In structural health monitoring based on the acousto-ultrasonic system with multiple transducers, a huge amount of data is generated due to a large number of possible actuator-sensor combinations. This leads to the difficulties in identifying a small set of useful features from these highly redundant signals with subtle variations due to slight structural changes and/or wave propagation path differences. Current approaches, which indicate the structural health by using a damage index computed based on changes in the acquired acousto-ultrasonic signals with respect to their baseline signals, often produce inconsistent results. In this paper, an alternative and new approach is proposed based on principal component analysis to establish the relationship between the observed data set and the underlying changes in structural health. Using the multi-path acousto-ultrasonic signals acquired at fixed intervals from a metallic aircraft panel under cyclic loading, the proposed methodology is presented in detail in the paper. It is shown that the first two principal components are sufficient for signal discrimination. With signals acquired from each path throughout the fatigue test projected to the feature space formed by the first two principal component axes, all signal trajectories are seen to progress in the same direction and most of the trajectories are seen to follow a similar arc-like curve. Furthermore, superposition of all projected signal points in the feature space shows a small number of clusters, which can be used to develop a traffic light based classification system for status indication of the current structural health.
Principal component analysis of acousto-ultrasonic signals for structural health monitoring
2008
8 Seiten, 3 Quellen
Aufsatz (Konferenz)
Englisch
Acousto-Ultrasonic Monitoring of Ceramic Composites
Online Contents | 1997
Acoustic Emission and Acousto-Ultrasonic Monitoring in High-Temperature Environments
Springer Verlag | 2023
|Ultrasonic structural health monitoring - An overview
Tema Archiv | 2014
|