Fault detection for wind turbine gearboxes is the pivot to ensure the sustainable and stable development of wind power. The planetary ring gear is the core component in this type of equipment. However, the challenge of ring gear fault detection is the coupling of pass-through effect and failure impulses with the same period in time and frequency domain. Traditional diagnostic methods lack the directional constraints, which leads to the incomplete decoupling. Therefore, a sparsity-assisted energy decoupling model is proposed to identify the different states of the planetary ring gear in wind turbines. Firstly, a sparse representation with a weight-modified generalized minimax concave penalty function is deduced to reduce the influence of background noise. Then, a hyperplane, defined by Gaussian distribution with an elastic parameter, is proposed to decouple the pass-through effect and failure impulses, which consists of the global-local fuzzy cost function and pseudo-gradient optimization. The proposed method is verified by simulation cases and on-site cases, which can be utilized to prove the effectiveness in the aspect of decoupling the states.

wind turbines; planetary ring gear; fault detection; energy decoupling