Search

This article presents a novel methodology to extract the bridge frequencies from the vibrations recorded on train-mounted sensors. Continuous wavelet transform is used to distinguish the bridge frequencies from the other peaks that are visible in the Fourier amplitude spectrum of the accelerations recorded on train bogies. The efficacy of the proposed method is demonstrated through numerical case studies. For this, a detailed three-dimensional finite element model that can capture the vibration characteristics of the bridge, track, and train is created, and each component of the model is separately validated. The train model used is a three-dimensional multi-degree-of-freedom system that can simulate the pitching and rolling behavior. The train was then virtually driven over the bridge at different speeds and under varying track irregularities to evaluate the robustness of the proposed method in extracting bridge frequencies from train-mounted sensors under different conditions. The proposed methodology is shown to be capable of identifying bridge modal frequencies even for aggressive track irregularity profiles and relatively high speeds of trains. [ABSTRACT FROM AUTHOR]

A new framework for long-term monitoring of bridges is proposed in order to negate (i) the impact of measurement uncertainties on damage detection in vibration-based structural health monitoring and (ii) the low sensitivity of damage indicators to low levels of damage. The framework is developed using three vibration-based damage indicators that have an intuitive physical correlation with damage: modal curvature, modal strain energy and modal flexibility. The article first quantifies the efficacy of these damage indicators when based on two observations, one from the undamaged state and one from the monitored state, in detecting and locating damage for different damage levels that are simulated on an 84-m long railway bridge. A long-term monitoring framework based on a new parameter defined as the frequency of the damage indicator exceeding the threshold value within a population of observations is developed. Impact of several factors including the damage location, damage indicator used in the framework, and the noise level on the success of the developed framework was investigated through numerical analysis. The new parameter, when used together with modal strain energy, was shown to provide a very clear picture of damage initiation and development over time starting from very low damage levels. Furthermore, the location of the simulated damage can be identified successfully at all damage levels and even for very high noise levels using the proposed framework. [ABSTRACT FROM AUTHOR]