Determining tolerance limits

[3DDave]

I figure there are several ways to determine tolerances for dimensions:

1) Use whatever was on the last drawing someone else did - seems very popular; see title block tolerances.
2) Get QC data for the last several hundred or thousand similar features, picked from an inexpensive process that isn't totally sloppy, calculate a 3 to 6 sigma value and tweak the design accordingly.
3) Get a stress report showing likely levels of force to push parts into place or stresses and deformations induced due to variations in manufacture and choose dimensional variation limits that are acceptable.
4) Make up some numbers because manufacturing will usually use a process that has a 1 sigma chance of making to print and will ask you to open the tolerances anyway, including then moving to a crappier process and asking to loosen them again, while reminding you you held back unnecessarily the first time.
5) Look at the assembly costs and choose values so the parts need no manual alignment, even though the manufacturing estimates are a little higher, saving big in assembly time and fixtures.
6) Dart board.
7) Ouija board.

Just for fun, rank these in order of desirability and then in order of likelihood. Feel free to include other methods that have worked, or might be expected to work better.

 

[drawoh]

3DDave,

I like the darts. It sounds much more fun than playing with an Ouija board.



--
JHG

 

[CheckerHater]

But when you determine tolerances for position, don't forget to multiply the result by 1.4

 

[drawoh]

3DDave,

I am not sure whether or not to answer your question intelligently. Your tone is not totally serious.

Tolerances ought to be assigned based on the requirements of the fabricated part. The first check on your newly assigned tolerance is whether or not it can be achieved by your intended fabrication process. If not, you are going to have to fix your design. Assign the loosest tolerances you can.

Manufacturing does not care about tolerances primarily because drafters and designers don't. If, over and over again, you agree that your tolerances do not matter, they will stop asking, and you will have no control over your process.

--
JHG

 

[3DDave]

The serious part is determining what the requirements are - which is what tolerance (i.e. to tolerate) supports. Methods 1 through 5 are actual methods I've seen people use; many might as well have used the dart board, and the Ouija board is a variant of the first method.

It continues to surprise me that tolerance determination is so often seen as a simple addition or geometry analysis task, without considering the effects on downstream costs or installation related distortions.

It used to surprise me that in-house manufacturing and QC saw no value in supplying engineering with as-built and as-received measurement data, but that company no longer manufactures much, so I am no longer surprised.

Back to the original question: where do you (and others) rate these methods and do you have others. Where do those methods rate?

 

[dgallup]

Option #1 can be a good start if you are just making a minor tweak to something that has been in production for some time with low scrap, low cost & low customer complaints. It does not work for clean sheet of paper design.

Cpk data is always a good source of information. Depending on your organization, it may be next to impossible to get or readily available. Asking the machine operators on the floor about which dimensions they have to constantly fight is also useful information.

The absolute worst thing (we have bought entire product lines that were done like this) is to just dimension everything nominal and put excessively small default tolerances in the title block. To quote a French "designer" who sent us a whole set of drawing like that "Tolerances, we don't need no stinking tolerances, we are French". I don't know the emoticon for sneer.

Have not used dart or Ouija boards but I did have a manager who regularly consulted a magic 8 ball.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[dgallup]

PS That magic 8 ball is right on the money. I asked if I should go home early today & it said ABSOLUTELY!

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[Belanger]

The best way is probably a combination of things. As mentioned, you can work from the functional requirements (i.e., what size opening must this fit into?). But at the same time, you should talk to the manufacturing folks or look at historical data to determine what the process is capable of.
Then it becomes an art to massage those two ideas into a workable tolerance. So instead of a purely top-down approach or a purely bottom-up approach, this is sort of a meet-in-the-middle approach. But it's just a general approach; many other factors could go into it.

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

 

[pmarc]

To support J-P's opinion, here is the link to the article written by Don Day. It can be found on Tec-Ease website:

http://www.tec-ease.com/pdf/gdt-dfa-2.pdf

 

[MikeHalloran]

<tangent>
I remember one company's first experience with GD&T, guided by an idiot drafting manager.
Many-holed aluminum plate, 'located' by ~23 pan head machine screws in counterbored holes.
The shank clearance holes, laboriously calculated, were big enough to let the heads pass completely through if every screw was at its basic location.
Said plate's location was so uncertain that assembly fixtures were required to fix it relative to adjacent parts.
Assembly fixtures had to be issued to field service.
Said idiot drafting manager was reassigned to lesser duties, after _way_ too long a delay.
</tangent>

Best advice for a newbie on a greenfield project with no mentors available (usual condition for a startup):
Find something similar and use similar tolerances.
Better, find several somethings similar, and use similar tolerances, but closely study the ones that are different and figure out why.
Take advantage of every opportunity to visit related fabrication and assembly shops, and pay attention.







Mike Halloran
Pembroke Pines, FL, USA

 

[3DDave]

Don Day's paper is interesting, but starts with a big error - "...national standard on Geometric Dimensioningand(sic) Tolerancing (ANSI Y14.5M-1982)" There is no "Geometric Dimensioning and Tolerancing" standard, The title of the Y14.5 series has been just "Dimensioning and Tolerancing." The term "geometric dimensioning" was removed from the body of the document in the 2009 revision, having gone undefined and without reference in the 1982 and 1994 versions.

