Speaker
Description
The increasing demand for refractory materials that can withstand high temperature environments necessitates the discovery of advanced alloys for aeronautic and energy applications. This study introduces a computational approach leveraging AFLOW's machine learning frameworks—specifically the Property Labeled Material Fragments (PLMF) and Molar Fragment Descriptor (MFD) methodologies—to accelerate the design of Mo-Nb-Ta ternary alloys with superior thermomechanical properties. Mo-Nb-Ta system was strategically selected for this investigation due to the similar body-centered cubic (BCC) crystal structures of these metal elements, which promotes favorable solid solution formation and structural stability at elevated temperatures. Our approach utilizes specialized software to extract essential structural data including lattice parameters and crystallographic positions from a large set of Mo-Nb-Ta compositions. This data is then fed into the AFLOW machine learning model to predict critical thermal and elastic properties including Young's modulus, thermal expansion coefficients, and heat capacity. The generated dataset was subsequently analyzed using scikit-learn for advanced material informatics. Specifically, we implemented isolation forest and local outlier factor algorithms for anomaly detection to identify promising Mo-Nb-Ta alloy compositions with superior property combinations, while Bayesian optimization techniques were employed to systematically navigate the compositional space and determine optimal alloy formulations that maximize application-specific performance metrics.
