Analysing Machine Learning Approaches for Lamb Wave-Based Damage Detection

One of the most efficient strategies to automate the structural health monitoring (SHM) process is the data-driven approach. It requires a data acquisition system, preprocessing techniques for turning the data into useful features, and a machine learning (ML) algorithm for creating a correlation between the structure’s final condition and the extracted features. In this study, a variety of ML classification algorithms such as, ‘Support Vector Classifier (SVC)’, ‘Random Forest Classifier’, ‘Extra Tree Classifier’, ‘Extreme Gradient Boosting Classifier (XGB Classifier)’ and ‘Artificial Neural Network (ANN)’ have been analysed to determine the best algorithm for identification of damage in a thin aluminium plate. SVC, Random Forest Classifier, and Extra Tree Classifiers have achieved an accuracy of 50%, whereas XGBoost classifier has achieved an accuracy of 58%. It is observed that ANN is able to successfully differentiate between the lamb wave responses of damaged and undamaged samples of thin aluminium plates.