On the Parameters Investigation of a Non-Intrusive Multiscale Framework for Structural Analysis

Resumo

IGL-GFEMgl is a multiscale framework proposed by H. Li, P. O’Hara, and C. A. Duarte in 2021 that combines the IGL strategy with the GFEMgl. In the Iterative Global-Local method (IGL), two different meshes are adopted. The global mesh is used to describe the global behavior of the structure. Local features are represented in the local mesh. The solution of the two meshes is coupled through an non-intrusive iterative algorithm that exchanges displacements and enforces the equilibrium between them. The GFEMgl considers two scales of representations, but the coupling is provided by the GFEM’s enrichment strategy. Finally, in the IGL-GFEMgl
framework, a third problem is defined, named mesoscale. The mesoscale works as a bridge between the two methods (IGL and IGL-GFEMgl), allowing a non-intrusive coupling of the global problem FEM solution and the meso-local scale solution provided by the GFEMgl. In this work, the commercial software Abaqus solves the global problem and is coupled with an in-house computational platform where GFEMgl is already implemented. A thorough investigation is performed over some IGL-GFEMgl parameters, such as the size of the mesoscale and the use of acceleration techniques to improve the convergence of the method.