Pointwise Damage Mapping using Guided Wave Signals for Structural Health Monitoring

Abstract

This paper presents a point-wise damage mapping algorithm for composite aerospace structures wherein we classify each location of interest on a structure as belonging to a damaged part of the structure or otherwise. The system employs an array of transducers attached to the structure for inducing guided waves in the structure and recording resultant waveforms arriving at the sensors. We present a baseline signal-based supervised learning algorithm in which baseline measurements on the structure are made prior to deploying it in the field and signals recorded during inspection are compared with their baseline counterparts. For each location of interest on the structure, difference metrics and other characteristics for M closest actuator-sensor paths to the location (M is typically much smaller than the total number of actuator-sensor paths on the structure.) are used as input features to a machine learning classifier. The classifier outputs for all locations of interest are further processed through a series of morphological filters to produce the damage map for the structure. The algorithm was experimentally validated on data recorded from a unidirectional composite panel impacted at five locations. Experimental results show that, for sufficiently high sensor densities, the algorithm presented in this paper is capable of detecting and characterizing multiple damages accurately and with high resolution and Sørensen-Dice index greater than 0.9 for all test cases, and is comparable to results obtained from traditional C-scan ultrasound non-destructive inspection systems. The point-wise damage mapping algorithm offers an attractive high-resolution alternative to traditional grid-based tomographic reconstruction approaches.