Estimation and prediction of bridge component condition states based on condition data availability to support the digitalized estimation of future interventions

Abstract

To estimate the future interventions required on a bridge portfolio, bridge managers need to have an overview of their current condition states of all bridge components and an understanding of how these condition states will evolve in the future. As the exact current condition states of all components on all assets do not exist, as inspections take time and are not done simultaneously on all components of all bridges, this information has to be estimated as best possible. This report introduces a methodology using Bayesian networks to estimate the current condition states in two situations that compliment the situation where the current condition state is indeed known. The first is when no data on the condition state of the component exists. This is done considering influencing factors such as environmental condition and traffic loading that would contribute to the deterioration of assets. The second is when historic data on the condition state of the component exists. This is done considering the same influencing factors as when no data exists and this additional data. The proposed methodology is illustrated by using it to estimate the current and future condition states of the components of 26 bridges of a 25 km railway network in Switzerland. As intervention planning processes are becoming increasing digital, the results are visualized in BIM. It is argued that this methodology provides a more consistent and comprehensive overview of the current condition and future condition states than currently exist in practice and that this overview is an integral part of digitalized intervention planning processes. The results of this work help facilitate the use of algorithms that can generate a complete and consistent overview of the type and time of interventions required in the future, an initial overview of the likely lengths and types of possession time required, and the likely costs of intervention.