A method based on kernel fuzzy c-means clustering (KFCM) optimized by particle swarm optimization is proposed for fault diagnosis of wind turbine gearbox. Firstly, the clustering model is built based on wrong classification rate of training samples. The training samples are classified by kernel fuzzy c-means clustering. Then particle swarm optimization is introduced for solving the clustering model while the initial clustering center and parameter of kernel function are chosen as optimization variables. The class centers of optimal clustering result are acquired. Finally, the similarity parameters in kernel space between new data samples and the class centers are calculated for diagnosing whether the new data sample belongs to knows faults. The results show that the proposed method can diagnose both the known faults and unknown faults effectively compared to traditional neural network based on supervised learning.