The design team tried many test methods and finally decided on a functional test. They created specifications using a manufacturing DOE (Design of Experiments) output and wanted to transfer the test method to the plant. This was where Quality stepped in to ask for a gauge R&R (Repeatability and Reliability.) The design team saw this exercise as needless busy-work and too academic for their purposes, so they pushed back on the need for the test.
Quality began by asking the design team to test 2 fixtures, 3 operators, 8 parts, and 3 repeats. We explained how the parts needed to encompass the entire range and by doing so would include some failing parts. We stressed that the operators must be well-trained, but not be able to identify the part so they would not to know the results of repeat testing.
Again there was pushback. Together Quality and the design team settled on 1 test fixture and the design team would choose the parts. When we analyzed the results we found a value of 95%. All the results were well within the specifications, but with over 95% variation, less than 5% was coming from part to part variation. The repeatability and reproducibility were both in the 40% range. Anyone that has done a Gauge R&R knows this result is unacceptable and shows the test to be incapable. The design team, lacking understanding of Gauge R&R asked whether the test was good because all the parts passed.
After more discussions about the purpose of Gauge R&R, the design team agreed to look for ways to reduce variation. We replaced a manual part of the test process with an automated process. We ran the gauge R&R again with 3 parts, 2 operators, and 2 repeats. The results showed reduced variation; however, the test was still considered incapable. The total gauge was above 45% with no part to part variation, and with the automated process, almost no reproducibility variation. Things were looking better, but the parts being chosen did not cover the whole range, or include possible failures.
A third Gauge R&R was conducted to test 2 more “failed parts” with 2 operators and 2 repeats to combine with the second gauge data above. Completed testing showed the results passed. The total variation was now 17% with the variation coming from repeatability and large part to part difference.
The test method can be considered a good test. The Gauge R&R led to an improved test: more automated processes reduced the test variation. The results show how important careful selection of the parts is to proper analysis of a gauge R&R. Using parts that are too similar can make even a good test method look bad.
Rachel Dowell got her ASQ-CQE 6 years ago. She works as a quality engineer in New Product Development at Baxter. She has over 10 years experience in quality and new product development in medical device and pharmaceutical companies. Rachel earned a BS from Brown University and a Masters from the University of Washington in Seattle in Chemical Engineering.