Nested Design for R&R Study in Chemical Industry (2000)
P. Grabov, QA&R, ALD Ltd., Israel
A. Korenev, Indigo, Israel

Standard R&R study implies repeated testing of some parts by several operators. This 'matrix' procedure is of Factorial Design nature, and works reasonably well in mechanics, electronics and many other applications. In chemical industry, however, the conventional R&R approach is not applicable due to presence of some preparation stages before the material testing. Since we are interested in total error evaluation, all these stages should be taken into account, because the final result is affected by an uncertainty associated with every stage. The special procedure based on so-called 'Nested Design' should be used in this case. The procedure implies total  error decomposition through subsequent purification of upper level error from contribution of error on lower levels. The results of the procedure application have indicated that the measurement error (EV) is the most critical contributor to overall R&R error. The problem of high EV could be solved either by purchasing more precise testing equipment, or by increasing the number of measurement. Due to budget limitations, the second alternative has been chosen. Some optimal algorithm has been developed to keep the number of tests at a reasonable level and to enhance the test precision in critical vicinity of specification limits. The algorithm states the dependence of number of measurements on accepted results: 'The closer the test result is to specification limits, the more additional tests are required', and sets the formula for number of measurements calculation.