failures due to the abnormal (if compared with 50/60 Hz sinusoidal voltage) stress profiles that voltage impulses induce in the insulation system. This paper will ...
The influence of PWM voltage waveforms on induction motor insulation systems: perspectives for the end user A. Cavallini, G. C. Montanari, D. Fabiani and M. Tozzi Φ Abstract – Inverter-fed motors may experience premature failures due to the abnormal (if compared with 50/60 Hz sinusoidal voltage) stress profiles that voltage impulses induce in the insulation system. This paper will briefly recall the reasons for these failures, focusing on the key role of partial discharges in wire-wound insulation systems. Furthermore, it will provide a short review of the actions taken by the International Electrotechnical Commission (IEC) to minimize the impact of such failures. Eventually, some solutions for testing online wire-wound motors, thus preventing premature failures to occur, will be presented.
Index Terms--.Partial discharges, converter, induction motor, reliability
Indeed, among all the factors listed above, the most serious one is partial discharges (PD). Once incepted, the loss of reliability will depend on the insulation type, and will be greater in all parts of the insulations where chemical bonds are weaker, i.e., the organic parts of the insulation. Therefore, in the following, focus will be on PD in electrical machine windings. The action taken by the International Electrotechnical Commission IEC to reduce the impact of these problems (i.e., drawing technical standards aimed at improving motor manufacturing quality), as well as some possible solutions to monitor the condition of the insulation system by online techniques will also be described. PD UNDER INVERTER VOLTAGES: GENERAL CONCEPTS
INTRODUCTION The use of power electronic converters is largely the major technological shift in the electrical industry. In the field of electromechanical conversion, the flexibility of power electronic drives allows to use AC motors in all applications where DC motors were once used. AC motors have several advantages with respect to DC ones (e.g., they are less complex and expensive, require less maintenance and cleaning and are more efficient). However, it was soon recognized that the interoperation of power electronic drives and motors could have detrimental effects on the reliability of the stator winding insulation system [18]. The root cause of this loss of reliability can be traced back to the electrical stress enhancement associated with inverter impulse voltages impinging on the stator windings. In particular, electrical stress is enhanced by (a) overvoltages generally in the range of 1.5-2 p.u. due to reflections at motor terminals induces by cable/motor impedance mismatch, (b) uneven voltage distribution between turns (similarly to lightning impulses) [1-7]. Thermal stress is also increased due to the higher dielectric losses. Furthermore, the large capacitive currents in the insulation associated with steep fronts can overheat the region where the slot coating overlaps the stress grading system, leading to an additional failure mode that is typical of form-wound medium voltage machines [8].
Effects of PDs PD are incepted within or at the borders of an insulation system whenever (a) a starting electron is available, (b) the field is large enough to increase, over the distance of a mean free path, the kinetic energy of electrons above the level needed for impact ionization to occur. If both conditions are met, the first electron becomes the seed of a multiplication process that produces an electron avalanche impinging on the dielectric surface at the anode. At the cathode, the dielectric will impact with an avalanche of positive ions. Both phenomena can be energetic enough to break chemical bonds within the insulation system leading, sooner or later, to breakdown [9]. Additionally, during the avalanche, ozone is generated in the air, so that the dielectric surfaces undergo a chemical attack also. The results of these phenomena on enameled wires can be seen from the optical microscope photos reported in Fig. 1, where (a) is a pit due to erosion (b) is a region where the enamel lost its transparency to ozone attack.
(a)
(b)
Fig. 1. Effect of PD bombardment on enameled wires: (a) pit due to erosion, (b) area where the enamel lost its transparency due to chemical attack. A. Cavallini, D. Fabiani and G. C. Montanari are with the Dept. of Electrical Engineering (DIE), University of Bologna, viale Risorgimento 2, 40036 Bologna, Italy. Marco Tozzi is with TechImp HQ Srl, 40069 Zola Predosa, Italy
At every avalanche, a given fraction of the total number of electrons (i.e., the so-called hot electrons) will have enough
978-1-4244-9303-6/11/$26.00 ©2011 IEEE
288
energy to interact with the polymer matrix, breaking bonds through Dissociative Electron Attachment events. The probability of DEA events depends on the energy of the bonds, of the electrons and on the cross section of the bonds, among other parameters [9]. Since bond breaking is the main reason for insulation degradation, it is understandable that the larger the frequency of PD events, the faster the degradation of the insulation. Inverter voltages produce train of square waves that can have, to date, frequencies up to 20 kHz. Whenever the field at some point of the insulation is large enough to trigger the multiplication process described above, there is a great chance that a couple of PD is generated at each ON/OFF cycle of the inverter. Assuming that, under 50 Hz AC or under inverter, the same area of the insulation undergoes equivalent PD phenomena (i.e., same number of electrons per PD event, same distribution of electron energy) both providing two PD events per cycle of the supply voltage waveform, the time to breakdown under inverter will be shortened by a factor 50/fc, being fc the commutation frequency of the inverter itself. For a state-of-the-art inverter, having fc=10 kHz, this implies that the life will be shortened by a factor 1/200. A more accurate assessment would require accelerated testing performed on twisted pairs or “motorettes” (“formettes” for form wound motors). As an example, Fig. 2 reports results of accelerated life tests performed in the laboratory on twisted pairs [10,11]. As can be seen, by increasing the frequency from 50 to 10000 Hz, life is reduced, on average, by 2 orders of magnitude in the absence of PD. In the presence of PD, the reduction is even more dramatic, being 4 orders of magnitude!
1000 50 Hz 10 kHz
Life (average value) [h]
100
10
1
10 kHz inverter voltage has to be questioned deeply. It can be shown, in fact, that the PD magnitudes tend to be higher under 10 kHz inverter sources rather than at 50 Hz. Repetitive Partial Discharge Inception Voltage The Partial Discharge Inception Voltage is a key concept in PD measurements: it is defined as the minimum level at which PD are first incepted in an equipment. Generally, at PDIV, only a few discharged are observed sporadically. Indeed, it was mentioned above that, under fast repetitive impulse voltages, the inception of a PD often implies that at least two PD per cycle are generated. Let us explain more in detail why. A PD is a self-quenching process, since it terminates when the (equal) amount of charges deployed by the electron/positive ion avalanches on the insulation surface is large enough to decrease the electric field down to a point where impact ionization is no longer possible. When the PD is extinguished, the electrical field associated with the two space charge distributions is large enough to prevent other PD to occur immediately. Using a simplified reasoning, this field equals in magnitude the field induced by the inverter, but has opposite sign, so that the field at the defect size is equal to 0. Assuming that nothing else happens, when the voltage is switched off the two space charge distributions induce in the air gap a field that is equal in magnitude to that due to the inverter, backfiring a PD event in the opposite direction which neutralizes the charge at the defect surfaces. When the electric field follows this pattern, there is a large probability that two PD are incepted per each cycle of the applied voltage. Indeed, between the PD event and the time the voltage is switched off, the two space charge distribution tend to recombine, so that the field in the gap starts gradually to increase (see sketch in Fig. 3). Given enough time, the field could recover to the level ascribable to the generator prior the PD. If this happen, at the moment the voltage is switched off, no further PD can be observed.
0.1
0.01
0.001
0.0001 PD
NO PD
Fig. 2. Comparison of accelerated life tests carried out on twisted pairs under AC and inverter voltage sources. The number of tests carried out to determine the average lifetime is n=5.
The huge difference between the simplified evaluation presented above and what found from accelerated life tests depends on two factors. First of all, the effect of thermal and electrical stress were not considered, but they are large enough to decrease the insulation lifetime even in the absence of PD (see the “NO PD” bars in Fig. 2). Then, the assumption that the same (in terms of bonds broken in each cycle) PD phenomena are incepted under 50 Hz AC or under
Fig. 3 Behavior of applied voltage and electrical field in the gap in the presence of PD.
