Frontiers in Statistical Quality Control 6

Book excerpt: In the 1920's, Walter Shewhart visualized that the marriage of statistical methods and manufacturing processes would produce reliable and consistent quality products. Shewhart (1931) conceived the idea of statistical process control (SPC) and developed the well-known and appropriately named Shewhart control chart. However, from the 1930s to the 1990s, literature on SPC schemes have been "captured" by the Shewhart paradigm of normality, independence and homogeneous variance. When in fact, the problems facing today's industries are more inconsistent than those faced by Shewhart in the 1930s. As a result of the advances in machine and sensor technology, process data can often be collected on-line. In this situation, the process observations that result from data collection activities will frequently not be serially independent, but autocorrelated. Autocorrelation has a significant impact on a control chart: the process may not exhibit a state of statistical control when in fact, it is in control. As the prevalence of this type of data is expected to increase in industry (Hahn 1989), so does the need to control and monitor it. Equivalently, literature has reflected this trend, and research in the area of SPC with autocorrelated data continues so that effective methods of handling correlated data are available. This type of data regularly occurs in the chemical and process industries, and is pervasive in computer-integrated manufacturing environments, clinical laboratory settings and in the majority of SPC applications across various manufacturing and service industries (Alwan 1991).