Bayesian approaches to the reliability assessment of existing structures: Fid-Bau Portal

Bayesian approaches to the reliability assessment of existing structures

Marsili, Francesca

Due to environmental, economic and socio-political reasons, significance and field of application of existing structures' reliability assessments extend rapidly, also in view of the preservation of cultural heritage. Over recent decades, the Bayesian analysis has been increasingly recognized as an efficient procedure for the definition of the probability models for the random variables governing the reliability of the structure. It is especially applied in case of material mechanical parameters, where a prior is firstly defined according to engineering judgments and then updated with the result of material tests. However, if the choice of the prior distribution is wrong, the result of the analysis might be incorrect. Material tests are usually destructive, leading to a loss of the integrity of the structure and even of its cultural value, in case of heritage buildings. Therefore it is of paramount importance to propose procedures according to which more sound prior distributions are defined and non-destructive approach to the Bayesian analysis are promoted. In this work, a general methodology that increases the robustness of the Bayesian analysis with respect to the prior distribution for the description of material mechanical random parameters is defined. The methodology focuses on the identification of homogeneous material classes considering certain mechanical properties, and the definition of the related statistical parameters, directly analysing the results of material acceptance tests. A practical application of this procedure is developed for the identification of concrete classes in Italy during the 1960s analyzing the results of compressive tests on standard cubes. On the other hand, it is shown by practical application on relevant case studies that a more realistic reliability evaluation can be obtained without touching the structure, refining the probability models of action or action effects. In the first case study (a concrete water tank from the 1960s) the Bayesian updating of the Finite Element model is carried out considering the results of static and dynamic tests. In the second case study (a masonry aqueduct built in the 16th century) the Bayesian updating of the wind speed is performed according to the results of field studies.