Simulation of Moving Objects on an Cartesian Mesh Using an Improved Alternating Direction Forcing Immersed Boundary Method

点击翻译按钮获取中文摘要

The main aim of this PhD project is to develop an improved alternating direction forcing immersed boundary method (AFD IBM) for turbulence-resolving simulations of moving objects on a fixed Cartesian mesh. Most of the immersed boundary methods suffer from spurious force oscillations (SFOs) when moving boundaries are simulated. One of the main sources of SFOs has been identified as the velocity discontinuity that occurs at the immersed boundary interface. The proposed method is based on a cubic spline interpolation to obtain an artificial internal flow within the solid domain while ensuring the continuity of the velocity field at the wall of the solid object. Even though the proposed method is intended for moving objects, it also provides significant improvements for static objects. The ADF IBM will be implemented in the high-order, finite-difference flow solver Incompact3D, which is a powerful tool to simulate turbulent flows using the most powerful supercomputers in the world. Once the ADF IBM is implemented and validated, it will be used to study energy efficient swimming by multiple fish through deep reinforcement learning. By combining state-of-the-art turbulence-resolving simulations of the 3D Navier-Stokes equations with reinforcement learning, this project will demonstrate that fish can reduce their energy expenditure by harnessing turbulence.