E-WEN HUANG

NATIONAL YANG MING CHIAO TUNG UNIVERSITY

Department of Materials Science and Engineering

Center for Institutional Research and Data Analytics

Professor, Deputy Director

About

E-Wen Huang is a professor of Materials Science & Engineering Department at the National Yang Ming Chiao Tung University, Taiwan. His research focuses on the mechanical behavior of the 3D additive manufacturing metals and high entropy alloys.

Research Disciplines

Materials Science, Synchrotron X-ray

Featured Presentation

See the full schedule here.

Title: Synchrotron x-ray-based combinatorial stoichiometry & microstructure high-throughput machine learning-assisted prediction & validation of a high-entropy-alloy hardness mapping

Keywords: high-entropy alloys; synchrotron x-ray; hardness; machine learning; nanoindentation; microstructure

Abstract: A supervised machine learning (ML) framework using different regression models was applied to predict localized hardness maps based on both alloying elements and microstructure in the Cu15Ni35Ti25Hf12.5Zr12.5 high-entropy shape-memory alloy, and the predictions were evaluated against nanoindentation hardness map. Favorable element distribution of Ni-Hf-rich dendrite and Cu-Ti-Zr-rich interdendritic regions was obtained using x-ray fluorescence mapping. The lattice characteristic distribution of specific orientations was determined using x-ray nanodiffraction mapping. The nanoindentation hardness map was performed and calibrated to correlate with predicted hardness maps. Using a cluster-wise modeling strategy combined with the CatBoost model enables accurate prediction of mechanical properties in heterogeneous microstructure. The predictive accuracy can be further enhanced by incorporating additional microstructural-related features in addition to the chemical composition as input data. The ML-based hardness prediction shows great promise in elucidating the correlation between localized microstructure and hardness at the highly spatially resolved micrometer scale beyond stoichiometry. The methodological development in this study enables scalable and spatially resolved prediction of mechanical properties, facilitating future studies of processing-induced heterogeneous microstructure effects in multicomponent alloys.