Fault Diagnosis And State Prediction Of Gravure Press Based On Neural Network

The position of printing press in the printing industry is very important.As the most core and key equipment in the printing industry,its operation reliability is directly related to the safety and stability of the printing production line.With the rapid rise of the printing industry,enterprises have a more urgent demand for the intelligence of the printing press.Due to the poor operating environment of the printing press drum,the drum bearing is eroded during the long-term operation.When the bearing has potential failure,it will lead to the unbalanced operation of the drum quality and greatly reduce the ink transfer accuracy.Therefore,the drum bearing has become the key vulnerable part of the drum.The accurate diagnosis of the drum bearing of the gravure press has important practical significance to improve the reliability of the gravure press.Aiming at the single source of fault diagnosis information of gravure press,which can not comprehensively evaluate the gravure press,this thesis puts forward the fault diagnosis basis of gravure press based on vibration signal and current signal.The principal component analysis method is used to extract the characteristics of the vibration signal and current signal of the drum bearing of the gravure press,and the effective information contained in the original signal is obtained.The results show that the feature set composed of principal components can better diagnose the fault types of cylinder bearings of gravure press.When creating the fault diagnosis model of self-organizing feature map（SOM）,aiming at the problem that the samples may be distributed in a certain area during the random initialization of network weights,this thesis proposes to obtain the original data distribution by using principal component analysis,and linearly combine the feature vectors corresponding to the two largest eigenvalues with the position of neurons in the output layer to initialize the weight vector.The accuracy of fault identification of roller bearing by the diagnosis model is further improved.Taking the roll to roll two-color overprint gravure printing machine as the research object,the vibration signal acquisition system and electrical signal acquisition platform are used to collect the operation data of the drum bearing of the gravure printing machine under different states,and the pca-som neural network diagnosis model is tested.The experimental results show that compared with the traditional SOM fault diagnosis model,the fault diagnosis model based on pca-som has higher discrimination of fault types,and the diagnosis accuracy has been greatly improved.