Strain-based autoregressive modelling for system identification of railway bridges

Abstract

Vehicular traffic represents the most influential loads on the structural integrity of railway bridges, therefore the design on dynamic criteria. This work explores the use of strain dynamic measurements to characterize the health condition of railway bridges under moving train loads. Specifically, the approach proposed in this work exploits the implementation of auto-regressive (AR) time series analysis for continuous damage detection. In this light, continuously extracted AR coefficients are used as damage-sensitive features. To automate the definition of the order of the AR model, the methodology implements a model selection approach based on the Bayesian information criterion (BIC), Akaike Information Criterion (AIC) and Mean Squared Error (MSE). In this exploratory investigation, the suitability and effectiveness of strain measurements against acceleration-based systems are appraised through a case study of a simply supported Euler-Bernoulli beam under moving loads. The moving loads problem in terms of vertical accelerations and normal strains is solved through modal decomposition in closed form. The presented numerical results and discussion evidence the effectiveness of the proposed approach, laying the basis for its implementation to real-world instrumented bridges.