Abstract: In order to improve the ability of feature extraction and fault identification of allocatable ball bearings in mechanical rotating system, A fault feature extraction method combined FCMMWPE and BSASVM was designed. and Isomap was used for fault identification. And the fault diagnosis case analysis was carried out. The results show that the FCMMWPE algorithm achieves the highest state entropy and forms a smoother entropy curve, and the generalized coarse-grained method has obvious advantages. When local faults occur, vibration signals with regular characteristics are formed, which indicates that FCMMWPE can meet the reliability conditions and has obvious advantages in extracting fault features of self-allot ball bearings. When FCMMWPE and Isomap feature sets constructed in this paper are used for operation fault identification, 99.9% accuracy is achieved, and self-aligning ball bearing fault identification is realized efficiently. BSASVM provides better fault identification performance, pattern recognition performance, and processing efficiency. The research can be extended to other mechanical transmission fields and has good application value.