# NDC (Number of Distinct Categories)

Discussion in 'Gage R&R and MSA - Measurement Systems Analysis' started by leftoverture, Apr 22, 2016.

Tags:
1. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
That is why there really should be an ndc as a function of tolerance to give the "statistical discrimination" as a function of tolerance. There really is no reason why their couldn't be...

2. ### ncwalkerWell-Known Member

Joined:
Sep 21, 2015
Messages:
216
127
Trophy Points:
42
Location:
North Carolina
That's an interesting thought. But I'm not sure how you would calculate it. % SV has PV as the denominator. % Tol has the tolerance as the denominator. NDC doesn't involve the tolerance at all. But, you could ratio the PV to the tolerance and multiply that against the "normal" NDC and get one scaled against the tolerance.

In other words, NDC in words means "Given the parts you have shown me and my gage discrimination, I can divide this stratification of parts up into X buckets." We could then say "And Y of these buckets will fit in your tolerance zone."

3. ### ncwalkerWell-Known Member

Joined:
Sep 21, 2015
Messages:
216
127
Trophy Points:
42
Location:
North Carolina
Bob, the more I think about this, the better I like it. It makes total sense. If the rule of thumb is you want your gage to be 20% of the tolerance, that would mean our new "NDCt" would need to be 5 or better. If I did a Gage R&R on an extremely accurate centerless grinding process that had a huge tolerance, I may have all my parts close together. Right at the center of the tolerance. My resulting NDC would be low (as would %SV). But I could then say "Even though my NDC was 3, 26 of these would fit in my tolerance zone, so I'm good."

In reality, it would be just another way of looking at %SV vs %Tol that wouldn't be a percentage. But (at least to me) it's way easier to visualize it.

Bev D likes this.
4. ### MinerModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
227
163
Trophy Points:
42
Location:
Greater Milwaukee USA
New metric = ndcTol

Bob Doering likes this.
5. ### Bev DModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
340
342
Trophy Points:
62
Location:
Maine
No walker. That is essentially what I do when someone can't see it in the youden plot and they need a mathematical quantification. Not sure that that the %tolerance values are equal though - I would challenge you to think about that. Since the true % of the tolerance that is consumed by the measurement error is much less when you do the math correctly than when you do it incorrectly, a 20% consumption would allow a fairly bad discrimination to pass.

6. ### ncwalkerWell-Known Member

Joined:
Sep 21, 2015
Messages:
216
127
Trophy Points:
42
Location:
North Carolina
Stop clouding the issue with logic.

That's not how automotive industry works.

7. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
Yes, you have more "statistical buckets" within your tolerance than within your process variation (or, more accurately, the amount of variation presented to the study). If you are using the gage for inspection, and not control, that should be good!

8. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
Although SPC control limits are related to process variation, I also prefer to see ndcSPC where you use UCL-LCL instead of PV, and you should have a value of at least 10 to assure adequate resolution for charting.

9. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
Use Tolerance (USL-LSL) instead of PV.

10. ### Plague DoctorMember

Joined:
Jan 19, 2016
Messages:
13
0
Trophy Points:
1
Location:
Ukraine
Hi all,

We have the same problem with "good GRR but bad ndc" when we do a study for molding purchased parts due to their low part variation.

1. What I understood from discussion above: ndc is not applicable (makes no sense) when you do %GRR to Tol?
2. If it is really so, could you point where it is stated? It would be good to have some clear statement like "When you use %GRR to Tol do not consider ndc" in MSA manual to avoid long discussions with auditors. Because in ourcase most of the auditors are not very deep into MSA, they just know that "there must be ndc >5".
3. Do you perform GRR for incoming inspection measurements for purchased parts?

Thanks!

11. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
1. ndc is calculated based on the variation of the samples you used in the study. It does not use tolerance. That is the only thing that shows that there is no relation. Want better ndc? use samples that have as much variance as your tolerance.
2. There is no "clear statement" in the AIAG MSA book that you are looking for. If you want to avoid the conversation with your auditor, use samples in your gage R&R study that represent the entire tolerance range. If that is less than 5, you really do need another gage.
3. Yes. It is the only way to prove the gage is the right one for the job.

Plague Doctor likes this.
12. ### Plague DoctorMember

Joined:
Jan 19, 2016
Messages:
13
0
Trophy Points:
1
Location:
Ukraine
Thanks for the answer.
For our processes, yes, we take the samples covering the entire tolerance range. We do %GRR to TV and we have good GRR and good ndc.
For purchased parts we cannot take the samples with whole tolerance range for obvious reasons. We can choose only from parts we are delivered (and e.g. for plastic molding parts the dimensions are pretty stable, often covering only 1/10 of the tolerance or less within sample group.) That's why we do GRR to Tol here. we always get excellent GRR (<5%) and very bad ndc (1-2).
The resolution of the gage is 0,001 and the tolerance is 0,1. I don't see that we need better gage here. The measurements accuracy by operators IMHO is acceptable. (GRR is the proof)
My main question will be:
In you opinion will it be Ok if on auditor's question "show me MSA for incoming inspection check" we will provide GRR to Tol without ndc value.
On predictable question: "Why no ndc?" the answer will be "the only purpose for incoming inspection check is to define whether part is good or bad so just GRR is enough"

If it's not Ok what would be your suggestions? Please no offers for better gage for described above and many other reasons

13. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
Try calculating the ndc using tolerance (TV= (USL-LSL)/6) instead of the PV from the samples. If that is >5 then you may be able to justify the data by stating as incoming material you do not have access to samples that represent to total ongoing expected process variance, and the tolerance is your best estimate.

