Audio fault detection for high-voltage cable terminations

This paper addresses the challenge of weak discharge characteristics being masked by strong background noise in the acoustic detection of partial discharges in high-speed train high-voltage cable terminals. It innovatively introduces the Minimum Entropy Deconvolution (MED) method for fault diagnosis. Traditional detection methods are difficult to apply rapidly in vehicle maintenance due to equipment complexity and environmental constraints. The proposed method collects acoustic signals from cable terminals using portable devices and leverages the core advantage of the MED algorithm—designing an optimal inverse filter to maximize the enhancement of periodic impulse components within the signal. This effectively extracts the pulse sequences associated with partial discharges from heavy noise. Combined with a 2000 Hz high-pass filter to suppress low-frequency interference, the method clearly identifies discharge signals characterized by 50 Hz and its harmonics in both the time-domain waveform and frequency spectrum. Experimental and field verification demonstrate that this method can accurately distinguish faulty cables. Furthermore, an integrated graphical user interface analysis program was developed based on this approach, automating and streamlining the detection process. This provides a novel, low-cost, and efficient on-site solution for the operation and maintenance of high-voltage cables in rail transit. • Innovatively applying MED to extract weak PD pulses from train cable terminations via entropy-minimized filtering. • A technical scheme: 2000Hz high-pass filtering suppresses low-frequency noise, MED highlights weak pulses, envelope spectrum analysis identifies fundamentals and harmonics accurately. • To overcome invasive methods, we use non-contact audio recording for adaptable and efficient in-situ detection on high-speed trains. • This MATLAB GUI implements the audio algorithm for importing, analyzing, and diagnosing faults, enabling rapid on-site testing across vehicle lifecycles.