| As an important part which is widely used in rotating machinery,when rolling bearing is damaged,it may lead to unplanned shutdown of equipment,which seriously affects production efficiency and economic benefits.Fault diagnosis of rolling bearings is of great significance to ensure the safe and reliable operation of mechanical equipment,reduce maintenance costs and the safety of human machines.Based on the theory of recurrence analysis,this paper constructs a recurrence plot for the vibration signals of rolling bearings,and carries out an analysis around the recurrence plot.From the perspectives of recurrence quantification analysis and image recognition,an indepth study is carried out for the purpose of realizing intelligent diagnosis of rolling bearings.The main research contents are as follows:1.In view of the limited ability of traditional signal processing methods to extract the characteristics of rolling bearing vibration signals,and the shortcomings of various complex operations on the signals which are needed to obtain the characteristics that can truly represent the fault state,recurrence analysis is introduced into the field of rolling bearing fault diagnosis.The vibration signals of rolling bearings are used to build a recurrence plot,and the image differences appearing in the recurrence plot of vibration signals of different fault types were qualitatively analyzed.On this basis,Recurrence Quantification Analysis(RQA)is used to obtain its recurrence characteristic parameters and then the recurrence feature extraction of rolling bearing fault is realized.2.A fault diagnosis method based on the combination of RQA and Sparrow Search Optimization Support Vector Machine(SSA-SVM)is proposed.In this method,RQA features with rich dynamic characteristics of rolling bearings is used as SVM inputs.Aiming at the problem that the improper value of penalty factor C and kernel function parameter G in SVM severely restricted the SVM classification performance,Sparrow Search Algorithm(SSA)is introduced to optimize the key parameters,and a fault diagnosis model of SSA-SVM is established.Then the SVM performance of different optimization algorithms and the model performance under different feature inputs are analyzed.The results show that compared with other methods,this method can accurately and efficiently realize the fault diagnosis of rolling bearings.3.In view of the incomplete quantization of dynamic characteristics of recursive graph revealed by recursive quantitative analysis,and the feature extraction relies on manual operation,a fault diagnosis method combining non-threshold Recurrence Plot(NTRP)and convolutional neural network is proposed.In this method,the vibration signals are reconstructed into a recurrence plot without threshold to avoid the influence of threshold on the recurrence plot,and the vibration signals are input into the convolutional neural network to realize adaptive feature extraction and fault diagnosis.Experimental results show that this method has a high accuracy in fault diagnosis.Then compared with other vibration signal visualization methods,it is proved that the combination of non-threshold recurrence plot and convolutional neural network can more effectively realize the fault diagnosis of bearings. |
