Probabilistic Design of Wind Turbines

During the last decades, wind turbines have been continuously developed with the aim of maximizing the life cycle benefits (production of electricity) minus the costs of planning, materials, installation, operation & maintenance as well as possible failure. In order to continue this development, new and more refined design methods must be developed. These methods can for instance be developed using probabilistic design where the uncertainties in all phases of the design life are taken into account. The main aim of the present thesis is to develop models for probabilistic design of wind turbines and the central topics considered are statistical load extrapolation of extreme loads during operation and reliability assessment of wind turbine blades.

Wind turbines differ from most civil engineering structures by having a control system which highly influences the loading. In the literature, methods for estimating the extreme load-effects on a wind turbine during operation, where the control system is active, have been proposed. But these methods and thereby the estimated loads are often subjected to a significant uncertainty which influences the reliability of the wind turbine. The uncertainty related to the existing methods for estimating the loads during operation is assessed by applying these methods to a case where the load response is assumed to be Gaussian. In this case an approximate analytical solution exists for a statistical description of the extreme load response. In general, the uncertainty is dependent on the method used for load extrapolation, the number of simulations and the distribution fitted to the extracted peaks. Another approach for estimating the uncertainty on the estimated load effects during operation is to use field measurements. A new method for load extrapolation, which is based on average conditional exceedence rates, is applied to wind turbine response. The advantage of this method is that it can handle dependence in the response and use exceedence rates instead of extracted peaks which normally are more stable. The results show that the method estimates the extreme load effects well and more consistent than the existing methods.

Blades for wind turbines are normally made of composite material which consists of fiber and matrix materials. The material properties of structures made by composite materials are often subjected to a significant uncertainty due to variations in the constituent materials and the manufacturing process. Additionally, methods for estimating failure of composites are subjected to significant uncertainties. The reliability of wind turbine blades are assessed in both ultimate and fatigue limit states. In the ultimate limit state, the dominating failure mode is often instability / buckling in the blade. In this thesis, the reliability is estimated using a response surface technique determined from nonlinear finite element analyses. In fatigue loading, the SN-approach in combination with Miners rule is normally used. The uncertainty on the SN-curves related to the material properties for the composite materials is determined from tests along with the uncertainty on Miners rule for linear damage accumulation. The reliability of the blade is estimated for exceedence of the fatigue strength in the composite material.