Abstract:Traditional simulation of complex mechanical systems relies on numerical solvers of Partial Differential Equations (PDEs), e.g., using the Finite Element Method (FEM). The FEM solvers frequently suffer from intensive computation cost and high running time. Recent graph neural network (GNN)-based simulation models can improve running time meanwhile with acceptable accuracy. Unfortunately, they are hard to tailor GNNs for complex mechanical systems, including such disadvantages as ineffective representation and inefficient message propagation (MP). To tackle these issues, in this paper, with the proposed Up-sampling-only and Adaptive MP techniques, we develop a novel hierarchical Mesh Graph Network, namely UA-MGN, for efficient and effective mechanical simulation. Evaluation on two synthetic and one real datasets demonstrates the superiority of the UA-MGN. For example, on the Beam dataset, compared to the state-of-the-art MS-MGN, UA-MGN leads to 40.99% lower errors but using only 43.48% fewer network parameters and 4.49% fewer floating point operations (FLOPs).

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the efficiency and effectiveness problems in the simulation of complex mechanical systems. Traditional simulation methods rely on numerical solvers for partial differential equations (PDEs), such as the finite element method (FEM). Although these methods are accurate, they are computationally costly and have long running times. In recent years, simulation models based on graph neural networks (GNNs) have improved running times while maintaining acceptable accuracy, but these models have problems of insufficient representation and inefficient message passing when dealing with complex mechanical systems. To overcome these problems, this paper proposes a novel hierarchical meshed - graph network (UA - MGN), which achieves efficient and effective mechanical system simulation through up - sampling and adaptive message - passing techniques. Specifically: 1. **Up - sampling - only GNN model**: This model first performs message passing on the coarsest meshed - graph and then gradually increases the resolution until the finest meshed - graph. This design avoids the need for down - sampling, reduces the number of message - passing steps, improves efficiency, and better captures global representations at an early stage. 2. **Adaptive message passing**: In the meshed - graph, edges are grouped according to their directions, and the number of message - passing steps for each group of edges is adjusted, thereby achieving adaptive message passing. This not only alleviates the problems of infinite message - passing loops and over - smoothing, but also adapts to mechanical simulations of different geometries and external conditions, improving the generalization ability of the model. Through these two key techniques, the experimental results of UA - MGN on two synthetic datasets and one real - world dataset show that its performance is better than the existing state - of - the - art methods. For example, on the Beam dataset, compared with the state - of - the - art MS - MGN, the error of UA - MGN is reduced by 40.99%, while the number of network parameters is reduced by 43.48% and the number of floating - point operations is reduced by 4.49%.

Up-sampling-only and Adaptive Mesh-based GNN for Simulating Physical Systems

Graph Neural Networks (gnns) Based Accelerated Numerical Simulation

Graph Neural Networks for Mesh Generation and Adaptation in Structural and Fluid Mechanics

Physics-informed MeshGraphNets (PI-MGNs): Neural finite element solvers for non-stationary and nonlinear simulations on arbitrary meshes

Learning-Based Finite Element Methods Modeling for Complex Mechanical Systems

A Multi-Fidelity Graph U-Net Model for Accelerated Physics Simulations

Learning Mesh-Based Simulation with Graph Networks

Multiscale graph neural networks with adaptive mesh refinement for accelerating mesh-based simulations

X-MeshGraphNet: Scalable Multi-Scale Graph Neural Networks for Physics Simulation

Learning to Simulate Complex Physical Systems: A Case Study

Mesh-based GNN surrogates for time-independent PDEs

Scientific Computing Algorithms to Learn Enhanced Scalable Surrogates for Mesh Physics

Learn to Simulate Finite Element Analysis Via Mesh-based Graph Networks

A Multigrid Graph U-Net Framework for Simulating Multiphase Flow in Heterogeneous Porous Media

A Microstructure-based Graph Neural Network for Accelerating Multiscale Simulations

Multi-scale Graph Neural Network for Physics-Informed Fluid Simulation

Deep Learning-based surrogate models for parametrized PDEs: handling geometric variability through graph neural networks

PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems

Multi-scale rotation-equivariant graph neural networks for unsteady Eulerian fluid dynamics

Convolution, aggregation and attention based deep neural networks for accelerating simulations in mechanics