Classification of adhesive bonding between thermoplastic composites using ultrasonic testing aided by machine learning

Adhesive bonding is widely used for joining light weight structures and surface preparation methods play a significant role in improving the bonding quality. Bond quality assessment of thermoplastic composites can be challenging due to the highly inhomogeneous structure. This article explores the use of ultrasonic testing to nondestructively characterize the bond state between thermoplastic composites with different surface preparation methods. Classification of bond states was carried out using a physics-based statistical method and machine learning based method. The results suggest that machine learning methods show a higher classification accuracy. The ultrasonics results were validated using destructive lap-shear testing.