Research On Fault Diagnosis Method Under Complex Working Conditions

Mechanical and electrical equipment plays a very important role in the development of modernization.As an important component of mechanical and electrical equipment,rolling bearing usually works in a relatively bad environment.Rolling bearing failure and performance degradation often account for as much as 30% of the total fault diagnosis.In order to ensure the stability and safety of mechanical and electrical equipment operation,prevent the occurrence of major accidents and reduce maintenance costs,it is of great significance to carry out real-time monitoring and fault diagnosis of the equipment in operation.Due to the high failure rate of rolling bearing,this paper takes the bearing as the main research object to carry out the research of fault feature extraction and diagnosis based on vibration analysis.The main contents are as follows:(1)Aiming at the problems of data distribution difference,low classification accuracy and high manual labeling cost in mechanical equipment under multiple working conditions.This paper proposes an intelligent fault diagnosis method(CWT-MADA)combining wavelet time-frequency graph and transfer learning multi-adversarial domain adaptation(MADA),which is used for fault diagnosis across working conditions.This method first builds a dual-stream deep convolutional neural network to learn the original information and time-frequency map features of the source and target domains.This combined feature helps to solve the problem of insufficient utilization of fault features;Secondly,the source domain sample clustering is used to mark the target domain samples with false marks,and the feature extractor is constrained during the multi-domain adaptation process to continuously narrow the distance between the same kind in different domains and reduce the difference caused by the change of working conditions;(2)Aiming at the problem of coupling between bearing compound faults and the difficulty of extracting the impact characteristics of compound faults by traditional fault diagnosis methods.In this paper,a fault diagnosis method based on empirical mode decomposition(EMD)and convolutional neural network with decoupler is proposed.Firstly,the fault vibration signal of a single fault category is decomposed into multiple IMF components by empirical mode decomposition(EMD);Secondly,the correlation between the IMF and the original signal is calculated,and the most important IMF components are selected to form a matrix;then,the IMF components are input into the convolutional neural network combined with the decoupler for network training;finally,the test set is used for testing.In this experiment,two kinds of bearing fault vibration signals and two kinds of gear fault vibration signals,single fault signal and compound fault vibration signal,obtained on QPZZ-II simulation fault diagnosis experimental platform,are the most experimental objects for experimental verification.The results show that the method can realize normal fault,single fault and composite fault,and the average accuracy of fault diagnosis is 92.115%.