Search papers, labs, and topics across Lattice.
This paper introduces a self-attention-based neural surrogate model designed to approximate the dynamics of free-surface flows simulated via the Particle Finite Element Method (PFEM). By leveraging attention mechanisms to effectively model node interactions and spatial dependencies, the proposed architecture addresses the computational challenges posed by evolving geometries and non-Newtonian rheologies. The results demonstrate that the model achieves accurate predictions of transient dynamics and final configurations while significantly enhancing scalability, making it suitable for engineering-scale applications.
Surrogate models using self-attention can dramatically reduce the computational cost of simulating complex free-surface flows while maintaining high accuracy.
High-fidelity simulations of free-surface flows using Lagrangian methods such as the Particle Finite Element Method (PFEM) are computationally demanding due to continuous domain updates and repeated solution of the governing equations. This challenge is further amplified by non-Newtonian rheologies, where material nonlinearities increase computational cost. These limitations motivate the development of efficient surrogate models to approximate PFEM dynamics at reduced cost. While data-driven deep learning approaches are promising, a key challenge is designing models that operate on arbitrary and evolving geometries. We propose a self-attention-based neural surrogate for PFEM simulations of free-surface flows. The architecture leverages attention mechanisms to model node interactions and capture complex spatial dependencies, while preserving the PFEM mesh discretization. This provides a geometric and topological framework for remeshing and node redistribution, maintaining high-quality spatial discretization during rollouts, improving long-term stability, and enabling reconstruction of derived mechanical quantities via standard finite element operators. Two attention formulations are considered: a standard self-attention mechanism and a linear variant that reduces computational cost and improves scalability. The models are evaluated on two- and three-dimensional free-surface flow benchmarks with evolving geometries, varying material parameters, and non-Newtonian fluids. Results show accurate prediction of transient dynamics and final configurations, with significantly improved scalability. The mesh-based formulation also enables direct reconstruction of quantities such as stress fields. Overall, the framework provides an accurate and scalable surrogate strategy for PFEM simulations in engineering-scale applications.
