Anti-interference damage localization in composite overwrapped pressure vessels using machine learning and ultrasonic guided waves

Abstract

This paper describes a structural health monitoring (SHM) technique for composite overwrapped pressure vessels (COPVs), which utilizes machine learning (ML) algorithms to achieve anti-interference damage localization based on ultrasonic guided waves. Firstly, this study developed a multi-channel guided wave signal automatic acquisition system and established a comprehensive dataset comprising multi-region damage/health signals from composite pressure vessels. Subsequently, based on the interaction between guided wave features and damage, a feature extraction-ML based anti-interference damage localization scheme was proposed. This scheme achieved robust damage localization through steps such as data cleaning, feature extraction, feature selection, and ML-based damage localization. Furthermore, this study introduced the guided wave-temporal convolutional network (GW-TCN) model based on deep learning method. The GW-TCN model integrates the temporal correlation and the multi-channel physical properties of guided waves, while leveraging the advantages of deep learning's adaptive feature extraction. It effectively mitigates the vibration and sensor failure interferences that mimic real operational conditions, enabling precise damage localization. Finally, through validation on multiple data samples, it was demonstrated that the model exhibits good generalization ability. The research findings will contribute to the advancement of composite pressure vessel SHM technology, paving the way for improved safety and reliability in industrial applications.