Proceedings of the
The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK
Statistical Analysis of Offshore Wind Turbine Failures
1Department of Mechanical and Industrial Engineering, NTNU, Norway.
2Department of Engineering Cybernetics, NTNU, Norway.
ABSTRACT
This paper aims to investigate the characteristics of offshore wind turbine failures. Four hypotheses on failure features are proposed and strictly examined by statistical tests. Cox model is chosen to model failure process. Three forms of co-variates are designed to research their influence on failures. Their coefficients are obtained by maximum likelihood estimation and Breslow estimator is calculated. Finally, goodness-of-fit tests verify the assumptions of Cox model. Results of long-term models show that wind significantly favors the growth of baseline hazard. However, temperature and production condition mildly reduce it. The effects will gradually become stable if accumulation time increases. Similar results are observed in models with principal components of co-variates. Comparison of models suggest the highest likelihood belongs to models with three accumulated co-variates.
Keywords: Hypotheses testing, Cox model, Offshore wind turbine, Failure process, Accumulated co-variates.
