Research On Fault Diagnosis Methord Of Printing Machine Based On Multi-source

Printing machine is integrated equipment which combines the principles of light,electricity,fluid,gas,control,chemistry and network,and has been widely used in the fields of knowledge dissemination,product packaging,securities,and printed electronics,etc.The working state of the key components was characterized by the extracted feature of the fault image marks,thereby recognizing and diagnosing faults.Based on multi-source information,such as sound information and vibration information,the extracting method of fault feature for positioning mechanism,main transmission gears,and hearings was proposed,and was validated by experiments.Moreover,an on-line testing device for fault marks was designed,and a fault diagnosis system based on image information was developed.The main contents and innovative points are as follows:(1)A rectangle detection method based on Radon transformation and gray projecting integral(GPI)method was proposed,by which the detection accuracy of the straight line for registering was increased,and deviation induced by mark skew was reduced.In this detection method,by measuring the distance from the straight line(registering mark)to rectangle and calculating the standard deviation,the working state of the positioning mechanism or grippers was determined.The work provides reliable information source for online fault diagnosis of positioning mechanism in printing machine.(2)The state detection method based on 50%dot images for press cylinder and ink roller has low extracting accuracy of dot area under uneven light.Present study proposed a new calculating method for dot area,which overcame the reverse adverse effect of uneven light.By comparing the testing results using this method and those using densimeter,it was found that this method has higher extracting accuracy in both light area and dark area.This method can realize online calculation,reducing the error and workload of manual calculation.It provides a new research thought for extracting global dot area.(3)Some transmission components in printing machine,such as bearing and gear,are prone to failure,and it is difficult to recognize these fault sources.Present study proposed a fault characteristic extracting method for bearing based on signifying intrinsic mode components.This method includes three stages.Firstly,collected original information were extended,and analyzed using empirical mode decomposition(EMD),and delaying parts of the information were intercepted.Secondly,on the premise of ensuring the reliability of IMF,the intrinsic mode components were signified,encoded,and reconstructed to characterize the changing trend and to extract information entropy.Finally,the faults of bearings and gears were classified and diagnosed combing model recognition method.This method not only retains the advantage of adaptive analysis,but also effectively reduces computation,improves calculation efficiency,and decreases redundant information.The work provides theoretical foundation for calculating the feature set of large-scale integrated printing equipment.(4)Under complicated conditions,it is difficult to diagnose the faults of some key components in printing machine based on single information source.Present study put forward to a multi-source feature fusion method based on manifold learning.Both vibration information and sound information were introduced to this method,forming multi-source information sources.Intrinsic mode signal entropy was used to fuse the multi-source information,establishing fused feature set.Manifold structure was introduced to realize the fusion of the intrinsic mode symbolized feature set.The feature extracting results of LMD-LE method was quantified combining support vector classifier model,thereby realizing the classification of faults of bearing under different conditions.This method has been used in the fault diagnose of bearing and gear in printing press,and it was found that the faults could be well recognized and classified according to the multi-source information.(5)Considering that there is a lot of mechanical state information in printing press and accumulated human empirical knowledge,a fault diagnosis system based on knowledge rules and image information was proposed.Through investigating printed image feature extraction method based on gray-level co-occurrence matrix(GLCM),redundant information removal method based on principal component analysis,and the decision network constructing method based on support vector machine,a fault diagnosis system based on graphics was developed.The reliability of the system was confirmed experimentally.In present study,the fault feature extraction and fault recognition methods for some key components in printing machine were investigated based on image information,vibration information,and sound information.A fault diagnosis system was developed,and it was proved feasible and reliable.The fault feature extracting method and fault diagnosis method establishes theoretical foundation for the fault diagnosis of large-scale complex integrated electromechanical equipment.