Novo paradigma em Física Computacional - software livre e computação em nuvem - aplicado à solução numérica de EDPs

• Ano de defesa: 2014
• Autor(a) principal: Oliveira, Marceliano Eduardo de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Física computacional; Software livre; Volumes finitos; Análise numérica; Equações diferenciais parciais
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/11277

The main goal of this work is numerically solving physics problems associated with transport phenomena, which are described by partial differential equations, whose numerical solution requires the use of computer codes. We use open source software like R and OpenFOAM, writing our own routines. The aim is to approach computational physics from a new paradigma, composed by open source software and cloud computing. To numerically solve a partial differential equation one needs some sort of discretization. We first study the frequent found numerical methods: Finite Difference (FD), Finite Volume (FV), Finite Elements (FE). Latter on, we implement a finite volume scheme for the solution of the diffusion equations in 1D and 2D using R and the two phase flow in a porous media using OpenFOAM. We choose the (FV) for its simplicity and because it has some conservatives properties resulting from vector calculus identities. Finally we perform some numerical experiments. During the work we keep in mind code reusability and portability. Our main achievements are: implementation of highly portable and reusable codes (specialized routines that can be used in different problems) into R and OpenFOAM that can be used from notebooks to cloud computing. We present also a scheme to specify domain heterogeneity through a blocking operator. We apply the studied methodology for numerically solving multiphase flows in porous media. We also show its application for solving classical problems like 1D and 2D diffusion.