Abstract - This paper investigates AC induction motor shaft voltageproblems, current flow thru motor bearings and electric dischargecurrent problems within bearings when operated under both puresinewave and Pulse Width Modulated (PWM) inverter sources.Recent experience suggests that PWM voltage sources with steepwavefronts especially increase the magnitude of the above electricalproblems, leading to motor bearing material erosion and earlymechanical failure. Previous literature suggests that shaft voltage -bearing current problems under 60 Hz sinewave operation arepredominantly electromagnetically induced. It is proposed thatunder PWM operation these same problems are nowpredominantly an electrostatic phenomenon. A system model todescribe this phenomenon is characterized and developed.Construction and test of a new Electrostatic Shielded InductionMotor (ESIM) verifies this model and is also a possible solution tothe bearing current problem under PWM operation.

I. IntroductionBearing currents and shaft voltages under 60 Hz sinewave

operation has been a recognized problem since 1924 [1-3].The bearing impedance characteristic largely determines theresulting bearing current that will flow for a given shaftvoltage magnitude and waveform present. A number ofsurveys have indicated that 30 % of all motor failures operatedwith 60 Hz sinewave voltage are due to bearing currentdamage [4]. All rotating machines potentially have a bearingcurrent problem whether it is DC or AC, and either large orsmall horsepower in size. These rotating machines have threebasic sources of shaft voltage - electromagnetic induction,electrostatic coupled from internal sources or electrostaticcoupled from external sources.

Electromagnetic induction from the stator winding to therotor shaft was recognized by Alger [1] and is more prevalentin long axial machines. The shaft voltage is due to smalldissymmetries of the magnetic field in the air gap that areinherent in a practical machine design. Most inductionmotors are designed to have a maximum shaft voltage to frameground of < 1 Vrms with recommended practice limits statedin [5]. The induced shaft voltages cause bearing current flowin a circulating path from the shaft, thru side A groundedbearing, thru the stator frame, thru side B grounded bearingand back to the shaft. The induced shaft voltage, although low

in magnitude, results in a high circulating current thru bothmotor bearings since the impedance of the circulating path islow. Modern day induction motors less than 250 horsepowerhave grounded bearings but have minimized steady state shaftvoltage to extremely small values. However, during transientstart and stop conditions across the AC line, magneticdissymmetries appear as increased shaft voltage, resulting inbearing current flow and reduced life [4]. This transientbearing current flow for line started motors wasexperimentally verified. The traditional electromagneticsolution to induced shaft voltage on larger frames is to insulatethe non drive end bearing. This does not mitigate shaftvoltage but rather the resulting bearing current.

Electrostatic induced shaft voltage may be present in anysituation where rotor charge accumulation can occur.Examples are belt driven couplings, ionized air passing overrotor fan blades or high velocity air passing over rotor fanblades as in steam turbine [6]. The electrostatic solution is tokeep the shaft and frame at the same potential by installing ashaft grounding brush to reduce electrostatic build up andreduce shaft voltage to 70 - 400 mV. This value is not enoughto cause damaging bearing current to flow.

Electrostatic coupled shaft voltage from external rotorsources, such as a static exciter in a turbine generator, ispossible and historically solved with the application of a shaftgrounding brush [6]. Electrostatic coupled shaft voltage fromexternal stator sources, such as a PWM inverter, isinvestigated in this paper.

