Statistical analysis of manufacturing defects on fatigue life of wind turbine casted Component

Wind turbine components experience heavily variable loads during its lifetime and fatigue failure is a main failure mode of casted components during their design working life. The fatigue life is highly dependent on the microstructure (grain size and graphite form and size), number, type, location and size of defects in the casted components and is therefore rather uncertain and needs to be described by stochastic models. Uncertainties related to such defects influence prediction of the fatigue strengths and are therefore important in modelling and assessment of the reliability of wind turbine components. The defect distribution is usually affected by the manufacturing process. In this paper, two methods of casting, sand casting and chill casting are considered. These are compared in statistical analyses of a large number of representative test samples using two basic stochastic models for the fatigue life, namely LogNormal and Weibull distributions. The statistical analyses are performed using the Maximum Likelihood Method and the statistical uncertainty is estimated. Further, stochastic models for the fatigue life obtained from the statistical analyses are used for illustration to assess the reliability of a representative component in an offshore wind turbine.