Datum shift + Stack-Up

[prdave00]

I just finished looking at the latest Tec-Ease tip (

http://www.tec-ease.com/gdt-tips-view.php?q=253)

regarding datum shift and its message that the modifier applied to a datum feature does not change the tolerance on the feature or features being toleranced, but should (or can) the modifier be included in the tolerance stack up?

In the attachment I provided an incomplete drawing of a part with coaxiality controls. The primary datum is an external thread, although a cylindrical boss could have been just as pertinent. I also included a tolerance stack up to look at wall thickness using both worse-case and statistical models. The handling of bonus tolerances and datum shift was based on a series of papers written by Ngoi et al (affiliated with the Nanyang Technological University) in lieu of having access to any stack up analysis textbooks, etc. 2-D CAD drawings were used to check my work based on the guidance from the aforementioned papers.

Should I have excluded the datum shifts? Can anyone refer me to a good text that discusses inclusion of geometric tolerances in a stack up analysis?

 

[Belanger]

Your stack should not include the effect of the modifier on datum A. It's true that there is datum shift coming from that MMB modifier, but since both of the geometric tolerances considered in your stack reference the same datum, there is no net effect on the distance that you seek in the stack.

Think of it this way: any shift available from datum A is simply looseness which would allow the part to jiggle a little bit around the threads. But that jiggling doesn't change the wall thickness between the ID and OD on the left end.

By my calculations, the wall thickness you are looking for would be .2075 max and .1575 min (worst case).

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

 

[prdave00]

Thanks Belanger. All I had to go by were these Ngoi papers that allegedly pull examples straight from Krulikowski's "Tolerance Stack: A Self Study Course, 1992, Vol. 1" to validate their method. The examples provided figure in the datum shift.

Here's a link to such a paper < pib.sagepub.com/content/214/3/235.full.pdf > that you probably have to cut and paste in your browser's address bar.

Any good resources for stack up analysis (besides course work, training courses, etc.)?

Looks I'll be going back to the tolerance analyses I recently performed for my actual parts!

 

[pmarc]

prdave00,
If you are trying to calculate wall thickness (amount of material between OD dia .750 and ID dia .375) datum shift should not be included in the tolerance stack.

Why? There is no datum shift between features that are gaged simultaneously (i.e. that are contolled by a geometrical tolerance using the same datum references, with the same material modifiers and in the same sequence). In your example both position controls references to the same A(M) datum.

I could recommend you 2 books on Tolerance Stacks with GD&T usage:
- Fisher B. - "Mechanical Tolerance Stackup And Analysis"
- Krulikowski A. - "Tolerance Stacks Using GD&T"

I especially like the second one. It is really comprehensive and offers very logical (at least for me) step-by-step explanation of whole process of calculating tolerance stacks for different geometrical characteristics.

 

[pmarc]

Agree with J-P.
min. wall thickness -- .1575
max. wall thickness -- .2075

 

[ak762]

Perhaps I am wrong
But what is purpose to specify M modifier with datum A for pos. requirements if it do not allow to float datum from ideal axis of RFS thread.
Contrary if it is allow to float datum then it will affect in variation with worst case.

 

[prdave00]

ak762: It originally made sense to me to include the material modifier for the datum since the modifier would theoretically allow the part to float and therefore contribute to the stack up. However pmarc's reminder about the simultaneous requirement, made me at least consider that it would only affect the position of one feature with respect to the datum but not the interrelationship between the ID and OD features. I've also been primarily looking at examples of stack ups for assemblies and then applying that process to a component.

pmarc: Does Krulikowski's publication address the datum shift quandary for stack up within a component? As I noted to J-P, the papers (see attached) I've been leveraging claim to use examples from Krulikowski's text to validate their method and the examples include the datum shift in the stack.

 

[pmarc]

prdave00,

Yes, Krulikowski's book offers datum shift calculations within single component as well as within larger assembly.

And your interpretation of my previous statement is right - for your example datum shift matters only when considering position of one feature (ID or OD) relative to datum feature A referenced at MMB (Maximum Material Boundary).

