Machine learning assisted material development for lithography-based additive manufacturing of porous alumina ceramics

Abstract

The transition of applying ceramic additive manufacturing (AM) from prototyping to mass production and from monolithic to multi-material (MM) components can be supported by continual development in materials and processes. Lithography-based ceramic manufacturing (LCM) can be used for MM printing of ceramics with high accuracy by introducing different approaches that enable discrete/smooth multidirectional material transitions. Adaptation of ceramic slurries plays important role for the successful printing and co-sintering. Especially for co-sintering, the shrinkage of the materials must be adapted so that no internal residual stresses occur. Machine learning (ML) offers promising opportunities for development of new materials and optimization of processes in AM by introducing relations between input features and output responses. In this article, ML algorithms were used for the prediction of shrinkage and porosity of alumina samples dependent on input features including material, printing, and thermal processing parameters so that defect-free porous/dense alumina combination can be achieved.