posted 29 September 1998 02:08 PM               

You have started stability tests using a ãregularä Xbar chart. However, you did not mention a chart for dispersion (R, s, sigma, etc.)?? Assuming a dispersion charts exists, then the calculation of the limits is straightforward for a ãregularä control chart. If a chart for dispersion does not exist, that is another matter.

Control limits do not give an accurate range of the readings. Rather, the purpose of control limits is to determine when a process is in or out of a state of statistical control. A process can be in control and still not be stable. Perhaps you are referring to process capability? It sounds, however, that what you are trying to do is not typical. Perhaps I am still missing something. Additional details may be required.

There are many SPC programs, but I prefer Statview with the QC module installed.

posted 04 October 1998 09:18 PM               

Be sure to discriminate the difference between ãstabilityä and ãstatistical process control.ä I would suggest that your customer does not know the difference, either. Rather than stability tests, would not repeatability and reliability (R&R) tests be more suitable? Maybe, Maybe not.

posted 15 December 1998 11:46 AM               

First, I must offer a very humble apology. My statement above: "A process can be in control and still not be stable" is not entirely correct. This goes back to my interpretation of SQC theory. "A process can be in control and still not be capable" is a more precise statement. While I agree (in part) with the definition supplied from the SPC reference manual you mentioned, there are still issues with the term "stability" I feel should raised.

While some prefer to use stability to refer to a process that is in a state of statistical control, it is not exactly precise to do so. When using SQC techniques, the ROOTS of the statistical theory should be remembered (all too often, the hammer gets forgotten in the process of driving a nail, that is, until it strikes your thumb).

We expect "almost all" points which represent samplings from a 'stable process' to fall inside 3-sigma limits. If they do not, we say "The process is not in control" and there is only a small risk of the conclusion being incorrect. If they do all fall inside, we say "The process appears to be in statistical control" or "The process appears stable"; which is not the same as saying "It IS stable." There is an analogy. A person is accused of a crime: The evidence may (1) convict the defendant of guilt, and we realize there is some small chance that justice miscarried, or (2) fail to convict, but this is not the same as believing in the person's innocence." From Process Quality Control, Ott and Schilling, 1990. This is, in a nutshell, a description of the null hypothesis.

IMHO, a more precise definition that should have been supplied in the SPC reference manual would be:

Stability: When a process appears to be in a state of statistical control and there is a lack of evidence of special cause variation.

I do not have, or have access to, the SPC manual mentioned. What is its source?

As you can probably tell by now, I prefer the term 'stability' not be used (EVER) to describe a process, or in this case, gage. Going back to Dawn's original post, all the stability tests ever conducted would not tell when a process is capable. The micrometers in question, for example. By 'fouling' the micrometer, I could present data set after data set that showed it to be very stable, but it would be 1" (made up number) OFF in accuracy. Would not R&R, MSA, process capability or Taguchi's Loss Function be a more suitable statement of the gage's fitness for use? When used correctly, SQC is a very powerful instrument. Of course, the inverse is also true.

Anyway, just my $0.02 worth.

posted 15 December 1998 07:43 PM               

Are they not the same? The theories are.

After all, a process creates a product. The stability of the gage produces a product. The product is the data in question.

Reference ISO 9000:

3.1 Product: Result of activities or processes.

From the notes: A product can be tangible (e.g., assemblies or processed material) or intangible (e.g., knowledge or concepts), or a combination thereof. Additional discussion is available from the additional notes of ISO 9000 and ISO 8402. I would think that the QS definition is similar. Maybe, maybe not.

I humbly suggest that whenever a gage is subjected to the ÎSo-Calledâ stability requirement, ask this question: ãWhat are you looking for? Stability or accuracy?ä

Reference my posts above (excluding the errors mentioned).