From trendcharts to control charts: setup tests for making the leap

The early stages of industrial SPC (statistical process control) programs are typically characterized by trendcharts at critical process steps. As processes are brought under control, these trendcharts are replaced by control charts. But premature replacement of trendcharts by control charts, before processes are under control, may undermine the credibility of nascent SPC programs, causing a barrage of special cause flags triggered by limits inappropriately calculated on unstable processes. The author describes an operational definition of the term under control as needed to decided which trendcharts can be implemented on control charts. Simple rules, which derive from Monte Carlo simulations of special cause flags from W.A. Shewhart (1931) and L.S. Nelson (1984), provide constant and low probabilities of inappropriate process rejection and maintain the high efficiency of time series analysis intrinsic to traditional special cause flags. The rules depend heavily on sample size; it is shown that any attempt to ignore the effects of sample size will invite wildly uncontrolled risks of rejection good processes. A table is provided to facilitate application of these setup tests at a constant and low risk of rejecting good processes.<
>