A comparative study of agent models via a MCS based stochastic model updating method

Stochastic model updating always involve huge calculation cost during uncertainty analysis of large, refined Finite Element (FE) models. This manuscript presents a Monte Carlo simulation (MCS) based model updating method taking uncertainties in FE models into account. In order to reduce the calculation burden, various kinds of agent models are employed during direct MCS. Particular emphasis is placed on the agent models' differences in fundamental principles and the advantages and disadvantages of their performances on model updating. In the proposed model updating method, direct MCS is improved with agent models, and the uncertain parameters are subsequently calibrated with an iterative test-analysis correlation. A set of precise parameters and a satisfactory forecast of features can be obtained with acceptable calculation and time cost because of the employing of agent models. A FE model of a satellite is given to fix ideas and illustrate the approach's application on complex engineering structures.