No Y14.5 version referencing Y14.5.1-2019. And, what is the "single solution"? 3

[Burunduk]

Since both the 2009 and 2018 versions of Y14.5 reference Y14.5.1M-1994 while Y14.5.1-2019 is described in the ASME website as "consistent with the principles and practices of ASME Y14.5–2009", who is supposed to be using ASME Y14.5.1-2019, and what for?

Also, what exactly is the "single solution" for datum stabilization mentioned in Y14.5-2018 and only covered in Y14.5.1-2019 (I don't have the 2019 document)? Is it based on Least Squares or something else?

 

[greenimi]

Burunduk,

Has been tangentially discussed here:

https://www.eng-tips.com/viewthread.cfm?qid=449536

Evan has some good input.

 

[Burunduk]

greenimi,
Thank you.
I read through the thread you provided the link for and it was an interesting read. But while there were some nice insights from the participants, I still remain clueless about what the answers to my questions above might be.

 

[greenimi]

Burunduk,
Just FYI, ASME has an online training class Fundamentals for 2018 standard. In that online class, Scott Neumann (from Geotol) gives a fairly decent explanation (however short) of his understanding of the constrained LSQ.
(Scott N. is in both committees, math 14.5.1 and in 14.5, so I assume he knows what he is talking about.)

 

[Burunduk]

greenimi,
Thanks, I will look into "constrained least squares" and consider that online class.

Regardless, I'm also trying to figure out what was the reasoning of releasing a new version of the math standard (Y14.5.1-2019) which remains unreferenced. It is not referenced by the version of Y14.5 it is supposed to support (2009) and neither by the newer version of Y14.5 (2018). There must be some explanation to this that I'm not aware of.

 

[pmarc]

Burunduk,
The reason why Y14.5-2018 does not reference Y14.5.1-2019 is that this version of the math standard was not yet available when 5 were releasing their standard.

From Y14.5-2018:
"2.1 INTRODUCTION
The following revisions of American National Standards form a part of this Standard to the extent specified herein. A more recent revision may be used provided there is no conflict with the text of this Standard. In the event of a conflict between the text of this Standard and the refer-ences cited herein, the text of this Standard shall take precedence."

Far from perfect solution, but at least it's something.

 

[axym]

Burunduk and All,

Yes, Y14.5.1-2019 references Y14.5-2009 and not Y14.5-2018. The original plan was to release the Y14.5.1 revision before the Y14.5 revision, but the order ended up being reversed for reasons that I can't get into here ;^). The release schedules are still not synchronized but we're catching up. I'm cautiously optimistic that the next Y14.5.1 revision (that will reference Y14.5-2018) will be published in 2024.

The "single solution" issue was made more complicated by the release schedule issues. In the mid-2010's, Y14.5 decided to move away from the "candidate datums" default for planar datum features and wanted to define a "single solution" instead. They asked Y14.5.1 for help with potential definitions for the single solution, and various methods were proposed and studied. Some of these are described in a nonmandatory apppendix in Y14.5.1-2019. The recommended definition for the single solution mentioned in Y14.5-2018 is the "constrained L2" simulator. This is a type of constrained least squares algorithm, that defines a reasonably equalized and stable datum plane for most configurations of as-produced geometry.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[3DDave]

The problem is that all the math that is done doesn't guarantee the same condition happens in assembly - a part may very well pass inspection and fail at installation because the orientation of the part at installation isn't the same as the calculation.

So the question is - should the referencing features be accepted based on:

A) Using a mathematically rigorous solution that doesn't represent the installation
B) Finding at least one solution to the datum simulator generation process that allows accepting the referencing features
C) Requiring that the features be accepted for all candidate datum simulators

With a fixed gauge choice B is automatic - it also matches how the part is installed, that is, the assembler is not trying all installation solutions and isn't performing a constrained least squares solution.

I am not surprised that CMM software writers and QA/QC personnel like choice A.

To be fair, it all papers over the desire for manufacturing and inspection to avoid installation force constraints getting specified on the component drawing - constraints makes for more work and they don't want more work. And engineering management doesn't want the engineers to put onto the drawing the forces they spent money to calculate for fasteners and the like for stress and strength reviews. Management sees that it adds time and, in manufacturing and inspection, more cost. Better not to know. Ignorance is bliss.

So everyone is just happy as can be by this mutual shrugging off of their various responsibilities. Besides - part failures are going to be the customer's problem when gaskets blow out or leak because they don't fit right (Thanks Ford/Mazda for screwing up the intake manifold gasket compression. Car can't idle if the air comes from some random hole.)

I suspect that L2 stuff can produce solutions that fail the 1/X criterion. Is that true?

 

[3DDave]

It further occurs to me that ignoring forces applied to simulate installation removes the requirement that the inspectors use FEA to evaluate the parts in the event that hard tooling is not created to hold the parts for measurements nor does there have to be any consideration of compliance of the actual mating parts during the application of assembly loads.

