Estimation of bridge component condition states with varying data availability

Abstract

For computer systems to estimate the type and timing of future interventions on a bridge, and more specifically, on its components, it is important for the bridge managers to understand their current condition states. That information, however, is almost never perfectly available. In this paper, a methodology is developed that accounts for the scenarios of having no or partial inspection data on the bridge components. A Bayesian network is used to estimate the probabilistic condition states of an asset, requiring the utilization of information that is external to an inspection campaign, including the component properties and environment. With partial information available on the bridge and/or component condition state, the Bayesian network takes advantage of the inference capability to draw conclusions on the condition state of interest. The methodology is used to estimate the condition of a railway bridge pier located in Switzerland.