A. Present Theory of Bearing Current with AC Line

The shaft voltage magnitude measured is commonly usedas an indicator of the possible bearing current that results. It isthe magnitude and passage of electrical current thru thebearing that results in ultimate mechanical damage [7].Bearing damage caused by electrical current is characterizedby the appearance of either pits or transverse flutes burnt intothe bearing race. Electrical pitting continues until the bearingloses its coefficient of friction, further increasing the lossesand breaking up bearing surface. Typical fluting results in awashboard like formation that appears on the race as shown in

IEEE APEC Conference Dallas. TX March, 1995

Effect of PWM Inverters on AC Motor Bearing Currents and Shaft Voltages

Jay Erdman, Russel J. Kerkman, Dave Schlegel, and Gary Skibinski

Allen Bradley Drives Division6400 W. Enterprise Drive P.O Box 760

Mequon, WI 53092(414) - 242 - 7151 (414) - 242 - 8300 Fax

Fig. 1. It has been proposed that the current density of the ballbearing contact area with the race is a better identifying factorfor permissible peak amps allowed without pitting or fluting.However, this contact area is difficult to analyze since it varieswith bearing speed and load, vibration, method of installation,viscosity and temperature of the lubricant. It is known that thecontact area increase is proportional to the bearing load raisedto approximately the 1/2 power [8].

Thus, it is important to characterize the impedance of thebearing under different loading conditions to determine theproblem severity. Surface contact is made in three ways: metalto metal, quasi-metallic surface contacts and metal pointcontact thru electrically insulating surfaces between the ballsurface roughness and race roughness.

The actual bearing contact zone area in a slow moving ornon-rotating bearing is large and consists mostly of

quasi-metallic surfaces. The lubricant film is only 50Angstroms (1 Ao = 10-10 m) while quasi-metallic surfaces havemetallic oxides of 100-120 Ao. Quantum mechanical tunnelingeffects enable the current to pass thru the contact zone withseries resistances < 0.5 Ω . This is evidenced by the lowbearing resistance measurement made at low speeds in Fig. 2.Reference [7] suggests that large current may pass thrunon-rotating bearings without damage.

The actual bearing contact zone area in a rotating bearingis smaller and depends on bearing surface roughness. Thecontact area comprises primarily of asperity point-like contactof ball metal to race metal as shown in Fig. 3a for low speedoperation. High speed operation in Fig 3b has fewer asperitycontact points. Asperity contact duration is typically 100 µs atlow speed and 33 µs at high speed. The increased bearingresistance with rotation shown in Fig. 2 suggests that thelubricant is introducing a partially insulating film betweenball and race at speeds greater than 10% of rated. Typicalsurface roughness of the race and ball from Fig. 4 is seen to bein the 1 - 10 micron (1 micron = 1 µm) range while the typicallubricating film of 0.1 - 2 micron depends on speed, lubricantcharacteristics and to a lesser extent on load [7]. Fig. 5 showsthe relationship between oil film and surface roughness in a

IEEE APEC Conference Dallas. TX March, 1995

Fig. 1 Fluting of AC Drive Motor Bearings

Fig. 2 Bearing Resistance vs. Speed

c) Perfect Bearing

b) High Speeda) Low Speed

Fig. 3 Asperity Contact Possibilities [8]

Fig. 4 Waviness and Vibration Spectra From Inner Ring With Accentuated Waviness [8]

bearing [8]. Percent film is the time percentage during whichthe "contacting " surfaces are fully separated by an oil orlubricant film while Gamma is the relationship of lubricantfilm thickness to rms value of contacting surface roughness.Most bearing applications operate in the Gamma = 1 to 2region. This implies that high quality bearings look like a highresistive impedance 80 % of the time with the oil film actingas a capacitor ready to charge to breakdown potential. Alower quality bearing will have low resistance metal to metalcontact a majority of the time and in the presence of highresistivity lubricant acts as a race to ball junction capacitorthat may charge only randomly during non contact peak tovalley points.

The magnitude of the shaft voltage will determine thebearing current present in lower quality bearings havingasperity contacts the majority of the time or high qualitybearings that use low resistivity lubricants. A high shaftvoltage causes increased current and pits or craters to formsince bearing current flows thru a number of points. Heatingcan occur at point contact to such a degree that the materialmelts creating craters, thus liberating wearing metal particlesinto the lubricant. A low shaft voltage has lower currentamplitudes but has been found to still cause corrosive type ofpitting due to grease decomposition.

