Correlation Between Electrical Signal Features and Physical Mechanisms of Early-Stage Foreign Object Contact Faults for Computational Pattern Discrimination

This study proposes a pattern discrimination method for early-stage foreign object faults that uses a “Pulse-Phase Distribution Map” and a set of five quantitative features to achieve identification independent of environmental parameters. The analysis reveals a fundamental difference between the dynamic evolution of tree branch faults and the static modes of iron wire faults. Quantitative results show that different fault modes possess unique feature vectors; for instance, an along-surface iron wire arc is uniquely identified by a combination of a low Pulse Count Rate (PCR), an extremely high Zero-Crossing Energy Ratio (ZCER), and a high Mean Pulse Amplitude (MPA). This demonstrates that the proposed multi-dimensional feature set can effectively distinguish various early fault modes, providing a robust computational basis for intelligent diagnostic algorithms.