Flexographic Pressure Prediction System Based On Convolutional Neural Network

Flexo printing pressure is light,and slight variations in the printing pressure can have a significant effect on the quality of the print.Usually the printing pressure is mainly determined by preprinting and checking the quality of printed products.This kind of pressure determination method needs to consume a certain amount of manpower and material resources,and is subject to human subjective factors.Currently,BOBST has developed an automatic prediction system of flexographic pressure,which can obtain a suitable pressure value that can be used for printing directly,according to the plate information in the paste section.Based on this,in order to achieve the prediction of flexographic pressure,this paper takes the Shaanxi BEIREN FCI300 flexographic printing machine as the research object,studies the influence of plate features on the contact pressure,and realizes the display of the plate pressure map on the Matlab platform,and establishes a pressure prediction model based on convolutional neural network.Achieves the predicted printing pressure transmission and storage from the computer to the plate cylinder through RFID technology.The specific research content is as follows:(1)The influence of plate features on the contact pressure is studied.A local finite element model of central impression cylinder and the plate cylinder(field version)is established by ansys.The influence of the graphic area characteristics and distribution characteristics on the contact pressure of the printing plate are analyzed under the compression of 0.22 mm.(2)A flexographic pressure prediction model based on convolutional neural network is established.The surface pressure of different printing plates at the initial position are taken as input samples of the neural network on the Shaanxi BEIREN FCI300 flexographic press.The pressure on the plate bearing strips in the normal printing are taken as the output samples of the neural network.The parameters of the convolutional neural network model are optimized,and the best convolutional neural network model is determined.Finally,the convolutional neural network model is verified by experiments.(3)The transmission and storage of the predicted pressure is studied.The CF-RH320 non-contact RF reader and s50 chip electronic tag are selected,and the reader interface is redeveloped to realize the automatic import of pressure data into the corresponding plate cylinder electronic tag.