Research On The Application Method Of Granular Computing In Fault Knowledge Discovery Of Rotating Machinery

In chemical and electric power industries,rotating machinery is a kind of important key power equipment.The research on the method and theory of fault diagnosis for rotating machinery has been the focus of attention at home and abroad.Rolling-element bearing is a very important component in rotating machinery.It is widely used and easily damaged,which causes the mechanical equipment to be unable to operate normally,and even causes huge economic loss or casualties.Therefore,timely and effective fault diagnosis for rolling bearings is of great significance.In view of the problem that the recognition rate of weak faults in rolling-element bearings is low in the early stage,the research and discussion on the sensitive fault feature extraction and attribute reduction algorithm based on the neighborhood rough set(NRS)theory under the concept of particle computing is studied in this paper.The main research methods are as follows:First,11 Intrinsic Mode Function(IMF)and 1 remainder are obtained by a set of Ensemble Empirical Mode Decomposition(EEMD)for the original vibration signals produced by the rolling bearing.The first 3 components of the decomposed IMF component are selected as the IMF components that reflect the fault information,and then the 9 time domain features of the IMF components are extracted and the original fault feature data sets are constructed by using the mean variance and euclidean distance.Then,the attribute values of the original fault feature data set are normalized,and the neighborhood rough set(NRS)algorithm is used to reduce the attributes of the original fault feature set after processing,and eliminate the redundant attribute information.The NRS algorithm can directly deal with continuous data without discretization of data,which avoids the lack of fault information caused by data discretization and affects the accuracy of fault diagnosis.Finally,the simplest fault feature data set after the reduction is put into the classifier as the input data,and the final fault recognition process is completed by the classifier.The accuracy of the attribute reduction hybrid algorithm is analyzed by adding the contrast test to verify the effectiveness of the proposed attribute reduction algorithm.The innovation of this study is to select the NRS algorithm to reduce the data set of the original fault feature data,and to improve the current situation of the rare experimental research in the field of NRS research,and provide a new idea for improving the accuracy and accuracy of fault identification.EEMD decomposition can extract key attributes of different types of fault features.NRS can accurately select sensitive features containing abundant fault information from a large number of original features,eliminate redundant and useless information,reduce the complexity of the classifier algorithm,improve the recognition accuracy of the fault status of rolling bearings,and make intelligent fault decision-making technology have a better development.