Resolução da equação da onda utilizando métodos multigrid espaço-tempo
Resumo
Resumo: Neste trabalho apresenta-se a avaliação de diferentes formas de solução para problemas modelados pela equação da onda, para os casos 1D e 2D. Utiliza-se para discretização espacial, o Método das Diferenças Finitas ponderado por um parâmetro n em diferentes estágios de tempo, para obter-se um esquema de solução implícito. Com isso, propõe-se a utilização de diferentes varreduras no tempo, a fim de gerar métodos robustos e eficientes, desde a clássica Time-Stepping, até outra varredura que envolve simultaneamente o espaço e o tempo, como Waveform Relaxation. Neste trabalho, combina-se o método dos Subdomínios com a estratégia Waveform Relaxation para reduzir as fortes oscilações que ocorrem o início do processo iterativo. Obtém-se excelentes resultados ao aplicar o método Multigrid para esta classe de problemas, já que, melhora-se muito os fatores de convergência calculados a partir das soluções aproximadas do sistema de equações resultante das discretizações. Na verificação das metodologias propostas e suas características, apresentamse simulações de propagação de ondas envolvendo problemas uni e bidimensionais, onde analisa-se os erros de discretização, ordens efetiva e aparente, fator de convergência, ordens de complexidade e tempo computacional. Abstract: In this thesis presents the evaluation of different forms of solution for probelms modeled by the wave equation, for the 1D and 2D cases. The Finite Difference Method is used for the spatial discretization, weighted by a parameter n at different time steps, in order to obtain an implicit solution. With this, it is proposed the use of different sweeps in time, in order to generate robust and efficient methods, from the classical Time-Stepping, to other less usual sweep as Waveform Relaxation. In this work, the Subdomains method is combined with the Waveform Relaxation strategy to reduce the strong oscillations that occur early in the iterative process. Excellent results are obtained when applying the Multigrid method for this class of problems, since the convergence factors calculated from the approximate solutions of the system of equations resulting from the discretizations are greatly improved. In the verification of the proposed methodologies and their respective advantages, simulations of wave propagation involving one- and two-dimensional problems are presented, where the discretization errors, effective and apparent orders, convergence factor, complexity orders and computational time are analyzed.
Collections
- Teses [100]