In high quality bearings with high resistance grease, thejunction bearing capacitor may discharge into a lowimpedance circuit when the electric field exceeds thebreakdown strength in the lubricant asperity points . Thebearing breakdown voltage threshold is 0.4 volts since mineraloil field strength is 106 v/m, a typical oil film is 0.2 micronsand there are two films in series. On occasion the bearingcapacitor voltage, charged by the shaft voltage present, becomes high enough ( > 0.4 volts) to break down the greaseand a short (nanoseconds) high current impulse flows from thecharged oil film capacitor within the bearing as shown in Fig.6. This discharge current pulse, if it occurs, is a prime sourceof bearing erosion and is commonly referred to as fluting orElectric Discharge Machining (EDM ). The washboard craters

of Fig. 1 are formed from the microscopic pits that softenunder repetitive heating of the race to its melting temperature.

Several authors suggest that shaft voltage < 0.3 volts issafe, while 0.5 - 1.0 volts may develop harmful bearingcurrents, and shaft voltages > 2 volts may destroy the bearing.The rotating bearing breakover threshold voltage (whenbearing current starts to flow) was measured under DC sourcevoltage to be 700 mv peak.

B. Proposed Theory of Bearing Current with PWM Inverters

The preceding analysis was based on steady state, lowfrequency and low dv/dt shaft voltage sources. However, PWMinverter modulation causes high frequency step-like voltagesource waveforms and high dv/dt's to be impressed across thestator neutral to frame ground. It is shown that a portion ofthis waveform is also present as rotor shaft voltage to grounddue to capacitor divider action. The preceding sinewaveanalysis applies to PWM operation but with the change thatthe experimental static breakdown threshold voltage on therotor shaft increases to 8-15 volts ( Fig. 6) vs. 700 mv for thesame bearing monitored under 60 Hz sinewave operation (Fig.10). This increase is explained using dielectric breakdowntheory for pulsed sources [9]. Fig. 7 shows that the impulsebreakdown strength of hexane (1.1 106 v/m) increasesdramatically over the static value for short step-like pulsedurations. The bearing voltage breakdown threshold alsoincreases as a function of shaft voltage rate of change [10].This increased breakdown level under PWM operation isundesirable since during bearing discharge the resulting EDMbearing currents are much higher than with sinewaveoperation. Fig. 8 shows that rough surfaces typically seen inbearings will have a statistical time lag of 3 us prior tobreakdown, which agrees with measured value of Fig 6.

It is theorized that the high quality bearings of Fig. 5(Gamma = 2 ) give long mechanical life when used undersinewave operation but may lead to premature bearing current

IEEE APEC Conference Dallas. TX March, 1995

Fig. 5 Percent Film vs. Gamma for a Bearing [8] Fig. 6 EDM Capacitive Charging Characteristics

failure under inverter operation due to the bearing junctioncapacitor being impulse charged 80 % of the time to higherimpulse shaft voltages. This will result in higher destructiveEDM discharge currents. The low quality bearings of Fig. 5(Gamma =1) give low mechanical life bearings when usedunder sinewave operation but may actually be better forinverter operation since the destructive capacitive EDMcurrents only occur 5 % of the time due to asperity contactresistance shorting the bearing.

Test results of a 15 HP motor ( with grounded motorbearings) under 60 Hz steady state sinewave operation showedno evidence of EDM current occurring, except on across theline starting. Test results on the same motor under BipolarJunction Transistor (BJT) and Insulated Gate BipolarTransistors (IGBT) PWM inverter sources however did showevidence of EDM and fluting on a continuous basis.

