Rolling Bearing Reliability Evaluation And Prediction Based On Feature Selection And LST

Rolling bearing is a key component in machinery and facilities,which widely exists in various fields.Its operation condition plays a decisive role in the health of the mechanical system.With the continuous operation of the bearing,its performance presents a degradation trend.Over time,it will cause the failure of the bearing and the accident of mechanical devices in serious cases.Therefore,need to apply technology to evaluate bearing state of degradation and analysis,predict the law of bearing performance degradation.Under the background of industrial automation and artificial intelligence,monitoring the operation of equipment by machine can obtain the state change information of equipment,which is also the cornerstone of analyzing the change trend of equipment reliability and service life.In this paper,bearing is taken as the research target,vibration data drives the operation of the model,and a prediction model is constructed according to the logistic regression model and long and short term memory network to evaluate and predict the change trend of bearing reliability.The main research contents are as follows:(1)This paper discusses the purpose,background and significance of the subject study,describes the research status of bearing reliability evaluation and prediction at home and abroad,expounds the basic structure and failure types of bearing,and introduces the method of vibration signal data processing.(2)According to the characteristics of the bearing in the reliability prediction to extract,can not accurately reflect bearing degradation process of the problem,select the monotonicity correlation and the characteristics of robustness of three evaluation indexes,and defined the restrictive indicators and characteristic parameters of screening,selected can accurate characterization of the parameters of the degradation process,constitute the degradation characteristic parameter set.(3)Aiming at the nonstationary and nonlinear problems of vibration signal,the dimension reduction algorithm of random proximity embedding(SPE)is used to reduce the dimension of degraded feature parameter set,which can reasonably deal with most dimension disaster problems and form feature parameter set.Then the wavelet denoising method is used to denoise the characteristic parameter set,reduce the noise in the data,and bring the denoised parameter set into the logistic regression model to evaluate the reliability,which is conducive to improving the accuracy of reliability evaluation.(4)A rolling bearing reliability prediction method based on logistic regression model and long and short-term memory(LSTM)network was proposed.The degradation characteristic parameter set and feature vector set were used as input data labels respectively in LSTM network,and the output data were used in reliability evaluation model for reliability prediction.The comparison between Elman and correlation vector machine(RVM)can effectively prove that the prediction method in this paper is better.