Need help with measurement system analysis. Would like some clarification as to measurement test designs…

I have a question as to whether I have selected a correct test design for measurement assessment. Also, as mentioned below, there seems to be a difference between NIST and Minitab v14 of test plan definitions. Any insight/suggestions as to a good approach to evaluations of this type will be appreciated. This process has never before been evaluated using actual data produced by the process itself. All that exist currently is the calibration hierarchy and uncertainty analysis using manufacturer specs. Also, surprise!, this is the first time I’ve been involved in a study of this sort.

Background:
I have been asked by my company to perform an assessment of the variability of a pressure calibration process. The lab concerned has a piston deadweight tester that is used to apply different loads to a pressure gage and obtain a load/response curve. I have been asked to ascertain repeatability of successive measurements and reproducibility between operators and days.

Test Plan:
After some investigation I found the NIST/SEMATECH web site which seemed to have the test design I needed: a 3-level nested. In fact, it is stated that operator effects can be determined from the 3rd-level pooled standard deviations. So what I have is three operators and three transducers of the same range to give me at least 2 degrees of freedom at those levels.

The plan is to have all three operators perform a calibration of a single transducer on the same test day. Due to the calibrator workload, it is impossible to have all three operators calibrate all three transducers on the same day. On the next test day, separated by at least two calendar days, a different transducer is calibrated, again by all three operators on the same day. The process will be repeated until all three transducers have been scheduled for six test days (which will be equal to 2-3 months calendar-wise).

Minitab vs NIST?
I've recently had the chance to look at Minitab’s v14 “Gage Study” stat functions. It looks to me what NIST calls a “nested” design, Minitab calls a “crossed” design. Also, it seems to me that Minitab pools the various level standard deviations differently than the NIST examples.

Again, thanks for any help/insight that can be provided.

Any takers on this one?

Camper,

While the Minitab route may be what you are looking for, I highly recommend you follow the NIST examples. The way Minitab pools thier data may provide inconclusive, or even inaccurate results. NIST sets the standard for the entire country. Otherwise your experiment sounds Good. I would be carefull to monitor temperature, and excitatation voltage if these pressure transducers are high accuracy as this can be a a sorce of error.

Giselle,

Thanks for the feedback. I looked at the AIAG MSA manual, but it did not seem to address my particular situation completely. I didn't mention that I also want to obtain a "Type A" uncertainty value for the process. Yes, I intend to monitor temperature and humidity; the transducer output will also be quantified in terms of the excitation voltage (these are bridge-types).

The idea is to create a generic "plan-of-attack" to evaluate different calibration processes, or maybe a better way to say it is that I want to prove the process is in statistical control; once proven, I then see the MSA approach as more applicable to an ongoing monitoring of the process.

camper