Bridges worldwide are failing at increasing rates as they heavily suffer from deterioration, leading to transport infrastructure system disruptions and millions of financial losses. For instance, the failure of the Polcevera bridge alone, costed €359 million in the immediate wake, while the estimated annual losses for the Italian economy are close to one billion euros. Despite the recent technological advancements in the field of structural health monitoring, involving digital measurements and remote data collection, bridge failures still occur. The reasons behind those failures are mainly due to the pertinent damages being not easily detectable and the scarcity of the available design drawings, especially for old bridges. Furthermore, the construction sequence for the asset of interest or the level of prestressing are often not known in detail, hence the as-built condition and the actual bridge properties are highly uncertain. This is a capability gap that can only be filled with meaningful data collection and application of advanced technologies in aid of decision-making. The paper aims to extend an existing methodology that was proposed before by Kazantzi et al. (2024a, 2024b) for identifying the damage state of bridges. The aforementioned methodology was initially showcased for balanced cantilever concrete bridges only. To further advance the methodology, application to other concrete bridge typologies, including beam and slab bridges, cast-in situ slabs, continuous box-girders, and arch bridges, is herein proposed. Additionally, the paper suggests validating the machine learning kNN (k-Nearest Neighbor) methodology utilizing Finite Element (FE) model updating. This process involves adjusting numerical model parameters to better match actual structural behavior, enhancing the methodology’s reliability and effectiveness for identifying damage states in concrete bridges of different types and scenarios.
Read full abstract