 

[ptruitt]

Great question. I was scratching my head over this same question recently. This morning, I came to the same conclusion offered, but next week I will probably need to get it thru my skull all over again. I see someone had a bad experience buying the Alex Krulikowski book thru Amazon. (They did not send the workbook, which should be included, apparently.) So I looked around the internet and found it here:

http://www.gdtsource.com/product.php?productid=92&cat=3&page=1

Peter Truitt
Minnesota

 

[Belanger]

PrDave: Your stack spreadsheet is actually a variation of Krulikowski's approach. (Our website has a similar template available for download with tabs at the bottom for both his method and yours.) Your version is sometimes easier for people who are used to doing stacks with everything delineated as equal bilateral around a nominal. The "Krulikowski" method chunks dimensions into max and min. I find method a little easier when geometric tolerances come into play, because bonus and shift (and some other GD&T stuff) aren't felt bilaterally.

The "nominal/plus/minus" method you're using can also factor in bonus and shift, but they must be rolled into the nominal and plus/minus numbers. More info if you need it.

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

 

[prdave00]

Thanks J-P. I guess I do need to give credit where credit is due. My spreadsheet is a marriage between the one downloaded from your website

http://www.gdtseminars.com

and a very busy one downloaded from

http://adcats.et.byu.edu

that caters heavily to using statistical stack-up models (RSS, Motorola 6-sigma). The latter "adcats" spreadsheet uses a nominal/plus/minus format, as do the Ngoi papers, so I stuck with that layout as opposed to a limits layout (albeit more straightforward). Being able to toggle between assuming a normal and uniform distribution intrigued me (even though my assemblies typically need to be 100% interchangeable, etc. so a worse-case stack up is typically utilized).

I guess I need to get the Krulikowski text to understand when and where datum shift should be included (I've found it's not always intuitive). However I was under the impression, and maybe a false one, that Krulikowski's method always factored in datum shift into the stack up. I just need more education (and time in a day).

 

[Belanger]

PrDave ... oh I didn't mean anything like that; these folks who make stack spreadsheets are all borrowing from one another. The one on my website is very close to Krulikowski's, which was then modified by my colleague John Stolter, and then I tweaked it even further. It's still a work in progress to allow the addition of Monte Carlo parameters.

When to include datum shift in a stack depends on how it's used and what the stack problem is. In general, there is no effect on a stack from datum shift if:

• the datum is referenced RMB
• the features are gaged simultaneously
• the shift acts in a direction different from that of the stack path (may be a partial trigonometric effect from shift, depending on the angle)

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

 

[ptruitt]

Pardon me for an off-topic question: It makes me nervous to see Datum A referenced at MMB by the two diameters. The extension lines go to the major diameter, but the 'dimension' is the thread call-out, which makes me think of the pitch diameter. From looking at the stack-up, I think that the math is based on MMB of the pitch. Right? From a practical standpoint, is that what will be achieved in inspection? It seems that the call-outs may lead to measurement bias. Am I wrong or missing something?

Peter Truitt
Minnesota

 

[prdave00]

Thanks J-P. I felt obligated to share with the Eng-Tips community where I've received "help" in the event someone else could benefit.I appreciate you offering it on your site.

 

[prdave00]

ptruitt: I see your point about my call out and it may be less ambiguous if I had instead used a leader line pointing to the body of the thread like as in Figure 5-62 of ASME Y14.5-1994 (whether I modeled the threads or drew a simple representation). Maybe I'm violating a tenant of ASME Y14.6.

However, regardless of the thread call out, the axis of the thread should be derived from the pitch cylinder unless MAJOR or MINOR DIA is specified by the datum flag per ASME Y14.5-1994 (& 2009). The one piece of info I left off is that all GD&T is per ASME Y14.5M-1994. My datum stack used the pitch diameter tolerance from ASME B1.1-2003 for the 5/8-11 UNC-2A thread.

I too considered measurement bias and thought maybe this is why one would add the datum shift into the stack. The only gauging scenario I could think of is attached, but I can see huge measurement bias creeping up given that the OD of the gauge would simulate the axis of the threads.

 

[ptruitt]

I have been finding myself referencing datums at MMB for the sake of inspection and not entirely satisfied that it is the best decision, given the importance of documenting design intent. I will simulate the datum at MMB to fixture the parts for inspection, which should be okay, but not wonderful. (The datums I am referring to are raw die casting surfaces.) For threads, I think someone makes a fancy collet-like gizmo that has balls inside that ride in the threads and do a good job of simulating the axis. The OD of the gizmo provides the same purpose as the OD of the inspection tool you illustrated. I don't recall what it is called, though.

