For offshore wind farm projects, especially in the early design stages, information on the subsoil is very little compared to the size and complexity of the task ahead. This can have significant effects on the pile design and subsequently the physical and financial risk of the project. Reliability-Based Design (RBD) methods lend themselves to this design task, as other than in a deterministic approach, the uncertainties can be individually addressed and quantified rather than accounted for qualitatively by selecting "cautious estimates". This paper investigates the effect of the uncertainty in soil property values within seemingly homogeneous soil layers as well as in the model error in transforming measurement data to constitutive variables on the design of monopile offshore foundations. Using site investigation data from Cone Penetration Tests (CPT's), borings and geophysical data across an offshore wind farm, the point and spatial statistics describing the subsoil at the potential pile locations are inferred via stochastic interpolation using Krigging. A RBD-framework is developed, which then allows for the assessment of the monopile performance at each location, that is, with respect to a set of given limit states. Employing a sensitivity analysis in conjunction with the reliability-based performance assessment subsequently provides a quantitative and transparent basis for subsequent pile design optimizations and for the planning of further uncertainty reduction measures (e.g. additional site investigation campaigns, laboratory tests, etc.).

Keywords: Monopile, p-y-curves, Reliability-Based Design, Uncertainty.