Métodos paralelos altamente escalables para modelado electromagnético de problemas a gran escala

Resumo

Surface integral equation (SIE) methods based on the method of moments (MoM) have emerged as a powerful and indispensable tool in computational electromagnetics (CEM) for the simulation and engineering of a wide range of applications. The use of this type of simulation tools can be critical in cutting-edge applications such as EMC engineering or the field of nanoplasmonics, where extremely large-scale problems that also suffer from problematic multiscale and multiphysics issues that complicate the accurate analysis of challenging systems. Schwarz preconditioners based on the domain decomposition method (DDM) are presented as one of the best alternatives to deal with the ill-conditioning of such systems. In this context, this doctoral thesis combines the latest breakthroughs and devises new methodologies in SIE algorithms, computing hardware, and parallelization skills to address the efficient and robust solution of the extremely large and ill-conditioned problems. First, a highly efficient domain decomposition method is developed and integrated with the advanced SIE technologies of the research group that endorses this thesis. Next, novel SIE advances in non-conformal discretization and preconditioning of deep multiscale and multiphysics problems are incorporated to improve the versatility of the proposed methods, providing a truly multipurpose methodology for the electromagnetic simulation of extremely challenging problems with real-life application.