore Pakistan on August 29 as Lecturer in Department of Electrical Engineering. He has successfully conducted and taught one Short Certificate Training Course of Power Distribution System Design in Spring 21. His interests include MEMS Design and Multiple faults in Machines. Engr. Shuja was awarded Higher Education Commission Pakistan Indigenous Fellowship Award for hid MS program at G.I..I He was also awarded two International Travel Grants for presenting his research paper in ICSCT 29 and ICMENS 21 held in South orea and China respectively by HEC Pakistan. ashif Imran received his elementary education from England where he obtained GCSEs and GCEs.He did B.Sc. and M.Sc. in Electrical Engineering from University of Engineering and Technology (UET), Lahore, Pakistan. His area of specialization during both degrees was Electrical Power. He started his career as a Lecturer at UET Lahore. Then he moved to SIEMENS where he worked as Engineer on project coordination of 132kV grid stations. Later he joined NESPA, leading engineering consultancy firm of Pakistan, as a Design Engineer in Power Distribution Section. In NESPA, his professional experience includes design of overhead and underground power distribution systems for a variety of buildings and installations. Currently, he is Lecturer at COMSATS Institute of IT, Lahore, Pakistan where he teaches Electric Machines. His book titled Power Exchange as a Deregulated Electricity Market has been published by LAP Lambert Academic Publishing, Germany. His research interests include Power System Economics, Restructured Power Systems Simulation, Energy Management Systems and Online Condition Monitoring of Electric Machines. 239