Two-stage stochastic model updating approach based on Bayesian framework and hybrid multi-population migrant genetic–Metropolis–Hastings algorithm for heavy-haul railway bridges

Abstract

This study proposes a stochastic model updating approach using an improved multi-population migrant genetic algorithm (MMGA) with the Metropolis–Hastings (MH) algorithm based on the Bayesian inference approach. To enhance the search capacity of the genetic algorithm (GA), the MMGA introduces heuristic: the promotion and elimination mechanisms of urban population migration. The stochastic model updating approach includes two stages. The optimal values identified by the MMGA without considering the parameter uncertainties in the first stage are used as the initial values for the MH algorithm to accelerate Bayesian inference for uncertain parameters in the second stage. A practical application, the Yellow River Bridge with a heavy-haul railway line, demonstrates that the proposed approach offers an increased probability of gaining the global optimum, and predictions of displacements and frequencies provided by the updated finite element model are comparable to those obtained from the field tests.