Reliability Analysis of Monopile Foundations for Offshore Wind Turbines

Abstract

Initiated by the increasing energy demands and instabilities in oil and gas prices there is an ongoing transition from the conventional to the renewable energy sources across Europe. The goal is to satisfy 20% of the energy demands with the renewable energy sources by 2020. In order to achieve this goal there is a continuous development of reliable and cost-efficient renewable energy sources. One of the renewable energy sources with a considerable potential is the offshore wind energy. The offshore environment introduces several technical and economic challenges which need to be resolved in order for the offshore wind industry to contribute significantly to the transition towards the renewable energy sources. This thesis examines some of the technological challenges related to the monopile foundations for offshore wind turbines.

A monopile foundation design is subjected to uncertainties due to limited soil investigations, inherent randomness of wind and wave loads, and the lack of knowledge about the processes driving the behavior of monopile foundations. This thesis examines the effects of uncertainties on the design of a monopile foundation. The uncertainties in the derivation of soil properties and classification of soil formations are extensively examined due to the reliance of monopile foundations on the surrounding soil to provide support to a wind turbine. A probabilistic transformation is developed to interpret soil design parameters from cone penetration test measurements (CPT) to account for the reliance of soil investigation procedures on a wind farm on CPT measurements. The probabilistic transformation model integrates the inherent variability of CPT measurements, measurement error, and the transformation uncertainty. The uncertainties in the classification of soil formations based on CPT measurements are evaluated by conducting a probabilistic classification with the Bayesian Mixture Analysis. The effects of uncertainties on the safe performance of a monopile foundation are investigated by conducting a reliability analysis. The reliability analysis if often a computationally demanding task due to computationally intensive reliability methods and simulation models of monopile foundations. One of the main task in this thesis is the implementation and development of efficient methods for reliability analysis of monopile foundations. In addition to the implementation of several state-of-the-art reliability methods, the Metamodel Line Sampling reliability method is proposed. The Metamodel Line Sampling reliability method attempts to reduce the computational expenses commonly associated with the reliability analysis of engineering structures by approximating the response of a structure or an engineering system with a metamodel. A correction coefficient is introduced in the Metamodel Line Sampling method to account for the error resulting from the approximation of the structural response with a metamodel. The Metamodel Line Sampling method demonstrated efficient performance on a reliability analysis of a monopile foundation.

The effects of uncertainties are integrated in the design of monopile foundations by implementing the reliability-based design optimization (RBDO) framework. In the RBDO framework, the performance criteria of a monopile foundation are optimized while explicitly accounting for the effects of uncertainties. The uncertainties are integrated in a design by means of reliability constraints, which specify minimum safety criteria of a monopile with respect to the uncertainties. The implementation and development of computationally efficient RBDO methods was one of the primary tasks in the thesis since the solution to an RBDO problem depends on the evaluation of computationally demanding optimization and reliability sub-problems. In addition to the implementation of several state-of-the-art optimization and reliability algorithms, the CE-LS RBDO algorithm was developed as a coupling between the Cross-entropy optimization and the Line Sampling reliability method. The CE-LS method implements a local weighted approximation of the reliability estimates to avoid the computational expenses related to the repeated evaluations of the reliability problem during the optimization process. The CE-LS algorithm demonstrated efficient performance on several classic RBDO problems and on an RBDO of a monopile foundation.