Research On Health Condition Prediction Of Flexible Material R2R Processing Equipment Based On Deep Belief Networks

Flexible material R2R(Roll to Roll)processing equipment is the critical equipment for processing and transferring flexible material between multi-position.R2 R processing technology has been widely implemented in the fields of flexible membranes,flexible electronics,etc.At present,the prediction of the health status of flexible material processing equipment is usually based on fault identification,which is a relatively passive maintenance mode.This mode cannot divide the complete period health status of R2 R equipment effectively and negatively affect the maximization of production equipment benefits for manufacturing enterprises.The flexible material R2 R processing equipment is taken as the research object of this dissertation.A health prediction method of processing equipment based on the deep belief network is proposed,specific to the problem that the performance of R2 R equipment decreases with the increase of working hours and the health status of roller shaft of R2 R processing equipment is evaluated by DBN network model.Based on the working peculiarity of the flexible material R2 R processing equipment,this dissertation analyses the processing technology of the enterprise's flexible membranes in the cause of obtaining the factors affecting the product quality,and further extracts the factors which affect the health status of the R2 R processing equipment.On behalf of maximizing the efficiency of equipment maintenance and solving the problem of cost waste caused by equipment excessive maintenance,the pre-processing and feature extraction of the vibration signal of the roller shaft are carried out,and the health status of the system is predicted by combining with the deep belief network.The main contents of the dissertation include:(1)This dissertation aims at extracting key performance fading indexes of flexible material R2 R processing equipment;analyzing the processing technology of the roller shaft under the normal working and the existence of various factors affecting the quality;reflecting the physical quantity of status change information explicitly.The sensors are used to set up a monitoring system for data acquisition of flexible material R2 R processing equipment.(2)Noise reduction and time-domain feature extraction are carried out on the original vibration data of the roller shaft,and the physical significance of common timedomain feature data is analyzed.The dimensionality of the eigenvalue matrix is reduced by the PCA algorithm.DBN neural network is trained on the principal component data after dimensionality reduction,and the data is reconstructed through the DBN network to learn the characteristics of roller shaft data.(3)Based on the realization of the DBN network model,the health evaluation system of the R2 R equipment roller shaft is designed.The steps for performance prediction of the R2 R processing equipment based on the DBN neural network are described in detail.Softmax classifier is introduced to extract the real-time features of vibration signals with field equipment,and the health status is classified and predicted according to the real-time signals of the roller shaft and DBN network model.