ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering

The bridge influence line (BIL) contains static information for each section of a bridge, which is an important tool for bridge design and condition evaluation. The current influence line identification procedure relies on periodic load testing using calibrated vehicles and always leads to long-time traffic disruption. To overcome this shortcoming and achieve online tracking of high-speed railway (HSR) bridge influence lines, this paper proposes an operational influence line identification approach for HSR bridges based on train-induced responses only. First, to consider the train load uncertainty during operation, a train load interval model is established based on field investigations. Then, the BIL intervals are determined by interval computations and accumulate to generate a continuously updated database. Finally, the influence line is identified from the BIL interval database using a binary classification algorithm. The proposed method is verified by an example of a three-span girder bridge, and a train–bridge interaction model is established to simulate the bridge responses induced by various train loads. The results show that the proposed approach has similar accuracy as the traditional load testing methods, which can achieve high-accuracy online tracking of HSR bridge influence lines.

Unreinforced masonry (URM) walls are commonly found in historic and legacy buildings around the world. The structural resistance of URM walls under in-plane (IP) and out-of-plane (OOP) loads is of primary concern to engineers, as their failure is generally sudden, with catastrophic loss of strength and structural integrity. Due to the complex behavior and inherent uncertainties of the masonry material, engineers opt for the use of low-fidelity (LF) resistance models with limited accuracy, such as design code models and other simplified analytical models in the literature. Models with enhanced prediction accuracy have attracted growing attention, particularly when uncertainty analysis (e.g., reliability evaluation) is needed. As such, high-fidelity (HF) models, such as nonlinear finite element models based on advanced computational mechanics, have been developed and used to characterize the structural behaviors and failure modes of URM walls, particularly the resistance, with remarkable success in terms of accuracy. However, their direct use for resistance prediction and uncertainty analysis is scarce due to the computational burden and technical complications. To address this issue, this study takes an efficient multifidelity (MF) approach that leverages both HF and LF models via information fusion to enhance LF models with only a few HF model evaluations for URM walls. The main research thrust is to develop an MF surrogate model to facilitate uncertainty analysis in the IP and OOP resistance of URM walls. The analysis results indicate that the MF surrogate models developed are capable of achieving significant improvements in terms of accuracy and efficiency in predicting the IP and OOP resistances of URM walls both deterministically and probabilistically, compared with the LF model and the surrogate model developed only based on a limited number of HF model runs.