Asymptotic Behavior of hp-HGS (hp-Adaptive Finite Element Method Coupled with the Hierarchic Genetic Strategy) by Solving Inverse Problems

摘要

The new hp-HGS multi-deme, genetic strategy for economic solving of the parametric inverse problems is introduced. The inverse problems under consideration are formulated as the global optimization ones, where the objective express the discrepancy between the computed and measured energy. The efficiency of the proposed strategy results from the coupled adaptation of the accuracy of solving optimization problem and the accuracy of hp-FEM direct problem solver. The asymptotic analysis allows to estimate the expected computational cost of hp-HGS and to show its advantage over the single population SGA algorithm as well as over the HGS strategy without the FEM error scaling.