Uncertainty Quantification and Management in Additive Manufacturing

Abstract

Additive manufacturing (AM) has shown immense potential in several industries. However, significant variability in the product quality currently hinders widespread use of AM. The focus of this dissertation is on uncertainty quantification (UQ) and uncertainty management (UM) in AM to help solve this challenge by integrating physics-model based prediction with probabilistic sensing and control strategies for the manufacturing process. This research develops a systematic UQ/UM methodology to quantify and control the variability in the AM process. Three objectives are accomplished: (1) formulation of the UQ methodology, (2) process design under uncertainty, and (3) process control under uncertainty. Various sources of uncertainty are considered, such as process and material uncertainty, model discrepancy, and measurement uncertainty. The uncertainty quantification methodology integrates heterogeneous information available from multiple sources and different models. Both forward and inverse problems are addressed. The forward problem (UQ) quantifies the overall uncertainty in the AM process, and the relative contributions of different sources to the overall uncertainty regarding the quality of the AM product. The inverse problem (UM) addresses uncertainty reduction through model calibration, and process design and process control. Efficient surrogate models are constructed to replace the expensive coupled multi-scale multi-physics simulation models for the uncertainty analysis. In particular, a methodology is developed for building surrogate models in the presence of high-dimensional field output. Laboratory experiments are conducted to validate the process parameter optimization results. The proposed methodologies are demonstrated with examples from metal and polymer-based additive manufacturing techniques.