Assessing the quality of adhesive bonded joints using an innovative neural network approach

Abstract

Purpose – The purpose of this paper is to assess the quality of adhesively bonded joints using an alternative artificial neural networks (ANN) approach. Design/methodology/approach – Following the necessary surface pre-treatment and bonding process, the coupons were investigated for possible defects using C-scan ultrasonic inspection. Afterwards, the damage severity factor (DSF) theory was applied in order to quantify the existing damage state. A series of G IC mechanical tests was then conducted so as to assess the fracture toughness behavior of the bonded samples. Finally, the data derived both from the NDT tests (DSF) and the mechanical tests (fracture toughness energy) were combined and used to train the ANN which was developed within the present work. Findings – Using the developed neural network (NN) the bonding quality, in terms not only of defects but also of fracture toughness behavior, can be accessed through NDT testing, minimizing the need for mechanical tests only in the initial material characterization phase. Originality/value – The innovation of the paper stands on the feasibility of an alternative approach for assessing the quality of adhesively bonded joints using and ANNs, thus minimizing the necessary testing effort.