huangyhg

During a conversation with a colleague the other day the subject of verification of out of tolerance dimensions came up.

Company requirements state that inspection tools used to verify dimensions must be capable of measuring 10X the precision of the dimension called out on drawings
Case 1: Drawing specifies a dimension to be .245 -.250 and you use a micrometer capable of measuring to .0001 increments. When inspecting the part you get a measurement of .2501, should the part be rejected or accepted?
Case 2: Drawing specifies a dimension to be .245 -.250 and you use a CMM to verify part dimensions. The CMM reports the part measurement to be .25001, should the part be rejected or accepted?
My stance is that to verify that the dimensional value meets the tolerance on the drawing it makes sense to use tooling capable of measuring 10X precision of the dimensions called out on the drawing. Without using 10X precision tooling then you get rounding error in the tool, correct? It seems to me that by being required to use 10X precision tooling I must verify out to that level of precision the part meets the drawing. My colleague feels that I have exceeded the intent of the drawing by measuring out to 4 places and should not reject the part. In the case when using the CMM I feel that you should truncate the dimensional values to 10X precision of the drawing dimensions and that any characters after the 10 precision should be ignored.
I have also been part of conversations with different opinions that if you specify a dimension on a drawing as .24 - .25 the criteria for part acceptance is different than if I specified the dimension as .240 - .250
Any thoughts?
Thanks



18 days ago

__________________
借用达朗贝尔的名言：前进吧，你会得到信心!
[url="http://www.dimcax.com"]几何尺寸与公差标准[/url]

huangyhg

Dennis Schwartz • ASME Y14.5-2009 Paragraph 2.4 INTERPRETATION OF LIMITS states: "All limits are absolute. Dimensional limits, regardless of the number of decimal places, are used as if they were continued with zeros." Examples: 12.2 means 12.20......0 : 12.0 means 12.00........0 ; 12.01 means 12.010.......0 . This can sometimes cause minor descrepancies between the CAD model and the engineering drawing where the model may be carried out to 15 decimal places but the drawing dimension is truncated at two or three decimal places. The criteria for acceptance is usually determined by internal company policies.
18 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion1 .
Nathan
Unfollow Follow Nathan
Nathan Weister • Dennis is correct. If you are complying to ASME standards, then yes, Case 1 and 2 should be rejected.
Conducting a MSA (Measurement System Analysis) and Gauge R&R could help you choose what precision might be needed.
Of course everything is relative to the criticality of the dimensions… I wouldn’t use calipers to verify measurements on my 2x4 framing.

There may be misconception that the .24-.25 vs .240-.250 criteria may be different. Decimal discrimination to determine tolerances (default tol blocks) are still used today, but were even more common in the past. It’s wrong, but it may be the case that it just “feels” like a looser spec having two decimals vs three. Just a thought.
17 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
FrankUnfollow Follow Frank
Frank Rodericks • I'm curious as to the nature of the feature(s) associated with the dimension limits in question. Are you dealing with a feature of size (most likely) such as a cylindrical surface or are the depicted limits associated with a locating dimension (hopefully not). If the latter is the case, you may have other issues to deal with.
17 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Mark
Unfollow Follow Mark
Mark Foster • You must deal with the issue of measurement uncertainty in order to address your questions. Ultimately, the only way that one can definitively say that a particular feature that you are measuring is within the stated tolerance specification is that to determine with your measurements, with whatever measuring equipment that you are using -- and their associated uncertainties -- indicate that the measured value is within the specified tolerance MINUS THE UNCERTAINTY associated with that equipment/procedure/etc. And, conversely, the only way that one can definitively reject a part is to ensure that the measurement results show that the actual dimension was outside of the specified tolerance PLUS THE UNCERTAINTY.

Thus the need, as Nathan stated, to conduct a MSA and Gauge R&R in order to quantify the uncertainty. If a measured value is indicated WITHIN the uncertainty range, it is then....well....uncertain. You cannot definitively say whether the actual value of the feature is within the specified tolerance range or not.

