Quantification of the out-of-plane loading fatigue response of bistable CFRP laminates using a machine learning approach

Abstract

This study proposes data-driven machine learning models to predict the nonlinear load-displacement response in constant amplitude high-cycle fatigue loading of unsymmetric, cross-ply bistable carbon fiber reinforced polymer composites. Four selected ML models are trained on experimental fatigue data with eleven unique frequency, temperature, and boundary conditions combinations. Stiffness and damage index values, which serve as additional evaluation metrics, are calculated using the predicted load data. The models capture the nonlinear load response with acceptable error for in-domain experimental conditions. Model expandability demonstrates the sensitivity of machine learning models to training features but suggests an economical alternative to extensive fatigue experiments.