A Data-driven Methodology for Damage Detection of Roadway Bridges Using Stress Data Distributions

Abstract

Knowing the health state of bridges and viaducts in complex infrastructures enables the structural risk management and optimization of maintenance actions. Real-time data collection from infrastructures subject to traffic loads allows learning about their behavior and detecting anomalies. In this study, a probabilistic approach for damage detection of existing bridges is proposed. The methodology makes use of stress data sets, which can be provided by innovative sensors, to identify anomalies in the static response of bridges and viaducts. More specifically, the local stress data allows the definition of a reference stress distribution, which is strongly related to the state of the structure. When damage occurs, a redistribution of stresses is identifiable by analyzing the evolution of local stress data. The validation process involves performance analysis at the scales of the individual element and the whole structure. Traveling loads are simulated using a Monte Carlo method, while stresses are estimated using a numerical model. The validation of the proposed methodology analyzes the numerical model of an Italian reinforced concrete (RC) arch bridge with stiffening deck, evidencing excellent damage detection capabilities.