Comparative investigation of electric signal analyses methods for mechanical fault detection in induction motors

Transcription

1 Comparative investigation o electric signal analyses methods or mechanical ault detection in induction motors Eltabach Mario, Charara Ali,. Faculté des sciences et de génie inormatique, Université de Saint Esprit de Kaslik, USEK; B.P.44 Jounieh, Liban. marioeltabach@usek.edu.lb. HEUDIASYC UMR 599; Université de Technologie de Compiègne; BP 59, Compiègne, France ali.charara@hds.utc.r Abstract: This paper present a comparative investigation o various media or non invasive diagnosis o mechanical abnormalities in induction motors. Stator voltages and stator currents as well as noises on these signals are simulated irst or a ault ree motor then or a motor with mechanical abnormalities. These signals are subsequently employed to compute the instantaneous powers P ab, P cb, P abc and last the current park vector modulus known as the Extended Park Vector Approach EPVA method. Waveorms o these simulated signals were analyzed using the power spectral density transormation. Commercially available diagnostics systems uses the act that the amplitude o some components known as ault characteristic requencies in these electrical signals, increase when mechanical abnormalities occur. The amplitude increase o these characteristic requencies is employed as a criterion in order to investigate noise immunity o the ive diagnosis methods thus making possible their classiication. Simulation and experimental results show that the extended park vector approach EPVA reveal the highest noise immunity. Utilization o the EPVA is thus enhancing the reliability o diagnostics o induction motor drives. I- INTRODUCTION Mechanical aults in electrical drive systems can shut down an entire process. Those unexpected shut downs have a cost, which result in loss production and inancial income. During the past iteen years there has been a substantial amount o research into the creation o new condition monitoring techniques or induction motor drives. This has reduced unexpected ailures, increased the time between planned shutdowns or standard maintenance and reduced operational costs. However the machine s operator must treat each induction motor drive as a unique entity. The potential ailure modes, undamental causes, mechanical load characteristics, and operational conditions have all to be taken into consideration when a condition monitoring system is being selected. In some cases there are several signals wich may contain inormation on the ailure mechanism. Generally we select a condition monitoring system based on evidence o its reliability to diagnose problems in industrial drives and on its applicability to the particular industrial installation. Vibration monitoring techniques are only usually installed on expensive and sensitive machines where the cost o such systems can be justiied. Moreover, the environmental sensitivity o the sensors can provide unreliable indications. In some situation, such as or electrical submersible pumps, sensor installation is not practical or prohibitively costly. Motor current signal analyses methods MCSA are obtained using only non-invasive sensors as current and voltages sensors. These methods are based on spectral analyses o electrical signals as stator currents or the instantaneous power.

2 This paper reports results o research on noise sensitivity o ive non invasive diagnostic methods and speciically, their spectral response to mechanical abnormalities. The diagnostic media studied included the ollowing: - The spectral analyses o the simple current, one o the three supply lines [] to [4]. - The spectral analyses o the partial instantaneous powers: Pab, Pcb and Pabc [] to [8]. 3- The spectral analyses o the current park vector modulus, the EPVA method [5]. A theoretical discussion o the diagnostic potential o these media is presented briely in section II, the simulation setup and the discussion results are presented in section III and inally experimental results are shown in section IV.. II- THEORETICAL BACKGROUND Considering a healthy induction motor, supplied rom a balanced three-phase source o sinusoidal voltages and driving a constant load, the ollowing waveorms o selected stator voltages and currents may be assumed: Vab = U m cos( w () π Vcb = U m cos( wt + ) 3 () π ia = I cos( wt α ) (3) π π ib = I cos( wt α ) 3 (4) π π ic = I cos( wt α + ) 3 (5) where U m, I, w, and α are respectively the maximum value o the line to line voltage, the maximum value o line currents, the supply requency in radians per second, and inally the power actor angle, see Fig.. As a unction o the three phases currents ( i a, i b, i c ), the motor current Park s Vector components ( i d, i q ) presented in a d-q plan are: = ( ia ib i ) () 3 = ( ib i ) (7) d ( c i q ( c i Considering a total balanced three-phase source, the current Park vector modulus can be computed as ollows: i = I (8) p id + iq =

