When is GRR (Gage R&R) required? AIAG MSA Manual Version 4

[tdvd101]

Hi, I am new to the forum - hope this has not been addressed before.. anyway:
On the bottom of page 77/78 of version 4 of the MSA manual it states that ‘when SPC is being applied for process control or collecting process data, and the control chart indicates that the process is stable and the total variation is acceptable, the measurement system can be considered acceptable for this use and does not require separate re-evaluation. If an out-of-control condition or non-conformance is found in this situation, the first thing that should be done is to evaluate the measurement system’.

Do we thus take it that it is not necessary (albeit beneficial) to perform a GRR study if the process is in a state of statistical control (providing of course the process limits are satisfactoriy within the product tolerance requirements?
Would appreciate thoughts on this, TQ

 

[bobdoering]

Re: MSA Version 4 when is GRR required?

On the bottom of page 77/78 of version 4 of the MSA manual it states that ‘when SPC is being applied for process control or collecting process data, and the control chart indicates that the process is stable and the total variation is acceptable, the measurement system can be considered acceptable for this use and does not require separate re-evaluation. If an out-of-control condition or non-conformance is found in this situation, the first thing that should be done is to evaluate the measurement system’.

Do we thus take it that it is not necessary (albeit beneficial) to perform a GRR study if the process is in a state of statistical control (providing of course the process limits are satisfactory within the product tolerance requirements?

That concept is goofy, because you should know your gage R&R before you choose it to use for SPC...not the other way around. Unbelievable, and plain sloppy. It is part of the Total Variance Equation you should evaluate as a part of capability, which also should have been done prior to SPC to determine which charting methodology is appropriate, and where the sources of variation are coming from. That is simply embarrassing. In fact, the reason why it may look stable (in normalcentric terms: normal distribution, random about the mean) is entirely from gage and measurement error masking the true process variation.

Their logic seems to be: If you have good data, then your gage must be good." We all know you can artificially get great "data" by using a gage with terrible resolution, so that most of the numbers look exactly the same. but, without gage R&R and a good ndc, how would you know? On top of that, if the characteristic is important enough to SPC chart in automotive terms, it should be the highest priority for gage R&R (and typically is), not the lowest.

I am sorry, but that is the most horrible advice I have ever seen in a AIAG document. I would advise to ignore it at all costs.

 

[debyang]

Re: MSA Version 4 when is GRR required?

I agree that MSA must be performed in order to implement SPC, but MSA is a costly tool, the reason to perform MSA is meant to improve quality. When process is already under SPC control long before MSA is required, and no product quality issue is encountered, I do not see the reason why MSA is needed? It will be a waste of resources, not lean at all.​

 

[bobdoering]

Re: MSA Version 4 when is GRR required?

I agree that MSA must be performed in order to implement SPC, but MSA is a costly tool, the reason to perform MSA is meant to improve quality. When process is already under SPC control long before MSA is required, and no product quality issue is encountered, I do not see the reason why MSA is needed? It will be a waste of resources, not lean at all.​

The costly approach is ignoring MSA. Waiting until there is a product quality issue to happen before acting on something really is not the point of lean, nor is it any accepted approach to "improve quality." The point is to understand your process to ensure it does not happen, and eliminate the excess overhead of sorting, rework, and rejection paperwork. Without an MSA you have no evidence that your SPC data is even valid. As I mentioned above, without a proper Gage R&R, you may be doing nothing but plotting gage error - and have no idea what your process is doing. In addition, you may be using the incorrect chart because of that. That is the reason why MSA - calibration and Gage R&R - is critical for SPC. That would make your whole SPC effort a waste of time...or 'not lean', if you will. It is a step you take long before implementing SPC, not the other way around...ever.

 

[Mandell]

<Quality Curmudgeon mode on> Harrumph. How long do we have to keep up this fiction that Gage R&R's are helping anything? It's easier and more realistic I think to use k=2 uncertainties to establish gauging performance, and easier to apply in differing situations. I'm beginning to think those guys at AIAG should go find honest work. </QC mode>

 

[bobdoering]

No matter what methodology you prefer - Gage R&R, Wheeler, k=2 - the point is ignoring gage error because you are already doing SPC and you think you are in control is a significant logical flaw. You may think you are in control because of gage error.

 

[cmi001]

Practically, I think that not all customers can accept skipping GRR even if you have good SPC results.

Observed process variation = actual process variation + measurement system variation.

If the SPC chart shows me the process is stable (in statistical control; data distribution is normal) and the observed capability study result is good (see Appendix B of the MSA manual 3rd edition), the actual performance of the process should be acceptable. Considering fit for use only, there should be no concern. Hence, the measurement system can be considered acceptable for this use only.

If the MSA does not performed, we have no idea about the error due to the measurement system. Then, the signal (actual variation) may be masked by the noise (measurement system error). This is a signal to noise ratio problem.

 

[bobdoering]

If the SPC chart shows me the process is stable (in statistical control; data distribution is normal) and the observed capability study result is good (see Appendix B of the MSA manual 3rd edition), the actual performance of the process should be acceptable.

Without a proper Gage R&R, you will never know for sure what the process data distribution is, since gage error is typically normal, and can mask the true underlying process distribution. Of course, the process distribution does not have to be normal to be stable (that is an over-generalization), but you do have to know what the process distribution is to determine its control.

Observed process variation = actual process variation + measurement system variation.

A more important equation for understanding the process variation is the full total variance equation.

 

[cmi001]

I have read the statement again and the TS-16949 standard requirements. The conclusion is that MSA cannot be skipped. Otherwise, this will violate the clause 7.6.1 of TS-16949. As MSA manual has used the word re-evaluation.in the statement, that means evaluation shall have been performed in the measuring system before. Therefore, under certain situation, re-evaluation of the measuring system can be considered to skip.

 

[LUV-d-4UM]

Re: MSA Version 4 when is GRR required?

I agree with TDVD101. We do 100% inspection of bulk products. The Gage R&R which the auditor demands us to do to check the equipment variance is useless because we are following the ASTMmethod and we are checking the ASTM R&R. Once we are within the ASTM R&R it is assume that the equipment is stable. How do I convince an auditor this? Any feedback is highly appreciated.