The design and optimization of offshore wind turbines (OWTs) require accurate predictions of their dynamic behavior, particularly the natural frequency, to ensure operational stability and structural integrity. Soil-structure interaction (SSI) plays a pivotal role in the dynamic response of OWTs, making it essential to incorporate its effects into predictive models. This research leverages Gene Expression Programming (GEP), an advanced computational technique combining genetic algorithms and symbolic regression, to develop an explicit, efficient, and practical model for predicting the natural frequency of OWTs while accounting for SSI. The monopile foundation is modeled as a beam on a nonlinear Winkler foundation, with soil behavior characterized by the spring model proposed by the American Petroleum Institute. The study constructs three distinct predictive models, rigorously evaluated using various statistical metrics such as MSE, RMSE, RAE, MAE, RSE, RRSE, and R2. Among these, Model-III demonstrates superior performance, with R2 values of 0.9817 for training and 0.9819 for validation data. Sensitivity analysis reveals that the length of the OWT exerts the most significant influence on its natural frequency, while the section thickness has a minimal impact. The results underscore the capability of GEP to model complex, nonlinear relationships, offering a powerful tool for the accurate prediction of OWT behavior and facilitating more reliable design and performance assessments of OWTs.
Abbasi et al. (Mon,) studied this question.