Bearing Performance Degradation Assessment And Residual Life Prediction Based On EMD And Logistic Regression

With the advancement of industry and science and technology,electromechanical systems are updating towards intelligent,efficient and sophisticated.Rolling bearings play a key role in carrying and transmitting power in many electromechanical systems.Rolling bearings in the event of a sudden failure,will not only seriously affect the production efficiency of equipment,and may even cause casualties.Therefore,to grasp the data of rolling bearing performance degradation,to predict the residual life of rolling bearings,to arrange for rolling bearing maintenance and replacement reasonably,should be quite significant for the assurance of electromechanical equipment normal operation and maintenance cost reduction of rolling bearings.During the life cycle of bearings,from the early trouble to the final complete failure,the bearings go through deepening,grinding,deterioration of the process.If the degradation of bearing performance can be monitored during the degradation process,and even reach the prediction of the remaining life,it would be well arranged for maintenance and reparation according to the current situation to ensure the safety and reliability of bearings and good operation of the equipment.In this dissertation,take the rolling bearing as research object.Extract the rolling bearing vibration signal feature according to empirical mode decomposition algorithm,achieve the quantitative evaluation of the performance degradation of rolling bearings treating logical regression model as an evaluation model,and finally use the self-regression moving average model and the gray model to achieve the signal eigenvalue and the residual life prediction.The main contents are:1.Establish the performance degradation assessment model for bearing.Randomly select sets of data of bearing in normal state and failure state,calculate the vibration signal intrinsic mode function as the eigenvector according to the empirical mode decomposition method.Establish a logistic regression model with training samples and obtain the model parameters.2.Performance degradation assessment and result verification for bearing.Extract the vibration signal characteristic vector of the bearing under test as the test samples.Input the test samples into the evaluation model to calculate the performance degradation assessment indicator of the bearing under test.Then verify the results of performance degradation evaluation by analysis method combined Empirical Mode Decomposition(EMD)and envelope demodulation.3.Residual life prediction for bearing.Use the statistical analysis method to forecast the vibration data or the degradation index of the bearing and estimate its remaining life.(1)Take the vibration signal data of the rolling bearing to obtain the vibration intensity,establish the autoregressive moving average prediction model with the vibration intensity time series,and then input the processed detection data into the established model to predict the vibration data of the bearing.(2)Following the evaluation of bearing performance degradation,establish the gray evaluation model by training with the obtained evaluation index as the degenerate feature sequence.Input the evaluation index into gray evaluation model established,calculate the bearing performance degradation index value for the future time,and then come to predict the residual life of rolling bearing.