Pattern hole position 4

[SeasonLee]

I thought a pattern hole’s measurement can be started from any one of the hole, and the position calculation result will be the same no matter which hole was chosen as origin, but the sample shown with a different result, what did I missed on the hole location measurement and position calculation?

Season

 

[pmarc]

In my opinion, the pattern can be measured from any point in space as long as that point is not derived from one of the features in the pattern. If you decide otherwise, that is, to measure 3 holes from the fourth one, like it is shown in your attachments, you have to be aware that you are loosing some portion of positional tolerance that could in the end help your pattern to pass the inspection.

But regardless of approach chosen, I am afraid this positional callout with only A referenced may require some kind of best-fit algorithm. So if your CMM is not capable to handle the data, you will not get to anything satisfactory.

I am just wondering, perhaps you could combine your CMM data with a CAD software in order to simulate the best fit procedure and obtain PASS/FAIL assessment.

 

[axym]

Hi All,

Very interesting thread. The determination of actual values for multi-feature systems is not currently well defined in the standards (but we are working on it). Different approaches have been used in industry, particularly when CMM's are involved. There are a lot of techniques being used that do not necessarily achieve the optimal result.

Because only datum feature A is referenced in the FCF, the DRF is not constrained in rotation about the Z axis and in translation in the X and Y directions. In the language of Y14.5.1M-1994, there are multiple "candidate DRF's", and for each one we can calculate actual Position values for each hole in the pattern. The sets of values that SeasonLee calculated are candidate actual values, for different candidate DRF's. Each set is valid, and equally correct. So the actual Position value for any given hole in the pattern depends on which DRF was used. For example, Hole 1 has a Position value of 0 in one DRF and 2.828 in the other. Many people find this concept counter-intuitive, but it represents reality. The Position requirement is like a template, that controls the relative position of the holes.

But what if we want there to be only one answer? As Paul illustrated in his presentation, rotation and translation can be optimized to achieve the best result. So we must define what we mean by "best". The optimization approach that I believe makes the most sense for conformance assessment could be described as "make the worst point in the system as good as possible". This is a generalization of how actual values work for single-feature characteristics like Flatness. We try to get all of the points on the feature (extremities and all) to fit into the smallest zone, as opposed to getting the most points within tolerance or near nominal.

The optimal candidate DRF for SeasonLee's example is the one that makes all 4 holes in the pattern conform to the smallest possible Position tolerance. To achieve this, we must make the worst axis endpoint(s) as good as possible. Using the terminology from Paul's presentation, this equates to maximizing the minimum residual tolerance.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[SeasonLee]

pmarc and Evan, Thanks for the valuable comments.

Actually, the concept of what I said earlier “a pattern hole’s measurement can be started from any one of the hole, and the position result will be the same no matter which hole was chosen as a origin” is from Al Neumann’s workbook page 9.4~9.7.

The attached two pages is an example of paper gage analysis on pattern hole’s position, you will notice hole #1 was chosen as the origin to collect data and the data is recorded in the chart, may I ask what is the differences in between the different measuring method? Or what did I missed?

Season

 

[PaulJackson]

SeasonLee,
Attached is 1) your pattern un-fitted (referenced from the center of the pattern)and 2) your pattern hand fitted (by eye) by rotating and translating the basic pattern and its known individual tolerance zones over a map of your measurements. Of course all of the tolerance zones are set to the minimum variable tolerance because you did not provide the oriented actual mating sizes of the four holes.

This is from a simple spreadsheet that I used to make the presentation above.
Paul

 

[axym]

SeasonLee,

Wow, that is confusing. The statement "a pattern hole's measurement can be started from any one hole, and the position result will be the same no matter which hole was chosen as an origin" is misleading when read outside of the context of Al's paper gaging example. Here's how I would try to explain it:
-The values from the Position Tolerance Actual column would be different, if a different hole was chosen as the origin
-The Paper Gage Evaluation and ACC or REJ columns would be the same, even if a different hole was chosen as the origin
-This is because the paper gaging technique allows the four measured points to be rotated/translated as a group, to optimize their relationship to the four tolerance zones (to bring the initially nonconforming holes 3 and 4 into conformance, at the expense of making the position values of holes 1 and 2 a bit worse).

So the post-optimization position result will be the same no matter which hole was chosen as an origin.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[axym]

Paul,

I did a further optimization of the data. Try a rotation of 2.446657 degrees, an X translation of 0.419371 and a Y translation of 0.015349. This should give three of the holes a Position value of 2.152, and the fourth hole a value of 2.093.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[axym]

Paul,

Hold the phone. That last optimization wasn't fully optimized - I've had too much coffee, and my hands weren't steady enough with the paper gage ;^). Try these numbers:

Rotation of 2.462486 degrees
X translation of 0.429652
Y translation of 0.009233

This results in a Position value of 2.148759 for all four holes, and should be the best possible fit.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[PaulJackson]

Well done Evan... your attention to detail is showing!
Now if we only had the feature sizes we would re-optimize and predict the precise amount of slop that the attribute gage could split in any given vector.
Paul

 

[axym]

Paul,

Thanks. The devil is in the details, as they say. We throw around statements like "optimize to get the best actual value" but doing that in practice can be very challenging. Inspectors are often left to fend for themselves with tools and training that are less than optimal (pun intended). You and I (and many others) have obsessed over this stuff for years and we still have difficulty.

This is one of the challenges of GD&T. All of the degrees of freedom that are available to get parts to conform during pass/fail gaging must be simultaneously optimized if a unique set of numerical values is desired.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[SeasonLee]

Thanks for all of your comments.

After position calculation, if we know all holes are within the tolerance, why do we still need to do the time consuming optimization study, what is the purpose of this work?

Season

 

[axym]

SeasonLee,

Good question! If the objective of the measurement is just to verify conformance, then doing a full optimization may not be worth the effort. Typically, the additional analysis is done on nonconforming or borderline parts, to try to get them to pass.

Some quality systems require numerical measurements for all characteristics on the drawing, for a sample of parts. Statistical values such as Cpk are then calculated for each characteristic to establish capability. If non-optimized values were used in these calculations, this appears as measurement variation and affects the Cpk values. This makes the production process appear to have more variation than it really does.

It may also be useful to know how close the part really is to being out of tolerance. Optimized actual values will be a good indicator of this, and non-optimized values will be more conservative.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[SeasonLee]

Got it, Thanks, Evan.
So, the function is same as paper gage analysis.

Season