1 On the Frequency Dependence of Harmonic Current Side-Band (HCSB) based Rotor Fault Indicators for Three-Phase Cage Machines C. Bruzzese, Student Member, IEEE, and E. Santini, Member, IEEE Φ Abstract—Inverter-fed cage induction motors can run at variable frequency and speed, and the diagnostic techniques to be applied for these motors must take in account frequency variations. Especially in variable frequency railway traction drives, bar breakage problems have been reported as a consequence of large sixth harmonic torques produced by the non-sinusoidal feeding. MCSA can be applied for motor diagnostics in these cases, but suitable indicators must be used for reliable fault severity assessment. Harmonic Current SideBand (HCSB) based Indicators (HCSBIs) for broken bars detection present a good rejection to frequency variations, as well as to motor load level, drive inertia, and motor parameters. In this paper, three industrial motors with broken bars have been tested for HCSBIs performance evaluation with different feeding frequencies, and the results have been reported and discussed. Index Terms—Bar Breakage, Fault Indicators, Frequency Rejection. T I. (fifth, seventh, eleventh, etc.) arise which produce harmful sixth harmonic torques capable to excite some cage resonance frequencies [5], [6], and to cause bar breakage. This kind of fault mechanism is not cited in [2], which mentions, about bar breakage causes, direct on-line starting, pulsating mechanical loads, and constructive imperfections of the cage. The same low-order current harmonics produce, however, some fault-related sidebands useful for fault detection. In [7] a theoretical demonstration was provided for HCSBIs derivation, and we mention here the proposed bar-breakage indicators: I (7 − 2 s )ω I , Γ (7 ) = (5 + 2 s )ω , I 5ω I 7ω I I Γ (11) = (13−2 s )ω , Γ (13 ) = (11+ 2 s )ω , I11ω I13ω Γ (5 ) = Φ C. Bruzzese is with the Department of Electrical Engineering, University of Rome “Sapienza”, via Eudossiana 18, 00184 Rome, Italy (e-mail: claudio.bruzzese@uniroma1.it). E. Santini is with the Department of Electrical Engineering, University of Rome “Sapienza”, via Eudossiana 18, 00184 Rome, Italy (e-mail: ezio.santini@uniroma1.it). (2) ………. INTRODUCTION HREE-PHASE induction motors are the most widely used electrical machines in industry, and motor condition monitoring is nowadays a major issue in the maintenance programs. Condition based maintenance (CBM) offers numerous advantages, since the maintenance can be dispensed when and where it is really needed; moreover, predictive monitoring permits to avoid unforeseen downtimes. MCSA (motor current signature analysis) is adapt for on-line non invasive motor diagnostics, in all the cases in which the fault signature can be properly individuated and evaluated for fault severity assessment. When applied for bar breakages diagnostics in cage motors, MCSA can suffer many disturbs such as load vibrations [1], [2], load level variations [3], and feeding frequency variations; moreover, drive features and motor parameters can vary from case to case, so complicating the problem. The authors proposed in [4] some original broken bars indicators, which present favorable performances as regard to disturb rejection and fault assessment affordability. Since they are based on the sidebands of the harmonic components of the phase current, HCSBIs can be used when a non-sinusoidal feeding is provided for motor driving. This kind of situation incurs, e.g., in railway drives: low-switching frequency GTO drives are not uncommon in railway applications, and notoriously low-order current harmonics (1) HCSBIs form a class of infinite elements, generally dependent on slip and frequency. Every element actually increases with the number of consecutive broken bars. Slip dependence has been found almost neglectable [4], whereas frequency dependence must be properly taken in account for variablefrequency drives. Experimental performances of (1), (2) have been addressed in Section III. II. THE EXPERIMENTAL TEST-BED Three motors (‘A’ in Fig.4, ‘B’ in Fig.6, and ‘C’ in Fig.8) with different powers (1.5÷3kW) and pole numbers (2÷4) were subjected to destructive tests. Motor parameters are reported here below: TABLE I. MOTOR ‘A’ D ATA (MEZ-MOHELNICE). 