Average Run Length Performance Approach to Determine the Best Control Chart When Process Data is Autocorrelated

Abstract

Most conventional Statistical Process Control techniques have been developed under the assumption of the independence of observations. However, due to advances in data sensing and capturing technologies, larger volumes of data are routinely being collected from individual units in manufacturing industries and therefore data autocorrelation phenomena is more likely to occur. Following this changes in manufacturing industries, many researchers have focused on the development of appropriate SPC techniques for autocorrelated data. This paper presents a methodology to be applied when the data exhibit autocorrelation and, in parallel, to evidence the strong capabilities that simulation can provide as a key tool to determine the best control chart to be used, taking into account the process’s dynamic behavior. To illustrate the proposed methodology and the important role of simulation, a numerical example with data collected from a pulp and paper industrial process is provided. A set of control charts based on the exponentially weighted moving average (EWMA) statistic was studied and the in and out-of-control average run length was chosen as performance criteria.The proposed methodology constitutes a useful tool for selecting the best control chart, taking into account the autocorrelated structure of the collected data.