Research On Register Error Prediction Method Of Gravure Printing Machine Based On Machine Learning

The printing unit is a key device for the operation of the gravure printing press,including the core unit of the gravure printing press.The normal operation of the printing cylinder in the printing unit ensures the stability of the entire printing process.With the rapid development of artificial intelligence technology,the intelligent identification of printing production equipment faults,diagnostic technology and fault prediction technology have gradually attracted the attention of many scholars.Fault prediction technology enables production personnel to predict possible system faults through historical operating data,which plays an important role in improving printing quality and reducing waste in non-printing processes.This paper takes the printing unit of the gravure printing machine as the research object,and based on the machine learning theory,studies the fault intelligent diagnosis and fault prediction methods of the gravure printing machine.The main research contents are as follows:(1)Research and analyze the fault prediction method of the printing unit of the intaglio printing machine based on machine learning,and the diagnosis method based on the machine learning support vector machine.According to the actual production situation and the failures that may occur during the operation of the machine,the causes of failures in the printing unit of the gravure printing machine and the failure data processing methods are analyzed.(2)Select the parameters of the support vector machine in a scientific and reasonable way,avoiding the increase in the number of samples due to a larger parameter selection range.A too large sample size will complicate the solution process.This paper uses the excellent global search ability of particle swarm algorithm.The parameters in the SVM are optimized,and the PSO-SVM fault prediction model is constructed.This method can optimize the parameters when the data form is more complicated,and improve the speed of data training.(3)Using the B&K vibration and noise measurement system,perform vibration testing on the printing unit of the intaglio printing machine.Transform the vibration signal measured by the acceleration sensor in the time domain and frequency domain,extract the fault feature value and feature vector from the tested data,and combine the PSO-SVM prediction model to build a vibration detection environment on the printing unit of the gravure printer.The pre-processed data is put into the model for training to predict the vibration waveform it will produce,and realize the function of predicting potential faults.(4)According to the vibration data collected from the historical operating conditions of the printing press,the faults of the printing cylinder are classified into categories such as health,printing cylinder bearing failure,and printing cylinder surface failure.In order to compare the applicability of the models in this paper,this paper compares the performance of the PSO-SVM failure prediction model,the traditional SVM failure prediction model and the BP neural network failure prediction model for the printing press failure prediction problem.The pre-processed data set is used as a training sample and input into the three prediction models.Use the same validation set to perform prediction accuracy experiments on the model.