II. Effect of PWM Drives on Bearing Current

A. Test Structure and Instrumentation

The measurement of the contributors to bearing roughnessinduced by PWM voltage source inverters requires detectingsignals within a noisy environment. The identification of thecontributors requires an experimental structure with testinstruments that provide isolation, but adequate sensitivity.Fig. 9 shows the test fixture and instrumentation employed forthe investigation presented in this paper. The motor was a 15Hp, 460 volt, 8 pole, induction motor. The drive and nondrive bearings were insulated. A grounding strap simulatednormal grounded bearings. A carbon brush sensed the rotorshaft voltage. The stator neutral was available for measuringthe stator neutral to ground voltage. High voltage probes withan isolation amplifier performed voltage measurements and acurrent probe detected the current through the groundingstrap. A digital sampling oscilloscope with mass storageprovided a tracking of the desired signals. A spectrumanalyzer detected the frequency and phase content of thevoltages and current.

B. Sine Wave Operation of the Induction Motor

Bearing and shaft currents are not specific to motorsoperating from PWM voltage source inverters. Algerinvestigated shaft and bearing currents in the 1920's. Excitingthe induction motor with sine waves provided a referencecondition. Measurements of the stator neutral to ground androtor to ground voltages and rotor current were made whileoperating the induction machine at no-load and 60 Hz. The

IEEE APEC Conference Dallas. TX March, 1995

16

14

12

10

8

6

4

2

0

Tim

e La

g ( u

s )

SmoothCathode

RoughCathode

RoughCathode

Step FunctionPulse

Fig. 8 Surface Roughness Effect on Statistical Time Lag to Breakdown [9]

AC Line460 Volt

Earth Ground

AC Motor

U V WGNDNeutral

CarbonBrush

GND U V W

L1

L2

L3

AC Drive

StatorNeutralVoltage

ShaftVoltage

ShaftCurrent

50 X Differential Probe

200 X Differential Probe

Current Probe

GroundingStrap

Oscilloscopeand

Spectrum Analyzer

Fig. 9 Test Fixture and Instrumentation

1.9

1.8

1.7

1.6

1.5

1.4

Pul

se S

treng

th (

MV

/ cm

)

Pulse Duration ( uS )

Pulse Shape

0 0.5 1.0 1.5 2.0

Fig. 7 Increased Dielectric Strength with Impulse Sources [9]

results of those tests are shown in Fig. 10. EDM currents werenot detected. The 60 volt stator neutral voltage induced a 1volt rotor voltage, a 60 to 1 reduction. This rotor shaft voltagelevel is at the upper end of the standards.

C. Evidence of Electric Discharge Machining (EDM)

Limiting the number of variables is essential in preventingunjustifiable conclusions from experimental results, especiallywhen investigating the effects of high frequency IGBTinverters. To accomplish this: The power cable was fixed to alength of ten feet with four conductors and the braided shieldgrounded at the drive end. A 4 KHz carrier frequency wasselected. Common mode chokes were not inserted in the inputor output of the drive.

Tests were performed on the drive system of Fig. 9. Thestator neutral to ground voltage, rotor shaft to ground voltage,and bearing strap current were monitored. Fig. 11 showsexperimental results when operating the AC drive at ratedvolts per hertz and 48 Hz. The stator neutral to groundvoltage displays the typical per carrier cycle waveformassociated with PWM voltage source inverters. The rotor

voltage, however, shows a quite different profile. For amajority of the time, the rotor is grounded, but occasionallythe rotor tracks the stator neutral to ground voltage. Thenquite suddenly, the rotor voltage collapses, producing a currentpulse. Fig. 6 is an expanded plot of an EDM discharge. Asthe stator to neutral voltage increases, the rotor voltageresponds with a capacitive charging characteristic. In fact, therotor voltage rises to a value fifteen times larger than themeasured value when operating on sine waves. At the instantof discharge, an impulse of current occurs with the rotorvoltage simultaneously collapsing.

A number of bearings were removed from motors operatingon AC drives and the AC mains. The bearings were examinedfor evidence of EDM fluting. Fig. 1 shows examples ofbearings from motors operated on AC drives after beingsectionalized. The fluting is quite pronounced. The outerbearing race on the left shows a random EDM discharge. Theouter race on the right shows a continuous etching of the racesurface.

