Can we use a measurement system with R&R% > 30% ? Paper Attached
I think that Mis-Conception in Calibration and MSA, led to huge loss in many organizations. In this paper we wanted to look at the measurement error from a dfferent eyelet and more practical.
Please read the attached file. it is typed in MS word 2000. If you have any problem to open the attached file, please inform me.
Some say over 10% is too much, some say over 20% is too much, some say over 30% is too much. All WRONG answers. Measurement error always increases the standard deviation of a sample of a measured parameter. It also creates a region of "uncertainty" at the specification limits - results in this gray area can not be confidently classified as "compliant" or "non-compliant." If you process variance is very good and appropriately centered (i.e. great process capability) a larger measurement error can actually be tolerated....that's if your making specification assessments. If you are using the measurement system to SPC chart a process with very little variance then the measurement system may cause a false "out-of-control" call. Basically what I'm saying...forget the goofy "rule of thumb" statements about measurement error and learn it's applicable impact.
Key Words
Measurement System Analysis, Uncertainty, Rule of ten or Rule of thumb
__________________
A Search is a terrible thing to waste! One Test is Worth 1000 Expert Opinions - The plural of anecdote is not data.
We can't solve problems by using the same kind of thinking we used when we created them. - Unknown
M. Seiedi, Welcome to the Cove!! This is the first time, to my knowledge, that someone with < 5 posts on the Cove has presented an article. Thanks for sharing a very informative piece.
K. Maruska Welcome to you as well!!! I think that M. Seiedi agrees with you, but there are situations where customers dictate a 10% rule, so we have to know the characteristics and limitations of such a rule, no? This paper goes a long way towards doing this, IMO.
Please, lets continue a discussion of this paper. Surely other members of the Cove have an opinion?
Last edited by Craig H.; 20th April 2005 at 09:11 AM.
Although there are problems in translation and use of terminology with the paper that make it difficult to evaluate (e.g., the authors seem to be confusing uncertainty with error,("In some standards such as Automotive Industry, acceptance criteria for a measurement system is proportion of uncertainty to tolerance interval") there is some useful information to be gleaned from it. MSA is easily the most widely misunderstood of all of the general automotive requirements. For example, I frequently see average-and-range reports with gross UCLr errors but R&R values <10 which are considered, apparently, unconditionally acceptable.
The authors also seem to have overlooked the importance of evaluating the components of the R&R value although they do say that their modified scheme "...is not a resolution of root cause." When the R&R number gets high, I want to know what the significant contributing factor is; it's almost always biased towards either operator error or device error, rather than a combination of the two (part-to-part variation ususally isn't a significant factor).
There is little doubt that the 10% threshold creates a false sense of security and causes people to proceed when there's good reason to stop and evaluate things. The relationship between capability studies and MSA is also vastly underappreciated; frequently I see a two or three standard deviation difference between the means of the two studies, and the submitters were blissfully oblivious to the fact that something was amiss (or the numbers were fudged).
R&R numbers have a tendency to decrease in importance in proportion to the width of the tolerance spread. As a practical matter, when tolerances are tight and measurement readings approach the limits, there is usually some redundant method (either in operators or type of devices, or both) that is used for confirmation, regardless of what the R&R study might have said. It's wrong to assume that production readings at or near the tolerance limits are accepted out-of-hand because of MSA results.
Customer requirements are what they are. Customers who insist on R&R numbers <11% regardless of the application are unknowingly making self-fulfilling prophecies. If the requirement is 10% or less, the results will meet the requirement, one way or another. I ask the suppliers I deal with to A) have an understanding of what they're doing and why it's considered important, and B) tell me when they think they have a better way, or when they believe there's good reason to modify the default requirements. And I always listen.
__________________
Some men are born mediocre, some men achieve mediocrity, and some men have mediocrity thrust upon them.-- Joseph Heller
Last edited by Jim Wynne; 19th April 2005 at 10:41 AM.
Can we use a measurement system with R&R% > 30% ? Paper Attached
30% Measurement Error CAN be OK!!
Application of unconventional - high variation measurement systems
Linear Mode
