Brain Teaser - nominally symmetrical parts - pass one way, but fail the other? 1

[Nescius]

Consider a round disk, let's say roughly the shape of a hockey puck. Let's name one flat face as datum feature A. Then, let's give a basic dimension for the thickness and control the other flat face with a profile tolerance.

To make the thought experiment cleaner, just assume that both faces are very flat and parallel, relative to the profile tolerance.

We pull the part out of a bucket, randomly oriented, and inspect the part by choosing one of the flat faces as datum feature A...the part passes. We then flip the nominally symmetrical part over and inspect again...but it fails.

I believe this is possible. If any of you agree, what then? Is the part "good" or "bad"?

 

[Vishal2015]

What is the tolerance value of flatness and profile of surface?
If flatness tolerance zone is larger then profile, then it could happen.

 

[Nescius]

Assume the profile tolerance is much larger than the flatness tolerance. If you want numbers, let's say that datum feature A has a flatness tolerance of 1 unit. Let's say the profile tolerance is 10 units.

Let's also assume that both flat surfaces meet the flatness spec of 1, so this is not the cause for failure when the part is flipped over. Going further, let's throw a parallelism tolerance of 1 unit on the top surface, relative to A. So, no weird effects from extreme orientation error.

 

[greenimi]

Per Mark Foster--on linkedin website:

"This symmetrical part discussion is not well-covered in the standard, so this is my opinion, not a Y14.5 edict. I believe that if you cannot positively identify which end is A and which end is B, then the part should check good from BOTH ends, and therefore each end would need to be held to the tighter tolerance. My reasoning is that if the part is designed to be truly symmetrical, then it should work in its assembly/function regardless of which way it is assembled -- the entire definition of a symmetrical part. I train people to *purposefully* destroy the symmetry of a part (in an inexpensive manner) just so that there can be no argument about which side is which and so that there is only ONE correct orientation of the part -- UNLESS it is the absolute design intent that the part be able to be assembled either way and work the exactly the same when assembled either way.
From the standpoint of the same mindset as "candidate datums" on a rocking datum feature, I agree that if you inspect from one end of our "symmetrical" cylinder in question and find a non-conformance, that you should then inspect from the other end to see what you get. And I further agree that (still in the same "candidate datum" mindset) if you find the part conforming when you hold it from the other end, then it is a conforming part. I think that these conclusions or methodologies can be best supported by what is currently in the standards.

There are a couple of problems to consider, though. One is that it may be impractical, impossible or even illegal (i.e. it may change the properties of the feature rendering it non-functional) to mark the feature that you found to create the conforming scenario. Another is, it may be the design intent that the part specifically function from either end. In other words, the design person wants assembly to be able to put together the assembly without worrying which way they are inserting the pin. Therefore, if *either* end used as the "datum feature" results in a non-conformance, design would like the part rejected.

I suppose in the latter case, we could recommend that the designer purposefully call *each* end its own separate datum feature and profile each opposite end relative to each opposing datum feature. Some would view this as double-dimensioning, even though I think that we have proven through our discussion that it would technically not be.

My bottom line question from where this thread has migrated, is: How should we definitively communicate that we *desire* a truly symmetrical part versus that it is ok for the part to only conform one way even though we cannot (for some reason) mark the part or destroy the symmetry on purpose?"

 

[CheckerHater]

Yes, it's possible.

On the enclosed picture both sides have the same flat/parallel variation. The trick is in highest points you use to establish the datum.

Even if you argue that part may be leveled, it will never be perfect, so part will always be "larger" in one position then in the other.

Naturally, this is purely theoretical example, one out of infinite number of imperfect shapes out there.

"For every expert there is an equal and opposite expert"
Arthur C. Clarke Profiles of the future

 

[JNieman]

When possible, features that serve no other purpose other than ruining symmetry, can be designed into the part.

If your hockey puck is a lathe turned part, a simple non-precise groove on the outside diameter can indicate direction, and also be referenced in assembly instructions further down the line.

Of course, designing it so that it will truly work either way would be simplest. Maybe that means it'll fail one way and not the other if it's at an extreme of form / surface condition - but that is the price for a symmetric part that can be used either-way. If it /can/ be used either way, it should be /able/ to be used either way. Much more eloquently stated by the quote from Mark Foster, which greenimi posted.

 

[Nescius]

Thanks, all! I stumbled upon the dilemma in a real-world application where adding another feature to the part was simply impossible. Even if I could've explained the situation to my boss and the customer, cost would have driven the decision. Furthermore, the parts would have been installed "backwards" anyway. The part is small and installed on an assembly line by a human. I doubt the customer wants to add that headache to their process.

As to my opinion on whether or not the part should pass in both orientations in order to be called "good", well, that depends.

If the failing "low spots" on one of the faces could be important to the part function, I suppose you'd better consider that. Even then, my initial sketches show that these low spots can never violate the profile tolerance by more than the flatness tolerance, and that's only if the face acting as the datum feature is perfectly flat and the hockey puck is absolutely as short as allowed in the "passing" setup. My sketches assumed good parallelism, though. So, I could be mistaken.

If the entire surface in question contacts a mating part, though, it might really be the tangent plane we're concerned with. This is the case with my part. Even if it technically fails inspection one way, if it passes when flipped over, it will work in the assembly. My part isn't shaped like a hockey puck, but the hockey puck example distills my issue nicely.

See the attached for my thoughts

 

[CheckerHater]

Even though symmetrical parts are not well covered in the standard, it doesn't mean they cannot be covered in your drawing.

Add the note to the drawing that part must pass inspection when either side is used as a datum "A" if the application is critical.

"For every expert there is an equal and opposite expert"
Arthur C. Clarke Profiles of the future

 

[3DDave]

1) What's the part made of? There is temporary and permanent marking that is smaller than the allowable variation. Magic marker or laser etch, for example.

2) Have the inspector place the part in a carrier that maintains acceptance orientation and have the repair instruction require discarding the item and replacing from the carrier. Billions of dust sized LEDs are installed this way every year.

 

[Nescius]

3DDave,

The part is made of steel. Simply "marking" the part is a good suggestion in cases where it matters.

Re: a part "carrier": In my case, the carrier will be a bin...containing 1000's of parts. So, no chance of maintaining orientation that way. That is another good suggestion, though.

 

[Sparweb]

Thanks, I learned something today!

STF

 

[KENAT]

I have deliberately added features to otherwise nominally symetrical parts in order to orient the datums etc. Classic is on a square or rectangular part or similar I might add a chamfer on one edge to aid orientation. Doesn't directly address your question but pointing out that I generally avoid such situations when possible due to the inherent ambiguity which arguably has no place in a pseudo legal document.

Posting guidelines faq731-376

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

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