1.5kW 380/220V TABLE II. 3kW 380/220V TABLE III. 1.5kW 380/220V 50Hz 3.5/6A 1410rpm cosφ=0.82 4 poles 28 bars MOTOR ‘B’ DATA (CAPRARI). 50Hz 6.5/11A 2800rpm cosφ=0.84 2 poles 23 bars MOTOR ‘C’ DATA (ELPROM). 50Hz 3.3/5.7A 2860rpm cosφ=0.88 2 poles 19 bars Measurements of the phase currents were done with progressive rotor damage (increasing number of broken bars), 2 for a complete characterization of motor current spectra under fault conditions. The three motors was fed by a square-wave inverter, to obtain the relevant harmonics and sidebands, Fig. 1. Then an increasing number of consecutive bars were drilled, and the harmonics were registered on a large load range, for every selected feeding frequency. Fig. 2 shows a functional diagram of the test-bed used for experimentation. Acquired data (a sampling frequency of 20kHz was sufficient, since the motor current Shannon frequency was found around 10kHz when the square-wave frequency is 50Hz) were automatically processed off-line by using a ‘script’ Matlab algorithm, that produced Fourier transformation and harmonic discrimination on the basis of the measured motor speed. Fault-related sidebands can be revealed around very high harmonic order frequencies, thanks to the square-wave feeding; however, every lowswitching frequency commutation technique can produce a typical spectral pattern useful for fault detection. was of concern during experimentations, Figs. 5, 7, 9. Fig.3 (obtained from motor ‘B’) well explains the better performances of the higher-order indicators, as far as concern the rejection to frequency and load variations, with respect to the classical ratio between LSB (lower sideband of the first harmonic) and the fundamental current (Γ(1)). Measures done on motor ‘B’ (2 poles), Fig. 7, were affected by inter-bar currents, that produced sidebands weakening (with the lighter fault degree) so producing a curvature [8]. For motor ‘A’ (4 poles), Fig.5, this problem was less remarkable. Motor ‘C’ (Fig.9) initially behaved like ‘B’; to overcome the influence of inter-bar currents, for motor ‘C’ bilateral bar interruptions were practiced, so obtaining more linear results. A remarkable result that rises from Figs. 5, 7, 9, is that four-poles motors (motor ‘A’ and the motor tested in [4]) presented indicators with analogous amplitude, and the same is true for the two two-poles motors (‘B’ and ‘C’), with amplitudes roughly halved. This leads to the definition of a criterion for fault severity assessment, as stated in (3): 2 N broken.bars = Γ (ν ) ⋅ N total .bars P ZOOM (3) (where P is the polar pairs number) that can be at least used for ν = 5, 7, in the range of the industrial frequencies and for two/four-poles motors. Equation (3) leads in turn to define an “electrical number of broken bars (per unit)”, ‘nel’, as in (4): nel = Fig. 1. Phase current spectrum for motor A. Four broken bars, 50Hz feeding frequency, 70% of rated load. Sidebands (5+2s)f and (7-2s)f have been evidenced. MAIN GRID Load regulation VARIABLE-VOLTAGE AUTOTRANSFORMER SQUARE-WAVE VARIABLE-FREQUENCY INVERTER Voltage regulation POWER METER voltage/power measure Frequency regulation INDUCTION MOTOR DYNAMOMETRIC DC-UNIT current measure OSCILLOSCOPE speed measure PC Fig. 2. Experimental test-bed (functional diagram). III. DISCUSSION OF THE E XPERIMENTAL WORK The frequency-dependence of the new indicators (1), (2) N broken.bars N bars . per . polar . pair = 2 ⋅ Γ (ν ) (4) that furnishes a measure of the degree of asymmetry caused by broken bars on the electromagnetic structure along the extension of one polar pair. This number can be retained as a ‘pure’ fault-gravity indicator itself; in fact, this can be easily understood thinking to the higher values of sidebands in the four poles motors with respect to two-poles motors. IV. CONCLUSIONS HCSBIs are a class of fault indicators for bar breakages detection and fault gravity assessment, and they are wellsuited for converter-fed motors. Their main features are good linear-like dependence on the number of consecutive broken bars and relative insensibility to motor operating conditions and drive features. Frequency dependence is not so much preoccupant around industrial frequencies (50-60Hz), but must be taken in account when a large variation range incurs, especially for low frequencies (below 25Hz). 