MSA Manual 3rd Edition - R&R, Linearity, Bias, Stability - Small Company Many Gages

[Martin Rudkin]

MSA and the 3rd Edition

We are a injection moulding and metal pressing company supplying 1st tier automotive suppliers - hence being pushed down QS9k.

I am at present trying to compile an MSA procedure/process. We have many parts which are 'live' within our company and to carry our MSA on all of them would not be feasible. (time/man power)

I would like to carry out Equipment Studies(micrometers, callipers, etc) but I am unsure how to include everything our auditor has been looking for - R&R, Linearity, Bias, Stability in a manner which is efficient and complies to QS9k 3rd Ed.

I have been slogging through the MSA booklet but it does not make it any clearer.


------------------

 

[TheOtherMe]

The MSA manual is primarily aimed at repetitive measurements of a specific dimension on a specific part. Take a look at page 37. Linearity = |Slope| x Process Variation. Take your equipment and do a few studies on a few different parts. Companre the part to part results. Now you have qualification of the instrument for measuring other parts by 'similarity'. Remember you are looking at a part to part situation and so you might want to do an R&R on one part as you need a way to address process variation. Some issues should be understood as well such as the differences you will find using the same instrument (such as micrometers or calipers) on different materials (you say you injection mold and you press metal).

What do you mean by " 'live' within our company"?

 

[Martin Rudkin]

Thanks for your reply, it has provided me with ideas.

What I meant by live is that they are in current production.

The problem is that if I was to carry out an MSA study on each 'live' part, I would have to carry out more than 1000 MSA studies - not exactly my idea of fun!

Also, if possible could you explain in layman terms 'Process Variation' So far, within the MSA manual, I can only find it given as a figure but not how to calculate that figure.

 

[Roger Eastin]

Basically, the process variation is the spread of the process or, a more inexact term might be RANGE of values (for a given parameter) you expect to see from your process. For a normally-distributed process, the process variation is described by the standard deviation (sigma). This metric allows you to describe your range of values in terms of, for instance, % of values within your tolerance. If you use a control chart method to do your MSA, then calculate sigma using the R-bar/d2 method(a list of d2 values is given in chapter 2, table 2). If you are not using this method, the easiest way to calculate sigma is the button on your scientific calculator (no joke!). Otherwise, you can use a hand calculation which is located in statistics text books. This should be a last resort unless you are a purist! If your process is not normally distributed, you can use other metrics. Hopefully this description is in layman's terms!

[This message has been edited by Roger Eastin (edited 03 September 1999).]

 

[Martin Rudkin]

Thanks Roger, sorry to be a pain but going back to TheOtherMe's reply about page 37 of the MSA handbook, the process variability has been calculated as 6.0 from the measurements taken.

Do you know how they have arrived at 6.0? Is it a sum the variances from each reference size/part trial?

 

[Marc]

I can't say where they derived 6 as the process variation for the example. This is not my forte - maybe one of the others can help who has the MSA manual.

[This message has been edited by Marc Smith (edited 03 September 1999).]

 

[Marc]

The problem is that if I was to carry out an MSA study on each 'live' part, I would have to carry out more than 1000 MSA studies - not exactly my idea of fun!

It's not? And all this time I thought this stuff was fun! Umm, here's where you have to use some 'calibration systems sense'. You can classify types and/or brands of equipment, for example. You have 10 B&S 5" calipers? Qualify 1 and 'by similarity' extend the results to the 'brothers and sisters'. Uncertainty for the instrument is expressed in the manufacturers stated tolerance for the instrument.

Roger obviously knows this subject a lot better than I - Comments, Roger?

[This message has been edited by Marc Smith (edited 03 September 1999).]

 

[Roger Eastin]

Marc - you are certainly more of the expert on calibration "common sense" than I am. On the question of where the "6.00" came from on page 37 of the MSA manual: I think it came from the nominal value in Table 5 on page 36 of the instrument considered in the example. The graph on page 38 is helpful here too. I believe they are doing this because they are comparing the linearity of actual measurements to a "best line" line (incorporating bias calculations).

 

[Alan Greatbatch]

Interesting eh!!! On page35 every time process variation is written it has (or the tolerance) written immediately after it. Now to me the process tolerance on a well contolled process that the tolerances(spec limits) are well calculated would be +/-3sigma (6sigma)possibly our figure 6.
In Fig.12 at the top of the graph it states the Process variation = 6.00, maybe it is just one of the assumptions some statistician has made but I would not get too hung up it as in the calculation on page37 it states;

Linearity=[slope] x process variation

%Linearity=100x [Linearity/process variation]

So, I think we can safely say;

%Linearity =100x [slope]

Sorry,I didn't mean to sound sarcastic but some of these calcultions take an hours reading of going round in circles to find some of the figures are unnecessary or meaningless.

PS. I agree with Marc, I would always group common equipments, products, processes for any kind of qualification, approvals or studies.



------------------
Alan

 

[Marc]

Sarcasm is welcome from time to time. We can handle it! We appreciate your post.