The normal dv/dt switching current is in the hundreds ofmilli-amp range and occurs with the rise in rotor potential. Areview of the technical literature does not indicate a consensuson the effects of this relatively small current. However, thelarge current following the rapid collapse of the larger rotorvoltage is believed to cause EDM. The value of the EDMshown is limited by the inserted grounding strap and its surgeimpedance. A standard drive system's bearing current wouldbe limited by the bearing short circuit impedance. Thiscurrent, its cause, modeling, and control, are the focus of theremainder of this paper.

III. An Equivalent Circuit for BearingDisplacement and EDM Currents

A. The Model

Fig. 12 shows the physical construction of the test motor.Both the drive and non drive ends of the rotor were outfittedwith an insulated bearing support sleeve, which isolated therotor bearings from the motor frame. This provided ameasurement of the rotor open circuit voltage, and whenshorted by the grounding strap, simulates an actual bearingmounting. In addition, the grounding strap provides amechanism for measuring the bearing to ground current. Fig.12 shows a carbon brush for measuring the rotor voltage andinvestigating solutions to the EDM bearing current problem.

The motor had 48 stator slots and 64 rotor bars. Fig. 13depicts the capacitive coupling relevant to the development ofthe model. The stator to frame capacitance (Csf) is adistributed element representing the capacitive coupling toframe along the length of the stator conductors. For mostinvestigations, magnetic coupling of the stator and rotor issufficient. But with the high dv/dt present with modern power

IEEE APEC Conference Dallas. TX March, 1995

Fig. 10 AC Line Operation

Fig. 11 AC Drive Operation

devices, capacitive coupling considerations cannot be ignored.Therefore, the stator to rotor capacitance (Csr) and the rotor toframe capacitance (Crf) are included.

The bearings, lubricating film, and insulating sleeve presenta combination of capacitances, resistances, and a nonlinearimpedance, Fig. 14. First there exists an inner and outer raceresistance. Then, depending on the physical construction, thebearing consists of n balls in parallel; each ball having aneffective resistance (Rball,i). In addition, each ball is immersedin the lubricating film; thus, each ball develops twocapacitances (Cball,i) linking the ball to the inner and outer

races. The ball portion of the bearing model, therefore,consists of n parallel combinations of (Cball,i) and (Rball,i).Between balls, the inner and outer races are separated by thelubricant, which forms a dielectric barrier. Therefore, acapacitance (Cgap,i) is formed between each pair of balls,resulting in n parallel capacitors. The nonlinear impedance(Zl,i) accounts for the mechanical and electrical abnormalitiesand randomness of the bearing.

Combining the individual components results in a reducedorder bearing model, which is compatible with the motor drivemodels employed in simulations and analyses. The reducedorder model consists of a resistance (Rb) in series with theparallel combination of an effective capacitance (Cb) and anonlinear impedance (Zl). Finally, the insulating sleeve adds aseries capacitance (Csleeve) that is shorted when the groundingstrap is employed.

Combining the bearing model with a simple inverter/motormodel yields the model of Fig. 15. Here, the inverter ismodeled as three line to neutral voltages with a neutral toground zero sequence source. This model allows the inverter'svoltages to be examined as positive, negative, and zerosequence sets. The motor is represented as two sets of threephase windings; one each for the stator and rotor windings.The capacitive coupling from stator to frame is lumped at theneutral of the stator winding and the capacitive coupling

