Research On The Method Of Rolling Bearing Intelligent Fault Diagnosis

The rolling bearing is one of the important components of rotating machinery,its operating state directly affects on the working performance of the whole machine.Rolling bearing acts the role that in sustaining and transferring load in rotating machinery.It would cause stopping,cause heavy economic loss even disastrous accidents if failure happens.Therefore,the condition monitoring and fault diagnosis of rolling bearing is always the focus on home and abroad in the field of fault diagnosis.With the increment of complexity and automatic degree of rotating machinery,fault diagnosis data that is required to process has a huge amount.In order to reduce the subjective influence of manager of rotating machinery,the demand of intelligent diagnosis of rolling bearing is increasing.For the problem of rolling bearings intelligent fault diagnosis at steady conditions,the S transformation,a new tool of time-frequency,has been used to extract features from the original vibration signal of bearing.At the same time,for the problem of dimension of the transformed matrices is higher,the singular value decomposition has been utilized to extract feature vectors secondly from the S transform,and using the diagonal elements by singular value decomposition formed feature vectors,and then put into radical basis function(RBF)neural network.For the samples of unknown fault type,the vibration signals of rolling bearings were firstly conducted S transformation,the singular value decomposition has been utilized to extract feature vectors from transformed matrices,and the radical basis functionneural networkis used to simultaneously judge bearing fault type.The results verified the effectiveness of the proposed approach and this method can extractmore fault information than the results of using Wigner-Ville distribution and singular value decomposition.At the same time,the comparison results of RBF neural network and Elman neural network shows that RBF has higher stability and accuracy.Intelligent fault diagnosis for rolling bearing variable load conditions,a novel approach for load robust incipient fault diagnosis of rolling bearings is presented by using the multiple domains feature extraction and Fisher feature selection.Using the wavelet packet node energy feature combining frequency domain characteristic statistics to construct the multiple domains feature extraction.It can effectively reflect the characteristics of bearing fault types,and robust to the load information.Using Fisher feature selection to reduce the dimension of the multiple domains feature in order to improve the classification accuracy,eliminate irrelevant features,and then put the feature vectors that has been reduced dimension into support vector machine(SVM)to diagnosis the unknow samples.The results verified the effectiveness of the proposed approach for judging rolling bearing fault type,and this method is better than the results of using the multiple domains feature extraction and distance evaluation technique(DET).Developped of intelligent fault diagnosis system of rolling bearing based on LabVIEW.The system has advantages such as friendly interface,complete functions,simple operation,and easy maintenance.