14. ### ncwalkerWell-Known Member

Joined:
Sep 21, 2015
Messages:
216
127
Trophy Points:
42
Location:
North Carolina
From the MSA 4th Edition, page 77:

"For measurement systems whose purpose is to analyze a process, a
general rule of thumb for measurement system acceptability is as
follows:
• Under 10 percent error – generally considered to be an acceptable
measurement system.
• 10 percent to 30 percent error – may be acceptable based upon
importance of application, cost of measurement device, cost of
repair, etc.
• Over 30 percent – considered to be not acceptable – every effort
should be made to improve the measurement system.
Further, the number of distinct categories (ndc)32 the process can be
divided into by the measurement system ought to be greater than or
equal to 5.
The final acceptance of a measurement system should not come
down to a single set of indices
. The long-term performance of the
measurement system should also be reviewed using graphical
analyses over time."​

You are not analyzing a process. You are checking parts against a tolerance. Also, these are rules of thumb. Also, it says right in the book you can't say yea/nay on a set of indices. Something EVERYONE forgets. Everyone wants the magic bullet, that's not how this works. You need lots of things to consider. And further, you learn a lot more from the graphs than the numbers.

From the MSA 4th Ed p. 116

"The results of this percent total variation needs to be evaluated to
determine if the measurement system is acceptable for its intended
application.
If the analysis is based on the tolerance instead of the process
variation
, then the Gage Repeatability and Reproducibility Report
form (Figure 25) can be modified so that the right-hand side of the
form represents the percent of tolerance instead of percent of total
variation.
In that case, %EV, %AV, %GRR and %PV are calculated
Property of General Motors Internal Use Only
Chapter III – Section B
Variable Measurement System Study – Guidelines
117
by substituting the value of tolerance divided by six in the
denominator of the calculations in place of the total variation (TV).
Either or both approaches can be taken depending on the
intended use of the measurement system and the desires of
the customer.
The final step in the numerical analysis is to determine the number of
distinct categories that can be reliably distinguished by the measurement
system. This is the number of non-overlapping 97% confidence intervals
that will span the expected product variation."​

I can't copy the formatting, but the blurb about % of tolerance is inset and flagged with an important point. What it is saying is that you report % Tol numbers (%EV, %AV, %GRR, and %PV) when you are querying a measurement system for suitability in checking against a tolerance.

You may get hit with this parargraph, earlier in the document p. 39

"The long-standing tradition of reporting measurement error only as a
percent of tolerance is inadequate for the challenges of the marketplace
that emphasize strategic and continuous process improvement
. As
processes change and improve, a measurement system must be reevaluated
for its intended purpose. It is essential for the organization
(management, measurement planner, production operator, and quality
analyst) to understand the purpose of measurement and apply the
appropriate evaluation.
"​

Which almost reads like %Tolerance isn't enough. But the context of this is in a continuous improvement environment. What they are saying is as you pursue your continuous improvement, you may need to improve your gaging to match it. If you don't, you won't KNOW if you CI efforts are working. Which is why they have added the final sentence. Doing your calculations on the incoming parts using %Tol and ignoring study variation and NDC is just fine. And your logic above is sound.

15. ### Bob DoeringMember

Joined:
Jul 30, 2015
Messages:
36
24
Trophy Points:
7
You are correct - if you were the manufacturer - and not the purchaser - of a product, then you may have more of an interest in using PV instead of tolerance to have enough resolution to detect shifts in your process or aid continuous improvement. As a purchaser, there is so much sampling error from product on a dock that it is IMPOSSIBLE to detect the original process distribution to even know what statistics are applicable. That is easy to prove, but beyond the scope of this topic.

16. ### karthikshastriNew Member

Joined:
Jul 20, 2017
Messages:
1
0
Trophy Points:
1
Hi
I am also facing a similar problem my MSA Rand R % is less then 10% but my NDC not more then 5
We use a 0.01 vernier caliper
I am also attaching my MSA report if any one can help me find whats going wrong
We make Plastic parts from injection moulding machines

We are using one piece from every hour out of 10 hours in a shift

File size:
129.1 KB
Views:
5
17. ### Atul KhandekarAdministratorStaff Member

Joined:
Jul 24, 2015
Messages:
239
165
Trophy Points:
42
Location:
Pune, India
For the data on the first sheet of your file, the total range of all the measurements is just 0.07 (Max=277.31, Min=277.24) against tolerance of 2 (=277 +/- 1). Since PV is too low, you will have low nDC (refer nDC formula). You need to select parts in such way as to span the entire tolerance range or at least the expected process variation. If there is not much variation in the parts produced over 10 hours, you may want to consider picking parts over a longer period.

OTOH, you will not get large/sufficient part variation if you process is highly capable. Are you using this measurement system for process control or is it used for final inspection ?

18. ### johnnymo77Member

Joined:
Sep 1, 2015
Messages:
29
7
Trophy Points:
2
I'm still learning about GRRs.

The training material I have from Ford specifically says this about GRRs:
When a process is highly capable and percent tolerance is used instead of percent study the criteria to have 5 distinct categories and 50% of the parts outside the blind spot of the gauge are not applicable.

Highly capable is described as the manufacturing process is stable, in control, parts pulled represent the entire range of expected variation, and Ppk>=2.5. (Ford underlines this phrase.)