Bayesian updating of existing concrete gravity dams model parameters using static measurements

Dams are fundamental infrastructures for energy production, flood control and agricultural-industrial sustenance. Most of them were built before the introduction of seismic regulations with no concerns about their dynamic behaviour. Nevertheless, in recent years, the international scientific community has been paying close attention to the seismic risk of existing dams. Concrete gravity dams have never failed during earthquakes, so no case studies are available and their seismic behaviour can only be explored and investigated using numerical approaches. For this reason, _nite element models must be calibrated through reliable procedures to obtain a sensible result. In this scenario, measurements acquired by a monitoring system and data from in-situ tests have taken on a major role as important sources of information. Methods usually employed for this purpose require a low computational burden, however they are characterised by a high level of uncertainty. Probabilistic methods may be suitable to solve this inverse problem, but they always require considerable computing power, due to the high number of analyses needed, especially when stochastic finite elements are involved. In this paper, a procedure for the model parameters calibration in a Bayesian context is proposed. The novelty of this study is the use of a proxy model replicating the mechanical behaviour of a dam, in order to reduce the computational burden. This approach also allows us to estimate the global model error. Two models, a single monolith and a complete 3D model of a large Italian dam, have been considered. After comparing the errors of different approaches, the best model simulating the observed behaviour of the dam was selected. The efficiency of the proposed methodology is also evaluated.