Abstract:This paper is devoted to exploring intelligent fault diagnosis methods which is based on pattern recognition of vibration spectrogram. Firstly, taking rolling bearing as an example, the GLCM extracted from SPWVD spectrogram and its characteristic statistic are described. Moreover a modified AIN algorithm is introduced and used in bearing fault diagnosis. Through the optimization of fault antigen sample,the memory antibodies sets are formed and classification is processed by the k-nearest neighbor method. A mass of fault sample are analyzed in the algorithm proposed and the results are compared with those obtained by BPNN. The comparison result indicates that the modified AIN algorithm has better classification ability as well as high diagnosis accuracy. As the intelligent fault diagnosis methods develop,methods based on spectrogram identification should be popularized,its practicability is proved through recognition of bearing fault in this paper.