Implementação não intrusiva do Método dos Elementos Finitos Generalizados Global-Local para simulação de problemas com trincas

Abstract

This master's thesis expands the implementation of the non-intrusive coupling strategy between the commercial software Abaqus and the INSANE (INterative Structural ANalysis Environment) computational environment, through the IGL-GFEMgl strategy, emphasizing the treatment of Linear Elastic Fracture Mechanics (LEFM) problems. The IGL-GFEMgl strategy combines the Iterative Global-Local (IGL) algorithm with the Generalized Finite Element Method with global-local enrichment (GFEMgl), dividing the analysis into three scales (a global, a meso and a local scale) and their respective models. In the implementation proposed, the global scale model is solved by the commercial software Abaqus via the Finite Element Method (FEM), while GFEMgl, implemented in the INSANE environment, performs the enrichment of the mesoscale model from the solution of the local scale model. The coupling with the global scale is iteratively provided by transferring displacements between the global scale and the mesoscale and the correction of the residual forces, using the IGL algorithm. Through the non-intrusive procedure of IGL, an iterative algorithm, and relaxation techniques were implemented in INSANE to couple the global scale model with the mesoscale model. INSANE's communication with Abaqus is carried out through scripts developed in Python programming language and interpreted with the help of the Jython project. The proposed expansion allowed the description and propagation of cracks at the local scale and implemented a more computationally efficient approach to the strategy. The implementation validation was carried out based on numerical simulations investigating parameters related to GFEMgl under the interactive strategy of non-intrusive coupling. The investigations revealed that the increase in the size of the buffer zone of the mesoscale improves the solution when compared with a reference model. It was also observed that there was no significant improvement in the accuracy of the results with the increase in the number of global-local cycles, such a fact can be explained by the repetition of the global-local solution process required for converging the iterative coupling strategy. It was also verified that the dynamic relaxation technique accelerates the convergence of the coupling strategies without compromising the accuracy of the results. This technique contributed to making IGL-GFEMgl comparatively competitive with standard solution procedures.