3 π/3 vb ib q π/3 + ia α π/ d va vab ic vcb vc Fig.. Current and voltage representations or induction motors Multiplying V ab by i a and V cb by i c yields partial instantaneous input powers P ab, P cb then adding the two partial powers will lead to the total instantaneous input power P abc : U mi π π Pab = (cos( α + ) + cos(wt α )) (9) U mi π 5π Pcb = (cos( α ) + cos(wt α + )) () 3 P abc = Pab + Pcb = U mi cos( α) () i a mechanical abnormality develops in a drive system, such as rotor cage ault, motor-load shat misalignment, broken teeth in the load gearbox, or vibration, harmonic torques are generated in the motor, accompanied by speed illation, and modulation o the stator currents, typically in a periodic manner. In the case o periodic disturbances, all three line currents ia, ib, and ic are simultaneously modulated with the undamental requency o the aultinduced illation o motor variables. For simplicity, it is assumed that a mechanical ault in the drive system causes sinusoidal modulation o amplitude o the stator current, while the phase modulation o the current is negligible. Thus, the three currents o the supply line may now be expressed as: i = i * ( M cos( w ) () a a + b ib *( + c ic *( + i ( = M cos( w ) (3) i ( = M cos( w ) (4)

4 where M is the modulation index and w is the modulating radian requency ( w =π ). The value o the modulation index depends on the severity o the abnormalities substituting (3) in () we get: MI π π ia = ia + (cos[( w + w ) t α ] + cos[( w w ) t α ]) (5) Clearly, in the power spectrum o current, two sideband components will appear about the undamental, at requencies =(w+w )/π and =(w-w )/π. With these assumptions, motor with a mechanical ault, the Park s vector components (id, iq) are shown below: = ( ia ib i ) () 3 = ( ib i ) (7) d ( c i i q ( c Substituting the values o ( i, i, i ) o equations (, 3, 4) in equations () and (7) yields: a b c i = i *( M cos( w ) (8) d d + q iq *( + i ( = M cos( w ) (9) i = id + iq = I *( M cos( w) () p + The expressions or the modulated instantaneous power, obtained by multiplying the voltages by they corresponding currents are: Pab = Vab * ia () π π π Pab = Pab ( + MUm I (cos[( w+ w) t α ] + cos[( w w) t α ] + cos( α + )cos( w) () 4 Pcb = Vcb * ic (3) 5π 5π π Pcb = Pcb ( + MVLLI L (cos[( w+ w) t α + ] + cos[( w w) t α + ] + cos( α )cos( w) (4) 4 Last we can show that the latest expressions can be expressed as below: P P P i = P *( M cos( w ) (5) ab ab + cb Pcb *( + abc Pabc *( + ( = M cos( w ) () ( = M cos( w ) (7) = i *( M cos( w ) (8) p p + t Our irst interested result can be expressed as below: With the assumptions that a mechanical ault in the drive system causes only sinusoidal modulation o amplitude o the stator currents, the simple currents, the instantaneous partial and

