Study On Multivariate Measurement System Capability Analysis Based On A Modified Region Approach

Measurement system analysis is used of high frequency in the modern quality management and it is an indispensable step before quality control or process capability analysis. Measurement system capability analysis is to determine whether the measurement system is capable for use. In modern industrial, product quality needs a collection of critical characteristics to describe. Therefore, more and more companies begin to use multivariate measurement system of which output is related multi-dimensional data. Because of the relationship between these measured quality characteristics, it is not proper to apply single-variable measurement system analysis in multivariate measurement setting. The existing research has been extended from univariate to multivariate cases. The multivariate analysis of variance(MANOVA) treats the volume of constant-density contours from different variation as the variability estimations and then constructs the approval criteria based on volume ratio. However, it ignores the fact that the distribution change of variations could affect the measurement system capability. The weighted principal components(WPC) method uses dimension reduction to reduce the computation complexity but is unable to build the precision-to-tolerance ratio because it cannot deal with the tolerances.This paper proposes a modified-region-based assessment method. It incorporates the variance-covariance matrix of the measurement errors to modify the constant-density contours of tolerances, total variation and process variation. The modified volumes are regarded as the estimation of variability to calculate the assessment indices including the multivariate precision-to-tolerance ratio, the multivariate percent of repeatability and reproducibility, and the multivariate signal-to –noise ratio. This new method can evaluate the measurement system capability completely and reflect the capability change caused by measurement errors variation.Finally, the performance of the modified region method is evaluated based on a dataset from the literature and a set of relevant simulation. The proposed method proves to be more complete and accurate compared with other methods.