Statistical control of complex geometries, with application to Additive Manufacturing

New production processes, as Additive Manufacturing (AM), allow the production of objects and shapes characterized by a growing complexity, especially when compared with those normally manufactured in traditional production processes. It is thus necessary to develop methods which can be applied to the study of the variability of a dataset whose elements are manufactured realizations of the same nominal model, which can carry a very high degree of complexity. In the present work, we propose a method allowing modeling a wide variety of possibly local geometric deviations of the manufactured object with respect to the nominal model. A way of identifying and monitoring these kind of deviations, based on Principal Component Analysis in Hilbert spaces, is proposed as well. The proposed method is tested on a real dataset of items produced via AM.