Temperature-Aware Recurrent Neural Operator for Temperature-Dependent Anisotropic Plasticity in HCP Materials

2025-08-26

AI Summary

Neural network surrogate models for thermomechanically coupled plasticity of polycrystalline magnesium suffer from slow training, strong dependence on temporal resolution, poor extrapolation capability, and difficulty in capturing pronounced crystallographic anisotropy and thermal sensitivity. Method: This paper proposes the Temperature-Aware Recurrent Neural Operator (TRNO), which integrates the nonlocal modeling capacity of neural operators with the sequential modeling strength of recurrent architectures, augmented by an explicit temperature embedding mechanism to enable generalization across temperatures, loading paths, and temporal resolutions. Contribution/Results: Validated on multiscale simulations of polycrystalline magnesium, TRNO achieves over three orders-of-magnitude speedup in training compared to conventional RNNs, delivers high predictive accuracy, and exhibits superior extrapolation performance—particularly under unseen thermal and mechanical conditions. TRNO establishes a novel paradigm for efficient surrogate modeling of complex thermo-plastic behavior in hexagonal close-packed (HCP) metals.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Temporal Planning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Neural network surrogate models for constitutive laws in computational mechanics have been in use for some time. In plasticity, these models often rely on gated recurrent units (GRUs) or long short-term memory (LSTM) cells, which excel at capturing path-dependent phenomena. However, they suffer from long training times and time-resolution-dependent predictions that extrapolate poorly. Moreover, most existing surrogates for macro- or mesoscopic plasticity handle only relatively simple material behavior. To overcome these limitations, we introduce the Temperature-Aware Recurrent Neural Operator (TRNO), a time-resolution-independent neural architecture. We apply the TRNO to model the temperature-dependent plastic response of polycrystalline magnesium, which shows strong plastic anisotropy and thermal sensitivity. The TRNO achieves high predictive accuracy and generalizes effectively across diverse loading cases, temperatures, and time resolutions. It also outperforms conventional GRU and LSTM models in training efficiency and predictive performance. Finally, we demonstrate multiscale simulations with the TRNO, yielding a speedup of at least three orders of magnitude over traditional constitutive models.

Problem

• Modeling temperature-dependent anisotropic plasticity in HCP materials
• Overcoming poor extrapolation and resolution dependence in neural surrogates
• Improving training efficiency and accuracy for complex material behavior

Innovation

• Temperature-Aware Recurrent Neural Operator architecture
• Time-resolution-independent neural network design
• Efficient multiscale simulation with speedup