[4Q1-IS-2c] Machine learning

[4Q1-IS-2c-02] Physics-Informed Neural Networks for Two-Phase Flow Simulations: An Integrated Approach with Advanced Interface Tracking Methods

〇WEN ZHOU¹, Shuichiro Miwa¹, Koji Okamoto¹ (1. The University of Tokyo)

キーワード：physics-informed neural networks, Navier-Stokes equations, interface tracking methods, two-phase flow

Physics-informed neural networks (PINNs) are emerging as a promising artificial intelligence approach for solving complex two-phase flow simulations. A critical challenge in these simulations are the accurate representation of the gas-liquid interface with different interface tracking methods. Therefore, this study aims to develop a robust and generic PINNs for two phase flows by incorporating the Navier-Stokes equations and three advanced interface tracking methods—specifically, the Volume of Fluid, Level Set, and Phase-Field method—into an improved PINNs framework that has been previously proposed and validated. To further enhance the performance of the PINNs in simulating two phase flow, the phase field constraints strategies and the time divide-and-conquer algorithm are employed for restricting neural network training within the scope of physical laws. The improved PINNs then is optimized by minimizing both the residual and loss terms of partial differential equation. The case of single rising bubble in two-phase flows is simulated to validate the robustness and accuracy of the improved PINNs. The accuracy of the simulations is compared with the velocity, pressure, and phase field against CFD solutions. The results indicate that the improved PINNs coupled with these interface tracking methods offers a satisfactory consistency in simulating rising bubble.