| Abstract Scope |
The ULTERA (ultera.org) data ecosystem accelerates the discovery of ULtrahigh-TEmperature Refractory Alloys through four live, tightly coupled, database-driven loops: distributed, version-controlled data ingestion and curation from almost 1000 publications, rapid predictive ML modeling, cGAN inverse design, and advanced experimental manufacturing for validation. Here, we demonstrate ULTERA’s robust capabilities across data quality control and materials discovery. The platform successfully models ductile refractory alloys using DFT-based data, predicts mechanical properties by extracting natural language-derived descriptors for processing conditions, and implements a novel approach to graph-based eutectic refractory alloys exploration using nimplex. Other key achievements include automated anomaly detection via PyQAlloy and compositional scope optimization via nimCSO. Furthermore, we highlight the ongoing advancements in inverse design powered by a thermodynamics-inspired Zentropy-Enhanced Neural Network (ZENN) models. By unifying automated curation, multi-scale modeling, and AI-driven optimization, ULTERA provides a comprehensive, closed-loop framework to overcome the complex challenges of designing structural materials for extreme environments. |