In order to improve classification accuracy and shorten calculation time in fault diagnosis, the optimal feature set can be selected from the original feature set. Therefore, a feature selection method for the axial piston pump based on wavelet and GA-PLS Algorithm is proposed. Firstly, the vibration signals are decomposed by wavelet transformation, and the decomposition coefficients of wavelet are obtained. Then, the optimal feature set is selected from the original feature set collected from original signal and wavelet coefficients, using the genetic algorithm-partial least square (GA-PLS) algorithm. Finally, the optimal feature set is used as input into a support vector machine (SVM) to diagnose and identify different faults. This feature selection method is applied in the fault diagnosis of axial piston pump. The experimental results verify the effectiveness of this method in comparision with current methods.