There are plenty of tools to conduct investigations within manufacturing. Some can be solved with a 5-why or an Ishikawa diagram. Others may require applying regimented knowledge of using a Red X™ approach. Regardless of the tool, these investigations are destined to find the multitude of root causes and to identify ways to prevent reoccurrence. But what are those triggers?
In a recent @Op_Empathy blog post, we reflected on these investigations as investments and discussed ways to avoid endless research. I have seen plenty of organizations seek to solve problems using the tools available, and sometimes get too infatuated in investigating the magnitude of problems versus putting in actions to solve. What if there was a way to align on a process control trigger that says we will only investigate when we are out of process control. Additionally, how could we trigger proactively when we might be on the brink of losing control? Do you know who Dr. Lloyd S. Nelson is?
Dr. Lloyd S. Nelson's 1984 article in the Journal of Quality Technology focused on interpreting a Shewhart Control Chart and challenging the definition of control. His writing seems inquisitive and content, with a heightened focus from an academic perspective. However, when I read it I imagine a passionate leader and less of a data scientist striving to take action versus continuing the death spiral of staring at the data. I read his writing as if he was tired of all of the organizations arguing over the definition of control, and instead of wanting his organization to take action based upon the data.
All of us can relate, I am sure. We have teams that make opinionated decisions on what is in control and what isn’t in control. We have groups of organizations that state that a condition is good, but another person in the group expresses that it is indeed not good. I want to imagine Dr. Lloyd S. Nelson standing up in front of a team of data analysts arguing over variation in the process from their siloed perspective. I want to imagine a leader exhausted over the endless banter to slam a table and proclaiming enough is enough. I want to imagine he said to this group of individuals overdosing on data, from now on this is the definition of control. Period!
Dr. Lloyd S. Nelson made eight rules to define this border of control versus non-control. Utilizing Shewhart Control Charts, he laid out his eight examples of what is out of control. His approach layered with reactive and proactive thought can be applied as triggers to begin investigations. Whereas, these investigations are aligned on predetermined triggers and defined investigational energy to bring the process back in control.
Test 1 - One point is more than 3 standard deviations from the mean.
Test 2 - Nine (or more) points in a row are on the same side of the mean.
Test 3 - Six (or more) points in a row are continually increasing (or decreasing).
Test 4 - Fourteen (or more) points in a row alternate in direction, increasing then decreasing.
Test 5 - Two (or three) out of three points in a row are more than 2 standard deviations from the mean in the same direction.
Test 6 - Four (or five) out of five points in a row are more than 1 standard deviation from the mean in the same direction.
Test 7 - Fifteen points in a row are all within 1 standard deviation of the mean on either side of the mean.
Test 8 - Eight points in a row exist, but none within 1 standard deviation of the mean, and the points are in both directions from the mean.
Test 1 is probably the most simple of the eight, indicating that we have an event extremely outside of control as an individual event. Things like a bad test or an extremely long delay can be examples that we can easily relate to. Test 2 is an indicating that there is a shift in the mean. Consider something like the wrong grease being put into an application witnessed in the sequential temperature readings. Test 3 is the deterioration test, showing that the process is degrading quickly. Consider things such as a pump’s effectiveness or a cutting tool’s capabilities pointing towards the end of life. Test 4 is that indicator that there is noise in the process creating the variation. Examples that come to mind are pressure performance from daily cycled pumps or a different operator operating a machine daily. Tests 1 through 4 are simple and should be easy to relate to within manufacturing processes to show variability.
Tests 1, 2, 3, and 4 should be applied routinely; the combined chance of a false signal from one or more of these tests is less than one in a hundred. Nelson (1985) describes these tests as “a good set that will react to many commonly occurring special causes.”
Transitioning into Tests5 and 6 are intended to guide an early warning. They are a little more sensitive to monitoring the process versus the first four. Test 5 is an indicator showing beginning signs of an event that could become a finding of Test 1 shortly. Whereas, Test 6 is zooming into the process a little more than Test 5, whereas we are attempting to flag a situation in the very early stages of degradation of control.
Tests 7 and 8 are the rules to group data points into events. Consider Test 7 where it is indicating that there is a sequence of events where we are in a control that isn’t to be expected. Consider something like the degradation of a working roll where a certain product type, that runs recently, removes material fast from the working roll versus other material. Test 8 is indicating nothing is within the boundaries of Test 7 indicating the possibility of bad training or a bad recent installation.
A common cause affects all of the points on the chart, as when a centerline is too high. A common cause is fixed by changing the system. A special cause is fixed by removing the perturbing influence that caused the out-of-control cause.
Consider how your organization uses triggers to launch an investigation. Then consider how you could apply Nelson’s Rules to launch investigations and more importantly, avoid investigating personal interpretations of the control charts. Using some or all of Nelson’s Rules can help remove the biased approach and put a process in control that triggers the investigation. Thank you Dr. Lloyd S. Nelson for your approach, and thank you for putting together rules to promote action versus just staring at data.
Comments