So, when you have measurements that are so borderline, then it becomes a business and risk management decision. Thus the need, as Dennis stated, becomes an internal company policy -- normally based on an assessment of the criticality of meeting the stated tolerance -- e.g. if the feature is actually out of the tolerance specified (not the measurement indication, the actual feature) will someone be angry, or will someone get hurt?

And, just to include all previous commenters in the thread, the point that Frank brings up is a completely valid one, and if you are interested in his latter case (orienting or locating dimensions/tolerances), then there is a whole bunch of other issues that will rear their ugly heads and probably deserves its own thread.
17 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion1 .
John
Unfollow Follow John
John Miles • Mark hit it on the head here. I worked Tool/Gage Engineering for 12+ years. We had a rule of 10X (gage design using 10X of engineering, inspection equipment using 10X of the gage for verifying the gage, etc. Everything flows down hill) but if the engineering tolerances were so tight that 10X was out of the question as it flows down hill, then State of the Art was accepted. All that said, measurement uncertainty has to be taken into consideration and subtracted from the measured result.
17 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MichaelUnfollow Follow Michael
Michael Cisneros • In my industry, the part without a doubt is rejected. However I do understand there are fit, form and function to consider. Those issues are addressed by the MRB and they will have to make the disposition. Hopefully a member of the MRB has an understanding of measurement uncertainty. This also brings into question the process's capability of holding the tolerance and what is the range in the lot of parts being inspected.
16 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MarkUnfollow Follow Mark
Mark Foster • @Michael, Oh, there is NO DOUBT that the part should be rejected based on the INSPECTION. I should have been clearer in my earlier post. If you measure something with whatever measurement equipment/personne/environmental conditions that exist, and you come up with a measured result that is outside of the specified tolerances, you should absolutely reject the part as a matter of procedure. THEN there must be a disposition made (This is where the company-specific stuff comes into play.) by MRB, a designated person, or whatever your company policies are, as to whether the part can be used or not and whether there needs to be some sort of corrective action.

When a measurement falls into the "uncertain" range, it should at the very least be investigated, even if the measured value falls barely INside the specified tolerances. If you measure within the range of uncertainty, then the dimension/feature in question is by definition borderline. Even if you are not truly out of spec, there is still something that probably needs to be addressed.
16 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion1 .
NathanUnfollow Follow Nathan
Nathan Corliss • and once you do all that work as you guys so eloquently explained, don't forget to ask manufacturing to quit manufacturing parts on tolerance limits! Yes I know that opens another can of worms about process capability.
16 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion1 .
Nathan
Unfollow Follow Nathan
Nathan Corliss • Another side note, using the 10X guideline is a good, safe practice. Many believe this is an industry standard. This all began from an old MIL STD, I don't recall the name, in regards to calibration, not measurement. Even that old mil-spec "recommended" a minimum accuracy of 4X for calibration, but ultimately, businesses decide what level of risk and uncertainty they are willing to accept when designing their measurement systems. Don't forget too, even measurement with a mic, or any other variable gauge is a sample. You could measure .2501 in one spot, but are you sure you've found the exact highest spot? Two point contact gauges may miss the functional extreme points that an attribute gauge may not accept. I love metrology and the joys of uncertainty.
16 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MarkUnfollow Follow Mark
Mark Foster • Nathan, since I know you a little bit, I will presume that I am "hearing" your email incorrectly. Would you care to elaborate? It "sounds" like you are saying that manufacturing normally would manufacture "on tolerance limits." Manufacturing almost never manufactures "on" limits. They normally manufacture to their own limits that they set "within tolerance limits," but not "on" the limits. If I specify .010 total tolerance and manufacturing chooses processes that are only just barely able to meet .010, then they will likely not be capable of producing my design -- at least not when mass-produced.

