Interpretable Fault Diagnosis of Shearer Power-Core Cables from Sensor-Accessible Terminal Electrical Responses

Shearer trailing cable faults are difficult to distinguish from terminal measurements because their signatures coexist with changes in load, source imbalance, cable temperature, and sensor error. This study presents a physics-guided categorical Mamdani framework supported by a distributed-parameter source, cable, and load model. A total of 180 simulations cover normal operation, phase-to-ground faults, A–B inter-phase short circuits, conductor-resistance degradation, and insulation-path deterioration over multiple severities, locations, motor loads, source imbalance levels, and temperatures. Paired pre-fault and post-fault changes in load-terminal voltage unbalance, source-terminal zero-sequence current ratio, and source-to-load voltage attenuation are mapped to fault path-based membership functions and diagnostic rules. Twenty repeated 70/30 holdouts grouped by base operating condition compare the proposed method with a deterministic threshold tree, radial-basis-function support vector machine, and k-nearest-neighbor classifier. Across these repeated grouped holdouts, the proposed method achieves 99.80% mean accuracy with a standard deviation of 0.63 percentage points on noise-free test cases. Accuracy remains 81.12% and 74.71% under 1% and 2% RMS waveform noise, respectively. These results demonstrate the effectiveness of the proposed framework for interpretable diagnosis of shearer power-core cable faults across the investigated simulated operating conditions and waveform-noise levels.