Is your SPC system performing at its peak condition
or does it have a common illness?
Not too long ago I was one of the unfortunates to be infected with the H1N1 virus. I was tested-by more than one method-and the doctor verified it.
Just in case you are wondering. No, I didn’t have the vaccine. No, I didn’t use hand sanitizer as often as I should have. No, I wasn’t taking any vitamins.
So while I was fighting it off, I began to think that a little preventive maintenance might have kept me healthy or at least made the recovery easier. Being the geek that I am, thoughts of statistical process control (SPC) began to invade my feverish days of recovery. Not only was I thinking about the virus I was fighting, but I also was thinking about the types of SPC viruses and illnesses that infect quality professionals and their process control systems.
Go ahead and snicker, but I’ve met plenty of quality professionals and seen quite a few SPC systems that, in my professional opinion, have been infected.
OK, keep snickering, but how healthy is your SPC system? Has your SPC system had an exam lately?
Sure, your SPC probably looks OK at a glance. You check it out once in a while, review a few control charts, look for “bad” data and check out the operator comments. Maybe you even give it a slightly longer look during audits. You may even go so far as to verify that operators are entering their comments, assignable cause codes and corrective actions.
But is the system performing? Are operators really using the system or are they simply going through the motions? Are you seeing real improvements in the process? Is your quality and productivity stagnant or even deteriorating?
If you are not getting the performance out of your SPC system that you would like, a quick examination will help determine if there is a problem.
Let’s review some of the common SPC illnesses to see if you have any of the symptoms. And don’t worry if your system is infected. We’ll also cover the doctor’s tips for preventive maintenance and treatment.
Control Limit Virus
This sneaky little control limits virus can even infect mature SPC systems. It is commonly transmitted directly from quality management to control charts.Symptoms include the use of control limits that are based on anything other than ±3 sigma or are calculated from sigma that is not based on an average or median dispersion.
The most common strain occurs when management sets control limits based on their opinion of where the process should be instead of the reality of the process. When this virus hits, management dictates control limits in an attempt to improve or control the process without understanding that the control limits are meant to show process behavior and point out assignable causes. Control limits should never be artificially set with the hope that they make a process perform better than its actual capability.
Another mutation results in adjusting the control limits to something other than ±3 standard deviations, such as changing to ±2 sigma in an attempt to improve the process.
Still another mutation results in a fever that causes management to think that control limits should be based on specification limits. The fever erases any knowledge that specification limits are customer-based product requirements while control limits are defined by process behavior. Once this fever has taken hold, control limits are simply an internal process specification and any chance of using them for a true process improvement is lost.
This virus may lead to a terminal loss of process improvement if left untreated.
Double Limit Vision
The double limit vision virus may be transmitted from quality management, but it is most often contracted from production management. There also is a theory that the disease originated from a “corporate” infection, although there is no documented proof.The virus settles in the optic nerve creating a vision distortion that causes one to place both specification and control limits on a control chart. Using both control and specification limits on control charts results in operator confusion, statistically invalid charts, no real process improvement and, in severe cases, may lead to death of the SPC system.
The purpose of SPC is to use control limits to improve the process based on the process behavior. When control limits are utilized, we only adjust the process when there is a true statistical signal.
But when specification limits also are used, there is a strong risk that the specification will be used to adjust the process, without regard for the true process behavior. If this occurs, the process variation increases and the capability of the process will decrease.
Obsessive Compulsive Data Gathering
Symptoms of obsessive compulsive data gathering include adding check sheets or new data collection to combat any production issue. Problems occur when there is an overabundance of data, mostly in paper form, and the data ends up in storage on someone’s bookshelf instead of actually being used.Another form of this disease occurs when automated data collection methods are used. More data must be better, right? So data collection is increased to, dare I say it, even 100%. Again, there are so much data collected so rapidly, that it is not possible create a rational control chart, much less utilize the data. Worse yet, try to plot the information on a control chart using data that wasn’t collected rationally.
Remember, data becomes useless if it is too overwhelming to properly analyze.
Tamper Fever
Symptoms of tamper fever include an unnatural desire to “improve” a process that is in control. Fever induces a sense of perfection that causes the patient to believe that every product can be produced exactly on target regardless of the process capability.Those not infected with this virus know that it is not possible to make every product perfect and that tampering with a stable, in-control process increases variation and actually creates an unstable and out of control process.
Treatment
Now that we have gone over a few of the most common illnesses, let’s see what the doctor says. Dr. Walter Shewhart that is, the father of statistical quality control and the originator of the control chart.Shewhart actually made it easy for us. He noted only four foundations must be met in order to use and benefit from control charts. If all four of these foundations are maintained, one will prevent and eliminate all of the illnesses covered here.
Foundation One
Shewhart’s charts always use control limits which are set at a distance of three sigma units on either side of the central line.Treatment is to always use control limits which are set at a distance of three standard deviation units on either side of the central line with the deviation being calculated from an average or median dispersion.
The three-sigma limits must always be based on process data. Since the control limits define when an action should be taken on the process, control limits are never based on any calculation using the specification limits.
Using anything other than three-sigma control limits will lead to a control chart that is either overly sensitive or not sensitive enough to process changes. Creating charts that do not adequately separate assignable cause and common cause variation will never lead to process improvement.
Foundation Two
One must always use an average dispersion statistic or a median dispersion statistic when computing three-sigma control limits.Using the average or median dispersion increases the robustness of the chart.
An average or median dispersion statistic of process data must be used to create control limits. Calculation of control limits based on anything other than true process dispersion does not work.
Foundation Three
The conceptual foundation of Shewhart’s charts is the notion of rational sampling and subgrouping.Rational samples are taken with regard to the way the process is measured: what, where, how and when it is measured. Samples must be taken frequently enough to monitor any changes in the process but not so often that autocorrelation between samples is present. Samples should be selected with the goal of keeping the process stream intact.
After the data has been obtained rationally, it must be grouped rationally. A rational subgroup is one where there is little possibility of having assignable cause variation between samples within the subgroup itself. If only common cause variation exists within the samples, then any differences within or between the subgroups will be attributable to assignable cause variation.
Process streams should not be mixed in the subgroup. If the subgroup includes output of two or more process streams and each stream cannot be identified, the sampling is not rational.
Foundation Four
Control charts are effective only to the extent that the organization can use, in an effective manner, the knowledge gained.Adequate training and support must be provided by management.
An overabundance of charts is worthless if they are not used. It is much better to focus on key characteristics and see real process improvement.
Operators must be allowed to respond to control chart signals. Any knowledge is only as effective as the ability to take action on it.
Take a moment to perform a complete checkup of your quality system to determine if there are any issues that are putting your products at risk. If you find that the system is on life support, do not fret, getting a clean bill of health may only require a little resuscitation and TLC.
Quality Online
For more information on SPC, visitwww.qualitymag.com for the following:“Making the Case for SPC”
Q-Cast Podcast: “An Overview of SPC and Justifying a Software Expense”
“Six Steps to Shop Floor Acceptance of SPC Software”