IEEE APEC Conference Dallas. TX March, 1995

Frame

Rotor

Csf

Csr

StatorWinding

Csr

CsrCsr

Csr

Csr

Csf

CsfStatorWinding

Stator Winding

Crf

Crf

Crf

Crf

Fig. 13 Motor Capacitive Coupling

n Ballsin Parallel

R outer race

C gap,i R ball,i

C ball,i

R inner race

C ball,i

,iZ

C sleeve

R b

C b Z

GroundStrap

Per Ball Model Reduced Model

Fig. 14 Motor Bearing Models

Drive Stator Rotor

Csr

Csf Crf

ZeroSequence

Source

Line to NeutralSources

Cb

Rb

Z

Fig. 15 Inverter / Motor Model

MOTOR FRAME

OuterRace

InnerRace

OuterRace

InnerRace

Insulating Sleeve

Insulating Sleeve

Stator Laminations

Stator Laminations

MOTOR FRAME

ROTORROTOR SHAFT

Carbon Brush

CurrentProbe

GroundingStrap

Fig. 12 Physical Construction of the Test Motor

between the stator and rotor connects the stator and rotor zerosequence networks. Finally, the rotor to frame capacitanceand bearing provide the paths to ground from the rotor shaft,here represented by the neutral of the rotor.

B. An Explanation of the Cause of Bearing Displacement andEDM Currents

Examining the bearing model in the context of theexperimental results shown in Fig. 11, the significance of thenonlinear impedance Zl is apparent. Because the bearingcapacitor normally exhibits a dv/dt or displacement currentwhen the stator voltage changes, the nominal dv/dt current islimited by the impedance given by the model of Fig. 15 with Zl

equal to a low non zero value. This corresponds to the bearingin a position of low impedance between outer and inner race.However, occasionally the bearing rides the lubricating film,which allows the rotor to track the source voltage with arandom duration. This condition corresponds to a substantialincrease in Zl . When Zl collapses, reflecting the preferredbearing position or the breakdown of the film, the capacitor Cb

is discharged and an EDM current occurs, with the currentthrough the bearing limited by the zero sequence or commonmode impedance. Thus, the bearing's impedance is statisticalin nature and depends on the position of the balls, thecondition of the bearing and its lubricant.

C. Model Parameter Values

Inputs to the model of Fig. 15 include relevant bearing andmotor parameters, and the zero sequence forcing function.Calculations and tests provided parameter values and thesource voltage. To calculate the stator to rotor capacitance,two parallel conductors were analyzed with a separation equalto the distance between the centers of the conductors. Thisvalue was modified to reflect the number of stator slots andslot opening area. To establish the rotor to frame capacitance,the rotor and stator were considered to be parallel cylinderswith an air gap. Fig. 16 shows the Crf as a function of

horsepower for 4 and 6 pole motors. The bearing filmcapacitance was calculated assuming a spherical constructionfor the ball with respect to the race surface. A typical valuefor the ball bearing capacitance is 190 pf [11]. The calculatedvalues for the test motor and bearing are contained in Table 1.

Tests were performed to establish the accuracy of the abovecalculations. With the stator unexcited and the rotor coupledto a drive motor, measurements of the effective capacitancefrom rotor to frame were made with a RLC meter at variousspeeds. The tests consistently produced a capacitance of 1400pf. This value represents the equivalent of Csr // ( Csf + (Crf //Cb)). Although the Cb depends on the speed of rotation, theinvariance of the measurement suggests Crf dominates. TheCsf is obtained by measuring the capacitance from the statorterminals to frame with the rotor removed. To establish theCsr, measurements were made of the effective capacitance fromstator terminals to frame with rotor shaft and frame connected.The Csr is obtained by subtracting Csf . Fig. 17 shows Csr forthe test motor as a function of frequency. Finally, the bearingimpedance Zl was measured as a function of rotational speed,the results of which are shown in Fig. 2. This in combinationwith the measured value of Cb allowed for the determination ofCrf . The measured values are included in Table 1.

