Overloaded Train Bearing Failure Of The Audio Signal Integrated Diagnostic Programs

Heavy train operation of the various security issues and directly affects the rail transport and the national economy, the main cause of accidents occurred lead to bearing failure, it has long been widespread concern at home and abroad. Take non-contact methods for diagnosis that the sound signals gradually become the focus of the study.In this paper, wavelet packet analysis of the bearing fault signals, and then extracted cepstral parameters, and finally to genetic algorithm based on BP neural network as a tool for the integration of its bearing fault diagnosis of a theoretical and experimental research, the main research work the following aspects:1,the analysis of heavy haul train causes of bearing failure, and failure of various types and characteristics of the difference between the frequency of the sound signal generated by the mechanism analysis, the corresponding sound signal collection and pretreatment.2,through the merits of the base wavelet analysis, this paper db5 selected as the base wavelet, for a more complete analysis of the whole fault signal characteristics of this wavelet packet analysis to the analysis of fault signals, strengthening the reliability of diagnosis, in the voice of the fault signal using wavelet packet processing, LPCC and MFCC were used in two ways on the fault signals collected cepstrum feature extraction.3,for the traditional BP neural network learning is slow and can not guarantee convergence to global minimum as well as learning and memory defects such as instability, this paper presents a genetic algorithm using the initial weights and threshold optimization and its applications Bearing Fault Diagnosis in the process of integration, through the fault signal Cepstrum analysis and to identify specific fault type, then MATLAB simulation results show its better than the traditional BP neural network the rapid and effective.