Accelerating the Printability of New Metal Additive Manufacturing Alloys: A Robust Calibration Approach

Abstract

Metal additive manufacturing (AM) or metal three-dimensional (3D) printing offers significant benefits for manufacturing parts with features and capabilities that conventional techniques cannot match. Meanwhile, it is widely accepted that metal AM processes come with their own challenges. Metal AM typically suffers from high degrees of variability in the properties of the fabricated parts and high complexity in the fabrication process, particularly due to the lack of understanding and control over the physical mechanisms during fabrication. Simulation models for AM are essential to enable process planning and accelerate qualification and certification of fabricated parts. One important task involves calibrating simulation models to ensure that predictions are in agreement with experimental observations.

Part of my dissertation works on integrating Bayesian model calibration to accelerate the development of metal AM processes such as Laser Powder Bed Fusion (LPBF). A framework is developed includes experimental design, multiscale modeling and simulation, uncertainty quantification, and experimental material characterization for fully characterizing parameter-process-property relations in LPBF for materials design, process standardization, part qualification, and discovery/innovation.

Other parts of this dissertation focus on statistical calibration of a computer simulation model with multiple outputs where experimental observations for one (or more) of the outputs are expensive to acquire. Bayesian multiple imputation method is used in a statistical calibration framework to help estimate calibration parameters in the case of lacking expensive experimental data. The proposed methodology is properly analyzed and validated by an analytical simulation model of melt pool geometry in LPBF process. The insights and understanding achieved by applying these methods advance the development of AM processes.