1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

CorrectSPC Gage Resolution Required for SPC

Inadequate gage resolution will destroy your SPC efforts!

  1. Bob Doering

    Bob Doering Member

    Jul 30, 2015
    Likes Received:
    Trophy Points:
    Bob Doering submitted a new resource:

    Gage Resolution Required for SPC - Inadequate gage resolution will destry your SPC efforts!

    Read more about this resource...
    Andy Nichols likes this.
  2. Gejmet

    Gejmet Member

    Jul 23, 2019
    Likes Received:
    Trophy Points:
    Thanks for sharing the above, here is an insight of my own.

    When trying to answer the question of whether a measurement process is adequate to be used to study process behaviour we have to think and have some knowledge about how the process behaviour chart is constructed. The point of the chart is to filter the noise (the everyday random variation) from the excessive special cause variation, to this end when a part is measured it contains many different sources of variation, many interactto create random variation and a few interact and dominate to appear as signals on the process behaviour chart.

    If the measurement process is consistent then its data will be homogenous and we can go on to characterise aspects which are of interest depending on what is of interest to us. It follows that if we see signals of excessive variation on a product process behaviour chart we can be pretty sure that the measurement process is adequate because the control limits already contain measurement variation, the purists might ask about bias but this is always relative and we can if we wish assess whether any detectible bias exists during inital studies.

    So, for the assessment of adequacy of the measurement process to track the process the 10:1 rule of thumb doesnt apply, its merely superstition. Ndc is interesting, it appeared in the late 1980's from a source where it was called the Classification Ratio, it was intended as a value to try to inform engineers about the adequacy of the X bar Chart following an EMP study.

    It was never initially intended to be used to depict the number of categories in the way its presented today and certainly the inclusion of target numbers is, as I described for the resolution, not based upon anything usefull or proven. If you require parts to be sorted in the inspection facility and grouped together then this is the most use you could glean from this number, but why would you want to sort when you can work on getting a uniform manufacturing process?

    If you want a carefully and completely proven empirical means of characterising measurement process usefullness use Dr Wheeler's Intraclass Correlation Coefficient (ICC). Anything else will tend to naturally over-emphasise to a large extent the effect of measurement variation upon a production process, this has lead to a lot of wasted time and cost when we could be getting on with maintaining and improving manufacturing processes instead of trying to find perfect measurement processes.

    You can find the ICC naturally residing within the data structure for a standard GRR study which demonstrates a consistent measurement process. Take the Product Variance and divide by the Total Variance, the answer will be the ICC. Compare this with the Ndc for a few studies and reach your own conclusions.

    More details for ICC can be found in Quality Digest under Don Wheeler's contributions and in his book EMP III.