ABSTRACT
To gain insight in the real deterioration state of reinforced concrete bridges, measurements can be performed. By application of Bayesian updating techniques, the resulting data can be used to update a model of the bridge, which should represent the real bridge as good as possible. However, constructing a bridge model for design always involves modelling assumptions and approximations. When using the approximate model in the Bayesian updating procedure, this leads to posterior distributions different from those obtained when using an exact model, possibly resulting in misleading conclusions. In this work, the influence of such a model error on the posterior distribution of the corrosion degree of an RC girder bridge with a skew angle of 7° is investigated. In design of such bridges, the skewness is often neglected. Measurement results are simulated based on the skewed model, while a straight model of the bridge is used in the Bayesian updating procedure. It is illustrated how the corrosion degree can be largely overestimated by using this straight model and how localization of damage becomes very hard. It is also illustrated how the (spatially variable) corrosion degree can be updated more accurately if the model corresponds to the actual structure.
