Position Bonus/MMC, Cpk, and Inspection Results, Nominal Geometry 4

[ptruitt]

I am interested in using statistical tolerancing on upcoming projects, but I am not sure what the Cpk will be based on if the 'size' nominal is not the same as the 'form' nominal if I use bonus tolerance. Is there a formula that averages the two or weighs them? Is it correct to say that 'form' is the area between the Virtual and Resultant conditions?

Since more and more software is using the CAD model for machining and inspection, should nominal CAD geometry be established at the nominal form geometry or at the nominal size geometry when using the MMC modifier on holes and similar features such as slots?

As others have said on another thread, 'position' information is often not reported on inspection reports. But if it is, is there a yes/no indication that the 'size' measured was allowed given the measured 'position'?

I prefer using profile call-outs whenever possible to keep my nominal size and form the same, but some of my suppliers are not comfortable with 'profile', so I need to figure out how MMC works with statistical tolerancing.


Peter Truitt
Minnesota

 

[pmarc]

Peter,
Have you tried this as an introduction to the topic?

http://www.tec-ease.com/gdt-tips-view.php?q=95

http://www.tec-ease.com/gdt-tips-view.php?q=172

http://www.tec-ease.com/gdt-tips-view.php?q=180

 

[ptruitt]

Thanks pmarc! I'll dig into this. I don't get into statistics too much, but it looks like I should take the medicine even if it gives my poor brain a headache.

Peter Truitt
Minnesota

 

[ptruitt]

I found the tec-ease tips helpful and interesting. Thanks, pmarc. Now I am wondering how to use the data to determine the overall process capability or to determine what needs correction. Repeatability of position and repeatability of size are nearly useless if the data does not demonstrate that the feature met both position and size simultaneously. The 'single limit control' concept for position Cpk seems to be fuzzy data relating one physical boundary (virtual condition) to a theoretical axis. I am wondering how I could determine that parts are meeting spec' given my supplier's Cpk data. Does anyone know how to make the best use of this information?

Peter Truitt
Minnesota

 

[Belanger]

If the MMC modifier is imposed, then the "fuzzy data" method of looking at the virtual condition is really the only way to go if you want statistical tracking. Otherwise you will have a multivariable problem (how can you do statistical tracking on goalposts that are constantly in flux due to bonus tolerance?).

In my opinion, the whole notion of statistics and capability studies flies in the face of the philosophy of the MMC modifier. If you use the "M" you are saying that slop is OK, right? So why the requirement to monitor statistics, as if it were a critical fit? Just my two cents...

John-Paul Belanger
Certified Sr. GD&T Professional
Geometric Learning Systems

 

[PaulJackson]

Peter,

See this presentation.

Paul

 

[dingy2]

Capability on positional tolerances at MMC are not practical since the virtual condition boundary is now paramount with the 2009 edition. The only practical way to confirm whether a feature of size is not outside the virtual condition boundary is the use of checking fixtures rather than a CMM. This is attribute data and not variable.

One cannot figure out the bonus tolerance and then apply it to the centre of the feature in addition to the tolerance stated in the FCF as we once did. Most of the time when a company was forced into providing capability studies on positional tolerances at MMC on features of size, it was not done correctly. The data of faked just to achieve a Ppk of 1.67. Some tried to convert a diametrical tolerance zone into coordinate tolerances but that was just wrong. The North American automotive suppliers spent a lot of $$ for showboating with no value added.





Dave D.

http://www.qmsi.ca

 

[PaulJackson]

Peter,
I don't know your what drives your desire to statistically tolerance.
Are your tolerances too small for your supplier to achieve the PPM defective levels expected "with specified limits"?
Are your suppliers disregarding the "variable portion of tolerance" in their capability predictions?
If you can get your suppliers data you can plug it into one of these two spreadsheets (PpkMMC if only size and Ø position deviation data) or (PpkMMCXY if size and coordinate deviation data). The later is more powerful since you can examine potential capability based on mean shifts to both the feature size and the independent coordinates given their demonstrated sample variations.
If you are considering specifying statistical limits with their targeted conformance ppm defective levels in a negotiated resolution with a supplier... you can use these spreadsheets to observe their potential capabilities with respect to the variable limit tolerances.

Dave,
We have been over this many times... I disagree! Enabling reliable prediction is very powerful economically. The difference between continuous and discreet variation control and conformance prediction is day and night.

 

[PaulJackson]

Here is the other spreadsheet

 

[ptruitt]

I'm not sure whether I should thank you or curse you, Paul. (Actually: Thank you very much.) My poor brain might just explode, though, trying to ingest this.

My concern is mostly that I should be able to audit a supplier and know when the supplier has a solid understanding of design requirements, proper process control, and appropriate inspection and data collection. (I am a designer, but I have significant input into QC activities, where I work.) Also, I struggle to apply tolerances that are appropriate to my designs, I need feedback from the QC data to improve my designs.

I have been looking into software that compares the CAD model to point clouds and reports how well a part conforms to the design. (SmartProfile, GeoMagic, PolyWorks, VeriSurf) I am guessing that they would somehow quantify HOW WELL a feature with bonus tolerance met the spec'. It would be interesting to see if all those software packages evaluated the feature the same way. Regardless of the answer to that, how can any of them claim that it is possible to give a number? (Maybe your presentation will clarify that, Paul.)

