A Comparative Study of Impact Localisation in Composite Structures Using Neural Networks under Environmental and Operational Variations

Abstract

In this study we compare the effectiveness of the Normalised Smoothed Envelope (NSET) method for Artificial Neural Network (ANN) based impact localisation under simulated environmental and operational conditions with respect to other ANN based localisation methods developed by other studies. It is shown that when the testing and training impact case is the same, most studies give comparably good accuracy of localisation irrespective of feature extraction method or structure geometry. However, when the testing and training impact cases are not the same, only the NSET method is able to negate the variations caused by various impact cases and provide good localisation accuracy for an ANN trained using only a single impact case thus allowing for smaller training data set size requirements and increasing feasibility for real life application.