A Radial Basis Function Neural Network Approach for Detecting Wind Turbine Blade Damage via Embedded Accelerometer Data

A significant number of wind turbines deployed across the Penghu Islands are exposed to challenging coastal and marine conditions.To address the operational risks associated with such environments, in this study, we introduce a real-time monitoring and fault detection framework aimed at identifying abnormal turbine behavior, particularly those linked to blade damage.The proposed system enhances turbine efficiency and reliability by continuously assessing blade integrity, swiftly detecting irregularities, and supporting near-instantaneous maintenance actions.This capability is especially vital in disaster-prone regions, where turbines may encounter conditions exceeding standard design specifications.The framework integrates sensor-based data acquisition-capturing input from accelerometers, anemometers, hygrometers, thermometers, and barometers-with signal processing and neural network techniques to analyze three-axis vibration data.By identifying distinct patterns associated with blade faults, the system enables timely detection and reporting of malfunctioning units, thereby facilitating effective repair and operational continuity. Research BackgroundEnhancing the operational efficiency of wind turbines and minimizing maintenance expenses are the shared objectives among wind farm operators.However, much like the corrosive impact of seawater and harsh weather on offshore and coastal turbine systems, mechanical components remain prone to degradation, leading to reduced energy output.Despite the significance of these challenges, they have not been thoroughly addressed in a comprehensive manner. (1)(2)(3)(4)(5) Reducing operating costs by avoiding the unnecessary replacement of core components remains a top priority.However, to achieve this, early and accurate fault detection is essential alongside routine