ROTOR BROKEN BAR DIAGNOSIS IN ASD USING INSTANTANEOUS POWER SPECTRUM AND MEAN ABSOLUTE DIFFERENCE APPROACH A COMBINED TECHNIQUE

Transcription

1 ROTOR BROKEN BAR DIAGNOSIS IN ASD USING INSTANTANEOUS POWER SPECTRUM AND MEAN ABSOLUTE DIFFERENCE APPROACH A COMBINED TECHNIQUE H.H. Hanafy A. M. Huien E. A. El-Zahab Electrical Power and Machine Department, Faculty of Engineering, Cairo Univerity, Giza-Egypt, Zip 12613, Hanafy_Haan@hotmail.com, eng.ahmed.huien@hotmail.com, zahab0@yahoo.com Abtract: Rotor broken bar i a common fault in quirrel cage induction motor. Rotor failure due to that fault, now account for a moderate percentage of total induction motor failure. Thi paper will deal with detection of broken rotor bar of an induction motor which i upplied from adjutable peed drive (ASD) equipped with AC reactor and ine-wave haping filter on the output ide. The experimental reult how that the intantaneou power pectrum (IPS) analyi method i only effective when the motor i upplied from inuoidal upply, in contrat for the demontrated ASD. So the intantaneou power pectrum of the healthy rotor cae and the broken bar rotor cae i proceed with mean abolute difference (MAD) algorithm to invetigate the diimilarity between them. The reult how that thi combined technique i highly effective in diagnoi of broken rotor bar fault in cae of ASD. Key word: Fault diagnoi, induction motor, rotor broken bar, intantaneou power pectrum, mean abolute difference. 1. Introduction The Squirrel cage induction motor (SCIM) are widely ued in the indutrial drive field today. In order to avoid reduced output, emergency maintenance cot, broken down equipment, and lot revenue caued by fault, it i important to detect an upcoming motor fault a oon a poible. Generally variou monitoring technique for fault detection were ued by reearcher. [1-2] reviewed a large number of publication covering the different type of induction motor fault. A comparion and performance evaluation of different diagnotic procedure that ue input electric ignal to detect and quantify rotor breakage in SCIM upplied by the main have been performed in [3]. Author in [4] ued the air-gap torque pectra a a potential ignature for online condition monitoring and fault diagnoi in SCIM. One of the mot well-known approache regarding the diagnoi of rotor fault in induction machine i baed on online monitoring and proceing of the tator current. The main advantage and drawback of motor current ignal proceing technique for induction motor rotor fault detection (mainly broken bar and bearing deterioration) have been preented in [5]. The Fat Fourier Tranform (FFT) method i uccefully ued for the rotor fault detection in the SCIM. G. Didier et al. have preented a new technique to detect broken rotor bar in poly phae induction machine, by employing the Fourier Tranform of the tator current to analyze it phae combined with the Hilbert Tranform [6]. Tooma Vaimann et al. in [7] analyzed the diagnotic poibilitie of three-phae quirrel-cage induction motor rotor fault through the implementation of FFT. Diagnoi of mixed broken rotor bar and eccentricity fault in SCIM uing intantaneou power pectrum ha been invetigated by Zhenxing Liu et al. in [8]. Diagnoi baed on the global fault index method applied to the intantaneou power ignal and line current ignal provide relevant reult for the detection of broken rotor bar ha been preented in [9]. Recently a powerful mathematical tool known a Wavelet Tranform (WT) ha been ued which can diagnoe rotor broken bar fault. M. Riera et al. introduced a new method for the diagnoi of rotor broken bar in induction machine baed on the application of the Dicrete Wavelet Tranform (DWT) to detect the tator current harmonic of a particular frequency variation during the tartup proce, when a rotor broken bar occur [10]. The detection of rotor fault in induction machine by applying a newly developed quantification technique baed on the wavelet tranform to the envelope of the 1

