Two-stage stochastic model updating method for highway bridges based on long-gauge strain sensing

Currently, the total number of highway bridges is growing rapidly. To ensure the safety, accurate evaluation of bridges is necessary. Among the existing methods, a finite element model which can reflects the bridge's actual condition is usually required. Thus, the bridge model updating is inevitable. Although many model updating methods have been proposed, there are still some limitations, such as the difficulty in acquisition of effective structural information from measured data and the need for time-consuming optimization simulations. Under these backgrounds, based on novel long-gauge strain time history, the study proposes a two-stage bridge model updating method by combining a radial basis function (RBF) neural network with Bayesian theory to increase its efficiency and accuracy on highway bridges. This method's feasibility was tentatively verified through a series of numerical cases. An indoor model experiment was also conducted to further investigate this method's performance. The results demonstrated that this method performs well under various conditions and has the potential to be applied in actual cases.