Stochastic degradation model of concrete bridges using data mining tools

Abstract

Bridges have a significant importance within the transportation system given that their functionality is vital for the economic and social development of countries. Therefore, a high level of safety and serviceability must be achieved to guarantee an operational state of the bridge network. In this regard, it is necessary to track the performance of bridges and obtain indicators to characterize the evolution of structural pathologies over time. In this paper, the time-dependent expected deterioration of bridge networks is investigated by use of Markov chains models. Bridges in a network are likely to share similar environmental conditions but depending on their functional class may be exposed to different loading conditions that diversely affect their structural deterioration over time. Moreover, the deterioration rate is known to increase with time due to aging. Hence, it is useful to identify and divide the bridge network into classes sharing similar deterioration trends in order to obtain a more accurate prediction. To this end, data mining tools such as two-step cluster analysis is applied to a dataset obtained from the National Bridge Inventory (NBI) database, in order to find associations among the bridge characteristics that could contribute to build a more specific degradation model which accurately explains and predicts the future condition of concrete bridges. The results demonstrate a particular deterioration path for each cluster, where it is evidenced that older bridges and those having higher Average Daily Traffic (ADT) deteriorate faster. Therefore, the degradation models developed following the proposed methodology provide a more accurate prediction when compared to a single degradation model without clustering analysis. This more reliable models facilitate the decision process of bridge management systems.