Different analytical models exist to predict the shear strength of reinforced concrete members. Generally, each of these shear strength models consists of a formulation based on certain underlying theory and fitted model coefficients. The model fitting parameters are usually established from the comparison with test data. Hence, the predictive value of a shear strength model depends, to some extent, on the quality and representativeness of the used test data.
The first immersed concrete tunnel in the world is heavily deteriorated due to reinforcement corrosion. Beside its great historical importance, the tunnel is situated in the heart of Rotterdam and is one of the most important city connections. Therefore, upon discovering structural damage, a major task has been to prove that the structural safety is at a reliable and acceptable level (Blom 2013). Consequently a repair plan has been worked out to preserve this tunnel for the future and to minimise disturbance of traffic in the city centre during repair closure. This paper shows the dilemma’s and choices in realising the repair strategy.

First the conceptual consideration is shown on why removal of chloride contaminated concrete in this case is preferred above cathodic protection. Secondly, it is shown how to deal with temporarily decreased structural safety and increased probability of cracking during the repair: i.e. in the stage when chloride contaminated concrete is being removed the tunnel is still fully loaded by the ground and water loading. Finally a design plan (including testing, measuring and optimising) is presented on how a repair concrete mix should be designed to meet substrate qualities and ensure good repair quality. Laboratory tests and in situ tests have already been carried out. The paper ends with conclusions and recommendations for future projects.
This work investigates the predictive capability of several shear strength models for reinforced concrete beams without shear reinforcement. Particular attention is given to the application domain of relatively low reinforced and high depth concrete beams where limited shear test data is available. The predictive capability of the models for this area of interest is analyzed with Bayesian Inference. This probabilistic technique calculates the posterior distributions of uncertain parameters, given a set of measured test data and some prior knowledge.

The predictive capability of each shear strength model is quantified by means of a calculated model uncertainty. Furthermore, the influence of the uncertainty in model parameter values on the calculated model uncertainties is evaluated. Bayesian Inference is also used to estimate the model evidences conditionally on the used data.