Abstract: Aiming at the feature extraction process of wind turbine’s gear box fault diagnosis, a SVM fault diagnosis method based on the optimal feature extraction algorithm of vibration signal is proposed. Firstly, the signal types with high adaptability of the three main feature extraction algorithms are analyzed. Then, according to the signal characteristics of different types of signals, the characteristics of incoming vibration signals are extracted and classified by signal analysis, and the different types of signals are matched with the feature extraction algorithm with high adaptability to achieve the optimal feature extraction of vibration signals. Finally, the matching algorithm and support vector machines model are combined to realize fault diagnosis. Three kinds of gear fault signals collected in practice are tested and verified. The results show that this method can effectively extract the optimal features and match the algorithm, and has a higher fault diagnosis accuracy than the unmatched algorithm.