Improved finite element model of a tall building in Santiago of Chile using seismic records

Abstract

Historically the structural models used for structural design have been developed using accepted parameters (damping, the level of cracking of reinforced concrete elements, the magnitude and distribution of masses and actions, among others). Some model parameters have been confirmed in laboratory tests and in the few instrumented buildings that have been under the effect of severe earthquakes. For a long time, estimate the parameters that we use in the models (mass, stiffness, and dissipation) was hard work. The previous engineers have backgrounds about the response (accelerations amplitude, displacements, among others), the modal parameters (modal periods, mode shapes, and modal damping). The model parameters that allowed the above conditions were selected through the trial and error process. From this procedure, hundreds of possible models were obtained that arrive at the observed response. The current thesis work continues the research conducted by Ebrahimian et al., 2018 on the parameters' estimation for non-linear models. The algorithm proposed is robust and determines with precision the basic parameters to derive an appropriate model. In addition, notice the uncertainties associated with the parameters. The possibility to improve our model techniques using the model parameters defined in structural design standards stays open. The methodology is validated using the Chilean Chamber of Construction Building (CChC). Since 1997, a monitoring network with 12 uniaxial accelerometers located at four different levels of the structure (including its base) has strong-motion records and ambient records of the building. Since its installation, the sensor system has recorded the building's dynamic response to slight, medium-intensity, and strong-motion earthquakes. The thesis work presented aims to improve the predictions of the finite element model (EF) of the Chilean Chamber of Construction Building (CChC) through updating its parameters (modulus of elasticity, mass distribution, etc.). For the above, a novel technique is used that consists of update the finite element model. The idea behind the algorithm is to model the uncertainty of the model parameters with a probability distribution function (PDF). The PDF is updated in an iterative process to reduce the discrepancy between the actual response of the structure and that estimated by the numerical model. The results show that the algorithm successfully manages to decrease the misfit between the measurements and the response of the numerical model. This methodology is an innovative and practical tool in the profession.