An Integrated SPC-EPC Study Based On Nonlinear And Nonparametric Time Series

Statistical process control (SPC) and engineering process control (EPC) are twotechniques originated from the process industry and parts industry. Although the pathsthey achieve their goals are different, they have the same goal of reducing thedeviation in the quality characteristics. With the development of modern productiontechniques, integrated SPC-EPC is now considered to be an effective quality controlmethod, and has received many attentions and applications.Along with the development of the complexity in the production processes andthe products, more and more products have complex nonlinear autocorrelationships oftheir quality characteristics in their production processes. Now the studies ofintegrated SPC-EPC are based on linear time series model to describe theseautocorrelationships. But these linear models have errors in the description ofnonlinear relationships. This will affect the final control result. In order to solve thisproblem, this dissertation proposes a method using two typical kinds of nonlinear timeseries model——the threshold autoregressive model (TAR) and smooth transitionautoregressive model (STAR) to describe the autocorrelationships and buildingcontroller and integrated SPC-EPC system based on these two models. Theperformance of this control method is studied and verified through examples andsimulations. The results indicate that the method based on nonlinear time series modelcan effectively control the process which has nonlinear autocorrelationships. Thisdissertation then proposes a method based on a kind of more advanced time seriesmodel. That is the nonparametric functional coefficient autoregressive model (FCAR)to describe the dynamic nonlinear autocorrelationships. Using the same procedure tostudy this method, the results indicate that the method based on nonparametric timeseries model can effectively control the process which has complex nonlinearautocorrelationships.Now the studies of integrated SPC-EPC are based on linear transfer functionmodel to describe the relationship between the input variables and output variables.But these linear transfer function models have errors in the description of nonlinearinput-output relationships, which are closer to modern manufacturing processes. Inorder to solve this problem, this dissertation proposes a method using thenonparametric transfer function model to describe the input-output relationships and building controller and integrated SPC-EPC system based on this model. Theperformance of this control method is studied and verified through examples andsimulations. The results indicate that the method based on nonparametric transferfunction model can effectively control the process which has nonlinear input-outputrelationships.