A deep learning-based framework for two-step localization and characterization of acoustic emission sources in metallic panels using only one sensor

Abstract

This study focuses on localizing and characterizing acoustic emission (AE) sources in metallic panels with rivetconnected doublers. In particular, a deep learning-based framework is proposed that first performs zonal localization with only one sensor and then depending on the zone in which the source occurs, either finds the coordinates of the source or characterize it based on its source-to-rivet distance. The performance of the framework is assessed in typical scenarios in which the training and testing conditions of the deep networks are not identical, and Hsu-Nielsen sources were carried out for validation.