This mirrors the typical FEA design analysis that ignores the effects of acceptable variation on stress distributions and, in some cases, the uniformity of contact pressure.

 

[sendithard]

Burunduk,

My understanding is that the constrained least squares for a plane in this example does a least squares fit to the planar part surface, and then forces the plane somehow outside the material. How that calculation is made is proprietary to pcdmis(cmm), but here is a quote from a member of that team:

"It does a least squared fit first, then shifts, external to material. I obviously can't share the exact algorithm on a public forum as it is Hexagon's intellectual property"

If I remember rifling thru the pcdmis manual a while back going over this topic, it is similar to a cylinder as well...least sqr first then force it externally with stabilization.

I don't like it being called constrained least squares, I think it should be called external least square so you know it is an external fit.

 

[axym]

sendithard,

The mathematics of the constrained L2 fit are summarized in a nonmandatory appendix in Y14.5.1-2019. It's generic mathematics. Regardless of what a Hexagon person may have said or implied, the constrained L2 fit is not based on intellectual property from Hexagon or anyone else. There may be proprietary algorithms in PC-DMIS for computing the constrained L2 from point data, but the definition itself is not proprietary IP.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[3DDave]

It is worth reading :

https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=926718

MSEC-6305
Toward a New Mathematical Definition of Datums in Standards to Support Advanced Manufacturing
Craig M. Shakarji
Vijay Srinivasan

Note that in Figure 11 it is possible that a usable datum feature extraction might be ignored in favor of mathematical perfection causing related, usable features to be rejected.

One can get a copy of A Constrained L2 Based Algorithm for Standardized Planar Datum Establishment by the above authors for $25 from

https://asmedigitalcollection.asme.org/IMECE/proceedings-abstract/IMECE2015/57540/V011T14A023/256381

or, for free, what is likely very similar:

from

https://www.nist.gov/publications/t...ishment-using-constrained-total-least-squares

https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=907559

which has discussions of several algorithms. "On the Enduring Appeal of Least-squares Fitting in Computational Coordinate Metrology"

 

[sendithard]

axym,

Well said, I've never seen the 2019 math definitions standard you stated, but at some point I'll have a look at it. I wasn't implying pcdmis was the gold standards Just they do a least squares first then go external...I assume the math standard 2019 says the same.

The pcdmis core manual is like 3000 pages, it is a monster. I remember reading they may do some void filtering somehow for areas that wouldn't contact a surface plate, so it may not be a purist least squares either...I'm not sure what defines a void vs a simple low spot, lol.

Thanks

 

[axym]

sendithard,

The datum establishment definitions and algorithms are complicated indeed - there are several variations on least squares that have various different "externality" constraints. I'm glad that we have experts on the Y14.5.1 subcommittee to "do the math" on those. I try to stick with simple topics like profile tolerancing ;^). I'm sure that you would find the appendix very interesting - it talks about and illustrates the behavior of the different algorithms with different cases of as-produced geometry. Issues such as void filling are discussed as well.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[Burunduk]

Thank you all for the valuable input.
Speaking of GeoTol and Scott Neumann who were mentioned above by greenimi, I found some brief commentary on their website about the new "Constrained Least Squares" default: "What's new in ASME Y14.5-2018?", that also includes this illustration.
According to GeoTol, "This can be done with shims on a surface plate or with an official algorithm further outlined in ASME Y14.5.1 Mathematical Definitions of Dimensioning and Tolerancing Principles to be released at the end of 2019."
Can anybody describe how the "single solution" can be achieved in manual inspection, using "shims on a surface plate"?
As mentioned, I don't have access to the newest version of Y14.5.1.

 

[3DDave]

Bite the bullet and spend the money. I know - the next version to eliminate the symmetry and concentricity explanations is due out in 3-5 years so there should be a public review draft and purchasing the current one would be a waste. But it's unlikely to be free like last time; they might charge 90% of of the cover price and black stamp all the critical words again.

 

[SeasonLee]

Better quality image

Season

 

[Burunduk]

SeasonLee,
Thanks for the clearer image.
Does your source include the Y14.5-2018 default too?

 

[SeasonLee]

Its a snap shot from GeoTol Pro 2020 textbook, the following image for those who are interested to read.

Season

 

[3DDave]

There was no need for the term "void filling" as that is already a function of creating the convex hull. It's right there - "convex" means no concave portions.

Another way to describe "single solution" is that all the feature control frames essentially have a simultaneous requirement to use the single solution - the component is not allowed to be repositioned during inspection for any datum references that match. All having, for example, "A" as the primary must use the same derived "A" regardless if the features are "SEPARATE REQUIREMENT" under the "single solution." What can be repositioned under that condition is the Feature Frame, but that's not what the standard implies.