So I think I'm simply missing your point. Yes, you're right, it does open up the process capability can of worms, but you opened it, so let the worms crawl.
16 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
NathanUnfollow Follow Nathan
Nathan Corliss • Mark, I meant it as you just put it, using manufacturing process that are barely able to meet the specified tolerance. I also may have posted with a bit of rum in me and hadn't contributed to a post in a while I got over zealous! As the original example showed, would the part be out of tolerance or not based on measurement uncertainty. In a perfect manufacturing world, why are the parts that close to the tolerance limits anyway? Yes, I know the other can of worms I'm opening there too. But I would hate to see people focus a lot of time and energy arguing over, in spec, not in spec, when that time could be focusing on a manufacturing process that can produce well with in the specification limits and reduce / eliminate the risk of manufacturing defects in mass production.
12 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MarkUnfollow Follow Mark
Mark Popovich • "The CMM reports the part measurement to be .25001, should the part be rejected or accepted? "

I assume you mean .2501 (not 25001) . If your quality group is reporting diameters from a CMM with precision of .00001 I would question the ability of a CMM to do that. Ours 5 axis machines are on their best day within 0.0002. Expressened in machinists venacular, the measuremenrt you indicated would be precise to .1 of a tenth. I have never heard that come from an actual machinist.

I also agree the part should initially be rejected, but upon review of the uncertainty measures determined by the gage RR (which they probably never did), may be considered acceptable, pretty much what everyone else said.
12 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
John
Unfollow Follow John
John Miles • @ Nathan - Look at MIL-HDBK-204A. That is the standard I used for years to provide requirements for inspection and functional gaging. I don't think it has been revised but I do believe it is still current (I believe the year revision is 1986 or so). Its fun to look at though. ASME Y14.43 is a current gaging std that provides some guidance. Have fun.
12 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MichaelUnfollow Follow Michael
Michael Unmann • Couple of questions. How many of the parts are suspect? Is the process producing parts from the high to the low? Typically when I run a job I will reduce my tolerance threshold 10% and expect the machining process to work with in that tolerance. If it does not hold that tolerance I will review the process and tooling. The process needs to be stable and not have the operator chasing the nominal.

Back to the original question, .250 is the maximum and .2501 is out of tolerance. I think most shops can accurately measure .0001 so if your customer specifies .245-.250 and your part is .2501 you have a part that should not be shipped to your customer. As to the .25001 well I think that you can take that part out of QC and that will change, if you can even realistically measure it. Too many variables come into play at that accuracy.
11 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
NathanUnfollow Follow Nathan
Nathan Corliss • @John, I couldn't find the old mil spec I was referring too. It was fun reading some old mil specs. I use the Y14.43 currently as my guide for gage design, but the tolerance strategies only have provide recommendations.
10 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MarkUnfollow Follow Mark
Mark Foster • Nathan, I think the old mil spec to which you are referring is MIL-STD-120.
7 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
NathanUnfollow Follow Nathan
Nathan Corliss • That wasn't it either. I specifically recall having to find this mil-spec for a DCMA inspector for him to allow our measurement system which was only at a 6:1 ratio. There was an explicite statement about 4:1 being the minimum, with 10:1 optimized. I did run across this article though http://asq.org/quality-progress/199...n-overkill.html referencing an ANSI spec too.
7 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Mark
Unfollow Follow Mark
Mark Foster • Mil-Std-120 does state that 4:1 is the minimum, but that 10:1 was preferred, but it also says, "should" for both of those extremes, not, "shall," so it was only a guideline, not a mandate. Perhaps Mil-Std-8 may have also addressed this issue, I'm not sure.
7 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Michael
Unfollow Follow Michael
Michael Adcock • This is a great question. And the myriad of responses – mostly biased to rejecting the part – show that there is value in bringing more attention to this subject.

From Daniel’s original posting: If there is a company spec. that inspection must use a 10X precision over the specified tolerance, then you must report to the 4th decimal place for the .245-.250 limit dimension. That being said the Absoluteness of ASME Y14.5’s definition of dimensional limits indicates that all unspecified decimal places are zeros. You can look at the 5th, 6th, and 7th decimal place if you have the capability to do so.

