Remaining Useful Life Prediction Of Key Equipment Driven By Data-driven

With the continuous improvement of industrialization,many industrial equipment are invested in industrial processes.Due to the impact of various operating factors,the performance and health of this industrial equipment will inevitably degrade,resulting in the final failure of the equipment.If there is an accident caused by failure,the resulting loss of personnel and property and even environmental damage are often inestimable.Therefore,prognostics and health management(PHM)technologies are widely used in industry,national defense,aerospace and other fields.PHM technology includes two parts: fault prediction and health management.Determining the remaining useful life of the equipment is the key step of fault prediction.Therefore,based on the existing data-driven remaining useful life prediction method,this paper studies the remaining useful life prediction of key equipment in the industrial process.Specific research contents include:(1)In order to improve the prediction accuracy of remaining life of lithium-ion battery,a prediction method of remaining useful life for the lithium-ion battery based on particle filter error compensation algorithm is proposed.This method is based on the particle filter algorithm,and introduces the error compensation algorithm based on the correlation vector machine to build the error model for compensation.Compared with the prediction results of particle filter algorithm,the prediction error of this method is smaller,which shows that this method can improve the prediction accuracy.(2)In view of the degradation of prediction accuracy caused by particle degradation,an improved particle filter algorithm is used to predict the remaining useful life of the lithium-ion battery.In this method,the regularized particle filter algorithm is combined with the extended finite impulse response(EFIR)filter.The EFIR algorithm is used to reset the particle filter.A new posterior particle set is generated after resampling,and then the final estimated value is obtained.Compared with the prediction results of particle filter algorithm,the prediction error of this method is further reduced,which shows that this method can improve the prediction accuracy.(3)In order to reduce the accuracy of local prediction caused by the capacity recovery phenomenon in the static stage of the lithium-ion batteries,a deep learning prediction method considering the intermittent time is proposed.In this method,the interval time series of the static stage of the battery is taken as the input of the long and short-term memory network,and a more perfect prediction model is obtained.Comparing the prediction result of this method with that of long and short-term memory network,the prediction error and root mean square error of this method are smaller,which proves the validity of this method.(4)In view of the low accuracy of the traditional single variable based remaining useful life prediction methods for complex systems,a deep learning fusion algorithm is proposed to predict the remaining useful life of coupled multivariable systems,that is,the long and short-term memory network algorithm with principal component analysis.This method analyzes the correlation of various monitoring variables and decouples them,the coupled principal component sequence is used as the input of the network to get a more perfect prediction model.Finally,comparing this method with the prediction results of the other three machine learning algorithms,and the predicted root mean square error of the method is smaller,which shows the effectiveness of the method.