Fully Convolutional Spatiotemporal Learning for Microstructure Evolution Prediction

2026-02-23

AI Summary

This work proposes a fully convolutional spatiotemporal learning framework to address the high computational cost of traditional phase-field simulations in predicting microstructural evolution, which often struggle to balance accuracy and efficiency. For the first time, fully convolutional neural networks are introduced into microstructure evolution modeling, leveraging self-supervised learning to capture dynamic processes—such as grain growth and spinodal decomposition—from spatiotemporal image sequences generated by phase-field simulations. By eliminating recurrent architectures, the method substantially reduces both training and inference overhead while maintaining high predictive accuracy. It effectively reconciles short-term local dynamics with long-term statistical characteristics and demonstrates strong generalization across varying spatiotemporal domains and simulation parameters.

Technology Category

• Machine Learning: Structured Learning
• Computer Vision: Learning & Optimization for CV
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Understanding and predicting microstructure evolution is fundamental to materials science, as it governs the resulting properties and performance of materials. Traditional simulation methods, such as phase-field models, offer high-fidelity results but are computationally expensive due to the need to solve complex partial differential equations at fine spatiotemporal resolutions. To address this challenge, we propose a deep learning-based framework that accelerates microstructure evolution predictions while maintaining high accuracy. Our approach utilizes a fully convolutional spatiotemporal model trained in a self-supervised manner using sequential images generated from simulations of microstructural processes, including grain growth and spinodal decomposition. The trained neural network effectively learns the underlying physical dynamics and can accurately capture both short-term local behaviors and long-term statistical properties of evolving microstructures, while also demonstrating generalization to unseen spatiotemporal domains and variations in configuration and material parameters. Compared to recurrent neural architectures, our model achieves state-of-the-art predictive performance with significantly reduced computational cost in both training and inference. This work establishes a robust baseline for spatiotemporal learning in materials science and offers a scalable, data-driven alternative for fast and reliable microstructure simulations.

Problem

• microstructure evolution
• computational cost
• spatiotemporal prediction
• materials science
• simulation acceleration

Innovation

• fully convolutional spatiotemporal model
• microstructure evolution prediction
• self-supervised learning
• computational materials science
• data-driven simulation