Univariate and multivariate control charts for monitoring dynamic-behavior processes: a case study

Abstract

The majority of classic SPC methodologies assume a steady-state (i.e., static) process behavior (i.e., the process mean and variance are constant) without the influence of the dynamic behavior (i.e., an intended or unintended shift in the process mean or variance). Traditional SPC has been successfully used in steady-state manufacturing processes, but these approaches are not valid for use in dynamic behavior environments. The goal of this paper is to present the process monitoring and adjustment methodologies for addressing dynamic behavior problems so that system performance improvement may be attained. The methodologies will provide a scientific approach to acquire critical knowledge of the dynamic behavior as well as improved control and quality, leading to the enhancement of economic position. The two major developments in this paper are: (1) the characterization of the dynamic behavior of the manufacturing process with the appropriate monitoring procedures; and (2) the development of adaptive monitoring procedures for the processes [for example, using trend charts (e.g., linear model) and time series charts (e.g., ARIMA models)] with a comparison between univariate and multivariate control charts. To provide a realistic environment for the development of the dynamic behavior monitoring and adjustment procedures, the cold rolling process is adopted as a test bed.