Bridge Management using Bayesian Condition Indicators in Bayesian Probabilistic Network

Publication SlJeFa02

Authors

Mette Sloth, J.S. Jensen and Michael H. Faber

Abstract

The present paper proposes to consider condition indicators within the framework of Bayesian Probabilistic Networks (BPN's) as a basis for bridge management decisionmaking. Condition indicators for individual bridge components and bridges are formulated in terms of time dependent conditional probabilities, i.e. the conditional probability of the bridge being in a state requiring maintenance given an observed condition of the bridge. The modeling of the quality of observation results combines subjective and quantitative information within the framework of Bayesian statistics. The BPN's provide a powerful basis for the representation of the consequence inducing event scenarios in terms of causal relations between uncertain basic variables thus facilitating an integral tool for the assessment of risk taking into account common cause effects due to e.g. common choices of detailing, concrete mix, exposure conditions, reinforcement type, etc. The proposed approach is illustrated on the level of management of individual bridges as well as for a stock of bridges.

Published in/by

Proceedings IABMAS'02, 1st International Conference on 'Bridge Maintenance, Safety and Management', eds. J.R. Casas, D.M. Frangopol and A.S. Nowak, Barcelona, Spain, July 14-17, 2002, pp. 75-76.