Alex Krulikowski workbook -"max dimension 2.15" shown on the minimum X distance 2

[aniiben]

If the objective is to calculate the minimum distance then why 2.15 maximum is value SHOWN has to do with the correct answer.

As far as I understood the correct answer for minimum distance is: 0.15.

The maximum X distance is 2.85. Then what 2.15 is good for? What is the meaning of 2.15 value?

 

[Belanger]

It's true that 2.15 is not the correct answer -- but notice that the answer cell for that column is X'ed out. So 2.15 is merely the sub-total for the maximum column.

Notice there are two rows down there: the sub-total and the true answer. It's helpful to still add up the maximum column to check the math (as displayed in the last column's sub-total). If the parts were permanently glued in the arrangement shown above, then 2.15 would be the real maximum because all shifting would be eliminated.

In order to find the true maximum, we'd have to create a sketch showing the shaft pushed to the left, and that would result in a different stack path and a different spreadsheet. In that case the maximum column's sub-total would be carried down to the answer row (2.85, as you said), and the minimum column's answer cell would be X'ed out.

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

 

[pmarc]

In order to find the true maximum, we'd have to create a sketch showing the shaft pushed to the left, and that would result in a different stack path and a different spreadsheet.

Or we could use the same sketch (with the shaft pushed to the right), but place X on the opposite side. This, of course, would create a need for a different stack path, thus for a second stack-up.

Side note: Technically, MIN and MAX values could be calculated in a single stack-up (with the assumption that both parts are centered relative to each other, and not shifted right or left), but that would require some "tricky" adjustments. As long as all that matters is worst-case numbers, the two methods (2 stacks vs. 1 stack) will yield the same results for MIN and MAX, but if the interest is in RSS numbers, the results produced by the two methods will not match.

 

[aniiben]

Thank you J-P and pmarc,

Should I understand that 2.15 value is for "check and balance" only with no relevance on the final answer/requested min distance?

 

[Belanger]

Aniiben -- yes.
Pmarc -- Yes, of course you could use the same sketch. But the one given above shows the parts touching on the right. That's why I suggested another sketch; it's much easier to draw the path on a sketch which reflects the parts in the actual configuration desired for the question.

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

 

[pmarc]

J-P,

True. It's just sometimes (in case of complex assemblies with many parts; and especially in case of axisymmetric assemblies) creating a "sketch which reflects the parts in the actual configuration desired for the question" may take more time than just placing the stack-up objective on the opposite side of the originally calculated gap.

 

[Getdan]

With the one column method, the min and max arithmetic values obtained do not correspond to the worst case values obtained with the dual column method.
Is it possible to use the one column method to calculate the worst case stack limits?
Thank you for the feedback.

 

[greenimi]

Getdan,

See this thread and the posts from pmarc

https://www.eng-tips.com/viewthread.cfm?qid=427871

 

[Getdan]

Thanks a lot greenimi,
I think the results are right in relation to the starting hypotheses.
In the field of worst case analysis, the hypothesis that seems to me to be the most correct is to consider all contributors (including bonuses and shifts) to their extreme conditions: dual column - Krulikowski exemple.
If one-column calculation is used, it is assumed that the bonus and shift are at their median value range: this is not exactly a worst case condition.

 

[pmarc]

For what it is worth, this is how MIN and MAX worst-case values can be calculated in a single stack using one column (+/-) method.

 

[pmarc]

I just noticed two typos in comments in lines D and H:
In D it should be 'max shift 0,15 = (10,6-10,3)/2'
In H it should be 'max sift 0,1 = (11,0-10,8)/2'

Fortunately, that doesn't change anything in the stack.

 

[Getdan]

Very interesting pmarc, if I use your spreadsheet with the RSS statistical approach (normal distributions for all tolerances and Cp = 1) would I get similar values compared to the classic spreadsheets used for stacks calculation?
Thanks for an answer

 

[greenimi]

Getdan,
Is your method of calculation posted in your embedded picture from 17 Mar 19 13:13, in a very close alignment with Brian Fischer’s method found in his Stackup Tolerance book?
If yes, my follow up question is how did you calculate “% Contrib”?

Pmarc,
Why “% Contrib” is not considered in your Method#1 and Method#2 (Two stacks and Single Stack, respectively) - previous discussion posted pdf’s - and how to include it (if all possible)?

Thank you

 

[Getdan]

greenimi,
The calculation sheet used is the Fischer one, the contribution percentages should be those relative to the linear tolerance values.

