Aviation DC Series Arc Fault Detection Method Based on Dual-Perspective Multidimensional Feature Fusion and Statistical Evidence

To address the variation of series arc fault characteristics under multiple load conditions, the limited adaptability of single-feature methods, and false alarms caused by load disturbances in aircraft DC power distribution systems, a fault detection method based on dual-perspective multidimensional feature fusion and statistical evidence is proposed. A dual-perspective representation is constructed using raw and enhanced current signals. Multidimensional features are extracted from the time domain, dynamic roughness, frequency domain, time-frequency domain, and dual-perspective coupling, followed by statistical-evidence-driven feature validation, selection, and weighted reconstruction to form a compact and discriminative feature space. Results show that all evaluated classifiers achieve detection accuracies above 95% under the unified feature input. The proposed method demonstrates strong cross-condition robustness, good disturbance immunity, and favorable adaptability across different classifiers.