Abstract:

Since crack damage of wind turbine (WT) blades is easy to occur and difficult to find, online monitoring of blade crack damage is carried out by collecting and analyzing blade vibration signals. Firstly, based on the theory of working modal analysis, an online identification method of blade modal parameters based on transmissibility is constructed, and a blade vibration physical experiment platform is built for the experimental verification of the method. By comparing the experimental results with the traditional hammer excitation method, the accuracy of the method is verified. Then, with a 5 MW WT as an example, the blade crack damage fault is simulated, and the damage fault characteristics are obtained through working modal analysis. Finally, blade vibration signals, modal parameters, and WT operation data are fused into multi-source data sets, and blade crack damage fault diagnosis is performed based on the LightGBM algorithm. The diagnosis results show that the LightGBM algorithm can achieve a better diagnosis effect than the conventional machine learning algorithm, and the accuracy of the diagnosis algorithm can be significantly increased by integrating blade modal parameters into the data set, so as to improve the accuracy of online monitoring of blade crack damage.

Key words: wind turbine, blade crack damage, working modal analysis, transmissibility, machine learning