For sixty hours over the next 20 days I’m going to be 100% sorting parts unless I can come up with another way to qualify or certify that the parts are good. My customer’s claim the parts are out of round, beyond tolerance, at a rate of 3 out of 10. The 200 new parts I have measured use less than half the tolerance. The customer has yet to provide me with any of the failed parts. I can’t validate the compliant. For years I have been accepting the roundness on a small C=0 sampling. How do I justify the hours of inspection?

For sixty hours over the next 20 days I’m going to be 100% sorting parts unless I can come up with another way to qualify or certify that the parts are good. My customer’s claim the parts are out of round, beyond tolerance, at a rate of 3 out of 10. The 200 new parts I have measured use less than half the tolerance. The customer has yet to provide me with any of the failed parts. I can’t validate the compliant. For years I have been accepting the roundness on a small C=0 sampling. How do I justify the hours of inspection?

Rick,

Before you engage in a costly 100% inspection excercise....:mg:

Do you and your customer have a common understanding/agreement what "round" means and how it should be verified?
Do the specifications (drawing) make a reference to Y14.5M-1994?

Stijloor.

For sixty hours over the next 20 days I’m going to be 100% sorting parts unless I can come up with another way to qualify or certify that the parts are good. My customer’s claim the parts are out of round, beyond tolerance, at a rate of 3 out of 10.

That is why measuring diameters should include measuring around the part, not just one diameter. The "Law of Diameters" is that there is an infinite number of diameters in a round feature. If you just measure one, you will measure a good one, and a customer will measure a bad one. Roundness is the problem. That is the one of the functions of the X hi/lo-R chart. No way to sample the problem after the fact. 100% is all you have to work with.

The customer has yet to provide me with any of the failed parts. I can’t validate the compliant.

Now that should have been done prior to accepting the complaint and beginning sorting. It becomes a real problem now.

For years I have been accepting the roundness on a small C=0 sampling.

Ouch...sampling error. Much better to control the process. If it is a precision machined product, I recommend reading: Statistical process control for precision machining (http://elsmar.com/Forums/blog.php?b=79)

Rick,

Before you engage in a costly 100% inspection excercise....:mg:

Do you and your customer have a common understanding/agreement what "round" means and how it should be verified?
Do the specifications (drawing) make a reference to Y14.5M-1994?

Stijloor.

1) No the customer has not returned the identified bad parts. My SQE has promised me the parts, yet demands production. A visual process audit shows parts are made to the established documented method. My dimensional inspect verifies the process.
2) “Roundness” is called out on the print as Spherical within .004”. The production release revision is date 6/97.

It’s a practice of this customer that when a problem or perceived problem is identified, they return all their stock rework or replacement, demanding acceptable parts in the morning to prevent their line-down situation

too common in certain industreis and certain customers.

unfortunately, without the "rejected" parts you can't develop an effective screen. even 100% might be a waste.

"That is why measuring diameters should include measuring around the part, not just one diameter. The "Law of Diameters" is that there is an infinite number of diameters in a round feature."

We, the customer and me have agreed to qualify the Spherical by subtracting two diameter measurement, the second taken after rotating the part 90 degrees. They use a comparator, I use a video machine (MicroVu)

One possible difference might be how the two spots to measure are chosen. You could choose one spot at random and then rotate 90 degrees. Or you could rotate until you find the high point, then rotate 90 degree. Clearly the second method will give a larger average difference.

But that wouldn't explain why they are suddenly seeing a problem.


Tim F

1) No the customer has not returned the identified bad parts. My SQE has promised me the parts, yet demands production. A visual process audit shows parts are made to the established documented method. My dimensional inspect verifies the process.
2) “Roundness” is called out on the print as Spherical within .004”. The production release revision is date 6/97.

It’s a practice of this customer that when a problem or perceived problem is identified, they return all their stock rework or replacement, demanding acceptable parts in the morning to prevent their line-down situation

Rick,

When you say "spherical", are you referring to ball-shaped parts?

