Offshore wind energy is expanding rapidly as governments aim for net-zero emissions, with monopiles and jackets being the primary foundation methods for offshore wind turbines (OWTs). Supply chains, heavier turbines and deeper waters influence the efficiency of jacket foundations. This research considers a jacket-founded OWT. The development of OWTs has sparked its expansion to building sites with a proclivity to earthquakes. Such loading involves soil-structure interaction of which more needs to be known to improve accurate computational modelling.

Current practice of Siemens Gamesa Renewable Energy (SGRE) is to perform aeroelastic simulations in their computational tool BHawC. The Foundation Designer delivers a foundation Superelement to maintain secrecy. The equation of motion for the jacket foundation can then be solved linearly with a reduced amount of unknowns. However, prevalence of nonlinearity in soil raises the question to what degree a linear model adequately captures the response.

This study investigates the impact of soil nonlinearity on OWT dynamic behavior under seismic loads by comparing linear and nonlinear soil models. The analysis involves performing seismic simulations using multiple earthquakes. Another distinction is made through soil models with different characteristics. Nonlinearity is introduced to the soil stiffness and energy dissipation mechanism under cyclic loading. The investigated variations are a linear (elastic), geometrically nonlinear (nonlinear elastic) or both geometrically and physically nonlinear soil model (nonlinear plastic).

The numerical model consists of a Rotor Nacelle Assembly (RNA), tower, transition piece, jacket, piles and soil springs. Beam elements are used for the tower, jacket and piles. The transition piece is simulated by stiffening the top jacket braces. The RNA is modelled using a lumped mass with rotational inertia. It is vertically eccentric to the tower top and connected with a rigid link. The earthquake is applied uniform over depth and only horizontal movements are considered.

The findings underscore a difference in results between the linear and nonlinear models. The evaluated results from simulations consist of forces, displacements and dynamic characteristics of the structure. Also noted should be that the computational time of the linear model is significantly lower. The results found in models can differ greatly due to the loading spectrum with highly varying frequency peaks. Another factor is softening of the stiffness. The frequency domain of the elastic model results consists of narrow peaks at the system's natural frequencies. The peaks for the nonlinear elastic model are wider due to softening of the stiffness. For the structure used in this research, softening introduces coupled modes with greater displacements along the height of the structure. This makes it possible for the evaluated results to have higher values, even with less energy put into the system. The plastic models' peaks are of a width in between the elastic and nonlinear elastic model due to the combined use of isotropic hardening and nonlinear stiffness. When the model falls back on its initial stiffness upon unloading, the eigenfrequencies related to that stiffness become more pronounced. To match the occupancy of wider frequency peaks, loading and unloading should both happen nonlinearly. This can be achieved by using kinematic hardening instead of isotropic hardening. Plasticity generally reduces peak displacement and sectional moment values and nonlinear stiffness broadens the response frequency spectrum. Careful consideration of cyclic material behaviour, eigenfrequencies and loading characteristics are essential for a realistic model.