Finite manufacturing primitives: A representation scheme for additive manufacturing quality assurance

Abstract

Product representation for additive manufacturing faces the issue of infinite geometric features. Furthermore, the same feature with different sizes can show nonlinear effects in deviation patterns owing to layer-wise fabrication. To address these issues, we propose the finite manufacturing primitive (FMP) representation scheme for quality assurance. Layer geometries will be represented by 2D FMPs such as curve segments and corners. 3D FMPs are generated by stacking up layers of 2D FMPs under a convolution framework to capture nonlinear effects and achieve dimension reduction in the 3D space. Application in small-sample quality learning and prediction is illustrated with an example.