Flatness / Straightness / Bow specification

[Proud Liberal]

I work in a plastic profile extrusion company that routinely deals with part bow. Our customers generally give us a specification for the entire length of the part.

Historically, parts were checked with pin gages being slid under the bow with the part on a surface plate or between two parts placed back to back. Both of these methods were ripe with error due to the "feel" required by the operator.

I designed and built a 12" fixture that is used on a optical comparator that eliminates the "feel" and increases the measurement resolution from .xxx" to .xxxx". The only problem was converting from bow/ft to overall bow specifications. To deal with that, we us a spreadsheet that:
• converts the bow (distance from the 12" cord to the high point of the arc/tangent) into the corresponding radius,
• then caculated the bow in THAT radius with a cord length = part length.

Our customer cites from GEOMETRICS by Lowell Foster the section on composite straightness. The example listed shows a straightness callout of .005" per 1.000" and the composite tolerance of .020" on part with a length of 4.00". Since the part just happens to be 4", their contention is that the .020" tolerance is calculated by multiplying the tolerance per inch by the length divided by per units measure (inches in this case).

I think that the example is an unforturnate carry over from the example in the ANSI DIMENSIONING & TOLERANCING Y14.5M standard where the math just happens to work our that way. But the standard goes further in §6.4.1.4 and explains:

"Caution should be exercised when using unit control without specigyina a maximum limit because of the relatively large throretical variations that may result if left unrestricted. If the unit variation apprears as a "bow" is allowed to continue at the same rate for several units, the overall tolerance variation may result in an unsatisfactory part."

The customer agrees this is an issue but contends that the second callout is unnecessary because of the example in the book. NOTE: see attached file for images from GEOMETRICS and ANSI Y14.5M.

I think that if the example would have been of a 96.850" part with a .002" per foot straightness and .020" overall tolerances, this confusion would have been eliminated since the overall tolerance wasn't .0161" (.002 x [96.850/12]).

I hope this isn't totally confusing.

Now, to my point. Which interpretation is correct? I have no problem either way but need a clarification for internal training purposed.

 

[Proud Liberal]

attachment

here the attachment

 

[Proud Liberal]

No takers? Should this have been posted to a different forum? I'm surprised there isn't a metrology/GDT forum?

Or is there a better (highly unlikely) web site to pose this question?

 

[Sam]

Zeno,
It sounds as if you are taking a sample length of the customers product and developing your inspection sample. If so, then I would expect that your acceptance criteria would be an equal percentage of the customers' requirement.
But what about that portion of the product you don't check.
That thought woukd work if the results of the "bow" of the product was a perfectly shaped chord each time.

I may be misunderstanding your question, but I would have to err on the part of L. Foster. The diagram you show is discussing virtual condition and its effect on a part; RFS.

Personnally, I would suggest a "go, no-go" fixture that considers the entire length.

 

[Proud Liberal]

Sam said:

Zeno,
It sounds as if you are taking a sample length of the customers product and developing your inspection sample. If so, then I would expect that your acceptance criteria would be an equal percentage of the customers' requirement.

— This is the point on which we disagree. The relationship, IMHO, is not linear (see ANSI example) and therefore not an equal percentage. —



But what about that portion of the product you don't check.
That thought woukd work if the results of the "bow" of the product was a perfectly shaped chord each time.

— For our process, this is a valid assumption.—



I may be misunderstanding your question, but I would have to err on the part of L. Foster. The diagram you show is discussing virtual condition and its effect on a part; RFS.

— Unfortunately, the example is of a feature of size. I would point to the clarification example from Y15.5M. —


Personnally, I would suggest a "go, no-go" fixture that considers the entire length.

— That is the bone of contention. What is the tolerance for the entire length? —

Another question, is there an ANSI forum that could make a ruling?

 

[Sam]

Zeno,
Try the attached GD&T forum.
http://www.datum-inspection.com - Link was: /forum_frm.htm

 

[Sam]

Let us know what kind of response you get.

 

[Ravi Khare]

Let me try and answer this.

I agree with Zeno on the fact that the total tolerance should not be extrapolated by multiplying the straightness tolerance by number of segments by which the part is longer.

Straightness per unit length and straightness across the entire length are two independent requirements which should be governed by the application the part is used for. As a matter of fact, the second callout in the frame is specifically meant to emphasize this. In case the total straightness was a result of a straight tolerance stackup of each of the running segments alitgned next to each other, the second callout would not have been necessary.

When the first callout specifies a 0.005" straightness in 1.000", it may not mean that you have to start from the begining, and measure each of the neatly divided running segments for a conformance of the 0.005" straightness. You would start anywhere within the length of the part and measure along 1.000" from there, and still should conform to a straightness of 0.005".

However Zeno, you point out that your process ensures that a perfect bow is produced, which means that the radius is uniform acorss the length of the part. In this case I would agree with your customer that the second callout is not necessary. In this case, you could go either way. You can measure a segment and extrapolate the way you have shown in the figure, or measure the virtual condition on the entire part as Sam suggested. It would amount to the same thing.

In this assumption, you will have to be careful that other cylindricity issues ( taper and roundness) don't confuse the determination of the axis and its straightness.

 

[Ravi Khare]

Zeno,

Are there any other interpretations that you could get? I would be interested in learning more.

Thanks

-Ravi

 

[Proud Liberal]

Although I think the ANSI standard is quite clear, my customer is convinced that there is a 1:1 linear relationship between a specified tolerance per unit length and an implied (by his interpretation) overall tolerance.

I have worked for companies in the '70's that also came to that conclusion, but, since I was very young at the time, I never even thought about questioning their logic.

I sent the question to ANSI but didn't get a response. And although the opinions expressed here have been helpful, I would like to be able to quote an "authority" on the subject to resolve this. To this point, what type of qualifications would justify calling someone an authority?