A fluid simulation system based on the MPS method

• Autor(a) principal: SILVA, André Luiz Buarque Vieira e
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UFPE
• dARK ID: ark:/64986/001300000f39d
• Texto Completo: https://repositorio.ufpe.br/handle/123456789/32144
• topic: Computação gráfica; Realidade virtual

Fluid flow simulation is a high active area in Computer Graphics and Virtual Reality, with applications in a wide range of engineering problems. In this scenario, meshless methods like the Moving Particle Semi-implicit (MPS) are a great alternative to deal with large deformations and free-surface flow, problems that usually impose the traditional mesh-based methods to perform inefficiently. This dissertation presents a stable, accurate and parallelized MPS-based technique which benefits from different advances in the MPS literature, and also from parallel computing, to obtain a method that can be adapted for a wide variety of scenarios. The proposed technique can simulate fully incompressible/weakly compressible fluid under different fluid behaviors such as two levels of compressibility, different fluid’ kinematic viscosity, turbulent flows and multiphase interaction. The method was evaluated under classical scenarios like Water Drop, Dam Break flow, R-T instability and Oil Spill, presenting comparable results to the State-OfThe-Art methods. The method and its variations are also integrated on a single solution which can switch on improvements such as better momentum conservation, more precise discretization of differential operators and less erroneous pressure oscillations through a user-friendly graphical interface. This enables a practical selection of models, approaches and parameter tuning, from, for instance, a stable physically coherent free-surface incompressible fluid flow simulation, to a GPU-accelerated multiphase free-surface weakly compressible flow simulation. Based on three different implementations (single-core CPU as the reference, multi-core CPU with OpenMP and multi-core GPU with CUDA for performance improvements), it is shown that the OpenMP-enabled weakly compressible approach achieves a speedup of 2.02 times and the fully incompressible approach of 1.82 times. The CUDA-enabled weakly compressible approach achieves a speedup of 3.15 times while the fully incompressible approach of 2.23 times.