The Study Of The Integration Of Statistical Process Control And Engineering Process Control For Autoregressive Processes

Reducing process variations is critical for industries, because it has a direct bearing on the quality of the products. Statistical process control(SPC) methods have been successfully utilized in discrete parts industry through identification and elimination of assignable causes. Engineering process control(EPC) methods have been widely employed in continuous process industry for reducing common causes. Recently, the integration study of SPC and EPC methods has significantly reduced process variations and has improved process outputs. However, when there is significant autocorrelation in process data, traditional methods of SPC should not be used. The autocorrelation should be modelled so that autocorrelated control charts can be constructed. For autocorrelated processes, when it is in the state of statistical stability, the process can be understood as certain stationary time series process. On the contrary, when it is in the state of statistical instability, it means that parameters of the time series model have changed or that the time series model is no longer suitable. Therefore, in order to tackle the quality problem of monitoring and adjustment for autocorrelated processes, this thesis provides three integrated methods of SPC and EPC for stationary time series model with arbitrary order.First of all, the steady-state auto regressive and moving average(ARMAST) control chart based on a stationary auto regressive(SAR) disturbance model has been integrated with the Grubbs’ harmonic rule for monitoring and adjustment of the setup error of autocorrelated processes. For Minimum-Mean-Squared-Error-(MMSE-) Controlled processes, in view of eliminating the autocorrelation of the process, a generic joint control chart is developed under auto regressive AR(p) stationary disturbance model with arbitrary order, which can jointly monitor process outputs and manipulated inputs. At last, a new generic framework for monitoring and adjusting the multiple input and single output(MISO) processes, based on the integration of response surface methodology(RSM), Shewhart control chart, and a newly developed minimum quadratic loss(MQL) controller, has been proposed under the auto regressive AR(p) stationary disturbance model with arbitrary order.Simulation experiments and case studies are conducted to demonstrate the validity of three newly generic methods of the integration of SPC and EPC. The experimental results suggest that the first integrated scheme can effectively monitor and adjust the setup error, the second scheme can eliminate the autocorrelation of the process, and the third integrated scheme can decrease the output variations of the MISO process.