MSA (Measurement System Analysis) for Characteristics with One Tolerance Limit

[Adryanna10]

Hello all,

I have the following problem: a need to make a MSA (Cgk and R&R) study for a characteristic that has just upper tolerance limit. Far as I know of you can not realize the MSA study for characteristics that has just one of tolerance limit. However, I would like to now if someone had this issue before and what I can do in this case?

Thank you in advance for your help.

 

[Jen Kirley]

Re: MSA for charatesristics with one tolerance limit

Welcome to the Cove!

My understanding of MSA is that it's not limited to 2-sided specifications. The studies are about measuring and analyzing error, which can certainly be done with a one-tailed test.

Does anyone disagree?

Raytheon put out a nice little tutorial and we have a number of templates in the Post Attachments List (see the green button in the header) for your use.

 

[Jim Wynne]

<snip> I have the following problem: a need to make a MSA (Cgk and R&R) study for a characteristic that has just upper tolerance limit. Far as I know of you can not realize the MSA study for characteristics that has just one of tolerance limit. However, I would like to now if someone had this issue before and what I can do in this case? <snip>

Whether the tolerance is bilateral or unilateral, it's the total tolerance spread that's used in GR&R. For example, it the tolerance is ?.005, the total tolerance is .010. If the tolerance is +0/-.010, the total tolerance is also .010.

You might be thinking of Cpk calculation which does assume a bilateral tolerance.

 

[Miner]

I posted a blog entry on my thoughts on MSA for one-sided tolerances.

 

[Uchiha]

You might be thinking of Cpk calculation which does assume a bilateral tolerance.

I don't think Cpk calculation assumes any bilateral tolerance. What is more, only Cpk (or Ppk) can be calculated in case of a unilateral tolerance. Cp (Pp) cannot...

Someone correct me if I am wrong.

 

[Jim Wynne]

I don't think Cpk calculation assumes any bilateral tolerance. What is more, only Cpk (or Ppk) can be calculated in case of a unilateral tolerance. Cp (Pp) cannot...

Someone correct me if I am wrong.

The purpose of Cpk is to provide an indication of how well a process is centered between upper and lower specification limits. Cpk/Ppk, which are dubious metrics to begin with, make no sense in terms of a unilateral tolerance.

 

[Uchiha]

The purpose of Cpk is to provide an indication of how well a process is centered between upper and lower specification limits. Cpk/Ppk, which are dubious metrics to begin with, make no sense in terms of a unilateral tolerance.

Exactly. And that is by calculating two Cpks to determine to which side is your process "shifting"...

So in case of, let's say, only an upper tolerance limit, calculating the upper Cpk will let you know how close is your data to the upper TL. Obviously the lower side Cpk is not defined in this case, neither is the Cp (Pp)...

 

[Jim Wynne]

Exactly. And that is by calculating two Cpks to determine to which side is your process "shifting"...

So in case of, let's say, only an upper tolerance limit, calculating the upper Cpk will let you know how close is your data to the upper TL. Obviously the lower side Cpk is not defined in this case, neither is the Cp (Pp)...

No. Cpk assumes a bilateral tolerance. If there is no upper and lower tolerance limit, there is no target to use as a reference point. The idea is to determine how far the mean is from the nominal value. How can you calculate upper and lower Cpk if only upper or lower exists?

 

[Bev D]

Well the Cpk/Ppk formulas don't include the target/nominal. htey only include the upper and lower spec limits. So Jim can you clarify your understanding?

 

[Jim Wynne]

Well the Cpk/Ppk formulas don't include the target/nominal. htey only include the upper and lower spec limits. So Jim can you clarify your understanding?

Instead of nominal I should have said "mean."