Detection and localisation of barely visible impact damage in fibre-reinforced polymer composites using a supervised deep learning algorithm

Composites are prone to internal damage, either during manufacture or throughout their service life, requiring non-destructive testing for detection, monitoring, and repair. However, some methods, such as visual inspection, may pose health and safety risks. The present work explores the use of a supervised deep-learning algorithm to identify barely visible impact damage in composite panels. The algorithm is trained on a small labelled dataset and tested on an unlabelled dataset. Results show that the algorithm could present a promising tool for automating structural health monitoring of composites, offering accuracy of 96% and 86% on the non-impacted and impacted surfaces.