SPC

Statistical Process Control

Statistical Process Control

Introduction

Statistical process control is based on statistical thinking using both statistical and non statistical methods for analyzing and solving problems in order to implement the actions required to achieve and maintain a state of statistical control processes and continuous improvement of their stability and reproducibility. An important form of statistical control used by manufacturers, hospitals, and other organizations is the process control chart. Many versions of the control chart have been proposed.

In these charts, the researcher generally plots some sort of average against time. The charts also include upper and lower control limits. The range between the two control limits indicates the bound of “normal” variability. For example, an individual can plot the mean time count of activities during the days to study the control across the weeks. These data are also used to compute the lower and upper control limits. Clearly, in a situation like this one, the upper control limit is more important. A dangerous or abnormal time rate is one that goes above the upper control limit, signaling a cause for alarm. The basic premise of the control chart is to try to discern between random and not-random causes of variability. Going above the upper limit generally indicates to the staff that the increase in the time rate is unlikely to be the result of random variation, a special cause. Of course, for control charts to be of value, they must be based on a substantial number of years. The control chart is also used in manufacturing to separate special from common (random) causes in variation. All processes have variability over time. The number of defects in a manufacturing process varies from one instance to the next.

Discussion

Manufacturers have learned that it is important to understand what causes a large spike, a spike that goes outside of the limits, on the number of defects. Going outside the limits indicates that the process is outside of the expected variability and possibly in the range of a special cause. There is no space to discuss here how the upper and lower limits are obtained in a control chart, but it is tantamount of computing a confidence interval. When computing the limits and the process is not stable, it is important to try to determine and eliminate the causes of instability. If a special cause for the variability is found, the data are purged and the upper and lower limits recomputed. Once a stable process is identified, one that contains only random variability, the upper and lower limits is obtained. These limits in turn can be used in the chart to look for future abnormal situations. If a point is found outside the limits, then it is probable that this is an unusual situation that requires special consideration.

The range chart, as seen in Figure 3, shows that the process range is in control. Nevertheless, the range is higher than expected with a mean (i.e. R-bar) of 13 minutes, which indicates that the process is not very ...