289
It is therefore clear that the average number of o PD per cycle will depend on the recombination time connstants for the space charge distributions [12] and on the frequency and duty cycle of the voltage impulses applied too the insulation. In general, for frequencies above or equal to 1 kHz, i.e., for square wave periods below 1 ms, the recom mbination times are sufficiently short to allow two PD per cy ycle to occur as soon as the PDIV is reached, that is, the mo ost unfavorable condition for the insulation.
Fig. 4 Behavior of PDIV and RPDIV in a crossed pair subjected s to square voltage impulses having rise and fall times of 300 ns, variable v width from 500 ns to 50 µs. Repetition frequency of voltage impulsess: 50 Hz.
For these reason, the IEC introduced thhe concept of Repetitive Partial Discharge Inception Voltag ge, RPDIV, that is: the minimum peak-to-peak voltage at which partial discharges occur with a repetition rate of 1 or more PD pulses per every 2 voltage impulses as a mean for the specified test time [TS41]. Of course, depend ding on the ON and OFF times of the square wave, the RPDIV V can approach the PDIV or it can be larger than it. Figure 4 shows, as an example, the dependence of PDIV (i.e., the voltage at which a PD is first observed) and RPDIV on the ON time of a square voltage having frequency of 50 Hz (in ( this specific case, the figure highlights how longer ON N times favor detrapping of a starting electron from the insulation). Summarizing, under inverter voltages, insulattion system can experience, at each defect site, two or moree PD per cycle even for applied voltages moderately abovee the PDIV. At switching frequencies of 10 kHz or more and d depending on the nature of the insulation system (energy reequired to cause DEA), PD can become very fast degrad dation process, leading to breakdown in very fast times. Add ditional heating due to dielectric losses act in a synergistic way, w and lead to dramatic reduction of insulation lifetimes. IEC TECHNICAL SPECIFICATION (T TS) In order to alleviate inverter-related problem ms, the IEC has defined techniques to perform qualification and a acceptance tests aimed at improving the quality of moto ors that are sold on the market for inverter drives. Qualificcation tests are employed to evaluate insulating materials and a processing technique. Acceptance tests are performed for completed
motors or coil sets. Since wire- and fform-wound insulation system are radically different, the IEC C has produced the TS 60034-18-41 [13] and the TS 60034-18-42 [14], dealing with the two system separately. The concept of stress category Electric stress within the motor windiing depends (a) on the overshoot due to reflections at motorr terminals, (b) on the degree of unevenness of voltage distrribution between turns. Therefore, motors can be more or lesss stressed depending on the combination of overshoot of tthe voltage at motor terminals and on the impulse risetim me (which controls the partition of voltage between turns) [ 15]. To avoid with all possible combinations of these twoo parameters, the IEC defines four stress macrocategoriess, listed in Table 1. Depending on the stress category a motor falls in (thus depending on the inverter rise time aand cable characteristic impedance and length), different pass/fail criteria are proposed in the TS 60034-18-41 and -42 (we are not going to examine in details these limits for the sake of brevity). It is important to observe that the reliabbility of a given motor will be defined by the interactioons among customer, manufacturer and “inverter drive integgrator”. Stress category A – Benign B – Moderate C – Severe D - Extreme
TABLE 1 Summary of stress categories Overshoot factor Impulse risetime (µ µs) ≥1 ≤ 1.1 ≥ 0.3 ≤ 1.5 ≥ 0.1 ≤ 2.0 ≥ 0.05 ≤ 2.5
IEC TS 60034-18-41 (wire-wound mottors) The key concept for this TS is that qualification tests (for these insulation systems) are designedd to ensure that PD will not occur within the insulation system m during the expected life. Indeed, the insulation of thesee motors is generally realized using purely organic mateerials (e.g., polymideimide). The chemical bonds (C-H H, C-C) within these materials are weak enough to be brroken by the electron avalanches induced associated with eaach PD event. Thus, PD should not occur (or at least occur vvery rarely) within the insulation system (note that, in theese motors, PD occur between adjacent turns, see Fig. 3). IIn practice, the RPDIV of the insulation system should be well w above the voltage experienced during operation. Provided that PD do not occur at the ttime the motor is put in service, the goal of qualification tests is to ensure that other stress will not weaken the insulation uup to a point where PD can start occur. Since the most important stresses are the thermal and mechanical ones, qualificcation tests are carried out on winding models (motorettes) inn exactly the same way as specified for conventional motors, but RPDIV values are checked after each aging cycle. If RPDIV falls below a specified level, the system will not be qualified. Once an insulation system has beenn qualified, acceptance tests are performed on freshly produuced randomly-selected stators to ensure that the manufacturiing procedure does not deviate from the one that has been quualified. These tests are
290
performed using either sinusoidal (50/60 Hz) and impulsive voltage sources. IEC TS 60034-18-42 (form-wound motors) In wire-wound motors, PD occur at the boundaries of the insulation system and can be detected easily, they also cause a fast loss of reliability, so that they are not allowed. In form-wound motors, contrarily, PD can occur in voids surrounded by micaceous material. Since it is very difficult to assess the degradation rate of these systems through PD measurements, the TS 60034-18-42 relies on life tests. To carry out life tests, one has to bear in mind that formwound motors insulation systems are far more complex than wire wound ones. The system can be subdivided into four subsystems: (a) turn insulation, (b) groundwall insulation, (c) stress grading systems, (d) conductive slot coating. Inverter surges affect all subsystems, but turn insulation and grading systems are the most affected parts of the insulation. Since it is quite difficult to test complete system in an appropriate way, the TS specifies that qualification tests should be carried out on insulation system models able to replicate the different features of an insulation system separately. Qualification thus involves accelerated life testing of these models. In particular, for turn/turn insulation and stress grading models, life tests are carried out using impulsive sources. For groundwall insulation systems, accelerated life testing is performed using sinusoidal voltage sources. Acceptance tests are carried out in a very similar way, i.e., by subjecting a set of complete stators to a sinusoidal 50/60 Hz voltage waveform with a magnitude of 2.5 times the rated peak-to-peak voltage at the motor terminals under the specified inverter operating conditions for as long as 250 hours. If one or more of the stator fails, the manufacturing procedure needs to be revised. It is worthwhile to observe that, likely, the turn/turn insulation and stress grading systems will not be stressed appropriately by the acceptance tests. DIAGNOSTIC SOLUTIONS FOR WIRE-WOUND MOTORS According to IEC TS 60034-18-41, the reliability of a wirewound motor insulation system can be ensured if and only if no PD take place within the insulation system. Thus, the problem of PD detection and interpretation becomes critical to prevent unexpected failures. Indeed, PD detection must be carried out with sufficient signal-to-noise ratio to ensure that noise is minimized and that PD are detected with enough sensitivity. Radiated signals from inverter commutations, however, can interfere with PD measurements. Indeed, each commutation can be exchanged for a PD, leading to erroneous conclusions. To overcome these problems, PD measurements are generally carried out in the UHF range using suitable antennas. The rationale is explained in Fig. 5: the PD impulse is a very fast transient, which irradiates in a broad range of frequency. Also the inverter irradiates, but generally the interference spectrum is limited to frequencies that are below 200 MHz. Therefore, using a wide band
sensor working in the UHF range, above the maximum frequency of inverter interference can ensure enough SNR to carry out measurements in a good way. This sensor is generally realized using a broadband antenna connected to the detector through a highpass filter of 400-500 MHz [16,17]. The detector itself can be either a UHF oscilloscope or a conventional PD detector, provided that suitable electronics is used to adapt the band of the sensor to that of the detector.
Fig 5. Strategy for detecting PD in inverter-fed motors rejecting inverter interference: the coupler should operate in the frequency range where inverter interference is minimal or absent.