2 tarting current of the machine, ha been preented in [11]. Induction motor fault diagnoi method baed on three-phae tator current envelope for broken rotor bar and inter-turn hort circuit ha been preented in [12]. A new Beirut diagnotic procedure (BDP) ue only two of the three motor upply current and doe not need any voltage enor intallation, ha been preented by Mario Eltabach et al. in [13], a a new non-invaive electrical diagnotic method for detection of broken rotor bar in induction motor. Experimental reult of enorle broken bar detection in induction motor baed on fluctuation of the tator current zero croing intant before actual breakdown occur have been preented in [14]. Boudinar et al. ued an improved Root-MUSIC (MUltiple SIgnal Claification) approach a powerful tool for extracting meaningful frequencie from the tator current to improve the diagnoi of rotor broken bar of the induction motor [15]. Thi paper preent an experimental tudy of intantaneou power pectrum (IPS) technique in cae of inuoidal upply and the ame technique combined with mean abolute difference approach in cae of pecial ASD for diagnoi of the broken bar fault. Two common rotor broken bar fault are conidered; one broken bar and two adjacent broken bar. In the propoed technique the intantaneou power pectrum of the healthy rotor cae and the broken bar rotor cae i proceed with mean abolute difference (MAD) algorithm to invetigate the diimilarity between them. 2. Intantaneou Power Signature Technique Auming a healthy motor with no peed ocillation, ha an ideal three phae upply voltage with intantaneou phae voltage of v (t) and an intantaneou phae current of ί (t), the intantaneou power P (t) can be written a follow [9]: v (t) i 2V co(wt) (t) 2Ico(wt ) (1) (2) P (t) VI [co(2wt ) co( )] (3) Where φ = the motor phae angle. A hown from (2) and (3), the current pectrum ha only one component at a frequency of f (f = ω/2π), while the power pectrum ha two component, DC component and the other component at a frequency of f c (f c =ω/π). When broken bar fault occur, a rotor aymmetry take place, cauing the appearance of additional two frequency component, the lower component i due to backward rotating field reulting from the broken bar fault and the upper component i due to torque pulation and peed ocillation. Thee component occur at frequencie given by [2], [9]: fbb ( 1 2KS) f S (4) Where: K = 1, 2, 3 S = operating per unit lip of the motor. The lower frequency component magnitude i not equal to the upper frequency component [8-9]. So in cae of broken bar, the mathematical expreion of the intantaneou power of a quirrel cage induction motor i given by [9]: P ( t) V I K 1 V I V I K 1 bpk bnk co co(2wt ) co 21 kwt bpk co2kwt bpk co 2 1 k wt bnk co(2kwt bnk) (5) Where: I bpk : RMS value of the current pectrum component at frequency (1-2kS)f I bnk : RMS value of the current pectrum component at frequency (1+2kS)f φ bpk : Initial phae angle of the current pectrum component at the frequency (1-2kS)f φ bnk : Initial phae angle of the current pectrum component at the frequency (1+2kS)f 3. Mean Abolute Difference Approach The Mean Abolute Difference (MAD) i a widely ued in ignal proceing application to invetigate the diimilarity between two vector [16]. It i often ued for real-time implementation, becaue it i computationally very efficient. In our propoed technique, the MAD i ued to invetigate the diimilarity between two vector, thee vector are the intantaneou power pectrum of the predetermined reference vector (healthy cage) and the examined vector (broken bar cage) at different operating peed of the ASD. The MAD i given by [16]: MAD n 1 1 n m0 A m B m Where: n: Length of the vector A m : Reference vector (Healthy cae) B m : Examined vector for each cae (6) 2

3 4. Experimental Reult and Dicuion The motor under tet ha been choen a 3 HP, 2- pole, 380 V and 50 Hz cage induction motor drive a water pump. The tet bed block diagram i hown in Fig.1, which conit of current and voltage tranducer from LEM and their part number are LTS 25-NP, LV repectively and the data acquiition ytem i NI PCI Bit, 1 MS/ with connection board connected to the IBM compatible PC, the diagnoi oftware i implemented on NI LabView; the reult documentation and formatting by uing NI Diadem. The tator voltage and current were ampled at 20 KHz individually for 10 econd of teady tate operation, which give a 0.1 Hz pectrum reolution with 10 KHz full cale pectrum. Voltage and current waveform are multiplied together to get the intantaneou power waveform and fat Fourier tranform i applied. The experimental data limited to frequency component (2±2KS)f, with K=1, 2. The dicuion will be demontrated in the following ection: the ideband component (for different fault cae there i booting in the ideband amplitude), which how that the IPS method i highly effective in diagnoi of the broken bar fault under different fault cae when the upply voltage i inuoidal. Fig. 2 The IPS of healthy cage for inuoidal upply Fig.1 The tet bed block diagram 4.1. Intantaneou Power pectrum of healthy and broken bar cage in cae of inuoidal upply Fig.2 how the intantaneou power pectrum of the healthy cage with exiting lower and upper ideband frequency component, thee component exit even in the healthy cae due to magnetic and electrical aymmetrie from the manufacturing proce. Fig.3-4 how the intantaneou power pectrum of the broken bar cage cae. According to the reult, a moving average cheme i applied which will be the reference intead of the zero point. Table (1) how the amplitude of the ideband component referenced from the calculated average and the peed of each cae. From Table 1 and Fig. 2-4, the broken bar fault can be eaily detected by noticing the amplitude of Fig. 3 The IPS of one broken bar cage for inuoidal upply Fig. 4 The IPS of two adjacent broken bar cage for inuoidal upply 3