This discussion likely wouldn’t have taken place – at least not as the conflict that I read – had your company employed dimensional measurement planning as defined in ASME B89 standards. The specifications on your drawing are absolute requirements, as so many have mentioned. However they’re only ½ of a conversation. The designer has asked for something and that something is absolute, but you cannot produce or verify to an absolute. As Mark Foster mentioned the inspector must always contend with measurement uncertainty. That’s one issue. The extent of the specification is another issue. Let say your .245-.250 spec is the thickness of a plate 10 inches by 10 inches; the absolute specification of the .245-.250 dimension applies at each actual local size (two opposing points). There are an infinite number of measurements in each direction for a grand total of infinite squared actual local sizes. Again the spec. from the drawing is absolute, but nobody actually expects the inspector to measure each of these – so what is a sufficient number? This is one role of the DMP (Dimensional Measurement Plan). Another is to assign a distribution for uncertainty or gage tolerances. In some DMPs the measured values of .2501 would be rejected, but in others they would be acceptable. Why this discrepancy may exist between DMPs is a longer conversation than I want to delve into now, but I felt it important to address that we’re not considering that the drawing specs are quite impractical. We need to do more to make sure that we’re participating in the other ½ of these conversations and not allowing inspectors to become inventors by developing the DMP they want.

I actually just took a phone call related to this very issue. Mexico is measuring 3 points on a hole and calling the parts out of spec. The supplier is measuring 12 points on the hole and calling the parts good. Who invented these numbers, and why weren’t they agreed to before the part was sourced? The DMP has the opportunity to bring a lot of value to both customer – by assuring minimum quality requirements, and the supplier – by assuring that products continue to flow out and payment continue to flow into the company.
5 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion2 .
Michael
Unfollow Follow Michael
Michael Adcock • John,
It's not just a measurement uncertainty thing. Both ASME Y14.43 and ASME B89.7.2 permit the same three basic dispositions to be devoped for part inspection.
Optimistic: you are permitted measurements significantly beyond the spec limits.
Tolerant: you are permitted measurements slightly beyond the spec limits
Absolute: No measurements are permitted beyond the spec limits

Y14.5's absolute nature for dimensional limits does not establish a disposition for gaging parts, and there are pros and cons to each.

My earlier point as without a DMP the spec .245-.250 could cause .245, .246, .249, and .250 to fail or it could permit .244 and .251 to pass.

We haven't even discussed algorithm based measurement vs. attribute gages and how differences in the applied algorythm can cause .244 to report .246 on the same CMM. Dimensional measurement planning is necessary if we are to operate on more than the illusion of quality.
5 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MikeUnfollow Follow Mike
Mike Matusky • Dimensional measurement planning is necessary if we are to operate on more than the illusion of quality.

Mexico is measuring 3 points on a hole and calling the parts out of spec. The supplier is measuring 12 points on the hole and calling the parts good. Who invented these numbers, and why weren’t they agreed to before the part was sourced? The DMP has the opportunity to bring a lot of value to both customer – by assuring minimum quality requirements, and the supplier – by assuring that products continue to flow out and payment continue to flow into the company.


Not just CMM dimensional planning, even a height gauge on a rock can have the same problems. In Detroit we used to do a lot of single part 10 trial Gauge R studies. We tried to measure every point that we would ever collect data on using that gauge. Though I can't claim that they revealed the Uncertainty of the Measurement they sure revealed to me how uncertain I should be of the measurements, Sometimes they drove changes in datum schemes all the way back to the assembly equipment.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Michael
Unfollow Follow Michael
Michael Adcock • Then where is the mental block?

"A .250 +/-.005 dia checks .2500006, is it bad??????? Based on this absolute rule of 14.5 the answer is yes but check it again and it might check .2499993, now its good."

B89.7.2 specificly states that best two out of three measurements is not grounds for rejecting a measurement as an outlier. We should also be in agreement that given a spec. .245-.250 measurements of both .2500006 and .2499993 might be acceptable or rejectable - it all comes down to the DMP.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Michael
Unfollow Follow Michael
Michael Adcock • Please elaborate.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Michael
Unfollow Follow Michael
Michael Adcock • John,
Thanks for the explanation. You're right I didn't get that because your post included the statement "Forget the measurement uncertainty thing for now." So I didn't consider that you wanted to discuss variations do to uncertainty (of course you know there are dozens of factors that contribute to measurement uncertainty). I do appreciate the explanation - Thanks.

