Multi-Physics Simulations via Coupled Fourier Neural Operator

2025-01-28

AI Summary

Existing Fourier neural operators (FNOs) struggle to model strong, dynamic couplings across multiple physical domains. To address this, we propose the Coupled Multi-Physics Operator Learning (COMPOL) framework. COMPOL introduces a novel multi-physics latent feature aggregation module that integrates recurrent mechanisms with self-attention, explicitly capturing nonlinear inter-domain couplings—e.g., between fluid dynamics, biological systems, and multiphase flow. Furthermore, it employs a unified joint encoding–decoding architecture for multi-physics fields, overcoming modeling limitations of conventional single-output or weakly coupled operators. Evaluated on diverse multi-physics simulation tasks, COMPOL achieves 2–3× higher prediction accuracy than state-of-the-art neural operators, while significantly improving generalization capability and interpretability for complex, tightly coupled systems.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Social networks and social learning
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Physical simulations are essential tools across critical fields such as mechanical and aerospace engineering, chemistry, meteorology, etc. While neural operators, particularly the Fourier Neural Operator (FNO), have shown promise in predicting simulation results with impressive performance and efficiency, they face limitations when handling real-world scenarios involving coupled multi-physics outputs. Current neural operator methods either overlook the correlations between multiple physical processes or employ simplistic architectures that inadequately capture these relationships. To overcome these challenges, we introduce a novel coupled multi-physics neural operator learning (COMPOL) framework that extends the capabilities of Fourier operator layers to model interactions among multiple physical processes. Our approach implements feature aggregation through recurrent and attention mechanisms, enabling comprehensive modeling of coupled interactions. Our method's core is an innovative system for aggregating latent features from multi-physics processes. These aggregated features serve as enriched information sources for neural operator layers, allowing our framework to capture complex physical relationships accurately. We evaluated our coupled multi-physics neural operator across diverse physical simulation tasks, including biological systems, fluid mechanics, and multiphase flow in porous media. Our proposed model demonstrates a two to three-fold improvement in predictive performance compared to existing approaches.

Problem

• Fourier Neural Operator
• Multi-physics Coupling
• Complex Interactions

Innovation

• COMPOL Framework
• Multi-Physics Phenomena
• Enhanced Prediction Accuracy