Model updating of a bridge-foundation-soil system based on ambient vibration data

In Structural Health Monitoring (SHM) applications, observed discrepancies between identified and numerically calculated dynamic characteristics of structures have to be carefully interpreted and evaluated. Structural damage is not always responsible for those discrepancies, which can also be attributed to the initial assumptions adopted for the numerical models. In the present work a model updating framework is applied for a structurefoundation- soil experimental test-case, where discrepancies were observed between identified and numerically computed dynamic characteristics of an undamaged bridge-foundationsoil system. The scope is to interpret the observed discrepancies, to validate alternative numerical approaches of simulating soil-structure-interaction and to investigate how the refinement of the numerical models influences the model updating results. To this end, a steel strucure comprising an equivelnt scaled system of an under-construction isostatic deck-piercaisson part of Metsovo bridge, in Greece, was studied in the laboratory. The Stochatsic Subspace Identification (SSI) method was implemented in order to identify the dynamic characteristics (natural frequencies, modeshapes, damping ratios) of the structure. Numerical models with different level of modeling complexity were developed and were then updated until correlation was achieved between identified and numerically calculated natural frequencies. The results indicated that the soil's initially adopted shear modulus G was overestimated thereby requiring appropriate calibration. After model updating takes place, all considered numerical models provide natural frequency estimates that are in good agreement with those identified based on ambient vibrations, indicating that less refined numerical models can be implemented in SHM applications in an efficient manner, provided soil parameters are accurately accounted for.