Leveraging machine learning for analyzing structural behavior of a composite wind turbine blade

The main objective of this paper is to develop models to quickly investigate the mechanical behaviors of an industrial case of a composite wind turbine blade based on probable variations in the mechanical properties of the constituent materials purchased from different suppliers. The wind turbine blade behaviors, including flexural natural frequencies and stiffnesses in both longitudinal and transverse directions, are predicted using four machine learning models. To construct each model, the blade's corresponding behavior is initially characterized using an appropriate low-fidelity model based on possible variations in mechanical properties. Subsequently, the required databases for training and evaluating each improved high-order machine learning model are created. After well-training the models and evaluating their performances through various error criteria, the sensitivity improved model is examined. The information presented in this paper substantially resolves one of the significant challenges in large-scale wind turbine industries.