Traffic hazard on main road’s bridges: real-time evaluation and management of the risk related to design load overcoming events

Bridges are among the most vulnerable elements of road networks because they undergo to failures that can undermine their serviceability and even produce a collapse. Since extreme traffic load hazard represent one of the main causes of bridge failures, the implementation of real-time risk management strategies is mandatory for a safe operativity of road infrastructures. Nevertheless, the risk induced by extreme traffic loads was seldom investigated in literature and the proposed assessment methods require computationally expensive elaborations, that are hardly compatible with a real time risk management strategy. Indeed, while optimized bridge maintenance plans were suggested to reduce risk, real time traffic management approaches based on site-specific data are missing. Moreover, the integration of emerging prediction tools, such as Artificial Neural Networks (ANNs), was never explored. This study fills the previous gaps by developing a three-block methodology to evaluate and real-time manage the risk related to design load overcoming events induced by extreme traffic load hazard. The proposed risk evaluation technique builds on the framework recommended by ISO 39001 for road safety analysis. Moreover, it integrates a simplified procedure inspired by the Limit State Method (LSM) to predict the frequency and severity of overcoming events of bridge design loads prescribed by Eurocodes. The method adopts: 1) Weight-In-Motion (WIM) systems to real time gathering site-specific traffic load hazard data, 2) a bivariate probabilistic Risk Prediction Model (RPM) based on traditional Generalized Linear Regressions (GLRs) and emerging Artificial Neural Networks (ANNs) to estimate frequency and severity components as a function of several safety factors; 3) an Intelligent Transportation System (ITS)-based architecture to implement real time risk management actions. The methodology was tested on a 2.5M+ vehicles raw data acquired during a five-months period in a WIM pilot station placed near a bridge along the heavy used ring road of the city of Brescia (Italy). As for extreme traffic load hazard, the results indicated that design load overcoming events on the bridge were significantly more frequent than that prescribed by Eurocodes. As for the risk prediction model, both GLRs and ANNs tools showed a good capability in predicting the frequency and the severity of overcoming events, although ANNs outperformed GLRs. Safety factors related to the compliance with mass limits prescribed by Traffic Code resulted those with the more significant effect on risk predictions. Specifically, the greater predicted frequency and severity were correlated to the highest vehicular overload ratio, percentage of overloaded vehicles and fraction of overloaded axles, as well as to the presence of extremely loaded vehicles following one another. As for the risk management strategy, the simulation based on two months of real traffic WIM data showed that approximately 3.7% of total passing flow would be interested by the more severe traffic management actions. The findings suggested the need for a greater caution by Road Authorities (RAs) when permits for extremely overloaded vehicles are issued, as well as for the implementation of enforcement strategies for sanctioning of illegal overloaded vehicles, and ITS-based architectures for the real time management of the risk related to the extreme traffic load hazard on bridges. Finally, this study recommends the future integration of new intermediate safety factors computed considering data acquired by sensors different than WIM (e.g., accelerometers, strain gauges, intelligent traffic cameras, etc.), as well as the extension of the analysis at network level to prioritize traffic management actions among bridges and optimizing the individuation of alternative routes.