Research On The Autocorrelated Process Based On EPC And SPC

In the process of enhancing product quality with Process Control, Statistical Process Control (SPC) mainly uses the control chart to test whether process parameters (e.g.,AVG) being affected by abnormality factors. Traditional Statistical Process Control (SPC) is based on the hypothesis of Independently&Identically Normally Distributed (IIND), but in some practical production processes, production data is significant autocorrelation, which results in shorten of the Average Run Length (ARL), increasing frequency of false alarm produced by Control Chart. Therefore, systematic research on processing theory and methods for autocorrelation data is carried out in this thesis.In order to look for effective SPC methods for autocorrelation process, first of all, this dissertation introduces Control Charts technology, autocorrelation and time sequence theory, EPC'basic theories in conventional statistic processes, and then gives two general processing methods for autocorrelation process. One is modifying control boundary method and the other is time series model method. Finally, we bring in a new treatment method, the method of Engineering Process Control (EPC) integrated with Statistical Process Control (SPC).For AR(1) and AR(2) autocorrelation data that is widespread in industrial production, we apply three technologies, Shewhart Control Chart, and CUSUM Control Chart and EWMA Control Chart in time series model method, and EPC integrated with SPC method in order to know the distinctions between the two methods. Research result shows that when data autocorrelation degree is lower, two methods are similar; When data is highly autocorrelated, EPC integrated SPC method is superior to time series model method. EWMA Control Chart and CUSUM Control Chart are also used to deal with autocorrelation data, and ARL value is less than that of Shewhart Control Chart, so EWMA Control Chart and CUSUM Control Chart are more suitable for autocorrelation data.