I won't cite your statement, but we both know it's not merit that gets one a spot at the table in the ASME commitees. I'm not saying you don't deserve your appointment to these committees, but "I'm on the committee" is hardly sufficient cridentials to settle an issue in this forum.

Your fun, I'll look forward to more dicussions.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Mark
Unfollow Follow Mark
Mark Foster • Michael, I know that you didn't mean it this way, but I can vouch for John. He is eminently qualified for being on the committees on which he serves, and is extremely knowledgeable regarding GD&T, Dimensional Management, and measurement uncertainty (and many other related subjects.

Let's all do our best to keep the discussions on topic and refrain from personal commentary on one another. It is always good to support your argument's position with facts and supporting credentials, but we all must remain professional.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion2 .
Michael
Unfollow Follow Michael
Michael Adcock • All,
I tried to be clear that I was not making a personal statement about John's appointment to the committees. I just re-read my posting. I was only remarking on the justification "I'm on the committee". I'm not asking anyone to agree with me publicly, but this is often an attempt to end a conversation, by cutting off the debate. I've nejoyed enough of these discussions which require sound logic, data, science, and citing standards to resolve. Without that we often end up agreeing that their is no solution and that more work must be done to develope a solution.

I may have come across as flipant with my last statement. I'm was just trying to convey that I'm not upset by any of this and I'm enjoying this discussion.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
Nathan
Unfollow Follow Nathan
Nathan Corliss • And all this time inspection / engineering is arguing about passing and failing a part, manufacturing is still making parts near the tolerance limits instead of offsetting the necessary tools to correct. I say this in gest to lighten the mood.

during my days as a quality manager of a defense contractor, any discrepancy between customer and supplier went into immediate containment, with MRB board to review the functional risks of these near limit / Out of spec parts. Consensus agreement to scrap, rework, use as is, or engineering change of tolerances was agreed upon. We always needed a corrective action plan to meet a determined CPK level moving forward or 100% verification based on customer agreed upon measurement technique.

Measurement uncertainty was always understood, along with understanding regardless of how many points taken, we can never truly measure every point to determine pass fail, plus the joys of form error. We would also introduce the functional gauges (go / no go), again understand that maximum extreme points either passed or failed, but this didn't tell us form nor help with process control, offsets to make.

Oh, MIL-STD--120 is for gauge design, the one I was referring to was specific to calibration but found the statement wasn't 4:1, but "25%" or 4:1, I say potato, you say freedom fries. MIL-STD-45662A was the one I used. There are so many MIL-STD's on the same subject. If anyone ever wants to relieve these treasures, you can go to www.assistdocs.com and review them for free.

Ultimately, businesses will make the final determination on what is acceptable amount of risk or uncertainty, hopefully making informed and educated decisions on the uncertainty or not, minimizing the risk of defective product getting to the public and avoiding serious situations.

Mark, I always remember and use your example of the defective latch that didn't latch.
4 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion1 .
MICHELUnfollow Follow MICHEL
MICHEL KUGLER • hi, I use "measuring tool sensitivity < TOLERANCE/10 " ; "Measurement uncertainty < TOLERANCE /4" (old school) ; there is some indications in: NF X 07-010:2001 ; NF EN ISO 10012 ; speaking of tolerances in inches it is not easy I prefer milimeters...
3 days ago • Unlike • Like • Reply privately• Flag as inappropriate • Flag as promotion0 .
MICHELUnfollow Follow MICHEL
MICHEL KUGLER • NF E 02-204 , NF EN ISO 14253-1

__________________
借用达朗贝尔的名言：前进吧，你会得到信心!
[url="http://www.dimcax.com"]几何尺寸与公差标准[/url]