A Reduced‐Order Surrogate Model‐Based Spatiotemporal‐Dependent Reliability Analysis for Wind Turbines

The continuous enlargement of wind turbines requires them to withstand a wide range of complex and intense dynamic loads. However, traditional time‐dependent reliability analysis often focuses solely on fixed hotspots, neglecting the potential for failure regions to change under spatiotemporally varying external loads. To address this challenge, this paper proposes a novel spatiotemporal‐dependent reliability analysis (STDRA) framework explicitly designed to track moving failure hotspots over continuous operational periods. To realize this framework efficiently, several established techniques are integrated as supporting components: An Ornstein–Uhlenbeck (OU) stochastic process combined with the expansion optimal linear estimation (EOLE) is developed to characterize and discretize time‐varying wind loads, whereas a reduced‐order model augmented with a residual neural network (NN) is created for the rapid and precise prediction of global spatial structural responses. The proposed STDRA framework is applied to a practical wind turbine tower. Results demonstrate that the framework successfully identifies dynamically evolving critical failure regions over time. Furthermore, it yields failure probabilities consistent with Monte Carlo simulation (MCS) benchmarks but requires significantly less computational time than conventional TRA methods. By revealing these spatiotemporal weak spots, this study provides a robust foundation for artificial intelligence (AI)‐driven and reliability‐based design optimization of critical components, such as long flexible blades and towers, in next‐generation wind turbines.