Eikonal Regularisation in PINNs for 3D Fluid Dynamics Explored

A new research paper explores the effectiveness of Eikonal Regularisation in physics-informed neural networks (PINNs) for simulating three-dimensional fluid dynamics. The study, authored by Muhammad Akbar Khan, investigates how the weight of this regulariser impacts accuracy and reproducibility across various benchmarks, including translating, rotating, and slotted spheres. The findings indicate that the optimal weight is highly dependent on the flow's geometric complexity and that while PINNs offer a promising approach, a fifth-order WENO solver demonstrated superior accuracy in the tested scenarios. AI

IMPACT This research could lead to more accurate and reproducible simulations in fluid dynamics using AI.

RANK_REASON The cluster contains a research paper detailing a novel application of neural networks to a specific scientific problem. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Eikonal Regularisation
• fluid dynamics
• Level set formulation of two-dimensional Lagrangian vortex detection methods.
• Muhammad Akbar Khan
• physics-informed neural networks
• WENO solver

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →