Smooth Particle Swarm Fluid Simulation Algorithm Based on Graph Neural Network

Smoothed Particle Hydrodynamics (SPH), as a fluid simulation technique, has significant advantages in dealing with complex boundary and large deformation simulation problems, and has been widely applied in the fields of film and television animation special effects, virtual reality, metaverse, and digital twins. However, traditional methods combining SPH with neural networks have limitations in capturing local area details. To address these issues, this paper proposes an improved model - Adaptive Search Smoothed Particle Network (ASSPN). ASSPN integrates deep learning technology, optimizes the neural network structure, and enhances computational accuracy and detail simulation accuracy. Compared to traditional SPH and Neural Particle Method (NPM), ASSPN demonstrates higher robustness in handling boundary conditions and complex application scenarios, offering greater spatial discretization flexibility in the field of fluid simulation, and is expected to become a powerful tool for fluid dynamics modeling.