Parsimonious modeling of hysteretic structural response in earthquake engineering: Calibration/validation and implementation in probabilistic risk assessment

Abstract

Under strong seismic excitations structural systems exhibit inelastic, hysteretic behavior. Their response can be accurately predicted through high-fidelity numerical models, an approach that typically involves a significant computational cost. An alternative simulation framework is discussed here, based on parsimonious modeling of the hysteretic behavior, established by describing globally the restoring force–drift relationship for each floor. Three different candidate models are considered in this study for the latter, ultimately corresponding to different types of hysteretic behavior, and their implementation using the versatile SIMULINK modeling environment in MATLAB is presented. The parameters of these models may be chosen to establish specific modal properties and yield displacement, ultimate strength and deterioration characteristics for each floor. The fundamental research question addressed is how one can select the model (along with its model-parameters) that will best describe the behavior of a given structural system, focusing on probabilistic risk assessment applications. For this purpose a comprehensive calibration stage for the parsimonious modeling is introduced using information obtained by the response of a high-fidelity structural model under simple excitation conditions (cyclic pushover analysis and sinusoidal loading). This stage determines the characteristics of each parsimonious model but also allows selection of the one providing overall the best fit. A case study for a three-storey concrete structure is presented, for which the accuracy and the computational savings of the parsimonious modeling is examined in two steps. The first step focuses on a detailed comparison over an ensemble of 25 ground motion records. The second step focuses on comparison of statistical measures (failure probabilities, mean response and fragility) within a probabilistic earthquake engineering setting, using a stochastic ground motion model to describe the seismic hazard. The results show that the calibration stage provides good guidance for selection of the most appropriate (in terms of accuracy) parsimonious model, whereas great computational savings are established through the proposed modeling approach, at the expense of a small only reduction in accuracy.