Fluid-solid Coupling by Neural Networks

In this study, the flow past a two-dimensional cylinder is adopted as the simulation case to investigate the capability of the neural network for the immersed boundary method. For the direct-forcing immersed boundary method, a dirac function is used for the velocity interpolation and force spreading with the information of neighbor points. However, this step can also be achieved by the machine learning model with a robust performance.

Publication

• Fang, Dehong, and Jifu Tan. "Immersed boundary-physics informed machine learning approach for fluid–solid coupling." Ocean Engineering 263 (2022): 112360.