About this Symposium |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
CANCELLED: Foundations for Autonomous Materials Design
|
| Sponsorship |
TMS Materials Processing and Manufacturing Division
TMS: Computational Materials Science and Engineering Committee
|
| Organizer(s) |
Megna Shah, Air Force Research Laboratory
Jeff P. Simmons, U.S. Air Force Research Laboratory
Stephen R. Niezgoda, Ohio State University
Veera Sundararaghavan, University of Michigan
Dennis M. Dimiduk, BlueQuartz Software LLC
Brian DeCost, National Institute of Standards and Technology |
| Scope |
This symposium is concerned with understanding, constructing, and optimizing ML/AI tools for processing-structure-properties relationships in materials science and engineering, to autonomously develop materials. The intent here is to go beyond combinatorics for materials discovery (i.e. active learning over chemistries to optimize a small set of properties) and rather into maturing candidate materials into deployable materials, which requires an iterative design process in which new constraints often emerge.
There has been considerable effort to develop physics-based models that map between input domains (processing) and output domains (structure and/or properties). Modern machine learning algorithms, for the first time, allow for exploring not just point solutions, but the effects of whole processing paths on the final output. These paths, themselves, can be engineered and manipulated to achieved desired results. For example, in a complex processing path, a cooling stage can be modified to explore its effects on structures or properties. If regions can be mapped out where small changes in the process results in a small perturbation of the structure/property – navigability - this allows for autonomous iterative design.
This symposium aims to bring together researchers interested in the various components of autonomous materials design. For this symposium, relevant contributions can include:
• Integrated Computational Materials Engineering models for mapping process-structure or structure-process and inversion of these models
• Generative AI for microstructure prediction
• Control theory and feedback design for process optimization
• Latent space representations of multi-dimensional (hierarchical and/or multi-modal) materials information
• Manifold hypothesis for data representation
• Generation of paired datasets for model training
• Stochastic models of microstructure and their representation
• Metric space and topological space representations of materials data
• Disentanglement and identification of latent space dimension in terms of materials relevant features
• Novel approaches for autonomous search for solutions with multiple objectives and constraints |
| Abstracts Due |
07/01/2025 |
| Proceedings Plan |
Undecided |