The above sensor can work well to suppress inverter interference. However, as PD are detected in the range going from a few hundred oh MHz to, possibly, 2 GHz, noise can be picked up from, e.g., cellular phones. It is therefore important to reject this kind of noise, that wound appear randomly. The most obvious solution for cellular phones is to use notch filters. However, other noise sources can exists at different frequencies. Therefore, alternative noise rejection techniques have been developed in the course of time. The first one [20] resorts to the dq transform of the instantaneous values of the voltages at motor terminals at the time the PD event took place. The PD events are thus plotted as dots in the dq plane, When the loci of the PD events form structures that can be explained in terms of phase/phase or phase/ground voltages, as shown in Fig. 6., then the recorded phenomena can be ascribed to partial discharges taking place within the insulation system. In a way similar to what is done under AC sinusoidal voltages, a PD pattern can be drawn plotting the PD magnitudes against the phase of the corresponding event in the dq plane. Since, PD magnitudes occurring at wire surfaces exhibit large dispersion, patterns ascribable to PD events tend to show vertical structures at some specific phase values. On the contrary, as shown in
291
Fig. 7 the pattern associated with random noise shows dots that tend to disperse everywhere, depicting the typical hexagon structure shown in Fig. 7. In this case, the recorded magnitudes (not associated with PD events, of course) show moderate dispersion.
ascribable to PD events (vertical structures at specific phase, corresponding to or in proximity of switch commutations), from a pattern ascribable to noise (horizontal structure).
Fig. 8. PD pattern in an inverter-fed motor affected by PD. The reference voltage is the 50Hz component of the applied voltage.
Fig. 6. Characteristics of dq loci and Hq pattern for phase/ground discharges in a coil..
Fig. 9. PD pattern in an inverter-fed motor that is PD free. The reference voltage is the 50Hz component of the applied voltage.
CONCLUSIONS
Fig. 7. Characteristics of dq loci and Hq pattern for random noise
Another tool that has been proposed to distinguish noise from PD is the use of measurements synchronized with the AC voltage fundamental component [21,22]. Figures 8 and 9 show the clear differences existing between phenomena
Since the mid-nineties of the last century, when inverterrelated failures of motor insulation systems were first understood clearly, putting in spotlight the role of partial discharge-induced degradation, the electrical industry has moved to take countermeasures to minimize the number of failures. The IEC has produced to Technical Specifications (TS) 60034-18-41 (wire-wound motors) and 60034-18-42 (formwound motors) that address the problem by ensuring that motors to be used under inverter voltages are manufactured with a quality good enough to guarantee that the lifetime will approach the expected one. Both TS specify qualification and acceptance tests. The first ones are aimed at qualifying materials and processing techniques, the second one to ensure that the manufacturing process will not deviate significantly from the one that has been initially qualified. There is no doubt that both documents will affect in a radical way customer/manufacturer relationship, also considering that the so-called “converter integrator”, i.e., who will
292
provide the inverter used in the drive, will become part of this process. For wire-wound motors, where PD-induced degradation is indeed the fastest degradation phenomena. For this reason, tools having large signal-to-noise ratio and the capability of distinguishing PD phenomena from noise have been devised to measure PD in online motors. These tools can be used in critical motors to signal the inception of PD phenomena and, therefore, trigger maintenance action to prevent unscheduled downtimes.
[2]
[17] A. Cavallini, D. Fabiani and G. C. Montanari, “Power Electronics and Electrical Insulation Systems - Part 1: Phenomenology Overview “ submitted to IEEE Electr. Insul. Mag., 2009. [18] Electrical Insulating Materials and Systems - Electrical Measurement of Partial Discharges (PD) Under Short Rise Time and Repetitive Voltage Impulses, IEC 61934 - TS Ed. 1.0, 2006. [19] D. Fabiani, A. Cavallini, G. C. Montanari, “A UHF Technique for Advanced PD Measurements on Inverter-Fed Motors”, IEEE Trans. Power Electr., Vol. 23, pp.2546-2556, 2008. [20]
REFERENCES [1]
[16] IEC 61934 – TS Ed. 1.0 Electrical insulating materials and systems Electrical measurement of partial discharges (PD) under short rise time and repetitive voltage impulses, 2006.
E. Persson, “Transient Effects in Application of PWM Inverters to Induction Motors”, IEEE Trans. Ind. Appl., Vol. 28, pp.10951101,1992. G.C. Stone, R. G. van Heeswijk, and R. Bartnikas, “Investigation of the effect of repetitive voltage surges on epoxy insulation,” IEEE Trans. Energy Convers., vol. 7, no. 4, pp. 754–759, Dec. 1992.
