A Machine Learning Model for Flaw Identification in Fibre-Reinforced Composites

Abstract

A Haar cascade classifier is a machine learning (ML) algorithm used for object detection. In this paper, the Haar algorithm is introduced in the context of a non-destructive evaluation of fibrereinforced composite (FRC) structures. The Haar learning model is used for flaw identification from thermal images. Thermal images are created from cross-ply (CP) carbon fibre-reinforced laminates with flat-bottomed holes (6–10 mm) of different depths from the surface (0.5–1.5 mm). After training is complete, the model successfully detects similar artificial flaws in previously unseen thermal images. In doing so, the feasibility of Haar classifiers for automatic evaluation of FRCs is established.