Abstract: Fluid simulation is a highly regarded research direction in the field of computer graphics, particularly in generating realistic models of fluid dynamics. When generating complex fluid control effects in large-scale scenes, existing methods often encounter issues such as particle accumulation and collapse, leading to excessive memory usage after surface reconstruction, which limits real-time applications in VR, gaming, and film. To address these issues and accurately control the fluid shape to match the target model while ensuring fluidity, we propose a simulation method that combines multi-layer sampling and multi-constraint controls. First, we introduce two sampling methods based on the uniformity of the model: for uniform models, we use dual-layer sampling, obtaining internal sampling particles through voxelization and surface sampling particles through blue noise sampling; for non-uniform models, we only use blue noise sampling for surface sampling particles. Next, we design various constraint models based on these sampling methods, including shape constraints, density constraints, half-density constraints, and weighted constraints, to ensure precise control of the fluid. Notably, our method can automatically adjust according to model characteristics, ensuring fluid flow in simple models and adherence in complex models. Experimental results across multiple scenarios demonstrate that compared to existing methods, our approach improves both fluid control accuracy and efficiency, achieving operational speeds of 60 to 120 frames per second in VR environments, thereby opening up new possibilities for real-time applications of fluid dynamics.