Based On The Abnormal Sound Freight Train Rolling Bearing Fault Diagnosis Method Of Research

Rolling bearing is one of the mechanical components of most widely used,most fragile and vulnerable to damage in the rail freight train.A large percentage of accidents are caused by the fault of the rolling bearing in the all safety accidents. Therefore, wavelet packet decomposition and BP neural network are used to bearing fault diagnosis of the freight train of sound signals. Details are as follows:The main fault form, the causes,and structure and statistical properties of the sound signal of the rolling bearing are analyzed.The sound signals of seven states of bearing are the objects of study. The traditional frequency analysis can only identify the bearing is faulty or not,and can not effectively extract the bearing fault features and locate the fault,therefore, wavelet packet decomposition is proposed to feature extraction.Wavelet packet based on Daubechies is used to decomposition,reconstruction and envelope spectrum analysis of acoustic signal.The method extracted the K factor parameters which can be reflect the failure exactly classified by pattern recognition in order to identify and position the bearing failure.The BP neural network as well as improved neural network algorithm are studyed to identificate and classificate the bearing failure.The sample data of K factor parameters which are processed by wavelet packet analysis are conducted as the input of the neural network,and then fulfill state recognition.Simulation results show that the wavelet packet decomposition of the acoustic signal is decomposed into different frequency bands for fault feature extraction which can effectively highlight fault characteristics; rolling bearing fault diagnosis based on the K factor and the BP neural network can accurately identify the fault, improving the validity and accuracy of the diagnosis.