Quality Quandaries: Interpretation of Signals from Runs Rules in Shewhart Control Charts

The Shewhart control chart is for many people the embodiment of statistical process control (SPC). In the past, many modifications and improvements have been devised (see Woodall and Montgomery [1999] for a brief overview). Several modifications or additions were aimed at more power for out-of-control situations. Among the earlier additions belong the runs rules from the Western Electric Company (1956). These rules stem from the idea that a graphical pattern in the chart may help in identifying an out-of-control situation even before the control limits have been exceeded. Runs rules are formalizations of such patterns and therefore easily understood by users. Although many rules have been devised, for all kinds of patterns, only a few should be selected in practice, because application of many rules simultaneously leads to an unacceptable number of false signals (Does and Schriever 1992), and users might easily get confused. Many alternatives for the Shewhart control chart have been designed with an increased power for specific out-of-control situations in mind (see Woodall and Montgomery [1999] for references). For example, a small change of the process mean will generally be detected faster with a cumulative sum (CUSUM) or an exponentially weighted moving average (EWMA) chart. Many processes, however, will get out of control not just for one reason but because of one of many. Each failure cause usually has its own specific effect on the process data: the mean may change more or less, or the variation, or both. SPC is aimed at finding and correcting permanent changes as quickly as possible, whereas feedback control is generally a better alternative for correcting gradual or temporary changes. Usually, there is no single optimal control chart for all possible failure causes. The Shewhart chart with additional runs rules might be considered as an omnibus test for many different permanent process disturbances. An effective SPC system requires an out-of-control action plan (OCAP; see Sandorf and Bassett 1993) to accompany a control chart. An OCAP is a guide for the user to solve the problem when an out-of-control signal is given. The form may range from a simple checklist to an extensive flowchart Edited by Ronald J. M. M. Does.