pqam-rmsadtandoc2023 0.1.0

PyQAlloy-compatible Model for RMSAD prediction based on [Tandoc 2023 (10.1038/s41524-023-00993-x)](https://doi.org/10.1038/s41524-023-00993-x)

Lattice-Distortion

This Fork

This small repository is a lightweight fork version of the original one by Tandoc. It is only slightly modified to fit automated pipelines of the ULTERA Database (ultera.org) infrastrcutre, which expects model.py script with predict(comp: pymatgen.core.Composition) function returning an ordered array of output numbers or a labeled dictionary of them.

Original README by Tandoc

This repository contains relevant code and data for "Mining lattice distortion, strength, and intrinsic ductility of refractory high-entropy alloys using physics-informed >statistical learning" by Christopher Tandoc, Yong-Jie Hu, Liang Qi, and Peter K. Liaw to be published in npj Computational Materials

RMSAD_tool.py is a linux command line script written in python that takes a chemical composition in the form of a text string and prints the lattice distortion in angstroms.

example usgage: ./RMSAD_tool.py Ti0.5V0.5

-This script uses pymatgen (https://pymatgen.org/) to process the input string and is thus a requirement for the script to work. Depending on the version of pymatgen you have >installed, lines 3 and 380 may need to be modified (https://matsci.org/t/python-problem-with-pymatgen/35720) -numpy (https://numpy.org/) is also a dependency -This tool is currently only able to make predictions for compositions containing Ti,Zr,Hf,V,Nb,Ta,Mo,W,Re,Ru and will return an error if any other elements are present in >the input -B2 and elemental feature data are defined in dictionaries at the beginning of the code

training.ipynb and training_data.csv contains code and data to reproduce the rmsad model training that was performed in the paper -jupyter notebook is needed to open training.ipynb, dependencies are numpy, pymatgen, matplotlib (https://matplotlib.org/), pandas (https://pandas.pydata.org/), and >sklearn(https://scikit-learn.org/stable/)