Peter Truitt
Minnesota

 

[pmarc]

Paul,

If I can comment your presentation, I see one - IMO very significant - simplification in your approach. Your data seems not to include effect of perpendicularity error of an axis of a hole relative to planar datum A. On page#5 you are showing hole axis as a point, which for me automatically means that you assumed it to be perfectly perpendicular to datum A. Do you think it is always true? Then on page#7 you are presenting X,Y deviations as a single pair of numbers for each sample, which in fact won't be a case if axes of the holes are out of squarness relative to A.

What I am trying to say is that if I had to do any statistical calculations, predictions or whatsoever on position at MMC (regardless if this is practical or not), I would rather invoke boundary (virtual condition) interpretation into play. Just like Don described in one of his tips:

http://www.tec-ease.com/gdt-tips-view.php?q=95

In the most comfortable situation (time resources) I would hit some points of hole surface at one level along 'z' axis the part, then repeat it on one or two levels more and try to check how far the points are from theoretically exact position of the hole. The smallest of the distances would be a size of related actual mating envelope (RAME) of the hole. Radial difference between RAME size and virtual condition size would tell me how far I was from the spec. limit. Repeating the probing routine on bigger population of parts (holes) would give me continuous data that I could easily work with.

 

[dingy2]

pmarc

Good analysis relative to perpendicularity and remember that the round hole may not be perfectly round which would affect the results. If the hole is long then one should also measure the centre distance to confirm that straightness is not a factor so, in fact, one should measure the top, middle and bottom of the angle.

Some people who tried capability where there was a diametrical tolerance zone of 0.5 mm or .020" just converted it to a coordinate tolerance of +/- 0.25 or +/- .010 which is wrong.

I have had this discussion many times with Paul that I came out with some thoughts at

http://www.qmsi.ca/capabilityindex.html

Dave D.

http://www.qmsi.ca

 

[PaulJackson]

pmarc, dingy2,

Your objection to the analysis procedure is… that it does not confirm that the maximum material boundary is not violated when only a sampling of the feature’s surface is used to generate data about its size and location… specifically disregarding its possible orientation deviation. Did I get that right?

So what!

Say that I have to demonstrate that my processes are capable to 1.67 Cpk before I am able to sell any product. If I use attribute gages I would have to demonstrate that less than 1 part in a million defective.
I could say that, as I was producing those parts that I am hopefully confident that my process is…
1.0 capable if it had no defects in 740.8 parts
1.33 capable if - no defects in 31574.39 pieces
1.67 capable if - no defects in 3488555.79 pieces
Of course I would probably need another instance of a defect at a similar interval to establish some marginal confidence that my prediction is valid.

My point is demonstrating predictive conformance with attribute gages is counterproductive and very expensive in terms of product required, gage purchase, gage maintenance, and scrap if defects are detected with moderate sample frequencies.

Let me say that…
All inspections are estimations and all inspections make assumptions to some extent… even the hard attribute gages aside (I refuse to call them functional until I verify that the tolerance modifiers were indeed specified according to the feature’s function).

With hard gages there are all sorts of problems that I have witnessed but a few that come to mind are details that enter the features before datum surfaces are contacted, incomplete pattern engagement, non-unified datum mobility allowance among feature patterns and most common worn gages.

With layout or dedicated bench variables inspection the assumptions are numerous and to the untrained, inexperienced inspector their application is seldom noticed. A trained inspector can examine the process, the fixturing, the sequence, the tooling method, and all other details, identify its strengths and vulnerabilities and adjust his assumptions to aggressively scrutinize vulnerabilities and abbreviate scrutiny of its strengths. He is aware of jaw squeeze, closely related features accomplished in separate operations, etc. To most others, engineer’s, designer’s, and the like the assumptions they employ when they inspect are transparent… ignorance is bliss.

With coordinate measuring machines in the hands of a programmer who is inexperienced and has never done layout inspection the assumptions are legion! Dangerous I might add. Does anyone question form when reading the size on a CMM inspection report? Only those aware of how many points were gathered, how may sample planes chosen, what algorithms were used… not to mention what should and should not be aggressively scrutinized or abbreviated according to the process vulnerabilities. Once again in the hands of an inexperienced or untrained inspector like engineer’s, designer’s, and the like… ignorance is bliss.

So all of these inspection methods have their weaknesses including the hard attribute gages… If I was checking a bore and I suspected that it was particularly vulnerable to an orientation deviation according to the observed process and I wanted always to use its “related actual mating envelope” in my variable limit position tolerance capability calculations… and I had not yet programmed that routine in my software… I just might move reference to one end of hole rotate orthogonal to the hole’s specified orientation, collect three eight point circles top, middle, and bottom, then use all 24 point’s X and Y values stripping off the Z (depth) values to figure a 2D maximum inscribed circle size and location to use in my capability equation.

The point is that it is not the tool that is the problem it is the unmerited assumptions used in the inspection process, the lack of attention to sources of process variation, and the lack of understanding of the tools available that make for poor process prediction and subsequent optimization.

