A Bayesian Optimization Framework for High-Fidelity Fatigue Testing of Wind Turbine Blades via Single-Point Multi-Frequency Excitation

Full-scale fatigue testing is a mandatory step to verify the reliability of large flexible composite structures. As wind turbine blades approach or exceed 100 meters in length, the coupling effects of lower-order modes become increasingly significant. However, traditional single-frequency excitation is limited. Constrained by a single mode shape, this method often fails to accurately replicate the damage distribution caused by real wind loads, leading to local over-testing or under-testing. To solve this inverse problem of structural dynamics, an automated loading framework is established in this paper. This approach integrates finite element simulation with Bayesian optimization. By coupling a parametric dynamic model with a Gaussian process surrogate, the framework automatically optimizes multi-frequency excitation amplitudes to approximate the target damage. A case study on a 91-meter commercial blade compares three strategies: traditional single-frequency, single-point multi-frequency, and multi-point multi-frequency. Results indicate that the proposed single-point strategy effectively circumvents physical limitations without requiring extra equipment. Specifically, the effective verification coverage (within ±10% tolerance) is expanded from 61% to 81%, and the global damage matching root-mean-square error is reduced by over 50%. This study confirms that algorithmically unlocking the potential of single-point excitation offers a cost-effective and rigorous solution for high-fidelity fatigue testing.