Surrogate-based model chains for establishing process-structure-property linkages with quantified uncertainties in metal additive manufacturing

Abstract

Achieving qualification and certification is the key milestone for the standardization of additive manufacturing (AM) technologies and their broader adoption in significant industrial sectors. However, the underlying knowledge gaps in complex AM processes hinder the whole progress. One critical step to advancing metal AM manufacturing readiness level is to build a comprehensive understanding of the complex physical phenomena during AM processing, specifically the process-structure–property (P-S-P) relationships. Although the significance of P-S-P linkage in metal AM has been highlighted frequently in AM community, relevant studies on physics-based models coupling are still at the early stage. In this work, we propose a framework to achieve fast prediction from process parameters to mechanical properties in laser powder bed fusion (L-PBF), and we investigate the model performance under the uncertainty of parameter variation and model inaccuracy via the methods of uncertainty propagation and sensitivity analysis. A case study on L-PBF 316L stainless steel demonstrates the proposed P-S-P surrogates are good replacements for computationally expensive physics-based model linkages. We identify the contributing variables to the variation of each output and present the P-S-P linkage performance under propagated uncertainty. This work paves a solid foundation for the continuing calibration work of the P-S-P surrogates and the thorough improvement of P-S-P prediction.