City Research Online

Abstract

Recently, we have developed a single-phase incompressible Smoothed Particle Hydrodynamics (ISPH) method accelerated by a graphic neural network (GNN) model (Zhang et al., 2023b) for predicting the pressure, which is trained using the ISPH results. The GNN model has shown as atisfactory generality by Zhang et al. (2023c), who demonstrated that the GNN model trained using the wave-only cases can be used to model wave-structure interaction problems with satisfactory accuracy and efficiency. In this paper, we train the GNN model for predicting the wave pressure using two-phase computational fluid dynamics (CFD)solutions. The trained GNN model is preliminarily tested using both the single-phase ISPH and two-phase CFD data. Satisfactory results are obtained. This builds the basis for further applications of the GNN model for either two-phase CFD practices or single-phase ISPH modelling.

Publication Type:	Conference or Workshop Item (Paper)
Additional Information:	The final paper has been published by ISOPE and it's available online at: www.isope.org
Publisher Keywords:	ISPH; machine Learning; GNN; CFD; wave-structure interaction
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Departments:	School of Science & Technology School of Science & Technology > Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads