cardiGAN A Generative Adversarial Network Model for Design and Discovery of Multi Principal Element

Abstract

Whilst a large number of MPEA compositions have been studied (empirically) in the past decade, the number explored to date is on the order of 103 alloys. Given that MPEAs can have anywhere from 3 to 8+ alloying elements –inclusive of the majority of the periodic table – the number of possible MPEA s is in well in excess of 1020 alloys. Traditional (i.e. empirical) alloy production and testing is both costly and time-consuming, partly due to the inefficiency of the early discovery process which involves experiments on a large number of alloy compositions. It is intuitive to apply machine learning in the discovery of this novel class of materials, of which only a small number of potential alloys has been probed to date. As a consequence, exploration of new alloys using methods such as generative models, is worthy of exploration. In this project, a model architecture for alloy design and prediction will curated on the basis that there is (in a relative sense) only a limited empirical dataset of MPEAs reported to date that can be exploited for model ‘training’, precluding a conventional machine learning approach.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 08, 2022
Source ID
N629092212017

Entities

People

  • Nick Birbilis

Organizations

  • Australian National University
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Neural Network Machine Learning.
  • Powder metallurgy of Titanium alloys.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks