Accelerating Computational Fluid Dynamics Simulations with Physics-Informed Neural Networks

Abstract Submission: Traditional Computational Fluid Dynamics (CFD) simulations rely on fine mesh grids to achieve high accuracy but are often computationally intensive and time-consuming, especially when dealing with complex geometries. This paper applies Physics-Informed Neural Networks (PINNs) to reduce the computational cost of CFD simulations by using a coarse mesh and leveraging PINNs to predict parameters such as velocity, pressure, and temperature within each cell. We demonstrate the effectiveness of this framework by applying it to a 2D square cavity as a benchmark case, where the PINN accurately learns the relationships between boundary conditions and internal flow properties. Once trained, this network is applied to other geometries’ grids, allowing for enhanced resolution and accuracy without increasing computational complexity. This approach not only accelerates CFD simulations but also maintains high fidelity in results, showcasing its potential for time-sensitive applications involving complex geometries by achieving high-resolution predictions with reduced computational resources.

Learning Objectives/Expected Outcome (Optional) :