Process Improvement with Control Charts
Statistical process control (SPC) charts were developed by Walter Shewhart and introduced to the world in his 1931 book "Economic Control of Quality of Manufactured Product". The purpose of control charts is to identify when an assignable cause of variation has entered your process stream.
Many of our manufacturing processes produce defects as a result of excessive variation. A goal of process improvement is to reduce variation and make incapable processes capable, as depicted below.
All processes have many sources of variation acting on them. When we measure products coming off our processes, these measurements vary. There is variation in raw material quantities and properties; variation in process run conditions as well as environmental changes. All of these interact and contribute to the variation in the measured output. Even though there are many sources of variation, we categorize them into one of two: assignable and random.
Assignable causes of variation are not inherent to the process. These sources act on the process periodically and cause shifts, trends, outliers and other nonrandom effects. When data indicate that an assignable source of variation has acted on your process, it is worth time to investigate.
Random sources of variation are always acting on your process. Even in the short term, consecutive parts coming off a production process still exhibit variation. When all you have are random sources, you will have a stable and predictable (within limits) output.
The image below shows two processes; one with random and assignable sources and the other with just random sources. The depiction of these two stages of process maturity helps to understand how to improve a process.
Detecting the assignable sources can be a challenge. This is where control charts are valuable - they help to identify when an assignable source has appeared. In the control chart below, an event happened at point 34 that caused a shift downward of 1.5 standard deviations. This event was identified by point 40. You won't always detect these things right away.
Points that fall outside the control limits are worthy of some investigation because they are likely to have occurred due to the influence of an assignable source. Another pattern that is unlikely is seven or more data points in a row on the same side of the centerline. The chance of being on one side is 1 in 2. Same as tossing a coin and getting heads. But when that happens seven times in a row, that seventh point is improbable, (1/2)^7 = 1/128, it happens less than 1% of the time.
In the control chart shown below we see highlights of two event types. The orange points occur after 7 (or more) points in a row on one side of the centerline. The red point indicates a data point outside the ± 3 sigma limits. The combination of these points shown below is insightful to the process owner who must try to determine the root cause of the process shift. Control charts are a valuable part of this discovery.
About the Blogger
I am an ASQ Master Black Belt with a 30 year career in manufacturing. I have managed many projects in industry and have been successful with solving complex problems through the use of analytics.
Through my business, Belfield Consulting & Training, I provide consulting and training services to manufacturers and can help your team develop their analytical skills.
Problem solving is my business and I am continually developing these skills.
Note: the control chart was created using the "qcc" package in R. This library was built by Luca Scrucca and is very easy to use. You input your data and it plots out a control chart with limits and also highlights the data points that are potentially meaningful as described above.
Scrucca, L. (2004). qcc: an R package for quality control charting and statistical process control. R News 4/1, 11-17.