As a small paper, it's OK, but it depends on straw-man arguments and includes extraneous themes. One is to clearly document the derived requirements using 'Y14.5. Another is to review said requirements by a multifunction team.

Then the following theme is thrown in: "Designers are led to believe..." which is a difficulty that neither using 'Y14.5 or having a multifunction review team can fix. After all, who is leading the designers to this belief? Others on the team. It then wanders into an emotional argument portraying designers as having a 'battle cry' and setting manufacturing into the position of being "destined to produce scrap."

The throttle pulley example is especially disturbing. The correct fix starts with identifying the reason the cable is climbing the pulley or changing the pulley so that the cable can't climb.

If the cables aren't following the pulley it is because the cable is climbing the edge of the pulley and walking off. The described change allows the guide pins to prevent the cable from walking off the pulley, but at the expense of wear on the cable and the pulley edge. This is a band-aid to cover a fundamental failure. Eventually the cable will fatigue and fail at the spot where the climbing begins. If the cable then frays and expands, it could wedge the throttle pulley and prevent it from moving, or separate and prevent it from opening, presuming the item the pulley is attached to is spring loaded to tend to shut off.

The motor example is odd. Given the large number of parts in the stack-up, including hole diameter change due to press-fit, there must have been a calculation or allocation for allowable angularity before binding. The suggestion is that of all the parts in the motor, the only item not dealt with was the bearing, and the engineer purposely ignored its contribution. This is a counterpoint to the earlier contention about making tolerances larger - obviously the tolerance was going to be smaller than the manufacturer typically produced, forcing the manufacturer to produce scrap.

It is more likely the bearing company came in with a low price. More dimensional control meant additional expense. Given the large numbers involved, greed at some level is the likely driver. A one cent change in the price is $60,000. How much would it usually cost to inspect 6,000,000 bearings for a few tenths on the diameter and a few tenths on position? Enough that a bonus check doesn't get written. I think the story is not how D&T can save the day, but how poor incentives can cause failures. Who knows - would the story have been different if the process for making the bearings was so consistent that the motors worked OK, proving the control was not required?

 

[pmarc]

The term "geometric dimensioning" was removed from the body of the document in the 2009 revision, having gone undefined and without reference in the 1982 and 1994 versions.

From the Foreword of Y14.5-2009 (3rd paragraph):
"To accomplish this it is becoming increasingly important that the use of geometric and (sic!) dimensioning (GD&T) replace the former limit dimensioning for form, orientation, location, and profile of part features."

Granted, this is not the definition, but the term "geometric dimensioning" is in the standard, whether we like it or not.

---

As for Don's article, I linked to it especially because of the paragraph on page 4:
"No one person can possibly have the answers to these questions. It requires the input of the multifunctional team to properly define a part's requirements. All too often these questions are not asked until the tooling produces a bad part, the fixture or gage isn’t working or the customer complains. When the questions are asked early in the design cycle, the answers can be documented at the cost of adding a symbol, line, number or note to a drawing or CAD file. If the questions are left unanswered until late in the product cycle, the cost of the answer is usually an engineering change costing hundreds if not thousands of dollars."

 

[drawoh]

I disagree with the costing diagram on page 3. I think it really looks like my attachment.

Doing a given process more carefully is not a good way to improve your tolerances. You need to switch to a more precise process.

--
JHG

 

[3DDave]

pmarc,

The foreword is not part of the body of the specification. It is incorrect to think otherwise, like it or not. Its contents have no value in discussions of product description.

I'm not certain who else submitted a comment to the Y14.5 committee to remove "geometric dimensioning" from the body of the standard. I submitted two. The committee accepted at least one of them, voted on the change or changes, and incorporated the removal of the phrase from the body of the standard. Perhaps someone will make a case for its return. Until then, it's out.

 

[pmarc]

3DDave,
Okay, the Foreword may not be part of the body of the standard, but it is a part of the entire book. And while I agree that "geometric dimensioning" isn't fortunate term, it is in the 2009 book along with other terms not mentioned in the body of the standard, like "Geometric Dimensioning and Tolerancing" and "GD&T", but commonly used by many people throughout industry.

I do not have Y14.5M-1982 in front of me, so am not able to verify if "Geometric Dimensioning and Tolerancing" was used anywhere in that version, but I am pretty sure this general term was known at the time when Don Day was writing his article. That is why I would not make big deal out of the "...national standard on Geometric Dimensioningand(sic) Tolerancing (ANSI Y14.5M-1982)". After all, this is not the same as: "...national standard ANSI Y14.5M-1982, Geometric Dimensioning and Tolerancing", is it?

 

[3DDave]

What is the first sentence in the foreword for the 2009 version?
In the 1994 version it is "(This Foreword is not part of ASME Y14.5M-1994.)" I'm guessing is similar for '2009.

The fact the phrase was used without basis then is no reason to find it acceptable now, particularly since it was removed from the standard.