 

[pmarc]

Getdan,
I am not sure what you mean by "classic spreadsheets used for stacks calculation". Could you clarify?

greenimi,
I did not include % contributions in any of the stacks because I simply had not considered them important from the final results perspective.

For worst case analysis a % contribution of each tolerance is the amount of that tolerance divided by the total amount of tolerance in the system and multiplied by 100.

For RSS analysis it is the square of that tolerance divided by sum of squares of toleances of all stack contributors and then multiplied by 100.

Does that answer your question?

 

[greenimi]

Pmarc,

Yes, it does. I figured out how % contribution is driven. Thank you very much.

Now I am trying to understand two things:

1.) Why you removed the bonus and the datum shift from your original calculations (method #2 single stack). I am trying to put them back (bonus and datum shift), but I am not able to get the RSS numbers from your first posted chart (the one which includes bonus and datum shift and you got RSS=0,689). I got RSS=1.015. Are you saying that a single stack cannot be done correctly with bonus and shift included? Or there are other “adjustments” needed?

2.)And maybe the second thing, is related (closely or not) with the first one. On first stackup chart (the one that includes bonus and datum shift) you are using datum feature A radius (-5.15/-5.30 row E and +5.50/+5.40 row G), but on the one that you posted yesterday (embedded picture) on ID row E you are kind of combining them together (±0.35 combo value). What I am not understanding how this “combination/combo” value will affect the component tolerance column and consequently the RSS calculated value.

I realize that you might have to go back and refresh your memory on these calculation to clarify my misunderstandings, so I am patiently waiting. No rush. Thank you again

Here is again the discussion referenced in this thread:

https://www.eng-tips.com/viewthread.cfm?qid=427871

 

[pmarc]

greenimi,
Yes, I have to go back to these stacks. If I have any questions, will let you know.

 

[pmarc]

1.) Why you removed the bonus and the datum shift from your original calculations (method #2 single stack). I am trying to put them back (bonus and datum shift), but I am not able to get the RSS numbers from your first posted chart (the one which includes bonus and datum shift and you got RSS=0,689). I got RSS=1.015. Are you saying that a single stack cannot be done correctly with bonus and shift included? Or there are other “adjustments” needed?

As I mentioned in the original thread, I removed bonus and datum shift from the calculation just to simply things. But it doesn't mean that a single stack cannot be done correctly with bonus and shift included:

https://files.engineering.com/getfi...a0a1bf0bcd47&file=shaft_housing_tol_stack.pdf

I think the results summary on page 2 nicely shows that it is possible to get to the same MIN and MAX worst case values in both methods (two stacks vs. single stack), but the same can't be said for RSS numbers.

2.)And maybe the second thing, is related (closely or not) with the first one. On first stackup chart (the one that includes bonus and datum shift) you are using datum feature A radius (-5.15/-5.30 row E and +5.50/+5.40 row G), but on the one that you posted yesterday (embedded picture) on ID row E you are kind of combining them together (±0.35 combo value). What I am not understanding how this “combination/combo” value will affect the component tolerance column and consequently the RSS calculated value.

Again, I think the results summary shows how different ways of including both datum features A sizes in the stacks affect the RSS numbers.

Does that help?

 

[greenimi]

Definitely helps.

Thank you for your revised stacks. As a matter of fact, I got RSS=1.015 (same as yours for method #2—Single/ Standalone Stack). YEY!! Somehow, I got confused with the calculations for independent (X max and x min, method#1, Two Stacks).
Interesting example regarding the RSS calculations.

Full disclaimer: I do not have access (or knowledge to use) for any variation and/or statistical analysis software like VSA, 3DCS, CeTol, etc. I am curious how the software(s) are handling these kind of RSS calculations (as for W-C calculations the results should be the same). Which method is used by which company? I would say this is extremely simple example, (only two parts involved), so should not be a big issue.

Pmarc,

Do you know if there would be differences between these packages (or any other packages you were expose to)?
I suspect there would be (because why not?).

I just would like to ask you (if it’s not a proprietary/secret information) are you using any of these variation analysis software?
Not sure if I should say, “I wish the company I am working for has one” or those analysis will just create more problems/ issues/ disagreements than solve.

 

[Belanger]

Also keep in mind that RSS might be "too good to be true." So there are actually formulas for "Modified RSS" -- I use a spreadsheet that also has a correction factor of 1.5 in front of the square root (the Bender formula). There's also a Gilson formula and maybe others.

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