摘要

A posteriori estimation of defeaturing error, that is, engineering analysis error caused by defeaturing, remains a key challenge in computer-aided design/computer-aided engineering integration; indeed, it is essential if analysis based on automatic simplification of the geometry of a complex design is to be reliable. Previous work has mainly focused on removing a single, negative, internal design feature (i.e. a void within the model's interior); effects of removal of multiple features are found by simple summation. In this paper, we give a second-order defeaturing error estimator that can handle multiple boundary features, either positive or negative (cutouts or additions on the model's boundary). This is the first such method to take into account interactions between different features. (Removing internal holes in 2D or through holes in 3D will change the model's topology and is not addressed here.) The proposed estimator is also the first to handle positive features without heuristics. Our approach uses second-order shape sensitivity, on the basis of reformulating defeaturing as a shape transformation process. The reference model used for sensitivity computation is carefully chosen, as is the associated shape variation velocity, to ensure efficient computation and simple sensitivity expressions. Mixed partial derivatives in a second-order Taylor expansion are used to account for interaction between features. The advantages of our approach are demonstrated using various numerical examples.