Abstract
Stator turn faults represent one of the most critical failure modes in electrical machine drives, necessitating rapid and reliable detection. This paper introduces a novel detection technique based on equivalent high-frequency (HF) impedance. By leveraging the inherent harmonics from pulse-width modulation (PWM) and the flux nullifying effect of shorted turns, the proposed method identifies early-stage insulation failures with greater sensitivity than traditional current-only monitoring.
The Challenge of Early Detection
Stator turn faults begin as minor insulation degradations but quickly escalate into catastrophic failures due to massive local currents and heat. Traditional Motor Current Signal Analysis (MCSA) often fails at early stages because low-frequency fault signatures are masked by load fluctuations and noise. While high-frequency injection methods offer better sensitivity, they typically require additional hardware and can introduce unwanted noise. Furthermore, most existing HF methods only monitor current ripple, ignoring the fact that a fault also distorts the phase voltage.
Proposed Methodology
The core of the proposed technique is the calculation of equivalent HF impedance ($Z_e$). Unlike conventional methods that monitor only current ripple, this approach accounts for the fact that a turn fault also distorts the phase voltage.
- PWM as a Signal Source: The method utilizes the high-frequency harmonics naturally generated by the inverter's PWM switching, eliminating the need for external signal injection.
- Signal Processing: Phase voltages and currents are processed through a second-order bandpass filter centered at the PWM sideband (e.g., 20 kHz).
- Impedance Calculation: The filtered signals are converted to RMS values ($U_{HF}$ and $I_{HF}$), and the equivalent impedance is derived as $Z_e = U_{HF} / I_{HF}$.
The Flux Nullifying Effect
The physical basis for this method is the "flux nullifying effect." When a turn fault occurs, the shorted loop behaves like a secondary winding in a transformer. The induced fault current generates an opposing magnetic field that nullifies the flux in the affected area, leading to a significant reduction in the high-frequency inductance and, consequently, the total HF impedance of that phase.
Experimental Validation and Results
The technique was validated using MATLAB/Simulink models and a 35 kW triple-redundant Interior Permanent Magnet (IPM) drive. This specialized motor allowed researchers to physically "tap" into the windings to simulate real turn-to-turn shorts in a controlled environment.
- Superior Sensitivity: The "fault signal ratio" using equivalent impedance was 30% to 50% higher than the conventional current-ripple method.
- Frequency Response: As shown in the magnitude plots, the impedance drop is clearly distinguishable across a wide frequency range, particularly at the high frequencies targeted by this method.
- Speed and Robustness: Detection was achieved in less than 25ms. The method proved robust against variations in motor speed and load, as the HF signals are decoupled from the fundamental power frequency.
Conclusion
By integrating both voltage and current signals into a single impedance indicator, this research provides a more robust and sensitive tool for early-stage stator turn fault detection. Future work will focus on adaptive filtering for variable switching frequency drives and more precise fault localization.
Summary updated