Measurement System Analysis (MSA) - Bias, Linearity and Stability

I am trying to set a procedure for the MSA in my company. So far, we have been doing GR&R studies without including part variation and total variation as we are using tolerance only as the denominator. Is this sufficient?
I read in one article saying that Bias, Linearity and stability is not neccesary in the MSA as there is no statement in the QS 9000 saying that we must include this. I would appreciate if someone could clarify this. If it is necessary, how do we know that the results we obtain are acceptable or not.

Well, ahem, QS9 says you have to follow the MSA manual. It is not technically auditable, or so it is said. Many auditors do not really understand all the mumbo-jumbo about stability and such. Ask an auditor about uncertainty. See the stare is the most common reaction.

I'm not an MSA guru but my understanding is it doesn't matter which you use. See page 32 of the MSA manual. My understanding is that R&R is to be addressed as a per cent of tolerance.

As far as linearity, bias, and stability - I have seen auditors ask about them and expect, in some cases, to see calculations. I would bet 80%+ of the auditors will not have a good understanding of l,b,&s. In the MSA they talk about bias of a system - but the MSA discussion is vague to say the least. They talk about system bias but confuse you with instrument bias.

Part of the problem here is going to be writing a procedure to address these issues. These things are generally looked at during early APQP as part of a knowledge base. That is to say in some cases linearity may be an important study and in others it is not 'significant'. I wouldn't write a procedure - I would say something like "the need for linearity and bias and stability studies will be determined by the APQP team. Where linearity or bias or stabuility is deemed critical or important studies will be done in consenance with the AIAG MSA manual.

But again, QS says you have to follow the MSA. The thing is - do you know enough to be able to discuss? For example, uncertainty is generally given by gage manufacturers - can you show and discuss these issues?

Part of my problem comes from the way the MSA manual is written. You keep reading about 'Reference Values' and they say '...the reference values of the parts can be determined with tool room or layout inspection equipment...' which doesn't explain much. And what do you do when your measurements are all done with the layout equipment (this is an issue of gage precision)? How can you do these studies when the most sensitive gage you have (maybe a CMM) is what you use to measure everything.

The MSA manual is not for every shop. Only the auditors think it is.

[Marc]

An FYI:

>From: Pfscott2
>
>List,
>
>I'm looking for practical examples of how a small machining company (50
>people, 2 in QA) can analyze the bias, linearity, and stability of its IMTE
>as required by QS-9000 3rd edition clause 4.11.4.
>
>Before you rush to suggest reading the MSA manual, I have......well, I have
>tried. The more I read it, the more I became confused.
>
>Thanks!
>Glover Enop

Date: Mon, 6 Mar 2000 10:07:40 -0500
To: jennejohnn
From: Philip Stein
Subject: Re: Q: Measurement system bias, linearity, and stability /Enop

I have cross-posted this on the ISO 25 (Testing and calibration) listserve in the hopes that you will get a wider range of answers.

Gage R&R will indeed analyze the bias, etc. of your M&TE but it is confusing. The basic principle for measuring your measurements is called a measurement capability study. Gage R&R is a subset of this kind of study but it is a poor choice. You measure a known (nondestructively) many times. The known does not change, therefore any variation you see in the measurement must be due to the measurement system. Statistical analysis of these data will give you information about linearity and stability if you have chosen your known well and do the experiment right. Bias can only be measured by calibration to an external reference.

[Marc]

Date: Mon, 13 Mar 2000 10:09:24 -0800
From: "Dr. Howard Castrup"
To: 'Greg Gogates'
Subject: RE: Measurement system bias, linearity, and stability

I agree with Phil. A gage R&R study is a useful thing to do, but you need to be careful in evaluating the results. Typically, such a study involves k technicians making n measurements of an attribute, each using m different instruments of the same make and model. After crunching the numbers, you get standard deviations (uncertainties) for technician bias, instrument bias and repeatability. You also get mean values for each, but there is no way to determine if the mean value for instrument bias is the same as the difference between what your instruments read and the "true" values of the measurands. To get close to this number, you need traceability to higher-level standards.

Howard Castrup
Integrated Sciences Group

I always explain this to auditors as follows:

Bias, Linearity and stability are covered by our calibration programme.
If the gauge is biased, this will show when it's checked against a traceable master - and we have a tolerance stated in our calibration procedure.

We check linearity by doing calibration checks through the working range of the gauge (i.e at 10, 20, 30 and 40 Nm)so if the gauge is out at one end of the working range we will know, and we have a tolerance.

If our guage is not stable over a long period we will notice a drift from one calibration check to the next. If it is not stable in the short term, our r & r system will identify this.
This is putting everything very simply - which is usually good enough for any auditor I've met.

Cheers

Brian

[Marc]

Also see:
http://Elsmar.com/ubb/Forum4/HTML/000005.html
and
http://Elsmar.com/ubb/Forum4/HTML/000142.html
and
http://Elsmar.com/ubb/Forum4/HTML/000143.html
and
http://Elsmar.com/ubb/Forum10/HTML/000075.html

[Dawn]

The MSA book says to use parts when doing a linearity study. Does this mean we must use parts or would it not make more sense to use standards traceable to NIST/ That is what I'm currently doing.

[Marc]

Originally posted by Dawn:

The MSA book says to use parts when doing a linearity study. Does this mean we must use parts or would it not make more sense to use standards traceable to NIST/ That is what I'm currently doing.

I would stick with the standards you are using. On page 35 of the MSA manual the second paragraph starts out : "...the reference values of the parts can be determined with tool room or layout insprction equipment..." which is just saying you are going to take a number of parts and make them standards by measuring them with a piece of M&TE which is much more accurate than say a dial caliper. I think the MSA manual sorta assumes you don't have higher standards available. If you have layout equipment, like a CMM (which is, of course, properly calibrated), it has to be at least 10x more accurate so the value you get should be good enough to 'standardize' (the book calls these reference values) a number of parts. This starts to fail when your equipment becomes more sensitive. Do you think you can do this for a laser mic if the laser mic is the same sensitivity as your CMM? How do you 'standardize' the parts you plan to use, if this is the case?