Manufacturing Breakthrough Blog
Tuesday February 24, 2015
As I finished my last posting, When Variation Is the Enemy, I told you we would discuss what the concept of being “in control” means and why it is so important. Let’s review briefly the basics of Six Sigma before we tackle what being in control really means.
The basics of Six Sigma
Six Sigma has become well known by the use of the acronym DMAIC. As can be seen in the above graphic, DMAIC stands for Define – Measure – Analyze – Improve – Control. What this acronym is telling us is that once we have defined a problem we must measure and analyze in order to develop a solution. And once we’ve implemented the solution, we must develop a way to control the process to prevent it from producing the same problem we originally identified. So let’s focus on Step 5, Control.
A real life example
In When Variation Is the Enemy, we discussed two different types of variation: common cause and special cause. We said that common cause variation is the natural variation that exists within all processes while special cause variation is not natural. So what’s the difference?
Common cause variation is typically characterized as being predictable variation within a process whereas special cause is highly unpredictable and comes into our processes as a result of a change that we probably don’t know about. For example, suppose we are cutting extruded plastic or rubber to a specific width. The extrusion passes through a guide as it is extruded and trimmed. As we take samples and measure the width of each sample, they are very close to each other in width, but are not exactly the same. The pattern of data forms a distinct pattern, typically in the form of a bell-shaped curve. But let’s say that one of the guides becomes loose—what happens to the data then? The width data changes dramatically with wide amounts of dispersion. In other words there is a clear shift in the variation and would be traceable back to the point when the guide became loose. This is the essence of special cause variation in that prior to the guide becoming loose, the data was very predictable, but after it became loose, the data is highly variable and unpredictable. Thus the central concept of being “in control” is that processes have no active special cause variation present and are therefore predictable. Obviously this is the preferred state that we want our processes to exhibit.
So how do we know that our process is in control or not? Is there a tool to tell us when our process moves from being in control and predictable to a state of unpredictability? The answer is, yes and the tool is referred to as a control chart. A control chart in its most basic form is a graph used to study how a process changes over time with data being plotted in the order in which they were collected. A control chart always has a central line depicting the average value with upper and lower control limits depicting the common cause variation limits. These limits are determined from historical data after all special cause variation has been eliminated from the process. So by comparing current data to these limits, you can draw conclusions about whether the process variation is consistent and predictable (i.e. in control) or is unpredictable (out of control, affected by special causes of variation).
Using the Control Chart
As such, the control chart has two parts, one for the process average (the top chart) and one for variation (the bottom chart usually depicted by the range of the data) as seen in the following graphic. Control charts for variable data are typically used in pairs or in subgroups. The top chart monitors the average, or the centering of the distribution of data from the process while the bottom chart monitors the variation. Some compare a control chart to a target in target practice. In doing so, the average is where the shots are clustering together while the range is how tightly they are clustered.
Both of these charts demonstrate a state of statistical control with all of the data points falling inside the calculated control limits. Going back to our extruded width example, this data is before the loose guide problem surfaced. Somewhere around data points 19 and 20, where the data points went above the upper control limit, the guide became loose and we see a shift in our X-bar chart in the following figure.
The control chart allowed us to see the change almost immediately, reset the guide and bring the data back into control. In this example, the process should have been stopped immediately after the first data point went “out
-of -control” but for whatever reason it continued to run, producing product that was too wide. Assuming that the control limits were well inside the specification limits, there may be a re-work opportunity.
Let’s go back to our R Chart for Mean Width:
In my next posting we’ll continue our discussion on various tools and techniques that can be used to produce better product as well as how the Theory of Constraints, Lean and Six Sigma can be successfully merged into a very powerful improvement strategy.
Thanks for reading. See you next time.