The method of feature extraction on analog fault data by integrating principal component analysis(PCA) and linear discriminant analysis(LDA) is proposed. Firstly, PCA on analog fault data is carried out. Then, LDA is implemented in the transformed PCA space for obtaining the optimal discriminant features. Finally, the acquired patterns are applied to the pattern classifier for fault diagnosis. The simulation results show that the proposed method can take advantage of the simple computation of linear methods, enhance the performance of feature extraction of single PCA or LDA method, acquire the intrinsic features in fault data set, simplify the structure of the pattern classifier and reduce the computational cost of the diagnostic system.