Damping Assessment of Railway Bridges Considering Nonlinear Decay Behaviour: Experimental Investigation Based on Train-Induced Vibrations

Abstract

A reliable determination of the damping characteristics of railway bridges is essential for realistic dynamic assessments, as damping strongly influences predicted vibration responses and resonance effects caused by train traffic. Conventional analysis methods assume linear systems with viscous damping, whereas in situ measurements reveal amplitude-dependent damping behaviour that is inherently nonlinear. Therefore, considerable parameter scattering can occur, depending on the time window under consideration, in an idealised linear system. Given this background, this contribution evaluates the suitability of different damping models for describing the measured decay behaviour of railway bridges. An idealised single-degree-of-freedom model is used to compare four nonlinear damping models with the conventional viscous model based on measured acceleration responses following train passages. The investigation comprises 53 decay processes recorded on nine single-span steel railway bridges. The related model parameters are identified by minimising the deviation between measured and simulated amplitude decay functions and are assessed with respect to parameter sensitivity, dependence on the selected time window, and overall model quality. The results show that basically all investigated models reproduce the measured decay behaviour with high accuracy. However, while several nonlinear models exhibit pronounced parameter sensitivity, the considered high-power damping model, based on a potential-function approach, demonstrates high robustness against parameter variance while providing a precise approximation of the decay behaviour. Although the viscous damping model is a practical and accurate choice in most cases, the proposed high-power model provides a reliable extension for capturing nonlinear damping effects and offers practical advantages, particularly when the linear model no longer provides a satisfactory fit due to high parameter variance.