Assessing Feasibility and Performance of Ultrasonic Guided Wave–Based Numerical–Experimental Methodology for Debonding Monitoring of Adhesive Joints: Application to an Internal Beam of a Battery Box

Abstract

Multimaterial solutions that combine adhesively bonded composite and metallic parts are being widely proposed as lightweighting strategies to reduce environmental impact. However, the introduction of adhesive interphases in components subjected to fatigue loads is a major concern in terms of durability, reliability and maintainability. Structural health monitoring (SHM) techniques can play a key role in providing structures with self-sensing capabilities. Although the use of ultrasonic guided wave (UGW) monitoring for predicting the damage of in-service adhesive joints has been proved feasible, several challenges remain, including the generation of large and high-quality data sets and the scalability of damage detection algorithms for real-world use cases. After a wide literature review of available algorithms and simulation techniques, the simplest yet accurate methods have been selected to build a methodology that may eventually be fostered with more complex models. In this work, a numerical–experimental integrative methodology is proposed to train predictive algorithms minimizing the need for extensive experimental campaigns, by creating synthetic data sets through physics-based simulation models. Although several features have been detected as damage-sensitive, simple regression models using the root-mean-square density (RMSD) have been trained and validated as damage indicators. The feasibility of this approach has been proven in a real subcomponent with an error below 2% in the debonding length prediction, calculated as the ratio of the Euclidean distance between actual debonding and predicted debonding to the total inspection length.