Finite element model updating of continuous girder bridge based on IMEA-BP neural network

Purpose This paper aims to update the finite element model of a pre-stressed concrete continuous girder bridge using a back propagation neural network optimized by the improved mind evolutionary algorithm (IMEA). Design/methodology/approach In this paper, we introduce particle swarm optimization in the convergence process of the mind evolutionary algorithm (MEA) and then dynamically adjust the inertia weights in the moving speed of the particle swarm through a nonlinear decreasing strategy, which is used to solve the drawback of the algorithm’s arbitrarily initialized population. Finally, the measured natural frequency of the Yunliang River Bridge is used as the input parameter, and the design parameters are obtained by updating in stages (pile foundation stage and bridge completion stage). Findings The relative error of the natural frequencies of the continuous girder bridge is significantly reduced from more than 7% before updating to less than 3%. Compared with the 2.31% average relative error of the natural frequency of the continuous beam bridge updated by the MEA-BP neural network, the error of the natural frequency after the update of the IMEA-BP neural network model is only 1.01%. The IMEA-BP neural network demonstrates stronger applicability and higher update precision. Compared to the bridge finite element models without phased updating, the phased updating presents a better performance. Originality/value This study provides an accurate and effective method of updating finite element models for continuous girder bridges, as well as a valuable reference for updating finite element models for other types of bridges.