Nonlinear combination of intensity measures for response prediction of RC buildings

Estimating the seismic demand as function of the earthquake intensity is a critical issue in performance-based assessment of structures and infrastructures. This requires proper functional relationships between intensity measures (i.e., parameters that characterizes the severity of the seismic ground motion) and engineering demand parameters (i.e., parameters that quantify the expected damage and losses in structural or non-structural systems). Existing models have been obtained by performing conventional linear regressions over data carried out from numerical simulations. Conversely, this work proposes a set of new nonlinear models obtained by combining multiple intensity measures, in order to enhance the prediction of the seismic response of fixed base or base-isolated reinforced concrete buildings subjected to ordinary and pulse-like earthquakes. Numerical data were first generated by means of non-linear dynamic analyses performed using OpenSees. The data-driven calibration of the final nonlinear regression models has been performed using the Evolutionary Polynomial Regression technique.