Is it acceptable to use traceable standards (VLSI, NIST, etc.) to complete stability, bias and linearity studies if these standards cover the operating range of the gage? For stability and bias, the AIAG MSA 4th edition (p. 85, p. 87) states, “Obtain a sample and establish its reference value(s) relative to a traceable standard…” For linearity, the AIAG MSA 4th edition (p. 96) states, “select g>= 5 parts whose measurements, due to process variation, cover the operating range of the gage.” Specifically for designating a master sample (from production) to assess stability, we have an issue with degredation or oxide growth on the master sample that introduces known variation in thickness measurements. In this case, would it be justifiable to use VLSI standards to assess stability over time? Thanks for your help and guidance!
The quick answer is “that depends”.
The purpose of measurement systems studies is to evaluate the entire measurement system which includes the equipment, method, appraiser and the within part variability. The problem with using standards is that they are too good in that their reference value will usually be on an incremental value / discrimination point and not between points which require either truncation or rounding. Using standards can end up with average ranges of zero which is interpreted that the measurement system does not sufficient discrimination for the task.
Consequently I am disinclined to support the use of standards in measurement system studies.
However, it appears that this question is motivated by the need to look at the stability of the measurement system and actual parts will degrade over time. In this case we are not interested in the common cause variation but the existence / occurrence of a special cause. If the measurement system shows itself to be acceptable using normal parts then the stability of the system can be monitored using standards (at least 2 at the limits, preferably 3 with one in the middle).
If a bias studies on the standards have the overall range of 0 then ANY reading in a stability study using these standards other than the reference value indicates that the measurement system has changed.
If a bias studies on the standards does not have zero variability then things get a little tricky. Statistical control limits will probably not work unless this variability is not small (then the question become understanding why). You may have to resort to using the “precision” limits provided by the equipment manufacturer. Remember the objective is to be able to identify whether something changed (a special cause is affecting) the measurement system.
For more on this topic, please visit ASQ’s website.