What does a tolerance really mean?

[breyman]

Hi,

We have debate going on in our company right now as to the true defintion of a simple tolerance.

Here's the very simple question:

Say you had a rod with a diameter dimension and tolerance of 1.000" +/- 0.005", what is the exact upper bound that that part can be and still satisfy this tolerance?

One group says that the part can be no larger than exactly 1.005", meaning that if it was actually 1.00500000001" (assuming you could measure to this accuracy), it is out of spec.

The other group says that as long as the first 4 digits aren't more then 1.005", than it is okay, i.e. 1.00599999999" would be "in spec"

I'm personally on the side of the first group and this has been my understanding all along, but there's oddly enough support/debate that I'm confused now. What are your thoughts?

 

[KENAT]

The first group are correct per ASME Y14.5M-1994 section 2.4 "All limits are absolute" - if this is your applicable standard.

I don't know for sure what Iso says.

Posting guidelines faq731-376

http://eng-tips.com/market.cfm?

(probably not aimed specifically at you)
What is Engineering anyway: faq1088-1484​

 

[powerhound]

I second KENAT's answer. The first group is correct per the ASME standard. No question about it.

Powerhound, GDTP T-0419
Engineering Technician
Inventor 2010
Mastercam X4
Smartcam 11.1
SSG, U.S. Army
Taji, Iraq OIF II

 

[ctopher]

I agree with KENAT.
IMO, it is a weird argument. I have not heard of others really caring about that type of accuracy.

Chris
SolidWorks 10 SP4.0
ctopher's home
SolidWorks Legion

 

[CheckerHater]

The first group is correct.
Regarding the accuracy: you use best tool available to you; If you have a 0.001 IN caliper, 1.005 is good, 1.006 is bad; if you have 0.0001 IN micrometer, 1.0050 is good, 1.0051 is bad.
But don't go by micrometer alone, as you also have to check Straightness, Cylindricity, Out-of-roundness, etc to satisfy Envelope requirement.
Good luck!

 

[TheTick]

Group B should be encouraged to find work with the competition.

 

[axym]

The first group is correct. End of story.

Regarding measurement variation, one could argue (and I would) that a borderline reading is cause for rejection. If the caliper reads 1.005, then the actual dimension could be anywhere from 1.00450000000 to 1.00549999999. We're not sure that the part is conforming, so we have to reject it (this is my QA background coming out).

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[MintJulep]

a borderline reading is cause for rejection

No, a borderline (indeterminate) reading means that you are using the wrong tool for the job.

If you are inspecting a requirement with a max limit of 1.005 then you need a tool that provides accuracy and precision beyond 3 decimal places.

 

[axym]

MintJulep,

Let's say I'm using a very accurate device that can measure to the nearest millionth of an inch. So the sixth decimal place is significant. If the reading is 1.005001, then we reject the part because it's definitely out of tolerance (larger than the upper limit). If the reading is 1.005000, the actual dimension could be anywhere from 1.0049995 to 1.0050005. We're not sure that the part is conforming, so we reject it.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[ajack1]

I guess that raises the question is the job of inspection to prove a part is right or prove it is wrong?

 

[MintJulep]

Put it the refrigerator for half an hour and measure it again then.

 

[weavedreamer]

Unless 30 minutes of refridgeration is specified on the print, measurements should be taken at precisely 68°F ±2°F if memory serves correctly.

 

[axym]

ajack,

The answer to that question depends on which side of the wall you're on - the manufacturing side or the QA side ;^)

MintJulep,

If we're not careful this could degenerate into the usual debate over what the impact on function is when a dimension is out of spec by an extremely small amount. Engineering drew their line in the sand, and QA enforces that line. It's not up to QA to judge whether a tiny amount over the line is reasonable. There's no end to that. If it's over the line by a millionth of an inch, it's over the line.

To get a first down, you have to advance the ball by ten yards from the line of scrimmage. If the offensive team thinks they gained 10 yards and the defensive team thinks it was under 10, they don't just round the gain to the nearest yard or make the decision based on the markings on the grass. They bring out the chains. If the tip of the ball is half an inch past the chain, it's a first down.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[MintJulep]

Axym,

I'm bored out of my skull today, a good debate sounds fun.

Which side do yo want?

 

[drawoh]

Let's say I'm using a very accurate device that can measure to the nearest millionth of an inch. So the sixth decimal place is significant. If the reading is 1.005001, then we reject the part because it's definitely out of tolerance (larger than the upper limit). If the reading is 1.005000, the actual dimension could be anywhere from 1.0049995 to 1.0050005. We're not sure that the part is conforming, so we reject it.

Reject it.

If .000001" over is acceptable, what distance over the specified value is not acceptable? Do you really want to work to a standard that states something like this?

A crude rule of thumb is that your measuring device should be ten times as accurate as your tolerance. If you are able to detect error with your precision, above, you probably are wasting time.

This sounds a lot like one of those philosophy 101 questionsl like "How far is up?".

JHG

 

[Belanger]

My vote goes with MintJulep.

"A borderline (indeterminate) reading means that you are using the wrong tool for the job.

If you are inspecting a requirement with a max limit of 1.005 then you need a tool that provides accuracy and precision beyond 3 decimal places."

While all instruments have some uncertainty, you shouldn't let the first digit with that uncertainty be the last decimal place given on the drawing.

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

 

[axym]

drawoh,

It kind of seems like you're agreeing with me, but I'm not sure.

I agree with the 10:1 rule of thumb. The more accurate and precise the measuring device, the fewer borderline good parts will have be rejected.

I'm not sure what you mean about wasting time if we're able to detect error with the high precision.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca

 

[breyman]

Thanks guys. I agree with "TheTick" - we should send those who thought that option B was correct to the competition.

It was one of those strange, floor-dropping-out-from-below you moments, when we actually discovered that we disasgreed on such a fundamentally simple issue

Anyway, problem resolved, and it's always fun to be right.

Thanks again!

 

[fsincox]

I am with the group, here, on this issue!
Frank

 

[axym]

I still disagree with the statement that a borderline (indeterminate) reading means that you're using the wrong tool for the job. No matter how precise the measuring tool, there will still be a borderline (indeterminate) reading.

If I'm using a caliper, the borderline reading is 1.003. If I'm using a micrometer, the borderline reading is 1.0030. If I'm using my hypothetical device from my earlier post, the borderline reading is 1.003000. All of these readings are indeterminate, because the actual dimension could be less than 1.003 or it could be more than 1.003. We don't know for sure.

Evan Janeshewski

Axymetrix Quality Engineering Inc.

http://www.axymetrix.ca