In the world of lean manufacturing, particularly in the Six Sigma orthodoxy, Statistical Process Control (SPC) can be seen as the C in DMAIC. Originally developed by Walter Shewhart in the 1920s, the process is designed to reduce variation in production as much as possible and easily identify outlying events. Above all SPC ensures stable and capable manufacturing, stable meaning consistent and predictable behavior, and capable meaning able to produce in a capacity and quality that meets the specs of the client. Statistical Process Control is considered a valuable tool for continuous improvement.
Perhaps the most powerful and widely used tool of SPC is the control chart, or 'SPC Chart'. Giving a visual representation of any production anomalies, the SPC chart has been recreated in many versions and variations to best suit the needs of manufacturers. But generally speaking, there are two primary types of variation an SPC control chart tracks; common cause variation (CCV) and special cause variation (SCV). CCV is defined as the unavoidable variation that is a function of living in an imperfect world. As long as manufacturing remains between the Upper Control Limit (UCL) and Lower Control Limit (LCL) the plant is functioning correctly. Specification limits are also tracked, which corresponds to a capable process being able to fulfill a client’s needs. An SCV is a large variation that stems from an external factor effecting production and pushing the line outside the UCL and LCL.
Statistical Process Control Chart - Courtesy AssemblyMag
One commonly used variation of the SPC Chart is the Cumulative Sum Chart (CUSUM). CUSUM charts are used to detect extremely minute changes in the process, applying to individual measurements or subgroup means. The chart uses the cumulative sum of deviations from the target to chart small shifts. This allows a very high degree of control over variation, and lets companies fine-tune processes on a level other statistical tracking models cannot. An example of this is how the CUSUM chart can be calibrated to detect out-of-control processes very close to start-up in order to avoid as much capital loss as possible. However, the level of specificity enjoyed by the CUSUM chart means it cannot track larger changes as well as the typical Shewhart SPC chart, so these charts are often used in tandem.
Another popular style of SPC control chart is the Exponentially Weighted Moving Average Chart (EWMA). The differentiating factor in EWMA charts is the way they track variables or attributes using data from the entire run of a process. The EWMA chart reduces the value of data points as they grow further back form the current time, but unlike other charts it tracks all prior sample means rather than treating rational subgroups individually. The benefit of the chart is that with its unique approach to analysis, it can be an even better indicator of long-term trends than a traditional SPC chart. This is especially true when normality cannot be assumed, and the average is likely to wander from the target a great deal over shorter amounts of time. And like the CUSUM chart, the EWMA chart is often used with the Shewhart chart for a more complete picture of the variation in a factory’s production.
For a more mathematically rigorous examination of these charts, you can find some great resources here.
