Calibration of Micromechanical Parameters for the Discrete Element Simulation of a Masonry Arch using Artificial Intelligence

Abstract

This study focuses on an old but still unresolved problem of automatically calibrating the constitutive parameters of discrete element models. Instead of the troublesome and time-consuming manual trial-and-error method, which is typical today, the authors suggest using artificial intelligence techniques. A masonry arch is analysed, whose experimental static load–displacement behaviour is known from the literature. An attempt is made to match this behaviour with discrete element models, through finding appropriate quantitative values for the parameters. Two methods (Genetic Algorithm (GA) and Particle Swarm Optimisation (PSO)) are tested and, since PSO turns out to be more reliable, a further improved version, ‘Trust-Based Particle Swarm Optimisation’ (TBPSO), is proposed. The results show that (1) TBPSO quickly leads to suitable alternative parameter sets that make the discrete element model match the behaviour of the real experiments and (2) the optimal values of the parameters strongly depend on the loading velocity and the discretisation method used.