Verification of the parameter values consisted of tests withthe insulating sleeve grounding strap open circuited and thedrive operating at various frequencies at no-load. The statorneutral to ground voltage and rotor voltage to ground weremeasured; the stator voltage from the neutral of the statorwindings and the rotor voltage from the rotor brush

IEEE APEC Conference Dallas. TX March, 1995

Fig. 16 Rotor - Frame Capacitance - Calculated

Fig. 17 Stator - Rotor Capacitance - Measured

Calculated Measured

Csr 100 pF 100 pF

Csf ----- 11 nF

Crf 1 nF 1.1 nF

Cb 200 pF 200 pF

Table 1 Motor Model Capacitances

attachment. Typical results of the tests are displayed in Fig.18. With the grounding strap open, the rotor voltage isstrikingly different from the rotor voltage of Fig. 11, where thegrounding strap was in place. The tracking of the stator toneutral voltage by the rotor voltage confirms the existence ofzero sequence paths as indicated by the model of Fig. 15. The stator to rotor voltage ratio confirmed the relativeweighting of the capacitors Csr and Crf in Table 1.

D. Simulation Results

For simulation and analysis purposes, the model of Fig. 15was reduced to a zero sequence approximation, which is theshaded portion of Fig. 15. A simulation was developed withthe parameters of Table 1 for the bearing model. Thesimulation provided an analytical tool for examining theeffects of PWM waveforms, verifying the system model andparameters by correlating simulation results with experimentaldata, and for evaluating various solutions to EDM. Fig. 19shows an expanded portion of Fig. 11 and a simulationemploying the zero sequence model. The forcing function forthe simulation was the stator neutral to ground voltage fromthe experimental results. The outputs include the rotor voltageand probe current as shown.

Comparing the simulation results to the experimentalresults shows good agreement. The dv/dt and EDM currentsare representative of experimental results. The rapid rise inrotor voltage at the point of EDM discharge is in very goodagreement with the data. To obtain this accuracy, an estimateof the nature of Zl is necessary. For the results presentedabove, Zl was modeled as a diac (Fig. 2); high impedance untilthe voltage threshold is met; thereafter it is voltage limited.The threshold voltage was experimentally determined. Thevalue of the impedance while voltage tracking, determinedfrom the rate at which the experimental rotor voltage of Fig.19 decayed, was found to be in good agreement with theresults of Fig. 2.

One area where the simulation fails to predict the observedresponse occurs in the transient response of the dv/dt andEDM currents. Close examination of the experimental resultsshows a 12.5 MHz oscillation in the measured current;however, the oscillation does not appear in the simulationresults. One explanation for this discrepancy is themeasurement technique. Inserting a grounding strap modifiesthe system impedance. The characteristic impedance of thegrounding strap alters the natural frequency and establishes anoscillation in the dv/dt and EDM currents.

IV. The Electrostatic Shielded Induction Motor: A Solution to EDM Bearing Currents

The previous section's experimental results suggestelectrical discharge as a principal contributor to bearingroughness. A bearing model was developed and interfacedwith the model for the electrical source and interconnectingnetwork. The model reflects the observed electrical behavior,which suggests the source of PWM induced bearing roughnessis the common mode or zero sequence voltage.

Using the model developed above, the task of proposingsolutions to EDM discharge becomes simply one of disruptingthe discharge either through the source voltage,interconnecting impedance, or the bearing design. Thus threedesign areas are available for investigation.

IEEE APEC Conference Dallas. TX March, 1995

Fig. 18 AC Drive Operation - Open Bearings

Experimental

Fig. 19 EDM Discharge Top) Experimental Bot) Simulation

Simulation

Because of the capacitive coupling from stator to rotor, themost likely candidate is the coupling mechanism from stator torotor - the Csr in Fig. 15. If an electrostatic shield is insertedbetween the stator and rotor, the coupling capacitance fromstator to rotor is defeated; thus reducing the dv/dt andpreventing voltage tracking by the rotor. Because theinduction machine generates torque through magneticinduction, the presence of the shield will not affect motoroutput ratings. A shield was constructed by inserting 1 inchadhesive backed copper foil tape strips to cover the stator slotarea. The shield was grounded to the motor frame.

