Part Variation, Unilateral Tolerance, Destructive Testing - Gage R&R Experts pls help

[Miner]

Re: Gage R&R Experts...Part variation, unilateral tolerance, destructive testing

As I said before, the same gages are used for Process control and Inspection.
I'm certain that I can't get parts which represent the entire process spread? So, How should I collect my samples?

Is it ok to collect those 10 parts without representing the entire process spread?

If so, does the standard deviation of part variation from the process capability study and the 10 parts that I select be from the same lot?

Thanks

Do you have data from a long term capability study available?

For the %Tolerance (inspection) decision, it makes no difference how you select the parts.

For % Study Variation (process control) decision, part selection is critical. You have two options:

• Option A: Find 10 parts that represent your process variation. Can you collect 1 part per production setup? This should represent your expected variation.
• Option B: Use existing information on the process variation from a long-term capability study (if you currently collect data for process control, use that to calculate the long term variation). Ignore the software calculated % Study Variation. Either manually calculate a new % Study Variation as described above, or if the software supports it (e.g., Minitab) use the supplemental metric that is calculated using manually input process variation data.

 

[aproddutoor]

Re: Gage R&R Experts...Part variation, unilateral tolerance, destructive testing

Do you have data from a long term capability study available?

For the %Tolerance (inspection) decision, it makes no difference how you select the parts.

For % Study Variation (process control) decision, part selection is critical. You have two options:

• Option A: Find 10 parts that represent your process variation. Can you collect 1 part per production setup? This should represent your expected variation.
• Option B: Use existing information on the process variation from a long-term capability study (if you currently collect data for process control, use that to calculate the long term variation). Ignore the software calculated % Study Variation. Either manually calculate a new % Study Variation as described above, or if the software supports it (e.g., Minitab) use the supplemental metric that is calculated using manually input process variation data.

Miner,
We decided to do the GRR on only 1 of 4 critical dimensions.
This dimension happens to have a unilateral tolerance.
The specification is .020 inches minimum.
I collected parts between .020 and .023 inches.
I haven't done the study yet but do I have to consider anything else?
Will it work?

For the same part, I started doing the Attribute Study. This has been a big problem because the parts are too delicate and there are about 14 people who need to take the GRR. I'm not even half way through and I've observed that the parts have gone so bad that I had to replace them. I can't keep on changing because it is difficult to find the same replacement. I started with 3 trials initially but later reduced it to 2 trials because the operators were failing the GRR.

The operators inspect the parts using a microscope. The wire is welded to a small ring. The ring and the weld should be inspected as per the print.

Do yo have any solution?

Thanks

 

[Miner]

Re: Gage R&R Experts...Part variation, unilateral tolerance, destructive testing

Miner,
We decided to do the GRR on only 1 of 4 critical dimensions.
This dimension happens to have a unilateral tolerance.
The specification is .020 inches minimum.
I collected parts between .020 and .023 inches.
I haven't done the study yet but do I have to consider anything else?
Will it work?

If you are using Option A or Option B, there should be no issues with calculating % Study Variation. However, the current MSA does not address one-sided tolerances. You would have to construct a Gage Performance Curve or use Donald Wheeler's watershed tolerance approach. (See files attached to my blog entries in the two links).

For the same part, I started doing the Attribute Study. This has been a big problem because the parts are too delicate and there are about 14 people who need to take the GRR. I'm not even half way through and I've observed that the parts have gone so bad that I had to replace them. I can't keep on changing because it is difficult to find the same replacement. I started with 3 trials initially but later reduced it to 2 trials because the operators were failing the GRR.

The operators inspect the parts using a microscope. The wire is welded to a small ring. The ring and the weld should be inspected as per the print.

Do yo have any solution?

Thanks

I would start by reducing the number of Operators by selecting three representative operators. This will tell you whether the test itself is acceptable. If you need to go on to qualify particular operators, you can add to the study with different parts.