A Winkler Model for the Seismic Analysis of Monopile Foundations

Abstract

The growing interest in the development of offshore wind farms in seismic active areas demands a better insight into the requirements earthquakes impose on the design of wind turbine foundations. The complexity of the soil-structure interaction (SSI) during an earthquake results in large uncertainties in the design process of a monopile foundation. The design codes do not provide a structured framework on how to deal with these uncertainties. On top of this, the design codes are for (offshore) structures in general and do not specify for the large diameter tubulars which are characteristic for the offshore wind energy sector.

This thesis presents the investigation into a 1D seismic Winkler foundation that is able to represent the SSI during an earthquake. The model, a beam on nonlinear Winkler foundation (BNWF) coupled with a nonlinear ground response model, is a fast method to determine the structural response for the stochastic seismic and offshore wind load cases. Semi-empirical formulations are used to implement cyclic loading effects into the model. A 3D finite element model of an embedded monopile is developed in parallel with the seismic Winkler foundation for tuning and validation purposes.
The resistance of the soil to lateral pile deflections is further investigated to address the uncertainty in the SSI. A 2D horizontal cross-section of the embedded monopile is considered to determine the linear dynamic properties (impedance) of the soil, and a plane strain finite element model is developed to investigate the effect of nonlinear soil behaviour. A combined frequency-time domain extension for the Winkler foundation is developed to incorporate the obtained frequency-dependent foundation properties into a time-domain analysis.

It was found that none of the 2D analyses resulted in SSI representation that could be directly applied in a time-domain analysis. Moreover, the large D/L ratio of a monopile activates a global soil response, making an uncoupled 2D analysis inaccurate for the frequency range of interest.