Fig. 20 shows the stator neutral to ground and shaft voltagefor an identical operating condition as shown in Fig. 18. Withthe shield in place, a rotor voltage of 18 volts peak exists whenthe outer race grounding strap is open circuited - a 56%reduction when compared to the 40 volts peak of Fig. 18.With the strap grounded (Fig. 21), the dv/dt currents werereduced from 500 ma to 50 ma. No EDM currents weredetected.

Employing the copper foil strips as indicated above reducedthe rotor exposure to the stator windings in the preciseproportion by which the rotor voltage is reduced. By

extending the Faraday shield to enclose the stator endwindings and duplicating the tests above, a near completeshielding of the rotor voltage was observed. As results of Fig.22 show, the rotor voltage with grounding strap open isreduced 98% when compared to the unshielded case.Connecting the grounding strap (Fig. 23), virtually zero dv/dtcurrent was measured and no EDM current detected.

The experimental results presented above confirm bearingcurrents, both dv/dt and EDM, are induced primarily byelectrostatic coupling. The stator to rotor capacitance couplesthe zero sequence or common mode source from stator torotor. The bearing provides a return path for the commonmode source, thus allowing dv/dt and EDM dischargecurrents.

V. Conclusions

The paper presented a review of electrically induced bearingroughness for AC machines under low frequency sine waveoperation. A theory was proposed for lubricant dielectricbreakdown under PWM excitation. Electrostatic coupleddischarge or displacement (dv/dt) and electric discharge

IEEE APEC Conference Dallas. TX March, 1995

Fig. 20 Stator Shield - Open Bearing

Fig. 21 Stator Shield - Sleeve Shorted

Fig. 22 Full Shield - Open Bearings

Fig. 23 Full Shield - Sleeve Shorted

machining (EDM) currents were identified and experimentallymeasured. Electrical models were developed andexperimentally verified for the source voltage, couplingnetwork, and bearing. An electrostatic shielded inductionmotor was described and experimentally demonstrated as asolution to the bearing current problem.

The technical literature and experience show unloadedmotors at high speed provide the worst case scenario forbearing currents. In addition, applications with coupled loadstend not to exhibit the problem because of parallel paths forelectrostatic discharge.

ACKNOWLEDGMENT

The authors wish to thank Mr. Steve Stretz for his researchassistance in the bearing current phenomenon from a motordesign point of view.

REFERENCES

[1] Alger P., Samson H., "Shaft Currents in ElectricMachines" A.I.R.E. Conf. , Feb 1924[2] Costello, M., "Shaft Voltage and Rotating Machinery",IEEE Trans. IAS, March 1993

[3] Lawson, J. ,"Motor Bearing Fluting",CH3331-6/93/0000-0032 1993-IEEE[4] Prashad, H., "Theoretical Analysis of Capacitive Effect ofRoller Bearings on Repeated Starts and Stops of a MachineUnder the Influence of Shaft Voltages", Journal of Tribology,Jan. 1991[5] NEMA MG-1 Specification Part 31, Section IV , 1993[6] Ammann, C. , Reichert,K., Joho, R., Posedel, Z., "ShaftVoltages in Generators with Static Excitation Systems-Problems and Solutions", 1987 IEEE Power Eng. SocietySummer Mtg.[7] Andreason, S. "Passage of Electrical Current thru RollingBearings", SKF Gothenburg[8] Harris,T. Rolling Bearing Analysis, Wiley, 1984[9] Alston,L., High Voltage Technology, Oxford Press ,1968[10] Prashad, H., "Theoretical Evaluation of Capacitance,Resistanace and their Effects on Performance ofHydrodynamic Journal Bearings , Journal of Tribology, Oct.1990[11] Prashad, H. "Theoretical Analysis of the Effects ofInstantaneous Charge Leakage On Roller Bearings Lubricatedwith High Resistivity Lubricants under the Influence ofElectric Current", Journal of Tribology Jan.1990.

IEEE APEC Conference Dallas. TX March, 1995