4 Table 1 The amplitude of the ideband component referenced from the calculated average and the peed of each cae. Motor Statu K=1 left (db) K=1 right (db) K=2 left (db) K=2 right (db) Speed (rpm) Healthy One Broken Bar Two Adj. Broken Bar Intantaneou Power Spectrum of Healthy and Broken Bar Cage in Cae of ASD The waveform of the inverter output voltage at 50 Hz i hown in Fig.5; mainly the difference in the waveform from the traditional inverter i due to the built in output ide AC reactor and Sin-wave haping filter. The main function of the filter and the reactor i to reduce the vibration in the motor caued by inverter witching waveform. Thi i the cae of high power motor in range of 200 HP inverter, where the vibration conideration i important. The reult hown in thi ection are the intantaneou power pectrum at the upply frequencie of 45 Hz in Fig.6-8, 50 Hz in Fig.9-11 and 55 Hz in Fig By noticing Fig. 6-14, one can conclude that, there are no identification ideband frequencie a in cae of the inuoidal upply. According to thee reult, the intantaneou power pectrum method i not effective in the cae of pecial inverter cae (ASD), in contrat with the inuoidal upply cae The Propoed Combined Technique In [16], the MAD wa applied a follow; extract appropriate feature that contain ueful fault ignature (upper and lower ideband) in the pectral information while uppreing other pectral information uch a fundamental and noie component. After the appropriate feature are extracted, the reference (healthy) and examined (broken bar) vector are compared uing (6) for detection of the broken bar fault. In the propoed technique, The MAD wa calculated a follow; the healthy, one broken bar and two adjacent broken bar cage intantaneou power wa acquired uing the propoed tet-bed the reference vector will be the intantaneou power pectrum of the healthy cae, and the examined vector will be the intantaneou power pectrum of the fault cae at the ame upply frequency and load condition, without removing any component from the pectrum. The MAD i ued to earch for hidden information in the pectrum. Table 2 how MAD magnitude veru upply frequency at different fault cae. Fig.15 how the calculated MAD for one broken bar and two adjacent broken bar cage veru the upply frequency. It i clearly that combined technique i highly effective in diagnoi of broken rotor bar fault in cae of ASD. Fig.5 Waveform of the inverter output voltage at 50 Hz Fig.6 The IPS of the healthy cage upplied from inverter at 45 Hz 4

5 Fig.7 The IPS of one broken bar cage upplied from inverter at 45 Hz Fig.10 The IPS of one broken bar cage upplied from inverter at 50 Hz Fig.8 The IPS of two adjacent broken bar cage upplied from inverter at 45 Hz Fig.11 The IPS of two adjacent broken bar cage upplied from inverter at 50 Hz Fig.9 The IPS of the healthy cage upplied from inverter at 50 Hz Fig.12 The IPS of the healthy cage upplied from inverter at 55 Hz 5

6 Fig.13 The IPS of one broken bar cage upplied from inverter at 55 Hz Fig.14 The IPS of two adjacent broken bar cage upplied from inverter at 55 Hz Fig.15 MAD magnitude veru upply frequency at different fault cae 5. Concluion In thi paper, the Intantaneou Power Spectrum technique in the diagnoi of induction motor broken bar fault i invetigated through inuoidal and pecial inverter ource (ASD), thi invetigation how the performance of thi technique in both cae. It i found that, thi technique i effective on the diagnotic for the inuoidal upply cae, in contrat for the pecial inverter cae. Thi reult, lead on uing one of popular digital proceing technique, which i the Mean Abolute Difference approach (MAD) to compare different intantaneou power pectrum at fault cae with the healthy intantaneou power pectrum at the ame operating condition to earch for hidden information in the faulty pectrum. The reult were atifactory, and how the potential of the propoed technique. The combined technique i very computationally efficient, and with the digital technology available now, an embedded ytem could be deigned for that purpoe with economical price. Reference 1. Arfat Siddique, Yadava G. S. and Bhim Singh: A Review of Stator Fault Monitoring Technique of Induction Motor. In: IEEE Tran. on Energy Converion, vol. 20, no. 1, March 2005, p ,. 2. Subhai Nandi, Hamid A. Toliyat and Xiaodong Li: Condition Monitoring and Fault Diagnoi of Electrical Motor A Review. In: IEEE Tran. on Energy Converion, vol. 20, no. 4, December 2005, p Bellini A., Filippetti F., Francechini G., Taoni C. and Kliman G.B.: Quantitative Evaluation of Induction Motor Broken Bar By Mean of Electrical Signature Analyi. In: Indutry Application Conference, Conference Record of the 2000 IEEE, Iue, 2000, vol.1, p Vinod V. Thoma, Krihna Vaudevan and Jagadeeh Kumar V.: Online Cage Rotor Fault Detection Uing Air- Gap Torque Spectra. In: IEEE Tran. on Energy Converion, vol. 18, no. 2, June 2003, p Mohamed El Hachemi Benbouzid, and Gerald B. Kliman: What Stator Current Proceing-Baed Technique to Ue for Induction Motor Rotor Fault Diagnoi? In: IEEE Tran. on Energy Converion, vol. 18, no. 2, June 2003, p Didier G., Terniien E., Capary O. and Razik H.: A new approach to detect broken rotor bar in induction machine by current pectrum analyi. In: Elevier, Mechanical Sytem and Signal Proceing, vol. 21, 2007, p Tooma Vaimann and Ant Kallate: Detection of broken rotor bar in three-phae quirrel-cage induction motor uing fat Fourier tranform. In: 10 th International Sympoium on Topical Problem in the Field of Electrical 6

7 and Power Engineering, Pärnu, Etonia, January 10-15, 2011, p Zhenxing Liu, Xianggen Yin, Zhe Zhang, Dehu Chen and Wei Chen: Online Rotor Mixed Fault Diagnoi Way Baed on Spectrum Analyi of Intantaneou Power in Squirrel Cage Induction Motor. In: IEEE Tran. on Energy Converion, vol. 19, no. 3, September 2004, p Gaëtan Didier, Eric Terniien, Olivier Capary, and Hubert Razik: Fault Detection of Broken Rotor Bar in Induction Motor Uing a Global Fault Index. In: IEEE Tran. on Indutry Application, vol. 42, no. 1, January/February 2006, p Riera M., Antonino J.A., Roger-Folch J., and Molina M.P.: Detection of Broken Rotor Bar in Induction Machine through the Study of the Startup Tranient via Wavelet Decompoition. In: Journal of Electrical Engineering (JEE), Vol.6, No.3, 2006, p Khadim Moin Siddiqui and Giri V.K.: Broken Rotor Bar Fault Detection in Induction Motor uing Tranient Current Analyi. In: International Journal of electronic & communication technology (IJECT), vol. 2, no. 4, Dec. 2011, p Aderiano M. da Silva, Richard J. Povinelli, and Nabeel A. O. Demerdah: Induction Machine Broken Bar and Stator Short-Circuit Fault Diagnotic Baed on Three- Phae Stator Current Envelope. In: IEEE Tran. on Indutrial Electronic, vol. 55, no. 3, 2008, p Mario Eltabach, Jerome Antoni, Galyna Shanina, Sophie Sieg-Zieba and Xavier Carniel: Broken rotor bar detection by a new non-invaive diagnotic procedure. In: Elevier, Mechanical Sytem and Signal Proceing, vol. 23, 2009, p Hakan Cali and Abdulkadir Cakir: Experimental tudy for enorle broken bar detection in induction motor. In: Elevier, Energy Converion and Management vol. 49, 2008, p Boudinar A.H., Bendiabdellah A., Benouzza N. and Boughanmi N.: Three Phae Induction Motor Incipient Rotor Fault Detection Baed On Improved Root-Muic Approach. In: Journal of Electrical Engineering (JEE), Vol.7, No.7, 2007, p Song M. H., Kang E. S., Jeong C. H., Chow M.Y. and Ayhan B.: Mean Abolute Difference Approach for Induction Motor Broken Rotor Bar Fault Detection. In: SDEMPED 2003, Sympoium on Diagnotic for Electric Machine, Power Electronic and Drive, Atlanta. CA, USA, Augut 2003, p Table 2 MAD magnitude veru upply frequency. Supply Frequency (Hz) MAD of One Broken Bar (db) MAD of Two Broken Bar (db)