Bayesian‐Based Mixed‐Integer Optimization for Resonance Fatigue Testing of Large‐Scale Wind Turbine Blades

ABSTRACT Fatigue testing of large wind turbine blades is a critical step in ensuring their operational reliability, however, repeatedly developing loading fixtures for different test phases (such as static and fatigue testing) often entails significant time and financial costs. To achieve reusability of test loading fixtures and comply with the strict load conservative envelope principle specified in the standards, this paper proposes a method for optimizing full‐scale structural fatigue test loading schemes for blades based on Bayesian optimization and mixed‐integer nonlinear programming (MINLP). This method treats the mounting positions of the exciter and counterweight in the fatigue test loading scheme as discrete variables, and optimizes the excitation force amplitude, the mass of the loading saddle at the excitation point, and the mass of the counterweight as continuous variables. In the mixed optimization of discrete and continuous variables, the static test loading points are treated as discrete candidate positions for the fatigue test equipment. The base mass of each static test loading saddle is used as the lower bound for the corresponding mass variable. In the dynamic model optimized for flapwise direction, the stiffness softening effects caused by pre‐bending and geometric nonlinearity is considered. The rationality of the optimized loading scheme was verified through full‐scale structural fatigue testing of 100‐meter‐class wind turbine blades. The results indicate that the optimized loading scheme achieved a global safety envelope for the target design bending moment in both flapwise and edgewise fatigue tests. The maximum positive deviation of the measured bending moment in the flapwise and edgewise directions was controlled at approximately 8.6% and 16.4%, respectively, meeting the certification requirements for equivalent fatigue damage specified in the IEC 61400–23 standard. This optimization method provides a technical framework for designing loading schemes for full‐scale structural fatigue testing of large wind turbine blades that combines theoretical accuracy with cost‐effectiveness. It can effectively reduce the preparation time for large‐scale blade fatigue testing and ensure the rationality of certification.