Multilabel Classification and Acoustic Signature Analysis of Defects in Outdoor Ceramic Insulators

Failures in ceramic insulators, primarily due to contamination flashovers and punctures, result in significant technical and economic losses. Early detection of these defects is crucial to prevent catastrophic outcomes. Insulator strings often exhibit multiple co-occurring defects, generating various discharge activities such as partial discharge, corona, and dry band arcing, each emitting ultrasonic signals. The overlap of these signals poses challenges for accurate diagnosis. This study develops a multi-label classification model to diagnose insulators with both single and multiple defects. Analysis of the acoustic signals for each defect identified unique temporal features, enabling precise characterization based on sound signatures.