5 total powers, the current vector components and inally the current Park vector modulus are modulated with the same modulation index, in other words these signals are aected by the same manner. III- SIMULATION SETUP This article treats a comparison investigation between methods that use only the electric signals o the induction motors. Signals are simulated according to a model o the electric signals or a ault ree motor and a motor with mechanical abnormalities. III-a Simulation setup In the simulation, equations () to (5) are generated to model the electric signals or a ault ree model and equations () to (4) are utilized to model the electric signals or a motor with mechanical abnormalities. A simulated motor o 3 KW, 38 V, 5 Hz two pole induction motor is used. The nominal current o the motor is 5 Amp as an rms value and the power actor angle is given by his cosine cos(α)=.89. Concerning the mechanical abnormalities the modulation index M is taken % and the ault characteristic requency is taken Hz. Additive noise is added to the simulated currents and motor voltages, these noises are assumed to be white noises, zero mean noises and with a variance o one. Thereore these noises are multiplied by two gains the K nc and the K nv in order to ampliy respectively the currents and the voltages noises. These noises determine the severity o the additive noises and take the values in the ranges: [-.5]Amp or the current noise gain and [-5]Volts or the voltage noise gain. Finally or each generated electric signals a 4995 values were recorded, with the sampling requency o Khz. III-b Signals presentation Spectral analyses o these signals can lead to better monitoring and diagnosis o the induction motors. First, with the assumptions taken to model mechanical abnormalities, (, 3, 4) we can clearly notice that in the case o mechanical abnormalities the spectrum o the simple current will contain, apart rom the undamental requency =5Hz, two sideband components at requencies ± while in the case o a ault ree motor the spectrum contains only the undamental requency. Second, in the case o a ault ree motor the electric signals P abc and current Park s modulus are constant that why the spectrum o these signals in the case o a mechanical aults contain only a spectral component directly at. This component is characteristic o the abnormality, not o the motor. Finally, concerning the instantaneous partial powers Pab and Pcb we can see rom equations (9,,, 4) that in the case o a ault ree motor the spectral analyses o these signals reveals the presence o only a undamental component at twice the power supply undamental = Hz. In another hand when mechanical abnormalities appear, the spectrum o these signals contains the component at, two sidebands components at ± and a ourth spectral component directly at. III-c Noise immunity and comparison Previous studies like in [9], have made a comparative investigation o some electric signals methods or the detection o mechanical abnormalities. This comparative approach treated only the instantaneous total and partial powers using the amplitude o ault characteristic requency

6 as a criterion to classiy these detection methods. In act rom equations (5), (), (), (4), and 7 we can extract respectively the amplitudes Ac, Acp, Aab, Acb, Aabc, o the ault characteristic requency d. This requency takes the value - or the simple current detection method and the value o or instantaneous powers and the EPVA detection methods. A c = MI (9) A cp = MI (3) A cos( ab = MU mi α + ) (3) A cos( cb = MU mi α ) (3) 3 A abc = MU mi cos( α) (33) Fig. shows the values o the ault characteristic requency amplitudes unction o the power actor angle. The range o the power actor angle is taken between and 8 degree but generally practical motors are designed that within a wide range o load torque, the power actor angle does not change signiicantly, staying at a low level to maintain a high power actor. However comparison between these amplitudes doesn t have a physical meaning i they aren t reerred to a reerence. Normally the amplitude values o spectral components at the ault characteristic requency in case o a motor with no mechanical abnormalities, is the reerence. We deine a criterion A which denotes the increase in amplitude, when mechanical abnormalities occur, o the spectral component at the ault characteristic requency rom its value when there are no aults. 3 P abc Amplitude P cb - P ab Power actor angle (degree) Fig.. Simulated ault characteristic component Amplitude unction o the power actor angle and the detection method or a motor with mechanical abnormalities.

7 Theoretically the equations (5) to (8) reveal that the increase in amplitude or all electric signals at the ault characteristic requency is perectly the same. Simulation results in Fig.7 show clearly the act. Thereore noisy signals perturb the diagnosis analysis. The spectrums o the electrical signals are very aected by noise coming rom measurement procedures done to the currents and the voltages. In general noises increase the low level o the spectrums so why our criterion A is aected. Figs. (3, 4, 5, ) present the spectrums o the simulated electric signals in the spectral range where the ault characteristic requencies may occur. All the igures show that when mechanical abnormalities occur, the amplitude o the spectral components at the ault characteristic requencies increase. This increase o amplitude A is directly linked to the severity o the ault so why it is used as a criterion in order to classiy the non invasive diagnostic procedures. Finally high amplitude o spectral components indicative o the ault is not the only quality expected or a good diagnostic method but the increase in amplitude rom the ault ree situation under noisy environmental is the criterion that seems more adequate to make a comparative investigation or the electric signal diagnostic methods. This paper investigates the noise immunity o each electric signal that can be used to elaborate mechanical abnormalities detection and diagnostic method in induction machine. The electric signals treated below are the simple current, the current Park vector modulus, the instantaneous partial powers P ab, P cb, and last the instantaneous total power P abc. To accomplish this work we calculated the criterion average value A over 5 simulated sample noisy electric signals or all the methods. The amplitude o the injected noise to currents and voltages is controlled by a gain K nc, K nv respectively or currents and voltages amplitude noises see Fig.8. The eect o the current noise amplitude is investigated into the range [.,.5] Amp using a step o. Amp while the voltage noise amplitude is investigated into the range o [.5, 5] Volts using a step o Volt. For each step o the noise current and voltage amplitude ive values o the classiication criterion A is calculated and saved corresponding to the ive diagnostic methods.

