Text

theses

ETD doctoral

Quality and reliability of additively manufactured (AM) parts using the Laser Powder Bed Fusion (LPBF) process are greatly affected by the thermal history during the manufacturing process. Prediction of thermal history, residual stresses, and distortions of a part during the LPBF process is critical to understand how the process parameters influence the process stability and mechanical properties of the part. Finite element modeling of the LPBF process at part-scale is challenging and requires massive computational time and resources. These models are computationally infeasible if they are not associated with simplifications in the mesh configuration and the heat source model, or with the reduced domain size. Due to the complexity of the computational model itself, uncertainties during the LPBF process are not systematically studied, and their effects on quality and reliability of the parts are not characterized.

The dissertation overcomes the computational expensiveness associated with modeling of the LPBF process on a part-scale level. It presents the use of different adaptive remeshing techniques that enable the thermal and thermomechanical simulations at the part-scale level without the sacrifice in accuracy. As a result, part-scale thermal and thermomechanical finite element modeling are computationally feasible. This is the first work where an adaptive remeshing framework was developed for the LPBF process, based on an existing general-purpose implicit solver and the tetrahedral mesh. In particular, the tetrahedral mesh can represent parts with complex structures using less elements than the existing remeshing technique. The thermal process modeling presents models for relatively large parts considering different process parameters, and the models can predict location-dependent melt pool size and the lack-of-fusion porosity. The thermomechanical process modeling predicts the thermally induced residual stresses, strains, and distortions for different parts. The model predictions find similar trends with the experimental results from the literature along with achieving a significant reduction in the computational time compared to the state-of-the-art models without using the adaptive remeshing. Furthermore, a general calibration and validation framework for the LPBF process was developed based on multi-fidelity models and limited experimental data. The framework enables the development of highly efficient and accurate models for melt pool predictions under various sets of process parameters through the seamless integration of finite element modeling, machine learning methods, and the model calibration and validation methods. Effectiveness of the framework is demonstrated by experimental data under different sets of process parameters available in the literature.

ETD_11034

Ph.D.

Includes bibliographical references

doi:10.7282/t3-g2sh-wm80

The author owns the copyright to this work.