Acoustic Pattern Recognition for Non-Destructive Testing.

Acoustic Pattern Recognition for Non-Destructive Testing. In our recent research we have investigated the suitability of acoustic pattern recognition techniques for non-destructive testing. In this paper we focus on the important application of structural health monitoring of airplane parts. Our experiments aimed at automatically detecting fissures and impacts in aluminum and carbon fiber reinforced plastic and at assessing the severity of such damage. The pattern recognition is based on ultrasound signals, and involves a feature extraction and classification using Hidden Markov Models (HMM) or Support Vector Machines (SVM). Both, HMM and SVM based pattern recognizers, proved to be capable of securely detecting all tested structural damages. The HMM is especially well suitable for the degree of damage, whereas the SVM is able to detect very small fissures or impacts.