You go ahead and use your attribute gages… I want to know what variation I have so I can either ignore it or fix it so that I can ignore it or continuously manage it for my customer’s benefit.

Paul

BTW… It is better to use both the related actual mating envelope size and its position separately in the continuous data capability equation rather than combining them in the resultant condition and comparing them to the virtual condition as is suggested in the Tec-Ease example. They are independent sources of variation and should be analyzed as such.

 

[dingy2]

Orientation could be two fold - oriented to the primary plane and also, if the feature of size is not round such as slots, rotational orientation. The last one is extremely difficult to achieve variable data while orientation (usually perpendicularity) to the primary datum is probable as long as one takes measurements at the top and bottom of the hole using a radial tolerance and virtual condition boundaries. Always report the worst situation.

Most secondary and tertiary datums are holes and if those holes are referenced at MMC, achieving variable data on the features may not be 100% valid. We could have a pattern of slots at MMC referencing secondary and tertiary datum holes also at MMC. Love to see the variable data in that situation.

If one wanted to proceed with a capability study of the centre of a hole using coordinate tolerancing, no problem but positional with a diametrical tolerance zone at MMC referencing datums at MMC, the data collected may be suspect.

Dave D.

http://www.qmsi.ca

 

[pmarc]

Paul,

You go ahead and use your attribute gages... I want to know what variation I have so I can either ignore it or fix it so that I can ignore it or continuously manage it for my customer's benefit.

I did not say anywhere in my previous post that I would like to use hard (functional) gages or see them as the only reasonable solution in checking position at MMC. I will keep insisting that method presented by Don Day is heading into good direction. By defining proper measurement routine CMM can be utilized quite nicely to gather all necessary continuous data for further capability study. It would require some work, but IMO it is absolutely possible.
-------

BTW... It is better to use both the related actual mating envelope size and its position separately in the continuous data capability equation rather than combining them in the resultant condition and comparing them to the virtual condition as is suggested in the Tec-Ease example. They are independent sources of variation and should be analyzed as such.

I agree with you that the size of a feature and its position are two different and independent things, however Don's example is only showing how to do a capability study on position tolerance at MMC callout using 'virtual condition boundary' approach without taking feature's size variation under consideration. And it may be a case when customer or somebody else is only interested in statistical verification of feature's ability to assembly. It would not make much sense to check statistically the size of the feature since the violation of its virtual condition is critical for this function. Verifying size of a feature (size of unrelated actual mating envelope) would require different probing procedure (at least having in mind what are the definitions of related and unrelated AME in Y14.5-2009).
--------

Let me say that...
All inspections are estimations and all inspections make assumptions to some extent...

Again I agree - all inspections are estimations and assumptions - we are not living in a perfect world, inspectors are not perfectly qualified etc., however disregarding feature's orientation relative to datum reference frame can lead to serious issues in some cases. If part is relatively thin, then, yes, you could omit it because perpendicularity error would be like almost nothing in 90% of situations. But if the part is thick, perpendicularity error might be significant and for sure I would not assume feature's axis as a single point with unique X, Y coordinates.

 

[PaulJackson]

pmarc,

I agree with you! Forgive me for the flippant comment "You go ahead..."

The point of my presentation was to demonstrate how to perform a variable limit tolerance statistical conformance prediction. It does not delve into the related-unrelated size issues... that was not the point of the presentation. It does show how to examine the potential of both constant and variable limit geometric tolerance data.

I will agree with you as well that Don's [virtual- Mean(resultant)]/3* sigma-hat(resultant) will give a prediction of the variable limit tolerance... but I do not agree that it is heading in a good direction because the actual variation of the processes generating the RAME and its location can be amplified or masked by using them "as oombined" in the resultant diameter variation. There is a better way!

Paul

 

[pmarc]

Correct me if I am not understanding your point well, but I do not see any combination of RAME size and its location. RAME is located perfectly at basic location from DRF as well as perfectly oriented relative to DRF. All you should to is to probe hole's surface in sufficient number of points and check whether any of them is violating virtual condition boundary.

 

[ptruitt]

Does "RAME" = Resultant ____ Mating Envelope?

Peter Truitt
Minnesota

 

[PaulJackson]

Here is what I mean:

The un-oriented actual mating envelope represents the size that the largest pin or smallest sleeve that fits the feature... It is the never perfectly square to the datum features. Its size cannot correctly be used for these calculations because it ignores orientation to the datum features. I call this the

The oriented actual mating envelope represents the size that the largest pin or smallest sleeve that fits the feature while perfectly oriented to the datum features. It is never perfectly located from the datum features. I call this the Related-AME.

The resultant condition is the boundary produced by the size (RAME) and its geometric location deviation from the specified basic location/orientation.

The combination I am talking about is the later two RAME and location deviation.

 

[SeasonLee]

Peter

RAME is a new term on 2009 standard, short for Related Actual Mating Envelope, there is another term UAME means Unrelated Actual Mating Envelope, you may get some idea from the tip here.

http://www.tec-ease.com/gdt-tips-view.php?q=77

SeasonLee