Binary Gaussian Process classification of quality in the production of aluminum alloys foams with regular open cells

Aluminum alloys foams with homogeneous and regular open cells have been frequently proposed and used as support structures for catalytic applications. In this kind of application, the quality of produced metal foam assumes primary importance. This paper presents an application of a classifier algorithm to predict quality in the manufacturing process of aluminum alloy foams with homogeneous and regular open cells. A data analysis methodology of experimental data, which is based on Binary Gaussian Process Classification, is presented. The proposed method is a Bayesian classification method, which gets away from any assumptions about the relationship between process inputs (the geometric design parameters of the regular unit cells) and process output (probability to obtain defective foam). We demonstrate that the proposed methodology can provide an effective tool to derive a model for the prediction of quality. An investment casting process, via 3D printing of wax patterns, is considered throughout the paper. Despite this specific case study, the methodology can be exploited in different processes in which the assumptions of traditional statistical approaches could not be easily verified, e.g., additive manufacturing.