Research On Fault Diagnosis Method Of Rolling Bearing Based On Rail Transit Train

Rolling bearings are an important part of rail transit trains.It is very important to realize the rapid and accurate identification of the types of rolling bearing failures to ensure the smooth,safe and reliable operation of rail transit trains.This paper takes the rolling bearings of rail transit trains as the research background,and realizes the diagnosis and recognition of normal state and three common types of faults: inner ring fault,outer ring fault,and rolling element fault.Mainly research and improve the fault feature extraction and fault type identification in the rolling bearing diagnosis process.In view of the characteristics of non-periodic and non-stationary vibration signals of rolling bearings,as well as the interference and poor effects in the extraction of fault features,Variational Mode Decomposition and Multiscale permutation entropy are proposed.The VMD method is a typical adaptive signal decomposition method,which can adaptively decompose the signal into K modes,and can effectively avoid the problems of mode aliasing and boundary utility.In this paper,the VMD method is used to decompose the collected vibration signal of the rolling bearing into four modal components,and the energy feature of each mode is extracted to construct the energy feature vector as the input vector for pattern recognition.In order to further quantify the energy characteristics of each mode,multi-scale analysis and calculation of each mode is performed by MPE method to more comprehensively extract the fault characteristics.Aiming at the problem of low accuracy of identifying fault types,this paper uses probabilistic neural networks in artificial neural networks to classify and identify fault types.PNN is widely used in the field of pattern recognition,but the value of the important parameter smoothing factor ? is still based on the traditional sample clustering method or empirical estimation method.The quality of ? directly determines the performance of the entire network.Therefore,particle swarm optimization(PSO)is used to optimize and solve? to improve network performance,and inertial weight ? and adaptive mutation are added to the PSO algorithm Enhance its ability to find the best.Through research,this paper has constructed four types of rolling bearing fault recognition models: VMD + PNN model,VMD-MPE + PNN model,VMD + PSO-PNN model and VMD-MPE + PSO-PNN model.The results of MATLAB simulation show the diagnosis results The correct rates are 88.75%,92.50%,95.00% and 97.50%.It can be seen from the experimental results that the vibration signal of the rolling bearing is decomposed by the VMD method,and the energy feature vector is further extracted by quantizing the energy feature through the MPE method,and finally input into the PNN network optimized by the PSO algorithm can effectively realize the classification and classification of the rolling bearing fault type The recognition classification rate is high.