Advanced Reliability Analysis of System’s Components through Parametric Modeling and Optimal Distribution Selection

Abstract

This study presents a comprehensive reliability analysis of a wind turbine system based on three critical components, the shaft, gearbox and generator. The primary objective is to evaluate the Time Between Failures for each component and to identify the statistical models that most accurately represent their failure behaviours. Four widely recognized statistical distributions (3-parameter Weibull, 2-parameter exponential, normal and 3-parameter lognormal) are assessed. Model selection is based on goodness of fit measures using the Anderson Darling coefficient, with parameters estimated through both Maximum Likelihood and Least Squares Estimations. A notable similarity between the best fit distributions for the shaft prompted the employment of Akaike and Bayesian Information Criterions to resolve this inconsistency and ensure optimal model selection. The analysis reveals significant variability in component reliability. The gearbox, characterized by early life failures and highest failure rates, emerges as the most critical component impacting the wind turbine's overall reliability. This is especially important given the turbine's series configuration, where the failure of any single component results in total system failure. However, the shaft exhibits a wear out failure pattern, with reliability decreasing steadily over time, while the generator demonstrates the highest and most stable reliability, associated with random failure behaviour. The results highlight the importance of component specific reliability assessment and statistical modelling in optimizing wind turbine performance. By integrating probability density, survival and hazard analyses with robust statistical fitting, it offers valuable insights into individual component behaviour and system level reliability, thereby supporting the development of targeted maintenance strategies and improved operational planning.