[3]
M. Kaufhold, G. Borner, M. Eberhardt and J. Speck, “Failure Mechanism of Low Voltage Electric Machines Fed by PulseControlled Inverters”, IEEE Electr. Insul. Mag., Vol. 12, No. 5, pp. 915, 1996.
[4]
Weijun Yin, “Failure Mechanism of Winding Insulation in InverterFed Motors”, IEEE Electr. Insul. Mag., Vol. 13, No. 6, pp.18-23, 1997.
[5]
G. C. Stone, S. Campbell and S. Tetreault, “Inverter-Fed Drives: Which Motor Stators are at Risk?”, IEEE Ind. Appl. Mag., Vol. 6, No. 5, pp. 17-22, October 2000.
[6]
J C. G.Wheeler, “Effects of converter pulses on the electrical insulation in low and medium voltage motors”, IEEE Electr. Insul. Mag., Vol. 21, No. 2, pp. 22-29, 2005.
[7]
H. Okubo, N. Hayakawa, G.C. Montanari, “Technical development on partial discharge measurement and electrical insulation techniques for low voltage motors driven by voltage inverters”, IEEE Trans. Dielectr. Electr. Insul., Vol. 14, pp. 1516-1530, 2007.
[8]
F. P. Espino-Cortes, E. A. Cherney, S. H. Jayaram, “Impact of inverter drives employing fast-switching devices on form-wound AC machine stator coil stress grading”, IEEE Electr. Insul. Mag., Vol. 23, No. 1, pp. 16-28, 2007.
[9]
L. Testa, “Physical model of damage inception and growth in polymeric insulating materials subjected to partial discharges under DC and AC voltage”, PhD thesis, University of Palermo, 2009.
A. Cavallini, D. Fabiani, G. C. Montanari, “A Novel Method to Diagnose PWM-fed Induction Motors”, IEEE Trans. Dielectr. Electr. Insul. , Vol. 15, pp. 1313-1321, 2008.
[21] M. Tozzi, A. Cavallini, G. C. Montanari, “Monitoring Off-line and On-line PD under Impulsive Voltage on Induction Motors, Part I: Standard Procedure”, submitted to IEEE Electr. Insul. Mag. [22] M. Tozzi, G.C. Montanari, D. Fabiani, A. Cavallini and G. Gao, “Offline and On-line PD Measurements on Induction Motors fed by Power Electronic Impulses”, in Proc. IEEE EIC, Montreal, Quebec, pp. 420424, May 2009.
[10] D. Fabiani and G. C. Montanari, “The effect of voltage distortion on aging acceleration of insulation systems under partial discharge activity,” IEEE Electr. Insul. Mag., vol. 17, no. 3, pp. 24–33, Jun. 2001. [11] D. Fabiani, G. C. Montanari, and A. Contin, “Aging acceleration of insulating materials for electrical machine windings supplied by PWM in the presence and in the absence of partial discharges,” in Proc. IEEE ICSD, Eindhoven, The Netherlands, 2001, pp. 283–286. [12] D. Fabiani, G.C. Montanari, A. Cavallini, G. Mazzanti, “Relation between space charge accumulation and partial discharge activity in enameled wires under PWM-like voltage waveforms”, IEEE Trans. Dielectr. Electr. Insul., Vol. 11, pp. 393-405, 2004. [13] IEC 60034-18-41 TS Ed. 1: Rotating electrical machines – Part 1841: Qualification and type tests for Type I electrical insulation systems used in rotating electrical machines fed from voltage converters, 2006 [14] IEC 60034-18-42 TS Ed. 1: Rotating electrical machines – Part 1842: Qualification and type tests for Type II electrical insulation systems used in rotating electrical machines fed from voltage converters, 2006 [15] R. J. Kerkman, D. Leggate and G. L. Skibinski, “Interaction of Drive Modulation and Cable Parameters on AC Motor Transients”, IEEE Trans. Ind. Appl., Vol. 33, pp. 722-731, 1997.
293