Peter Truitt
Minnesota

 

[ptruitt]

I could not stop myself from tweaking the drawing. I am working on stack-ups, too, and I think I'll go crazy and make mistakes trying to figure out which way all of the modifiers move the stack. I don't know how the part is used, so I realize that I may be off base.

I also noticed that the mean of the inner and outer boundaries for the O.D. do not coincide with the nominal size of the O.D. which means that the fabricator will need to decide whether to aim for the size or the position. I'll guess that they would aim for the size nominal, but there is no law that says that they must.

So I moved datum 'A' to the .375 diameter. Then I determined the outer and inner boundaries of the O.D. based on the .1575 (min) and .2075 (max) walls, calculated by several folks. The mean of these boundaries is .74 Dia.

Then I went out on a limb a guessed that, since the original drawing allowed the O.D. to float within larger boundaries, the design might allow the basic O.D. nominal to be the same as the mean of the boundaries.

Then, I set the second segment of the feature control frame at .015 (to cover all of the .750+/-.005 sizes) thinking that more size tolerance could probably be allowed if the nominal size was at the mean of the boundaries. (Another assumption, I know.) But the stack-up didn't look good, so I backed off on the .015 and ended up at .009.

I would be pleased to hear comments on all my assumptions and concerns about making the size nominal the same as the mean of the boundaries. Also, does my stack-up add-up? Do you find the stack-up easier or harder?


Peter Truitt
Minnesota

 

[Belanger]

Here's my two cents on a couple of the things you mention...

The mean of the inner and outer boundaries for the OD doesn't coincide with the nominal size of the OD because of the MMC modifier. Reason: the bonus tolerance allows more position error when the OD is smaller, so it's not equally distributive.

This is why you gotta be extra careful when using a stack spreadsheet that tries to lump everything into a "nominal/plus/minus" format. Here's how I would do a stack for your original drawing using that Excel template: First, we all agree that there is no shift tolerance But to account for the bonus tolerance, you can simply roll it into the size and use the "boundary" dimensions in the stack. Take line 1 for example: instead of having .375 followed by a position tolerance line and then applying the bonus line separately, we could just have one line: a positive NOM of .3725 (half of the OD's LMC) and the ± column as .015 (half of the total possible position tolerance). I think this is really what you were saying: the midpoint of the boundaries (my .3725) is not the same as the midpoint of the size alone (.375, radially speaking).

Line 2 on that spreadsheet would then simply reverse things for the ID: the negative NOM would be .19 (half of the ID's LMC) and the ± column would be .010 (half of the total tolerance happening to the ID). That's it! Only two lines in that stack, and the resulting answer will still be .2075 max wall thickness and .1575 min wall thickness.

The sketch from you last post is a bit different so I'll have to digest that one a bit... But I would say that while moving the datum around can make the stack numbers easier, in a real design the datum should be selected based on functional interface. For that reason, maybe keep the datum on the threads?

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

 

[ptruitt]

My guess is that the hole and threads are like yin and yang. To say that the function of the part starts at the treads and ends at the hole may feel good, but the truth is probably that either one is useless without the other. (Another assumption on my part, of course, since I did not design it.)What do you think, prdave00?

Would the part be any different if someone were to remove datum 'A' and all reference to it (from the hand sketch)? Would the stack-up change? Would it violate ASME Y14.5? Is this datum implicit or explicit in what it does for the design?

Peter Truitt
Minnesota

 

[prdave00]

ptruitt: I'll have to go back and study your and J-P's last posts, but at least wanted to mention that my drawing was just to provide an example and is not a real part I designed. However I do have a similar part with an external thread and a blind hole on the same end that runs coaxial to the thread and then an annular thread relief. In this case the relationship between thread, bore, and thread relief is very functional. I can post the actual part but I don't think it would change the question or answers regarding adding datum shift to the stack up.

I understand (and appreciate) J-P's comment about simplifying everything to 2 lines, but the contributions to the stack are "hidden" requiring further explanation to someone trying to make heads or tails of the analysis. Having a line item for each contributor is better in that regard. What column each bonus (or shift) tolerance goes into is another matter (i.e. less intuitive) and I had to rely on 2-D sketches to make sure I was putting everything into the right column.