Insulation health monitoring in electric drives based on PWM common mode voltage and current Integrated Drive Theme Igor Tsyokhla University of Sheffield 1 Key Targets Integrated Drive Theme • Aim to achieve ambitious, aspirational targets of: A power density of 25kW per litre for a drive of around 100kW. Halving the commercial cost for an industrial drive. One quarter reduction of the system losses of current drives. Increase in reliability by an order of magnitude. Operation at an ambient of 300 degrees Fault classification Insulation degradation Partial discharge Neutral/ground/bearing current Inter-turn short circuit Open-circuit failure Partial demagnetisation Bearing failure/eccentric Drive failure Capacitor Sensor failure Devices/gate drive failures Control failure Capacitor failure All failure modes have links to temperature 3 Resistor Inductor Connections Devices Gate Drivers Other Insulation health degradation Capacitance and dissipation factor of motor ground-wall insulation 4 Insulation Test Methods DC Resistance C/DF Test Partial Discharge Pros: Pros: Pros: • Cheapest test Cons: • • Poor diagnostic tool Requires stoppage • • • Overall indicator Most problems detected Runs Online Cons: • • Expensive Requires stoppage Attempted Method • • Runs Online Most problems detected Cons: • • Expensive Subjective interpretation 5 C/DF Test Traditional Procedure: • • • • Dissipation Model High voltage injected between Winding and Ground Magnitude and phase measured C and DF calculated Dissipation can be represented as a resistor Health Indicators: Thermal Degradation: • • • Organic insulation is made out of polymer chains • • Vibration of chains dissipates energy Thermal degradation splits and multiplies chains More particles, more vibration 6 more dissipation • C increase = moisture ingress C and R decrease = thermal degradation 6 Other degradation modes possible • Common mode (CM) voltage and current in drives Typical Servo Drive Connection CM Current Path • • CM voltage measured close to the machine CM current goes through ground wall insulation 7 Insulation Health monitoring Common mode (CM) Leakage current measurement High sensitivity current transformer Common mode leakage current paths 8 New Way to Measure C/DF Sample C/DF Traditional Measurement: • • • Sine wave injected Measurement made at 50Hz @50Hz Ceq is 1.4nF @50Hz Req is 500 MOhm ! Novel Measurement: • CM voltage used (below) • Voltage is rich in switching harmonics • This range of harmonics is used, Req is lower than 1 MOhm 9 Main objectives • Continuously monitor winding ground-wall insulation and provide earlier warning of degradation • Establish degradation model for predicting remaining useful lifetime (RUL) through on-line monitoring First detectable Vibration sign (1-9)months Insulation (1-6)months Thermal degradation (1-12)months Good Preventative maintenance Audible noise Excessive heat Few weeks Predictive maintenance detects problem early Failed Few days Equipment fails Corrective maintenance 10 Experimental setup Machine: • 3kW 3000 RPM 480V SPM • Standard servo machine • Small, low voltage, random wound Drive • Standard commercial drive, switching frequency at 6kHz s 11 Measurement Box Sensors In the Test: Main Sensors: • Differential Probe 25 MHz Measures zero sequence voltage via Artificial Neutral R Network • Bergoz current sensor 1MHz Aux Optional Sensors: • External digitally isolated 10MS/s ADC Op amp Vabc summation Total cost = £20 • • LEM Fluxgate sensor (£10) LEM hall effect sensor (£10) 12 Current (A) Raw Data 13 Initial Test Results With Emulated Degradation Raw Impedance Data Emulated Degradation: • Resistance added in parallel to common mode path • Different values used to test sensitivity Simulated Health Progression 14 Accelerated Test Methodology Why Accelerated Aging ? • Ambiguity of realistic Req Ceq progression • Need to clarify required sensitivity • No end point criterion in for C/DF in current standards End Point Criterion What information is expected ? • • • Observation of parameter change in real time Is there a sharp change at end of life ? Correlation between temperature and end of life 15 Aging curve from IEEE std 98 Accelerated Test Schedule Sample in Oven: • Full Stator assembly tested Schedule (days) Sample 1 2 Oven 1 Sample 3 Sample 4 Oven 2 16 Cold Test Results Results Discussion: • • High consistency of data Good agreement with offline measurement for Ceq Conclusion: Ideal sensors show good measurement of C/DF parameters, online • • • Req trend is the same Absolute value disagreement Consistency remains, Indicates calibration issue 17 Degradation Progression Ceq Ceq Plot: • Progression of the 6kHz harmonic shown • Ceq decrease expected • Hot sample shows faster progression as expected @6kHz 18 Degradation Progression Req Req Plot • • • • Correlated noise in both ovens Increase in Sample 1 Decrease in Sample 2 Trend clearly visible @6kHz 19 Summary and Future work • Novel insulation health monitoring method outlined • Method verified with ideal sensors • Aging experiment in progress • Initial results shown • Data analysis and prognostic model to be established 20 Thank you for your attention Any question? 21 Cheap Sensor Results LEM Fluxgate and 10MS/s ADC Results: • • Capacitance shows good agreement Dissipation shape is wrong Usable Not good enough 22