Abstract
Composite materials have garnered significant attention in industries such as automotive, aerospace, and defense due to their advantageous properties. However, concerns about the potential deterioration of mechanical performance over time have led to a growing interest in continuous structural health monitoring (SHM) during load-bearing applications. While there has been extensive research on damage sensing in composite materials under quasi-static conditions, limited attention has been given to dynamic impact loading conditions due to experimental challenges, particularly in measuring electrical resistance. In this study, the piezo-resistance response of hybrid composite materials is investigated under dynamic shear and mode-I fracture loading conditions, and machine learning models are applied to predict the piezo-resistance response. Two data training methods are implemented to train the models, and four widely used machine learning models are chosen for prediction. The results demonstrate that among the models, Gaussian Process Regression (GPR) and Neural Network (NN) models show the lowest Root Mean Square Error (RMSE) values and better prediction accuracy for all composite types under dynamic shear and mode-I fracture loading conditions. This research provides insights into the prediction of piezo-resistance response in multi-functional hybrid composites, offering guidelines for selecting appropriate machine learning models.