Foundation Model for Polycrystalline Material Informatics

2025-12-07

AI Summary

Addressing the challenges of scarce labeled data and poor generalizability in modeling microstructure–property relationships for polycrystalline materials, this work introduces the first foundation model tailored for 3D polycrystalline materials. Our method employs voxelized microstructural representations and a physics-structure-aware self-supervised pretraining framework: it leverages masked voxel reconstruction coupled with an orientation-aware deep material network (ODMN) to learn physically interpretable microstructural embeddings. On homogenized stiffness prediction, the model significantly outperforms non-pretrained baselines. Crucially, it achieves accurate inference of nonlinear stress–strain responses for unseen microstructures using only a few labeled examples. This study pioneers the adoption of the foundation model paradigm in polycrystalline materials informatics, demonstrating strong cross-structure and cross-task generalization as well as few-shot transferability—establishing a novel paradigm for data-efficient intelligent materials design.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Other Foundations of Computer Vision
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

We present a 3D polycrystal foundation model that learns a physically structured representation of voxel-based microstructures through large-scale self-supervised pretraining. The encoder is trained on a dataset of 100,000 FCC microstructures whose crystallographic orientations span the texture hull, using a masking strategy that forces the model to infer latent features from incomplete spatial information. The quality of the learned representation is evaluated through two downstream tasks with distinct physical characteristics. (i) Homogenized stiffness prediction: the pretrained encoder consistently outperforms the non-pretrained baseline across all masking ratios. (ii) Nonlinear response modeling: the encoder is coupled with an orientation-aware interaction-based deep material network (ODMN) to infer complete sets of network parameters, enabling accurate stress-strain predictions for previously unseen microstructures. In both tasks, the pretrained encoder demonstrates markedly stronger generalization capability. These results underscore the strong transferability of the proposed framework and its suitability for data-scarce scientific settings, where labeled microstructures are limited and physics-consistent generalization is essential. The foundation model provides a scalable route toward integration with experimentally derived microstructures, offering a new basis for microstructure-property reasoning in practical materials design.

Problem

• Develops a 3D foundation model for polycrystalline microstructures
• Evaluates representation via stiffness prediction and nonlinear response modeling
• Enables accurate property predictions for unseen, data-scarce microstructures

Innovation

• Self-supervised pretraining on 100,000 microstructures
• Masking strategy forcing inference from incomplete spatial data
• Encoder coupled with orientation-aware deep material network