3 Fig. 3. Γ(1), Γ(5) trends for motor ‘B’, three broken bars. V. [1] [2] [3] [4] [5] [6] [7] [8] REFERENCES A. H. Bonnett, G. C. Soukup, “Cause and analysis of stator and rotor failures in three-phase squirrel-cage induction motors“, IEEE Transact. on Industry Applications, vol.28, No.4, July/August 1992, pp. 921-937. W. T. Thomson, M. Fenger, "Current signature analysis to detect induction motor faults", IEEE Industry Applications Magazine, vol.7, pp. 26-34, July/Aug. 2001. C. Bruzzese, O. Honorati, E. Santini, “Real behavior of induction motor bar breakage indicators and mathematical model”, Proc. of the ICEM 2006 Conference, September 2-5, 2006, Crete Island, Greece. C. Bruzzese, O. Honorati, E. Santini: “Harmonic current sideband-based novel indicators of broken bars for on-line evaluation of industrial and railway cage motor faults,” IEEE 2007 International Symposium on Industrial Electronics, ISIE 2007, Vigo, Spain, June 4-7, 2007 (accepted for publication). C. Bruzzese, O. Honorati, E. Santini, “Rotor bars breakage in railway traction squirrel cage induction motors and diagnosis by MCSA technique. Part I: Accurate fault simulations and spectral analyses”, IEEE SDEMPED 2005, 7-9 Sept. 2005, Vienna, Austria, pp.203-208. C. Bruzzese, O. Honorati, E. Santini, “Minimization of harmful cage torsional resonances in traction motors by a combined mechanicelectronic optimization,” in Proc. of 12th European Conference on Power Electronics and Applications, EPE 2007, 2-5 Sept. 2007, Aalborg, Denmark (approved for pubblication). C. Bruzzese, C. Boccaletti, O. Honorati, E. Santini, “Rotor bars breakage in railway traction squirrel cage induction motors and diagnosis by MCSA technique. Part II: Theoretical arrangements for fault-related current sidebands”, IEEE SDEMPED 2005, 7-9 September 2005, Vienna, Austria, pp.209-214. R. F. Walliser and C. F. Landy, "The influence of interbar currents on the detection of broken rotor bars,” in Proc. 1992 International Conference on Electric Machines, ICEM’92, pp. 1246-1250. Fig. 4. Motor ‘A’ (MEZ-MOHELNICE). Fig. 5. Experimental results for motor ‘A’. 4 Fig. 6. Motor ‘B’ (CAPRARI). Fig. 8. Motor ‘C’ (ELPROM). Fig. 7. Experimental results for motor ‘B’. Fig. 9. Experimental results for motor ‘C’ 5 VI. BIOGRAPHIES Claudio Bruzzese (S’2005) received the laurea degree (M.S.) cum laude from the University of Rome “Sapienza”, Italy, in 2002. He is currently working toward the Doctoral degree at the Dep. of Electrical Engineering, University “Sapienza”, Italy. He is a member of the AEIT (Italian Electrotechnique Association), of the IEEE Industry Applications Society, and of the IEEE Power Electronic Society. He is a Registered Professional Engineer in Italy. His interests cover diagnostics of induction and synchronous machines, FEA, railway and naval power systems, electro-mechanical design, and diagnostic computer systems. Ezio Santini (M’91) received the Degree in electrical engineering (cum laude) from the University of Rome “Sapienza,” Rome, Italy, in 1982. From 1983, he was with the Department of Electrical Engineering, University of Rome “Sapienza,” where currently he is full professor of Design of Electromagnetic Devices. His research activities concern electrical machine models, transient phenomena in transformers and rotating machines, FEM analysis of electromagnetic fields in electrical devices, and numerical methods in electrical engineering. Prof. Santini is a member of the Italian Association of Electric and Electronics Engineers and of the IEEE Power Engineering Society.