Process Monitoring and Uncertainty Quantification for Laser Powder Bed Fusion Additive Manufacturing
MetadataShow full item record
Metal Additive manufacturing (AM) such as Laser Powder-Bed Fusion (LPBF) processes offer new opportunities for building parts with geometries and features that other traditional processes cannot match. At the same time, LPBF imposes new challenges on practitioners. These challenges include high complexity of simulating the AM process, anisotropic mechanical properties, need for new monitoring methods. Part of this Dissertation develops a new method for layerwise anomaly detection during for LPBF. The method uses high-speed thermal imaging to capture melt pool temperature and is composed of a procedure utilizing spatial statistics and machine learning. Another parts of this Dissertation solves problems for efficient use of computer simulation models. Simulation models are vital for accelerated development of LPBF because we can integrate multiple computer simulation models at different scales to optimize the process prior to the part fabrication. This integration of computer models often happens in a hierarchical fashion and final model predicts the behavior of the most important Quantity of Interest (QoI). Once all the models are coupled, a system of models is created for which a formal Uncertainty Quantification (UQ) is needed to calibrate the unknown model parameters and analyze the discrepancy between the models and the real-world in order to identify regions of missing physics. This dissertation presents a framework for UQ of LPBF models with the following features: (1) models have multiple outputs instead of a single output, (2) models are coupled using the input and output variables that they share, and (3) models can have partially unobservable outputs for which no experimental data are present. This work proposes using Gaussian process (GP) and Bayesian networks (BN) as the main tool for handling UQ for a system of computer models with the aforementioned properties. For each of our methodologies, we present a case study of a specific alloy system. Experimental data are captured by additively manufacturing parts and single tracks to evaluate the proposed method. Our results show that the combination of GP and BN is a powerful and flexible tool to answer UQ problems for LPBF.
Mahmoudi, Mohamad (2019). Process Monitoring and Uncertainty Quantification for Laser Powder Bed Fusion Additive Manufacturing. Doctoral dissertation, Texas A&M University. Available electronically from