Importance of the global sensitivity analysis in development of metamodels for geotechnical back analysis

Abstract

This paper investigates the use of global sensitivity analysis during the development of metamodels for back-analysis of geotechnical problems. Variance-based global sensitivity analysis was combined with Particle swarm optimization algorithm and POD-ERBF metamodel to backcalculate the hypothetic (synthetic) problem of anchor-supported excavation. PLAXIS 2D FE code was used for numerical simulations. The results of the back analysis for different sets of model parameters were compared to emphasize the importance of global sensitivity analysis before performing the back analysis and the recommendations for the robust modeling of presented geotechnical problem were given.