This talk will review two recent projects that cover: 1) microstructural modeling of Ni-based super alloys during additive manufacturing and 2) image-driven-machine-learning techniques for characterizing nuclear materials during a complex series of processing steps. The microstructure modeling work focuses on linking length scales from individual dendrites to macroscopic heat flows during laser processing. One of our advances in this project is to link pre-existing thermodynamic databases and GPU based parallel processing into the finite difference schemes. The machine learning work lays a foundation for "hybrid image analysis" that couples traditional methods and automated feature detection for assessing where the microstructure is in the processing sequence. We report what learned from this study in an effort to develop more general approaches in the future.
Prof. Lewis has experience in solidification processing, microstructure modeling and microstructure characterization of materials. He has focused on metallic materials for much of his career. His characterization experience extends to light-optical, scanning and transmission electron microscopy and microanalysis. He has used standard image analysis techniques and computer-based algorithms to understand microstructure. Most recently, Lewis and his research team have adopted and further developed MMSP (the Mesoscale Microstructure Simulation Project) to compute microstructure evolution using Phase Field techniques on Rensselaer's supercomputers. Projects have included the solidification of eutectic structures and grain growth by mean curvature. Current research involves prediction of grain size for the purpose of controlling microstructure evolution via closed-loop thermal control, prediction of grain growth kinetics under the influence of solute adsorption and study of texture evolution in microstructure models of the Earth's core. He teaches courses on mathematical methods in Materials Engineering, electron microscopy, and materials design.