<ul><li>Neural network architectures combining physics-informed neural networks with immersed boundary method proposed for solving fluid-structure interaction problems.</li><li>Approach features two architectures: Single-FSI network with unified parameter space and Eulerian-Lagrangian network with separate parameter spaces for fluid and structure domains.</li><li>Empirical studies on 2D cavity flow problem show that Eulerian-Lagrangian architecture outperforms Single-FSI network.</li><li>Use of adaptive B-spline activation functions enhances accuracy near boundaries, while challenges remain in pressure recovery due to lack of explicit force-coupling constraints.</li></ul>

Learning Fluid-Structure Interaction Dynamics with Physics-Informed Neural Networks and Immersed Boundary Methods

Discover more