UPCommons. Portal del coneixement obert de la UPC
Publicación
Artículo científico (JournalArticle)
Preconditioning iterative solvers via the Empirical Interscale Finite Element Method (EIFEM)
This work explores the use of the Empirical Interscale Finite Element Method (EIFEM) as a novel preconditioner for iterative solvers in the linear elastic regime. Originally developed as a multiscale Reduced Order Model (ROM), EIFEM captures the kinematics of a unit cell and leverages static condensation with a reduced basis to construct a coarse approximation. We integrate EIFEM as a coarse space within a preconditioned conjugate gradient solver, incorporating pre- and post-smoothing steps. Numerical experiments confirm that EIFEM-based preconditioning significantly reduces iteration counts, particularly for near-solid structures. Although non-solid structures also benefit from reduced iteration counts, the convergence rate is lower compared to solid or nearly-solid cases. This behavior stems from the sensitivity of the solver to the accuracy of the ROM approximation, and we show discontinuous edge functions and discuss other potential strategies as a way to enhance this accuracy. In terms of scalability, the preconditioner is robust, maintaining a nearly constant iteration count as the problem size increases. This work is the first application of EIFEM for preconditioning purposes, demonstrating its potential as a scalable and efficient preconditioner for iterative solvers in 2D and 3D geometries., R.Rubio and A.Ferrer gratefully acknowledge the support of the following projects: ”Optimización topológica con compuestos y materiales flexibles” (PID2023-153213NA-I00), ”Disseny computacionalòptim de bateries amb materials compostos sostenibles per a una mobilitat més eficient” (ACE122/24/000030) and ”Design of flexible materials, structures via shape and topology optimization” (PCI2024-155060-2). J.A. Hernández expresses gratitude by the support of the “MCIN/AEI/10.13039/501100011033/y por FEDER una manera de hacer Europa” (PID2021-122518OB-I00)., Peer Reviewed