Intelligent Identification Of Surface Damage Fault Of Gravure Printing Cylinder Based On Virtual Simulation

Intaglio printing is the most widely used printing technology in the flexible packaging industry.Intaglio printing equipment accounts for more than 90% of flexible packaging printing equipment.The printing cylinders are the key parts of intaglio printing equipment.The surface damage of the cylinder will lead to insufficient local printing pressure and blank spots on the surface of printed matter,which will affect the quality of printed matter.This thesis is oriented to the field of flexible packaging printing and takes the gravure cylinder of intaglio printing equipment as the research object.Intelligent identification of surface damage faults on gravure printing drums has been achieved based on virtual simulation and deep learning technology.The innovation of this thesis is to carry out the joint simulation of ABAQUS and ADAMS to establish the dynamic simulation model of Intaglio printing cylinders,design the drum sanding damage experiment,and propose amplitude normalization and a user-defined recognition network for intelligent recognition with an intelligent recognition software developed.The specific work is as follows:(1)Comparing and analyzing the dynamic characteristics changes after the occurrence of rotor surface damage based on the occurrence mechanism of surface damage,the production characteristics of intaglio printing and rotor dynamics theory,the dynamic model of micro rotor and eccentric micro rotor are established.The theoretical analysis results indicate that the amplitude of the drum vibration signal will undergo a sudden change if the surface damage occurs.(2)This thesis conducted mapping and proportional modeling for the intaglio printing equipment of a printing intelligent manufacturing demonstration enterprise in Hebei.Based on multi body dynamics and rigid flexible coupling theory,a virtual simulation model of the printing unit was established.Based on Hertz theory and on-site process cards,the operation input conditions were calculated,and surface damage faults of different sizes were defined.The joint simulation of ADAMS and ABAQUS was conducted and the simulation results were compared and analyzed.The virtual simulation results indicate that there is clutter interference in the drum vibration signal after surface damage occurs.(3)In this thesis,the reliability of theoretical analysis conclusions and virtual simulation results were verified through experiments and an experiment was designed to collect the vibration signals of the drum with surface damage.With the assistance of on-site engineers in the workshop,sanding damage was created on the surface of the printing drum.The vibration acceleration signals before and after the grinding damage of the drum was collected and a database of running signals for the surface damage of the drum was constructed.The collected signals were denoised by wavelet transform,envelope spectrum analysis and Fourier transform,and the value of frequency concentration ratio function was calculated.The reliability of theoretical analysis results and virtual simulation results was verified based on experimental results.(4)In order to identify whether the equipment has surface damage,the amplitude normalization method was proposed,and the vibration signal was integrated into the device status image.An intelligent identification network was designed based on the multi-layer perceptron model and the weight sharing principle.This thesis conducted two-stage intelligent recognition.In the first stage,simulated signals with less noise doping and obvious classification features were identified.By comparing the performance of several different recognition algorithms,the effectiveness and innovation of this algorithm were confirmed.The second stage was to identify on-site collected signals with complex classification features and confirm that this method has good adaptability to actual production situations.The development of intelligent recognition software was completed in this thesis.A surface damage intelligent recognition software is developed based on MATLAB GUI,local models generated by learning and damage datasets collected from physical experiments.The running environment and programs are encapsulated,and software migration is implemented at different terminals.