City Research Online

Abstract

As the mesh-free approach, the incompressible Smoothed Particle Hydrodynamics (ISPH) method is emerging as a potential tool for simulating the wave-structure interaction problems. The pressure in the conventional ISPH method is obtained by solving the pressure Poisson’s equation (PPE), which is the most time-consuming part. Recently, the machine learning (ML) techniques have been widely used in the fluid dynamics. In this paper, the graph neural network (GNN) is combined with ISPH and used to predict the fluid pressure instead of solving the PPE directly. The GNN supported ISPH method (ISPH_GNN) with one trained GNN model based on training data generating from relatively simple wave propagation cases without any structure will be extended to simulate different relatively complex cases of wave interaction with fixed structures. It will be demonstrated that the ISPH_GNN method does not only give satisfactory results, but also shows good generalization properties. In addition, this method will be demon strated to requires much less computation time than the conventional ISPH for estimating pressure for cases with a large number of particles that is usually needed in the large-scale simulation using ISPH.

Publication Type:	Conference or Workshop Item (Paper)
Additional Information:	This paper has been published in its final form by ISOPE, the International Society of Offshore and Polar Engineers and it's available online at: https://www.isope.org/
Publisher Keywords:	neural network, simulation, journal, khayyer, machine learning, isph, numerical simulation, particle, gnn, ppe
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TA Engineering (General). Civil engineering (General)
Departments:	School of Science & Technology School of Science & Technology > Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads