CMM Guard Banding: Improves OCC or DET?

[TroubledTom]

We are in the midst of a debate on what gets the "credit" for guard banding.
We have a process with a Cpk of 1.25 which is undesireable, but due to the tight tolerances and limits of our equipment, it is the best we can do for now.
All product is being 100% inspected on a CMM.
The CMM has shown a GR&R of 29%, which isn't good either.

The above has resulted in an RPN that indicated the need for corrective action. As we are at our limits on the process (OCC) and the measurement system (DET), it was decided to "guard band" the part tolerance. This guard banding is done by progamming the CMM with reduced part tolerances. The advantage is removing the borderline and defective parts to greatly decrease the chance of an out-of-specification part reaching our customer. The disadvantage is we're scrapping good parts. To us, the advantage outweighs the disadvantage.

Now that all the details are laid out, the question is: When we re-rank the factors of the RPN number to reap the benefits of the guard banding, which column is affected? Some say the gage detection has improved as it's culling all the bad parts. Others think that the capablitiy of the "modified process" has improved by virtue of the high conformity or "virtual Cpk".

Any thoughts?

 

[Miner]

It is really splitting hairs. While I tend to lean toward the Detection side of the argument, I understand the rationale of both arguments. As long as you take one approach only (not both) and are consistent, the end result, a reduced RPN is the same.

 

[Ron Rompen]

with all due respect, I disagree with Miner. The OCC score indicates how frequently in your process the defect occurs. Using the CMM for 100% sorting is a detection process, and does not affect how frequently the defect is produced.

Depending on how you have your detection station configured, this would (probably) result in a DET score of 4 at best.

 

[TroubledTom]

Thanks for the input gang.
We have resolved the issue by not re-ranking the OCC. An example of the reasoning was given by a manufacturing process that produced a part and could measure the weight of the part to determine if it was in spec. Having the ability to automatically reject the parts as made does not make a good process. Bad parts are still being made, so OCC should rate poorly.
Its tough to quantify the change in DET rating, by adding the guard banding. We can reference the GR&R, but just how much better the guard banding makes the rating change is a tough one. We still have some disagreement to resolve on that one.