Assessment of bridge safety due to scour by Bayesian network

The presence of inherent randomness as well as parametric uncertainty in related hydraulic models is known to be a weakness in using deterministic empirical equations for calculating bridge pier scour. The concept of system reliability assessment is a key to overcome this problem and the main focus of this concept is to determine a desirable safety margin for bridges against scour. Among different classical methods for reliability assessment, the Bayesian network (BN) has shown promise in systems with event chains such as scouring at bridge sites. Graphical charts and reasoning relations between all events in addition to a Bayesian formulation are the main characteristics used in this paper to conduct a new application of a BN in an assessment of bridge failure due to scour. This analysis permits examination of a very complex system of interactions and processes that are not well understood. The method is illustrated by analysing a combination of local and contraction scouring at a bridge section. Increase in reliability by using riprap around a bridge pier is also included in the analysis. A comprehensive comparison of BNs and fault trees is also presented.