Five Ways Statistical Tolerancing Can Fail and What to Do About Them

Abstract

In this article we explore the general non-robustness of traditional root sum of squares statistical tolerancing and describe, in particular, how it can fail. These are [1] deficiencies in the functional model, [2] lower process capability in inputs that what is desired of outputs, [3] biases, [4] correlations, and [5] non-normality. We also show that statistical tolerancing is extremely non-robust to the first four types of causes. Moreover we provide examples of each type and discuss what to do about each.