Ultrasonic Signal Processing for Structural Health Monitoring

Abstract

Permanently mounted ultrasonic sensors are a key component of systems under development for structural health monitoring. Signal processing plays a critical role in the viability of such systems due to the difficulty in interpreting signals received from structures of complex geometry. This paper describes a differential feature‐based approach to classifying signal changes as either “environmental” or “structural”. Data are presented from piezoelectric discs bonded to an aluminum specimen subjected to both environmental changes and introduction of artificial defects. The classifier developed as part of this study was able to correctly identify artificial defects that were not part of the initial training and evaluation data sets. Central to the success of the classifier was the use of the Short Time Cross Correlation to measure coherency between the signal and reference as a function of time.