Review of Understanding and Applying Statistical Process Control in Manufacturing, Part 4
In Part 4 of this series on SPC, I presented the fundamental definitions for control charts, my perspectives on auditing them, and how to evaluate the capability of processes.
In the final installment of this series, I will complete our discussion on statistical process control by looking at a way to further reduce the variability of processes and further improve process capability. I will do so by briefly discussing a very important tool known as design of experiments (DOE). It is not my intent to make you an expert on DOEs, but rather to make you aware of them and how they can be used in conjunction with SPC to improve your processes.
Understanding design of experiments
What if, after working hard to implement control charts, you discover that your process capability is not acceptable? That is, your process capability index, Cpk, is less than 1.0. Having a Cpk this low tells you that the variation within your process is much too high. It indicates that if you don’t correct the problem, you will produce high levels of defective product. When this situation arises, there is a tool that can be used to significantly reduce the variation of the process in question. This tool is known as design of experiments (DOE).
While a control chart is static in nature in that it waits for a change in the process to occur and then measures the effects on variation, a DOE forces controlled change and then measures the effect. In a DOE, the response variable is measured as a function of ever-changing process factors with the response variable (in this application) being variation, usually in the form of either the range or the standard deviation.
Understanding the fundamentals of the DOE
A DOE uses statistically designed matrices (i.e., design arrays) that include the numerous factors that could affect variation. It identifies various run conditions to isolate the effects of each factor. As each run condition is implemented under controlled conditions, the variation of the response variable is measured. We then use statistical analysis tools like analysis of variance (ANOVA) to determine which of the many factors tested in our screening test design significantly impact variation.
In essence, we separate the vital few factors from the trivial many. After we have identified which of the study factors significantly impact variation, we then run a confirmatory study to validate the findings from our screening study. As a final step, we run an optimization study to determine the optimum process setting for each significant factor that results in the least amount of variation. These few significant factors are the ones we focus our attention on through the use of control charts.
DOEs are used to understand the effects of different factors that impact the output of a process. If your Cpk is less than 1.0, then you have excessive variation within your process that must be identified and reduced. If you don’t reduce your variation, the quality of your product will be unacceptable and unpredictable.
The DOE helps you isolate those few factors that must be improved if you are to produce acceptable product. If you want your control charts to be effective, then you must always be monitoring your process and react to out-of-control situations.
Coming in the next post
In my next installment, I will begin a new discussion on continuous improvement. In the meantime, if you would like to learn more about this topic, check out my recent series, A Practical Guide for Manufacturing Process Improvement.
Until next time,
Don't miss out!
Stay on top of the latest business acumen by subscribing to the Manufacturing Breakthrough blog.