ABSTRACT
Vibration-based methods for damage localizations often rely on a damage feature defined in terms of changes of modal or operational shapes. These methods allow detecting variations of the damage feature that can be ascribed to damage. Many of these methods are based on the detection of irregularities in the deformed shape of the structure. The accuracy of the methods proposed in literature is often verified on numerical models that, by their nature, cannot reproduce all the sources of uncertainties - environmental, operational, and experimental - that affect responses recorded of the structure. The availability of data recorded on real structures provides precious material for the verification of the real capabilities of the proposed damage identification algorithms. Due to several sources of uncertainty the values of the damage feature may undergo changes even in absence of any damage. This variability has to be properly accounted for in order to reduce the number of false or missing indications of damage. In this paper is investigated the influence of the uncertainties on the results provided by the Interpolation Method for damage localization. Specifically several different probabilistic models of the marginal probability of the damage feature are compared in order to verify select the one able to provide the highest accuracy in damage localization. The investigation is carried out with reference to the real case study of a prestressed concrete road bridge, the S101 Bridge in Austria. The bridge, built in the early 1960, is a typical example of a European highway bridge. Responses to ambient vibration have been recorded both in the undamaged and in several different damage scenarios artificially inflicted to the bridge. Damage was introduced by lowering one of the bridge piers and by cutting prestressing tendons of one beam of the bridge deck.
