A new multivariate gage R&R method for correlated characteristics

This article explores how measurement systems having correlated characteristics are analyzed through studies of gage repeatability and reproducibility (GR&R). The main contribution of this research is the proposal of a method for multivariate analysis of a measurement system, a method that considers the weighted principal components (WPC). To prove its efficiency, what was first evaluated were the measurements of the roughness parameters obtained from AISI 12L14 steel turning machined with carbide tools. This GR&R study considers 12 parts, 3'operators, 4'replicates, and 5'responses (R a , R y , R z , R q and R t ). The data set has a correlation structure that determines 86.2% of explanation for the first principal component. As another step in proving the method's efficiency, the study generates simulated data with different correlation structures for measurement systems classified as acceptable, marginal, and unacceptable. The proposed method is compared with classical univariate and multivariate methods. It was observed that, compared to the other methods, the WPC was more robust in estimating the assessment indexes of a multivariate measurement system.