Ultrahigh TEmperature Refractory Alloys (ULTERA) Project

ULTERA Database is not simply a dataset but features a robust set of data processing, curation, and aggregation tools we built for the last 3 years. These tools allowed us to remove around the 5-10% erroneous data we identified in datasets available in the literature, ranging from non-obvious duplicate data points resulting in underestimating test errors to systematic composition parsing errors preventing model convergence or throwing off extrapolation capabilities. Most of our tools are not published yet, as the project is ongoing (they give us a competitive advantage), but some have been published open-source as a Python package called PyQAlloy, which will eventually contain all of them. They will work on all datasets of alloys, ranging from complex, concentrated solutions, i.e. metallic glasses / High Entropy Alloys (HEAs) / Multi Principle Element Alloys (MPEAs) / Concentrated Complex Alloys (CCAs) to more traditional alloys such as steels, nickel-based superalloys, etc. It should also work very well for complex non-metallic systems, such as High Entropy Oxides (HEOs) and High Entropy Nitrides (HENs), and we would be interested in supporting such deployment.

On this website, you can find:

1. Statistics on the current state of the literature-extracted dataset (95% experimental). If you are interested in something, shoot an email to ak@psu.edu and we will add it.

2. Our team (as of Mar 2023)

3. Overview of our data infrastructure (see details on how it fits into our Generative deep learning as a tool for inverse design of high entropy refractory alloys doi.org/10.20517/jmi.2021.05)

4. Initial (2021) team presentations describing details of our approach in regard to (1) inverse design, (2) computational thermodynamics, and (3) experimental validations.

 ULTERA HEA DATABASE STATISTICS

Our Team

Our Data Infrastructure

Project Introduction (TEAM PRESENTATION)