8 Amplitude( db) -5 ia Amplitude( db) A -5 A Pcb (bk) Pab (mag) Frequency (Hz) Fig.3. Spectrum o the simple current or a ault ree motor (dashed line) and a motor with mechanical abnormalities (straight line) 4 8 Frequency (Hz) Fig.4. Spectrum o the partial instantaneous powers or a ault ree motor (dashed line) and a motor with mechanical abnormalities (straight line) Amplitude( db) 5 Pabc Amplitude( db) -5 EPVA - -5 A -5 A Frequency (Hz) Fig.5. Spectrum o the total instantaneous power or a ault ree motor (dashed line) and a motor with mechanical abnormalities (straight line) Frequency (Hz) Fig.. Spectrum o the current Park vector modulus or a ault ree motor (dashed line) and a motor with mechanical abnormalities (straight line) 4 Amplitude( db) Pabc (green) Pcb (black) Pab (mag) EPVA (cyan) ia (red) Frequency (Hz) Fig.7. Spectrums o ive electric signals at the range o the ault characteristic requency or a ault ree motor (dashed line) and a motor with mechanical abnormalities (straight line)

9 Current generation or a ault-ree motor (i a, i b, i c ) and or a motor with a mechanical ault (i a, i b, i c ) Voltages generation or a ault-ree motor (V ab, V cb ) and or a motor with a mechanical ault (V ab, V cb ) White noise zero mean, Variance = White noise zero mean, Variance = K nc Computation: 3- noisy voltages(ault + ault-ree) 4- noisy currents K nv Computation: - Current Park vector modulus - Partial Powers K nc =K nc +S c Electric signals spectral analyses. Computation o the average value o A or every method over Nb=5 samples o 3 seconds each Knv =K nv +S v Y Nb 5 N Save all «A» Y Y K nc <K nc N K nv <K nv N END Fig.8 Flow chart or computation o the amplitude increase average when mechanical abnormalities occur, o the ault characteristic requency unction o the detection method and the current and voltages noises

10 Current noise (A) Voltage noise (V) Current noise (A) Voltage noise (V) 5 Fig.9 The amplitude increase in the simple current spectrum o the component at the ault characteristic requency 5- unction o voltages and current noises Fig. The amplitude increase in the current Park vector modulus spectrum o the component at the ault characteristic requency unction o voltages and current noises Current noise (A) Voltage noise (V) Fig. The amplitude increase in the Partial power Pcb spectrum o the component at the ault characteristic requency unction o voltages and current noises Current noise (A).5 3 Voltage noise (V) Fig. The amplitude increase in the Partial power Pab spectrum o the component at the ault characteristic requency unction o voltages and current noises Current noise (A).5 3 Voltage noise (V) Fig.3 The amplitude increase in the total power Pabc spectrum o the component at the ault characteristic requency, unction o voltages and current noises 4 5

11 EPVA Voltage noise :.5 V 5 Pabc Pcb 5 ia Pab Current noise (A) Fig.4 The amplitude increase o the spectral component at the ault characteristic requency, unction o the detection methods and current noises, (voltage noise =.5 V) EPVA Pabc Voltage noise : 9.5 V Pcb 5 ia 5 Pab Current noise (A) Fig.5 The amplitude increase o the spectral component at the ault characteristic requency, unction o the detection methods and current noises, (voltage noise = 9.5 V) 45 4 EPVA Current noise:. Amp Pabc Pcb ia 5 5 Pab Voltage noise (V) Fig. The amplitude increase o the spectral component at the ault characteristic requency, unction o the detection methods and voltage noises, (current noise =. A)

