A Bayesian network framework for evidence-based risk analysis of offshore wind turbines

This study presents a comprehensive methodology for assessing the risk of offshore wind turbine failures using a probabilistic framework based on Bayesian Networks. Drawing on a dataset of 1753 failure events from offshore wind farms, this study provides failure patterns through systematic statistical analysis, comprising failure frequencies, criticalities, and shutdown distributions across offshore wind turbines’ sub-components. Chi-square tests are conducted to identify statistical relationships among fault indicators, maintenance actions and subcomponents. A Bayesian Network model is created, integrating expert knowledge and evidence of failure data to develop relationships among components, sub-components, and failure causes. Sensitivity analyses identify key drivers of system failure, highlighting the influence of critical units such as generators, converters, and cooling systems. This study has two facets, statistical characterisation and probabilistic modelling, which allow accurate diagnosis of failure scenarios and support informed, evidence-based decision-making. The findings are directly applicable to improving operational reliability and guiding maintenance planning in offshore wind farms.