Bayesian model updating and nonlinear response estimation accounting for modeling uncertainty

During their lifetime, structures and civil infrastructures are subjected to overloads such as earthquakes, which can jeopardize the functionality and particularly the safety of these structural systems. Due to these adverse loads, structures frequently experience the development of nonlinearities that, in turn, lead to structural damage and permanent degradation of the capacity. Hence, it becomes of vital importance to monitor the health of the structures and to be able to assess the locations and extent of the damage after these events. Nowadays, vibrations sensors are installed in a few locations of a structure. The recorded data (e.g., acceleration measurements) can then be coupled with smart algorithms such as Kalman Filters, to estimate the complete response of the structure during the extreme loads and assess, in detail, the damage. For this purpose, a nonlinear Finite Element (FE) model of the structure is required, but uncertainties related to the selection of the model form and the values for critical parameters may undermine the accurate condition assessment. Therefore, Bayesian techniques can be a remedy to quantify and reduce modeling uncertainties by leveraging recorded responses. Along these lines, this study focuses on nonlinear response estimation combined with a Bayesian scheme that updates the probability distributions of model parameters and accounts for uncertainty in the model formulation itself.