PART-TO-PART VARIATION - WHY WAIT?

Abstract

Purpose: This presentation will outline a procedure for estimating gage variation against total tolerance variation when you don't want to (or can't) wait long enough to obtain product of certain dimensions.

Quoting from the AIAG manual "Measurement Systems Analysis" on page 39:

"The sample parts must be selected from the process and represent its entire operating range." They want you to obtain product that represents the full range of product variation. They state this may take "several days". What if you can't wait that long or your process doesn't run often enough for you to obtain that full operating range? You won't see the variation you need.

One way to obtain the part-to-part variation you need is to mathematically force part-to-part variation to obtain the full range of product variation and then make your assessment. Before you can even think about doing this, you have to fully understand your measurement equipment. If you know that your gage department has established that your measurement equipment has little or no bias, is linear through the range of measurements and is stable, you can continue on. Otherwise go back and do your homework.

One other important point must be made. You should also establish that your product variation over your operating range is fairly uniform - if that could impact your gaging methods.

Once you have established that you meet the criteria above, you can conduct your MSA. If you do not or cannot obtain product that spans your product tolerance, choose one part from your analysis to "move" towards your high limit and one towards your low limit of your specification.

To do this, you take the average (all trials and all operators) of all readings for the chosen part. You then subtract the lower specification limit from the average of the part you are moving in that direction. You then take that value and subtract it from every reading for that part (once again, all trials and operators). Take the upper specification limit and subtract the average (all trials and all operators) of the part chosen to move in that direction. Then add this value to every reading for that part (once again, all trials and operators). What you end up with is part variation that spans the specification limits WITHOUT impacting the gage variation whatsoever. Your gage variation is now compared against your specification limits.

One other comment to make, we utilize Minitab for our gage studies which gives you outputs as % Contribution, % Study Variation and % Tolerance. The only value we look at is % Contribution because it is based on variance, which is a linear calculation allowing the Total Gage R&R and Part-To-Part variation to total 100%. The % Study Variation and % Tolerance are ratios based on standard deviations, which are not linear - they do not add up to 100%.

This may be controversial to many purists but it does allow one to judge gage variation against your specification without having to wait to obtain parts that span the range of the specification.

Bio

Jack Meagher is the Manufacturing Excellence Coordinator for the HiTech Division on New Hampshire Ball bearings in Peterborough. NH. His duties include coordinating HiTech's Six Sigma program and training others to become Six Sigma Green Belts. Jack is a graduate of UMass Lowell with a B.S. in Nuclear Engineering and is a Certified Six Sigma Black Belt and an ASQ Certified Quality Engineer. He is currently the Vice-Chair of the NH Chapter of ASQ.