| Abstract Scope |
A unique integration of data driven approaches is carried out and used to design high entropy alloys given the challenges in trade-offs between strength, ductility and various environmental effects. A new methodology is developed and demonstrated which combines non-linear mapping, random walk analysis, and unsupervised modeling, while incorporating inputs as broad as density of states spectra, processing, and phase stabilities. This work addresses an on-going challenge in supervised modeling, such as neural network-based analyses, which require large amounts of data that capture an evenly distributed variance while also being sensitive to selected tuning parameters. This developed approach captures the interplay between chemistry, microstructure and phase stability, which allows for identifying chemical design rules for improving properties with minimal trade-offs. Given the unsupervised aspect of the work and the inclusion of uncertainty in the development of design rules, this approach provides a new framework for robust multi-component material design. |