Discussion in 'Gage R&R and MSA - Measurement Systems Analysis' started by Plague Doctor, Jan 20, 2016.
Is it allowed to use samples with out of spec. values for GRR study?
It depends on how your gage will be used. If the gage will be used as an inspection device and if you will use % Tolerance as the metric for acceptability, then it is allowed because the part variation is not part of the calculation for % Tolerance.
However, if the gage will be used for SPC and the % Study Variation metric is used then the parts must be representative of your process regardless of whether they meet the spec or not. This is because the Study Variation is part of the calculation of the metric. A work around would be to use the % Process Variation metric instead. The Process Variation would be from an independent capability study.
if you naturally make out of spec parts then yes it is not only allowed but desired to use out of spec parts. your parts should always be representative of the actual process spread.
Thanks for answers. But that is why I asked: in fact our process is quite stable, we are able to produe parts e.g. with range 40% from the range of spec limits.
But if I understand properly, the less range of the values of samples taken for GRR,the worse GRR result we will receive at the end. In other words we will receive better GRR result if we take 10 samples with values covering the whole spec area (from lower limit to upper limit) than if we take 10 samples with almost the same value. Is it true?
So to make a conclusion: can we take samples for GRR with wider range of values (even out of spec) than we normally have in our production?
Re-read my post. It depends on how you intend to use this gage. If you use it for inspection, you should use %Tolerance to determine whether the gage is adequate. Part variation plays no role in this metric, so it does not matter to your results how the parts are selected.
However, if you use this gage for SPC, you MUST select parts that are representative of your process because the % Study Variation metric is affected by the part variation. Deliberately selecting parts with greater variation than the actual process would inflate this metric and be deliberately misleading as well as unethical. The ideal solution is to use the information from a process capability study to calculate % Process Variation instead. It is much more accurate than %SV.
Separate names with a comma.