12 EPVA Pabc Pcb ia Current noise: Amp Pab Voltage noise (V) Fig.7 The amplitude increase o the spectral component at the ault characteristic requency, unction o the detection methods and voltage noises, (current noise = A) Current noise(amp) 3.5 EPVA Pabc Threshold:.5 db Pcb ia Pab Voltage noise (V) Fig.8 Methods Classiication by ixing a threshold value o.5 db or the amplitude increase o the spectral component at the ault characteristic requency..8 Current noise(amp) Pabc EPVA Pcb ia Threshold: db Pab Voltage noise (V) Fig.9 Methods Classiication by ixing a threshold value o db or the amplitude increase o the spectral component at the ault characteristic requency.

13 III-d Discussion o the results Fig.9 presents the amplitude increase A in the simple current spectrum o the component at the ault characteristic requency unction o the voltages and currents noises. This characteristic requency is ound at 5- which is in our case 48 HZ. We can see clearly that this criterion is normally not aected by the voltage noises. Starting with a current noise amplitude o. Amp the A criterion have a value o 4 db then this value decreases to zero when the current noise amplitude increases. Fig. represents the A criterion applied to the EPVA detection method. Once more this criterion is not sensitive to voltages noises because the last method does not use the supply voltages. The increase o the current noises decreases the value o the comparative criterion. Fig. presents the amplitude o the comparative criterion A or the partial power P cb. We can notice that this criterion is sensitive or both current and voltage noises. The criterion value will tend to zero either when voltage noises or current noises increase. Figs.(, 3) show the comparative criterion value unction o the currents and voltages noises respectively or the partial power P ab and or the instantaneous total input power P abc. In order to make a comparison between the diagnostic procedures we have group all the values o the comparison criterion or the dierent diagnostic methods. Figs.4, 5 shows the values o the comparison criterion or ive non invasive diagnostic methods unction o the current noises at respectively a voltage noise value o.5v and 9.5V which correspond to.3% and 3% o the nominal voltage. We can see that the EPVA method exhibit the highest values o the comparison criterion. This means that this method have more immunity against noises and thereore give better detection o the mechanical abnormalities. Next comes the total input power P abc, P cb, the simple current and last the partial power P ab. Figs., 7 shows the values o the comparison criterion or ive non invasive diagnostic methods unction o the voltage noises at respectively a current noise value o.amp and Amp which correspond to 4% and 4% o the nominal current. We notice in Fig. that only the instantaneous power detection criterion decreases when the voltage noises increase. The value o the detection criterion can be used to classiy the methods. More the value o the criterion is pronounced more the method present noise immunities and thus is more eective. In Fig given that the current noises value is ixed to. Amp and supposing the voltage noises is less then 3 volts the order o the non invasive detection methods rom the most eective to the less one is : EPVA, P abc, P cb, the simple current and last the P ab method. This order change i the voltage noise is greater then 3 volts and in this case the simple current will be more important the all the instantaneous power detection methods. Fig.7 shows that in case o great values o current noises, here about 4% o the current nominal value the classiication o the detection method is the same and do not change with the voltage noises, in this case the EPVA method show the most reliable values o the detection criterion, second comes the total input power, the instantaneous partial power P cb, the simple current, and last the partial power P ab. Figs.8 and 9 show the contour o the criterion A by ixing respectively two thresholds o.5 db and db. The detection criterion contour o all the ive detection methods unction o both voltage and current noises is presented thus rendering very clear the classiication. The Fig.8 invests the case o a threshold o.5 db, in another words we have mechanical abnormalities detection when the amplitude increase o the ault characteristic requency is more than.5 db. We can see in this case that all the non invasive methods show their detection aptitude even with great values o amplitude noises. In act the EPVA method shows little sensitivity to current noises. Mechanical abnormalities are detected even when current noises reach Amp which correspond to 8 % o the nominal current. The instantaneous powers P abc and P cb can deal with current noises up to Amp which correspond to 4% o the nominal current, the simple

