Simulação numérica do escoamento com superfícies livres empregando o método Weakly Compressible Smoothed Particle Hydrodynamics (WCSPH) e o processamento em paralelo (CUDA)

• Autor(a) principal: Góes, Josecley Fialho
• Data de Publicação: 2016
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13704
• topic: Computational fluid dynamics; CUDA; Particle Methods; Smoothed Particle Hydrodynamics; GPU/CUDA; Dinâmica dos Fluidos Computacional; Métodos de partículas; Smoothed Particle Hydrodynamics; Escoamento instável (Dinâmica dos fluidos); Fluidos newtonianos; Cálculos numéricos Programas de computador; Métodos de simulação; Partículas Escoamento Métodos de simulação; CNPQ::CIENCIAS EXATAS E DA TERRA::MATEMATICA::MATEMATICA APLICADA

In this work we introduce the Lagrangian meshfree Smoothed Particle Hydrodynamics (SPH) method oriented to the numerical simulation of compressible and weaklycompressible flows of Newtonian fluids. In this method, the problem domain is discretized using particles which carry physical properties that are calculated via interpolation, using the kernel function, from the properties of its neighboring particles. Depending on the algorithm used, the search for neighboring particles can reach a computational cost of the order of N², where N is the number of particles used on the numerical simulation. Therefore, we sought to develop a two- and threedimensional numerical simulator parallelized employing the C language and the CUDA parallelization library, Which enables parallel processing using multi-core processors of NVIDIA Graphics Processing Units (GPUs). To increase computational efficiency, access to global memory and data transfer between CPU and GPU was minimized, and float4 type variables were used to allocate the vectors, ensuring the size and alignment requirements of the CUDA library. In addition, the data was referenced to the texture memory. The problem of the Dam Break was simulated in 2D and the results compared with the sphysics, using the developed algorithms. The numerical results were also validated considering the resolution of the flow problem (bi and three-dimensional, respectively) in a spillway with steps. Thus, for the dam break problem it can be observed that there is a good qualitative agreement between the flow patterns and velocity values calculated with the numerical code developed in this work and the results obtained with sphysics for the same instants of time. In addition, a gain in computational efficiency was obtained in the version paralleled with the CUDA library, when compared to the serial version. In relation to the problem of flow in a spillway with two-dimensional steps, it was verified that the numerical simulator accurately reproduced the velocity field and captured the variation of the pressure field in order to prevent the phenomenon of cavitation on the stairs due to the appearance of negative pressures and the three-dimensional version was simulated realistically.