Statistical Process Control, or SPC, refers to a series of process monitoring and quality control techniques. Originally developed for the manufacturing industry in the 1920s, SPC is applicable in a wide range of settings and processes, including business, industry, and the public sector. Although it is rooted in statistical principles and concepts, the graphical tools it uses are easy to construct and interpret, and do not require in-depth statistical knowledge.
- The key to understanding Statistical Process Control lies in understanding the principle of variation. All processes have variation, but not all variation indicates a problem that requires corrective action. SPC considers two types of variation: common cause and special cause. Common cause refers to the random variation that exists in all processes, whether in industry, health care, government services, and others. A process that displays only common cause variation is in control and does not require special attention. Special cause variation, in contrast, refers to unexpected situations or problems that cause processes to go out of control. Special cause variation suggests the occurrence of something that is not part of the process design and should be eliminated.
Sources of variation -- common cause and special cause -- in processes include people, materials and measurements themselves. All people and materials are different and thus contribute to variatio n. Measurements and samples drawn at certain times or from certain areas can bias results. The key is distinguishing natural random variation (common cause) from special cause variation that requires intervention and improvement. - Statistical Process Control provides a set of graphical tools for monitoring processes and identifying special cause variation. The methods are rooted in statistical concepts, but a person need not be a statistician to understand and use SPC. The most popular SPC tools are the run chart and the control chart. They are easy to construct and do not require computer software.
A run chart is a temporal sequence of data, with horizontal and vertical axes, scaled to represent points in time and the range of values in the data, respectively. The run chart features a horizontal centerline, which represents the mean (arithmetic average) of the values of the data. For example, if a machine produces an average of 10 widgets per hour, the centerline would be at the value of 10. With a run chart, managers can plot the number of widgets produced for each hour in which measurements were taken over a period of time, then connect the individual points with a line. - Control charts extend the run chart to include upper and lower limits for variation from the mean. These limits are based on the standard deviation, a measure of variation in a set of data. The size of the upper and lower limits may vary, but many SPC users set a range of 3 standard deviations above or below the mean as the upper and lower limits, respectively. Limits can be narrower than 3 standard deviations at the discretion of the manager or monitor. For example, a health care administrator monitoring surgical errors will likely set a much narrower range.
- Interpreting a run chart to identify special cause variation, if any, requires examining the temporal plot of measurements, paying special attention to plot points that differ from the mean. A trend of multiple data points above or below the mean, a zig-zag pattern in the line joining data points, or wildly different individual points could suggest special cause variation and thus, instability in the process.
With a control chart, if variation in the process stays within the upper and lower limits (generally, plus or minus 3 standard deviations), the process is considered to be in control. Variation beyond these limits suggests instability, requiring intervention or improvement.
评论