Damage detection using the signal entropy of an ultrasonic sensor network

Abstract

Piezoelectric ultrasonic sensors used to propagate guided waves can potentially be implemented to inspect large areas in engineering structures. However, the inherent dispersion and noise of guided acoustic signals, multiple echoes in the structure, as well as a lack of an approximate or exact model, limit their use as a continuous structural health monitoring system. In this work, the implementation of a network of piezoelectric sensors randomly placed on a plate-like structure to detect and locate artificial damage is studied. A network of macro fiber composite (MFC) sensors working in a pitch–catch configuration was set on an aluminum thin plate 1.9 mm in thickness. Signals were analyzed in the time-scale domain using the discrete wavelet transform. The objectives of this work were threefold, namely to first develop a damage index based on the entropy distribution using short time wavelet entropy of the ultrasonic waves generated by a sensor network, second to determine the performance of an array of spare MFC sensors to detect artificial damage, and third to implement a time-of-arrival (TOA) algorithm on the gathered signals for damage location of an artificial circular discontinuity. Our preliminary test results show that the proposed methodology provides sufficient information for damage detection, which, once combined with the TOA algorithm, allows localization of the damage.