Any reference to a standard?

Stijloor.

The customer has yet to provide me with any of the failed parts. I can’t validate the compliant. No parts, no evidence, the complaint isn't valid. I have never done a sort or anything until a customer could provide me with the defective part (or at least one if more than one are claimed to have been found).

We, the customer and me have agreed to qualify the Spherical by subtracting two diameter measurement, the second taken after rotating the part 90 degrees. They use a comparator, I use a video machine (MicroVu)

Still, the two diameters you choose may have been different than the two the customer measured. Two out of an infinite number of diameters can have you chasing your tail, and can get differing data. Sampling has limited effectiveness to begin with, and sampling with this measurement error will make sampling even less effective. Spherical is just that much more of a pain on top of it all.

You should be running to a restricted tolerance to accommodate lab to lab measurement variation. One approach to control spherical dimension is to use a template on a comparator at 75% of the tolerance, rotate the part in a concentricity gage and ensure the part fits entirely in the template - or use a similar approach on your vision system. I did that with ground compound curves, and it worked very, very well.

Rick,

When you say "spherical", are you referring to ball-shaped parts?

Any reference to a standard?

Stijloor.

Opps, guess that an important part of the puzzle, yes, the part is round rubber ball.

There's no standard reference

Oops, guess that an important part of the puzzle, yes, the part is round rubber ball.

There's no standard reference

Whoa! A full sphere is even more frightening as far as lab to lab correlation. At least you are using non-contact on rubber. Without the rejected parts, who knows what they saw? A dimple?

I still think fixturing the ball (between two centers) and rotating it to get more diametric data would be a better failsafe for roundness issues.

Opps, guess that an important part of the puzzle, yes, the part is round rubber ball.

There's no standard reference

Rick,

Then, any two-point measurement (micrometer, caliper or other) is incorrect. Spherical means that the variations of all points on the surface of the ball must be equidistant from a common centerpoint within .004".

It requires in actuality a 3-D type measurement.

Stijloor.

Rick,

Then, any two-point measurement (micrometer, caliper or other) is incorrect. Spherical means that the variations of all points on the surface of the ball must be equidistant from a common centerpoint within .004".

It requires in actuality a 3-D type measurement.

Stijloor.

I think he can use his vision system, as long as he he can spin the part on an axis and take more than one slice with his diameter with roundness data. Surely, if his customer is using a comparator, he should be able to verify enough circular sections to protect himself. He just has to be sure he is not deforming the ball from holding between centers - have to be very careful.

One other approach from rubber manufacturers is freeze the part to measure it.

Rick,

Please let us know how this situation "turned" out for you.
I'm curious....;)

Stijloor.

Rick,

Please let us know how this situation "turned" out for you.
I'm curious....;)

Stijloor.

Will do...thanks for the discussion

Can you go to the customer?

If you can get to their plant to see how the part is used, and agree how you both will measure the part, a lot of pain could go away

If travle cost is too much, can you exchange video, teleconference, photos, etc.

Go to gemba

Can you go to the customer?

If you can get to their plant to see how the part is used, and agree how you both will measure the part, a lot of pain could go away


Sounds like they tried to do that, but there eeded to be a better understanding of the measurement technology that they both missed with the two measurements at 90 degree trying to define an infinite array of diameters.

The failed parts are in transit to me, I should get them Friday. Can't wait to measure them. I suspect the balls are warped or distorted, they were assembled. The claim of being “out of round” is just a nail to hang the complaint on.

Rick - do you primarily manufacture rubber parts? Do you ever freeze the parts (as in liquid nitrogen or such) to allow you to measure them with minimum deflection?

As suspected the balls are not out of round, they are undersized. Sorting for undersize is much faster. The error occurred when final inspector failed to account for the error of uncertainty in using a hand held caliper. A ring gage will replace the caliper.