Reporting Profile given multiple points

[jeannev]

As a standard rule, when given multiple inspection points taken via CMM or other method, how do you report the data to the customer? Do you report the worst case, the average or the range of data?

ie, if a profile of unilateral distribution is .006, and 9 inspection points are taken along the surface with results that range from .0016 to .0031 from nominal; do you report the maximum profile of .0062, the average profile of .0047 (averaging all 9 profile points), or the range of .0032 to .0062?

I can't seem to find a true standard, and I've always reported the worst case, but have recently found that others within my company report the average value, and some report the range of min to max. The way the profile is reported actually has an effect on whether or not the surface is reported as out of tolerance, so I want to make sure I get this right.

 

[HGMorgan]

Worst case or range is acceptable.
Range is better in most cases, stated as:
(Design Tol,I.E.; Profile sysmbol, Tol, Datums {Note Unilateral Outside/Inside}) 9 ea PT's on profile ck from +.0032 to +.0062.
Average is not acceptable, that would show your example as being with/in tolerance(.0047), when it is not(.0062).

 

[powerhound]

I agree with HGMorgan. Average is never the way to go because it does not give you the true functional picture. My opinion has always been that this is a major drawback to ISO because it allows the use of averages.

Powerhound, GDTP T-0419
Production Manager
Inventor 2009
Mastercam X3
Smartcam 11.1
SSG, U.S. Army
Taji, Iraq OIF II

 

[dingy2]

I also agree with HGMorgan that the worst case actual readings are reflected.

Dave D.

http://www.qmsi.ca

 

[powerhound]

I just re-read the OP a little more closely and noticed something. Please correct me if I'm wrong but if the profile tolerance is unilaterally disposed then I don't think you should double your numbers like you are. If the highest point from nominal is .0031 then that's what the number is. You shouldn't double it to .0062. Unilaterally disposed means the tolerance only goes one way from the true surface. Did you happen to mean equally disposed?


Powerhound, GDTP T-0419
Production Manager
Inventor 2009
Mastercam X3
Smartcam 11.1
SSG, U.S. Army
Taji, Iraq OIF II

 

[jeannev]

Yes, I actually did mean to say that the profile is bilateral distribution, not unilateral.

Thank you for all the replies which confirm my opinion that the average value does not reflect the true shape of the profile, especially given that one point is measured outside of the tolerance zone.

 

[MechEng2005]

Yes, reporting only the average would not be of any value. If I required a plate 1" thick +/-1/32, I certainly would not expect to receieve a plate that varies from 1/2" to 1-1/2". However, the average in that case would be right-on at 1".

It occured to me that perhaps you meant the average absolute deviation, in which case the above would be 1/2" and not allowable anyways. However, I would still argue that a plate with one point at 1/2" thickness would not be acceptable, even if you had 1,000 additional points at exactly 1".

-- MechEng2005

 

[chaingang]

My understanding for a bilateral distribution I take the CMM reading with the highest variation from nominal and double it to get my actual profile.
An example that I've been working on with a supplier:
I have a R10 (basic) with a profile to a line callout of 0.2 (or a +/- 0.1 band). My individual readings are: 10.101, 10.105, 10.110, 10.113. My largest deviation from nominal of 10 is .113. Based on my understanding my actual profile is .226 or .113 X 2. I'm trying to explain this to a supplier but cannot find it in any GD&T books or in ANSI Y14.5. Is there something available that I can show the supplier?

 

[PaulJackson]

I used to think that doubling the worst deviation from the basic profile was the best way too... until I was pressed by others on another forum quite a few years back to consider the best way to report all types of profile. You see doubling the worst deviation works for equal-bilateral profile tolerances but it doesn't work for ones specified un-equal bilateral or unilateral.

There was an engineer from Hewlet Packard that insisted that producers and customers should address the profile reporting issue in a policy agreement between themselves because there wasn't any standard defined method and the most popular practice didn't work for some specifications.

He said their policy agreement method was to report the actual deviation from basic plus or minus where a plus value signified the deviation in a "more material" direction from basic and minus a "less material" direction from basic... therfore a plus deviation could make an ID smaller or an OD larger.

After trying unsucessfully to defend how I thought measurements in relation to un-equal bilateral or unilateral profile specifications could be presented as a doubled value from the "equivalent equal-bilateral basic reference"... I gave up and agreed with him. Their method is simple, accurate, universal(with regard to zone depiction), informative, and statistically friendly.

There is a group of volunteers addressing this and other measurement data reporting methods in a proposed new standard... but the bread is still baking.

Paul

,