Research on the Application of Physical Information Neural Network in Multiscale Fluid Dynamics Simulation

Jinyu Liu; Yinxiao Yan; Junhui Wang

doi:10.62051/ijcsit.v5n3.11

Authors

Jinyu Liu
Yinxiao Yan
Junhui Wang

DOI:

https://doi.org/10.62051/ijcsit.v5n3.11

Keywords:

Physically informed neural networks, Multiscale fluid simulation, Deep learning, Turbulence simulation, Microscale flow, Multiphase flow, Adaptive sampling

Abstract

In this paper, a multi-scale fluid dynamics simulation framework based on physical information neural network (PINN) is proposed, which achieves efficient simulation of multi-scale flow characteristics by means of a multi-resolution network structure, physical constraints on scale decomposition and adaptive sampling strategy. Experimental results show that the framework outperforms traditional methods in microscale flow, turbulence and multiphase flow problems, with a 37% improvement in the average prediction accuracy and a 75% reduction in the computational resource requirement. In particular, it shows excellent generalisation performance when dealing with complex flows with large scale spans, providing a new method for fluid dynamics simulation with both accuracy and efficiency.

Downloads

Download data is not yet available.

References

[1] Wei L, Freris N M. Multi-scale graph neural network for physics-informed fluid simulation [J]. The Visual Computer, 2025, 41(2): 1171-1181.

[2] Yang Z, Dong Y, Deng X, et al. AMGNET: multi-scale graph neural networks for flow field prediction [J]. Connection Science, 2022, 34(1): 2500-2519.

[3] Cao Y, Chai M, Li M, et al. Efficient learning of mesh-based physical simulation with bi-stride multi-scale graph neural network [C]//International conference on machine learning. PMLR, 2023: 3541-3558.

[4] Hu M, Yang Z. Perspective on multi-scale simulation of thermal transport in solids and interfaces [J]. Physical Chemistry Chemical Physics, 2021, 23(3): 1785-1801.

[5] Gnanasambandam R, Shen B, Chung J, et al. Self-scalable tanh (stan): Multi-scale solutions for physics-informed neural networks [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(12): 15588-15603.

[6] Liu X, Zhang Y, Guo T, et al. A multi-scale hybrid attention Swin-transformer-based model for the super-resolution reconstruction of turbulence [J]. Nonlinear Dynamics, 2025: 1-30.

[7] Chen M, Niu R, Zheng W. Adaptive multi-scale neural network with resnet blocks for solving partial differential equations [J]. Nonlinear Dynamics, 2023, 111(7): 6499-6518.

[8] Carapezzi S, Delacour C, Boschetto G, et al. Multi-scale modeling and simulation flow for oscillatory neural networks for edge computing [C]//2021 19th IEEE International New Circuits and Systems Conference (NEWCAS). IEEE, 2021: 1-5.

[9] Migus L, Yin Y, Mazari J A, et al. Multi-scale physical representations for approximating pde solutions with graph neural operators [C]//Topological, Algebraic and Geometric Learning Workshops 2022. PMLR, 2022: 332-340.

[10] Du Q, Ellingsen K, M’Hamdi M, et al. The integration of neural network and high throughput multi-scale simulation for establishing a digital twin for aluminium billet DC-casting [J]. Materials Transactions, 2023, 64(2): 360-365.