14 current can give detection with current noise amplitude up to. Amp and inally the P ab method can deal only with current noise amplitude o.4 Amp. In order to invest the case o a great values o thresholds, Fig.9 show the contour o the criterion A or a threshold o db. In this Fig we can see that i the detection is ixed to a value o db all the treated methods shows critical situations or the detection o the mechanical abnormalities. The EPVA method gives detection only i current noises amplitudes are less then. Amp. The P abc and the P cb detection methods give detection i current noises amplitude is less then.35 Amp and less then 38 volts or voltage noises in the zone delimited by their contour. The simple current method will show no detection i current noises amplitudes are greater then.5 Amp and last the P ab detection method is no useulness i the current noises exceed. Amp or the voltage noises exceed 8 volts. IV- Experimental Results The experimental tests were carried out using data rom the Laboratory o images and signals LIS (Laboratoire des Images et des Signaux) Grenoble-France. The purpose o our benchmark GOTIX is to detect mechanical abnormalities in rotating machinery and especially deects in a gear box using vibration, acoustical and / or electrical supply signals. The benchmark GOTIX, see Fig., is ormed by a 55 kw Leroy Somer asynchronous motor, by a gear box with a multiplicative ratio o (57/5) see Fig., by a speed variable controller and by a DC motor in order to simulate variable loads. GOTIX is equipped with three voltage sensors, three current sensors, eight accelerometers, torque sensor, temperature probes and inally the benchmark is equipped with a speed sensor. The data acquisition system is used to acquire synchronously twenty instantaneous signals with KHz maximum sampling requency per channel. The detection tests were perormed with the equipment described above, or a healthy mode without any mechanical or electrical abnormalities taken as the reerence time (time=), ater 8 hours o continuous work with a load o 7Nm (mode ault) and inally ater 3 hours (mode ault). Obviously ater 8 hours o continuous eort the gear box is exhausted and encloses deects. Fig. shows the gear box equipped by some vibration sensors. Fig. Global view o the benchmark GOTIX o the Laboratory LIS Grenoble France

15 Fig. The gear box with a multiplicative ratio o (57/5). Laboratory LIS Grenoble France. For every o the three operation modes a number o eight signal o ten seconds each are acquired. These signals are re-sampled at requency o.5 KHz. The spectrums o the simple current, the current Park vector modulus and the three instantaneous powers are computed then averaged over the eight experimental set o electrical signals or the three operation mode. The simple current spectrum is normalized with respect to the amplitude o the undamental supply requency. The spectrums o the current Park vector modulus and the three instantaneous powers are normalized with respect to their direct component ( HZ). Fig. shows a comparison between the spectrum o the simple current or a ault ree machine and or the two unctioning modes ault and ault. The spectral analysis clearly shows that when a mechanical ault, gear atigue, is present several components appear at the ault s characteristic requencies in the current spectrum «e ±». These ault s characteristic requencies is directly related to shat rotating requency and its sub harmonics. Fig. 3 shows a comparison between the spectrum o total instantaneous power or a ault ree machine, and in case o ault and ault unctioning modes. The spectral analysis clearly shows that when a mechanical ault is present a component appears at the ault s characteristic requencies which in our case are 3.,. and 9. Hz. Figures 4 and 5 show the spectrum o the partial power Pcb or the three dierent unctioning modes respectively in the low band requencies [,5]Hz and or the band around twice the undamental supply requency [7,3]Hz. The values o the comparison criterion «A» unction o the diagnostic procedure and or two gear ault s severity level are given by table and presented by Fig..

16 Fault Characteristic requencies Fault Characteristic requencies Fault Characteristic requencies Frequency (Hz) Fig. The simple current spectrum or a ault ree machine (doted line) and or a machine with gear ault (straight lines) Frequency (Hz) Fig.3 The total power spectrum or a ault ree machine (doted line) and or a machine with gear ault (straight lines). - Fault Characteristic requencies Fault Characteristic requencies Frequency (Hz) Fig.4 The partial power Pcb spectrum or a ault ree machine (doted line) and or a machine with gear ault (straight lines) in the requency range [, 5Hz] Frequency (Hz) Fig.5 The partial power Pcb spectrum or a ault ree machine (doted line) and or a machine with gear ault (straight lines) in the requency range [7, 3Hz]. Table : A criterion or the characteristic requency 9. Hz, unction o the detection methods and or two gear ault s severity level Gear atigue (8 hours) Gear atigue (3 hours) EPVA Simple current SC Partial power Pcb Partial power Pab Total power Pabc