In the 1994 and 1982 versions 'geometric dimensioning' is contrasted with 'coordinate dimensioning' but, if quoted accurately, it would now not be the same as 'limit dimensioning.' Both of these statements are nonsensical.

In the '1994 version: Decimal inch dimensioning, millimeter dimensioning, angular dimensions, reference dimensions, baseline dimensioning, tabular dimensioning, polar coordinate dimensioning, and rectangular coordinate dimensioning are described, but not 'geometric dimensioning.' It didn't even make it as an entry in the index. That's how meaningless it was.

'Limit dimensioning' is included under the direct tolerancing methods. The '2009 foreword suggests limit dimensioning is a former, hence no longer used, method. This is unlikely.

tl;dr 'geometric dimensioning' has never been a supported concept in any version of the Y14.5 standard and now is no longer part of the latest ('2009) version of the standard. This fact is only important in determining the care put into understanding the entire standard.

 

[pmarc]

At the risk of beating this subject to death...

3DDave, please don't get me wrong, I understand your concern about 'geometric dimensioning' in the light of the terminology stuff. All I did, was posting a quote from the Foreword of the standard, trying to show that this term is used in the book.

If you also don't feel comfortable with 'Geometric Dimensioning and Tolerancing' or "GD&T", I respect that. I just don't think it is so big issue when someone uses these terms. After reading first paragraph of the Foreword for the 2009 version (especially the second sentence), I believe the Y14 committee shares this opinion.

You made comment on Don Day's: "...national standard on Geometric Dimensioningand(sic) Tolerancing (ANSI Y14.5M-1982)", naming it "big error", because there is no "Geometric Dimensioning and Tolerancing" standard. And I wholeheartedly agree that there is no such standard. But you did not answer to my question: is what Don wrote really the same as "...national standard ANSI Y14.5M-1982, Geometric Dimensioning and Tolerancing"? In my opinion it is not. My understanding of his sentence is - ANSI Y14.5M-1982 national standard is about Geometric Dimensioning and Tolerancing. It does not say the standard is titled "Geometric Dimensioning and Tolerancing". And since the term 'Geometric Dimensioning and Tolerancing' has been so commonly used over last 20-30 years as a contrast to old-fashioned dimensioning and tolerancing, I find nothing seriously erroneous in his statement.

Side note 1: There is no "(This Foreword is not part of ASME Y14.5-2009.)" in the Foreword for 2009.

Side note 2: While I agree that the use of 'limit dimensioning' in the 2009 Foreword is one of quite numerous examples of poor word choice, I disagree that "the '2009 Foreword suggests limit dimensioning is a former, hence no longer used, method". It suggests something much more important - limit (or plus/minus) dimensioning should be "reserved primarly for the size dimensions for features of size". Everything else, that is form, orientation, location, and profile of part features, should be described through geometric tolerances. It is so pitty that this statement shows up in the part of the book that is not considered to be part of the body of the standard.

Side note 3: I don't work for Don Day ;-)

 

[3DDave]

When there is a phrase, initial letters capitalized in the same manner as a document title is ordinarily capitalized, followed by a document number, the implication is that the phrase is the title of the numbered document.

"Horse Grease Specification, NAT-X000", indicates to most readers that there is a specification identified NAT-X000 titled "Horse Grease Specification."

Nice of the Committee to drop the note excluding the Foreword. I think they were hard pressed to keep the baseless phrase to support all the mistaken marketing material. It's a wonder they didn't re-title the document or just make up a definition and throw it in to justify using GD&T so many times in the foreword. Must have been quite upset.

I've long wondered where the phrase first kicked in. Some have suggested it originated with Lowell Foster. He published a book "A treatise on geometric dimensioning & tolerancing " in 1968. I just happened across its predecessor:

http://www.ebay.com/itm/A-Treatise-...ing-Torlerancing-Honeywell-1963-/161085842104

It is interesting as an early work because the title suggests a difference in meaning between "Geometric" and "Positional" I'm interested in seeing what that difference is when it arrives.

 

[MikeHalloran]

I learned GD&T from a traveling tutor at Ford in 1967.
Much of his published material was in no-print blue, so you had to keep buying it from him.
I forget his name, but he definitely called it Geometric Dimensioning and Tolerancing.

The opening of his presentation stated that it was a new name for something that had been going on for a very long time in the gage making trade.



Mike Halloran
Pembroke Pines, FL, USA

 

[3DDave]

Hi Mike,

I wonder how the traveling tutor would deal with the Internet were he in the business today.**

Any thoughts to share on the original topic?

By way of example, I recently came across a problem situation where an allowable free state condition became a locked-in condition due to welding the part in place. This would come under #3 of the original post. In this case, not understanding that extrusions aren't always straight and twist-free led to a drawing where only the extrusion section was dimensioned/toleranced and a design that worked right only when all the parts are perfect.

**Not that it matters much, a yellow filter on a copier should make no-print blue stand out. I expect the amount of yellow light reflected by the blue is small compared to the yellow reflected by the white background. More recently I found that placing a black page behind what I am scanning prevents printing on the reverse side from showing through.