Linear Lamb Wave Based Technique for Damage Detection in stiffened composite plate using Machine learning

Abstract

For real time damage detection in composite structure, data driven machine learning (ML) algorithms are more preferred as it provides better decision making from the acquired sensor data. Implementing lamb wave propagation data obtained from complex structures with ML algorithms can be more effective to extract damage-related features. Convolutional Neural Network (CNN) has the ability to discover abstract features which can classify damage zones and damages between the plies of the composite. In this paper CNN algorithm is performed on response voltage signal data obtained from the sensor through Finite Element simulation in ABAQUS in order to classify delamination zones in complex composites structure having two stiffeners. The antisymmetric and symmetric components of the damage signal were pre-processed by subtracting from the undamaged signal, followed by the Hilbert transform, to feed the network as an input and to enhance the performance of the CNN model. For training, diverse database were created by varying delamination length and changing the delamination positions between the layers of the composite plies. Based on the above study, obtained results show high accuracy and can indeed detect delamination in composites structures with stiffeners using guided lamb wave technique in realistic situations.