Quality Dynamic Monitoring Method Based On SPC Control Chart Pattern Recognition For Automatic Production Line And Application

With the development of computer technology, the quality management system hasbeen given widely attention, which is regarded as an important measure.to ensureproduct quality, and promote the competitiveness of the manufacturing enterprise It is ahot spot in the field of Manufacture Information Engineering that is how to effectivelyuse the computer and communication technology to achieve quality management andmonitoring in mechanical parts processing process.The SPC (Statistical Process Control) is the mainstream technology formanufacture process quality control. On one hand, the SPC control charts can be used todetect the potential anomalies in the process of production, so as to achieve the aim ofprocess quality control and process monitoring. On the other hand, the recognition ofcontrol chart patterns can effectively identify the abnormal types which is the auxiliarydiagnosis of abnormal causes. In this thesis, combining with the national major scienceand technology subject "gear processing automatic production line networking sitemanagement and intelligent monitoring and control system", we studied the SPC controlchart patterns based quality dynamic monitoring method for automatic production line.And the main research contents are as follows:First of all, we make a study of the SPC application strategy aiming at automaticprocessing production line. Considered of the mass produced and the high automationdegree characteristic, we deeply analysis the theory and application of control charts forautomatic processing production line. Then we proposed the SPC quality controlstrategy that is useful for mechanical parts automatic processing production line.Secondly, we proposed two SPC control chart patterns recognition based modelsfor automatic manufacturing process monitoring. The one is an integrated model basedon binary-tree support vector machine (BTSVM) for control chart pattern recognition.The integrated model firstly detect the whether the abnormal patterns arise, and then theBTSVM is used to classify the detected abnormal control chart patterns. The other is ahybrid control chart patterns recognition model with K-Means based feature extractiontechnology. The hybrid model utilizes the K-Means algorithm to extract distancefeatures which are then served as the input of classifier. Simulation results show theaverage classification accuracy of the two models is98.75%and99.24%respectively. Compared with the results of some previous research, both the two models haveachieved better average classification accuracy for8basic abnormal types.Finally, we analyzed the quality monitoring system requirements of someautomobile gear manufacturing enterprise in chongqing. Then, we established theframework and functional models of the system and provided a solution for datacollection. A quality monitoring system, aimed at gear production line, wasimplemented based on the research.