Comprehensive review on the use of machine learning techniques applied to the ultrasound data for the characterisation of porosity across carbon fibre reinforced polymer layers

Abstract

Carbon fibre reinforced polymers (CFRP) are increasingly being used in different industries, including the Automotive and aerospace sectors. One important reason for this is because they have interesting structural and mechanical properties compared to metallic materials. Their high strength-to-weight ratio makes them a preferred choice for high-stress applications. However, CFRPs are often subjected to various defects during their manufacturing that can significantly alter their structural integrity and durability. Amongst these defects, the occurrence of void formation (known as porosity) is the most common. Many methods have been developed for the characterisation of porosity including the ones based on the use of ultrasound data. The present work aims at providing a comprehensive review of the application of machine learning (ML) techniques to the mapping and characterisation of porosity across CFRP composites. The types of ML used, and their potentials for improving the accuracy of porosity detection are presented and discussed. It is particularly noted that ML techniques can extract unique features from CFRP complex ultrasound data with a relatively good level of accuracy. This result suggests that these techniques, particularly the convolutional neural network (CNN), would overcome the limitations of traditional signal processing techniques.