Objective With the increasing scale of offshore wind turbine installations and the continuous expansion into deeper waters, fault warning and maintenance of offshore wind turbines face significant challenges. The pitch system is one of the components with the highest failure rates in offshore wind turbines, and traditional fault prediction methods can no longer meet the operational requirements. Therefore, it is essential to implement health assessment and fault warning for offshore wind turbines.

Method This paper systematically reviewed potential failure points throughout the entire process of the pitch system, along with their primary manifestations and the evolution patterns of fault characteristics, providing a mechanistic foundation and interpretability for the development of a health assessment model for the pitch system. Based on mechanism analysis, data analysis, machine learning algorithms and Mahalanobis distance calculation were applied to present the health status of the pitch system in real time.

Result By leveraging health assessment and sliding window techniques, light faults and severe faults are effectively distinguished. Additionally, corresponding health warnings and maintenance recommendations are provided based on the probability of light faults occurring.

Conclusion Through early warning and proactive maintenance, the reliability of equipment is improved, major failures are prevented and downtime and operation and maintenance costs are saved.