A machine learning-driven surrogate modeling approach for the structural optimization of hybrid composite wind turbine blades

Abstract The global transition to renewable energy, driven by climate change, has catalyzed the widespread adoption of wind turbines as a key contributor to emission reduction. To meet the escalating demand for sustainable energy, enhancing the efficiency of wind turbine blade design is critical. A primary challenge in this design process is achieving an optimal balance between high structural strength and minimal weight. This study focuses on the development of hybrid laminate composites, comprising unidirectional (UD) glass, UD carbon, and twill carbon-reinforced polymers, for the blade structure. Using a 61.5-m, 5-MW wind turbine blade as a reference, a parametric finite element modeling (FEM) framework was developed in Abaqus™ to automate the simulation of various composite stacking sequences. The mechanical properties of each composite were determined through experimental testing to provide accurate input for the analysis. Surrogate-based ML models were trained to predict displacement, Tsai–Wu failure index, and weight, enabling rapid and cost-effective exploration of the design space. The ML-based optimization achieved a 15.6% weight reduction relative to the baseline factorial design approach while satisfying all structural constraints. To further enhance economic feasibility, a second-stage, section-based optimization was introduced, selectively applying hybrid laminates to high-load regions and glass fiber composites to less critical zones. This refinement reduced the total blade cost by 19% with minimal impact on stiffness and safety. The proposed two-stage framework demonstrates a robust, data-driven pathway for developing structurally efficient and economically viable hybrid wind turbine blades. Graphic Abstract