Rolling Bearing Fault Diagnosis And Remaining Life Prediction Based On Data-driven

In the context of the rapid development of urban rail transit,driving safety has become a top priority.Once an accident occurs,it will seriously threaten the personal and property safety of passengers.The vehicle is the most important factor affecting the safety of rail transit operations,and the safety of key components of rotating parts is of paramount importance.Therefore,studying the fault diagnosis and remaining life prediction technology of key components such as urban rail train bearings to achieve accurate prediction of the performance of key components can provide a powerful way to develop flexible equipment maintenance strategies from passive maintenance to active maintenance based on equipment status.reference.In this paper,rolling bearings are set as the research object,and the research based on fault diagnosis and remaining life prediction is carried out.The main research contents include:(1)Established a bearing fault diagnosis model based on acoustic emission technology.Acoustic emission monitoring and acquisition experiments under different faults were carried out,and Hit driving parameters and time domain characteristic parameters were extracted.The fault detection is realized by observing the distribution of the time-domain waveform and the frequency-domain fault frequency,and the comparison of the vibration signal reflects the advantages of the acoustic emission technology in the detection of sudden signals and early faults.At the same time,based on the Lib SVM model,the acoustic emission RMS characteristic value is classified into multiple faults,which verifies the accuracy of different faults.(2)A two-parameter exponential Bayesian bearing residual life prediction model was established.It is proposed to optimally memorize the latest posterior parameters of the phased prediction based on the original algorithm,and use it as the initial value of the next parameter.For degraded degraded bearings,based on the fault diagnosis results,the peak value is selected to expand the degradation trend of the fault type,the probability distribution of the remaining life is calculated,the average value and the confidence interval of the remaining life are output,and the maintenance work of the actual project is carried out.Provide more information.(3)Established the RVM bearing remaining life prediction model based on gray wolf algorithm optimization.A correlation vector machine model with multiple input variables is established based on multiple monitoring historical values.The gray wolf algorithm is used to optimize the weights and parameters of the hybrid kernel function,and information entropy is used to determine the best embedding dimension m represented by the number of historical values..It is proposed to add multiple eigenvalue input models that can characterize the bearing degradation process,which can effectively improve the accuracy of RVM prediction results and effectively reduce the width of the confidence interval.(4)A two-layer LSTM bearing remaining life prediction model was established.Combining the time series characteristics of bearing monitoring data,a LSTM network model is proposed to reconstruct multiple historical eigenvalues of RMS,which verifies that LSTM has the advantage of feature mining for long-order data under a single eigenvalue,and improves the remaining life of the bearing in the later stage of deterioration.The percentage of forecast accuracy.On the other hand,it is proposed to use the predicted value to carry out continuous multi-step prediction,and through the degradation trend fitting graph,it verifies the effectiveness of the three-step advance prediction of the degraded faulty bearing.