Bayesian FE-Model Updating of a Curved Approaching Span of the Indiano Bridge in Florence

Reliable predictions of the dynamic response of existing bridges and the assessment of their structural health require accurate finite element models. Unfortunately, their development is not straightforward due to the inevitable uncertainties on boundary conditions (e.g., friction of the supports and their behavior with the traffic loads, out of plumb of pylons, soil-structure interaction, etc.) as well as the absence of analytical and numerical methods for damping estimation. Full-scale measurements are then commonly used for updating the initial FE model through deterministic or emerging probabilistic approaches, the latter should be recommended due to the non-negligible uncertainties present in both modeling and measurements. This work deals with the Bayesian FE-model updating of a curved approaching span of the Indiano Bridge (Florence, Italy) through full-scale vibration tests. The case study is represented by a steel/concrete composite deck slab bridge with a span of about 21m, which has been equipped with both a wired and a wireless accelerometer network. A procedure based on the Bayes theorem has been developed and tested on the considered case study incorporating both model uncertainties and measurement errors. Results obtained from wireless sensors have been compared with those of wired sensors to quantify measurement errors, while model uncertainties have been defined according to expert judgment.