17 3 Fault 5 5 Fault 5 SC EPVA Pcb Pab Pabc Methods Fig. The amplitude increase o the spectral component at the ault characteristic requency A, unction o the detection methods and the ault s type. The results in Fig. show clearly that the A criterion is directly linked to the gear atigue severity. In act the A values increase in mode ault relatively to mode ault which correspond o an addition o 3 hours o use. We can also see that the EPVA method exhibits the highest values o the comparison criterion or the two unctioning modes, which signiies that this method has more reliable detection capability o gear abnormalities. Next the total instantaneous power Pabc method comes. The instantaneous power Pab occupies the second position. Next the simple current diagnostic procedure comes. Finally we can notice that the instantaneous power method Pcb exhibits relatively low values o the detection criterion A. These experimental results match very well the simulation results. However, we can see a degradation o the detection eectiveness o the instantaneous partial power Pcb that s why more research must be provided to explain this phenomenon. V- CONCLUSION This paper had reported results o research on noise sensitivity o ive non invasive diagnostic methods and speciically, their spectral response to mechanical abnormalities. Simulation results have investigated the noise immunity o the mechanical abnormalities detection methods thus making possible the classiication o these diagnostic methods. Simulations results show the great noise immunity o the EPVA detection method over all the other non invasive methods. Then comes the instantaneous total input power P abc, the partial power P cb, the simple current, and last the instantaneous partial power P ab. Experimental results were carried out using the GOTIX benchmark o the laboratory o Images and signals LIS Grenoble France, treating two severity gear box severity atigue. The experimental results show a quite good agreement with the simulation results. Presently, we are concentrating on the study o the relationship between the angular displacements o the current sidebands components which it seems that this one is directly related to the type o the rotating machines aults. Finally, we

18 don t ail to say that utilization o the EPVA is thus enhancing the reliability o diagnostics o induction motor drives. REFERENCES [] G.B. Kliman and J. Stein, «Methods o motor currents signature Analysis», Electric Machines and power systems vol., n 5, 99, pp [] D. Rankin, «The industrial application o phase current analysis to detect rotor winding aults in squirrel cage Induction motors», IEEE power engineering journal, April 995, pp [3] R. Yacamini, K.S Smith, L. Ran, Monitoring torsional vibrations o electro-mechanical systems using stator currents, journal o vibration and acoustics, vol, 998, pp [4] M.E.H. Benbouzid, «A Review o induction Motors signature Analysis as a medium or aults detection, IEEE Trans. Indust. Electronics, vol. 47, n 5,, pp [5] S.M.A. Cruz, A.J.M Cardoso «Rotor cage Fault Diagnosis in three-phase induction Motors by Extended Park s Vector Aproach», Electric Machines and power systems, vol. 8,, pp [] A.M. Trzynadlowski, M. Ghassemzadeh and S.F. Legowski, «Diagnostics o Mechanical Abnormalities in induction Motors using Instataneous electric Power», IEEE Trans. Energy Conversion, vol. 4, n 4, 999, pp [7] F.L. Stanislaw, A.H.M. Sadrul Ula and A.M. Tryzynadlowski, «Instantenous Power as a medium or the signature analysis o induction Motors», IEEE Trans. Ind. Appli., vol. 3, n 4, 99, pp [8] S.M.A. Cruz and A.J.M. Cardoso, «Rotor cage Fault diagnosis in three-phase induction Motors by the total instantenous power spectral Analysis», IEEE Industry Applications Conerence, vol. 3, 999, pp [9] A.M Trzynadlowski, Ewen Ritchie «Comparative Investigation o Diagnostic Media or Induction Motors: a case o rotor cage aults» IEEE transactions on Industrial electronics vol : 47, 999, pp.9-99