Comparison of Natural and Synthetic Spectrum Compatible Accelerograms Obtained by Ground Motion Selection and Stochastic Simulation

The accurate modelling of seismic load is a major topic that has raised particular interest in the literature in the recent years. One of the reasons is the advent of performance based earthquake engineering (PBEE) that has become the state of the art for both civil and nuclear structures. The PBEE generally requires transient analysis in order to evaluate the seismic fragility of structures and components. In consequence, a set of hazard consistent ground motion time histories is needed for the transient dynamic analysis to be performed. In this paper, a simple procedure for simulating artificial earthquake accelerograms matching the statistical distribution of response spectra, as given by the ground motion prediction equations (median and the standard deviation) and correlation coefficients, is presented. The approach, Zentner (2014), follows the general ideas of the natural ground motion selection algorithms proposed by Baker (2011) and Wang (2011) but using simulated (artificial) “spectrum-compatible” accelerograms. This contribution proposes to perform a number of comparative studies in order to assess the capabilities of the simulated accelerograms. Ground motion intensity measures will be compared to the target from GMPE and to the ones of selected natural accelerograms. Finally, the impact on structural response will be evaluated. In particular, we compare a set of natural accelerograms, selected according to the Baker et al. procedure, to a set of simulated time histories. Introduction The selection of pertinent input ground motions for transient structural analysis is a difficult task that has to be addressed thoroughly in the framework of performance-based earthquake engineering. It is now widely acknowledged that not only the mean or median spectrum, but also the accurate spectral shape is an important issue. Several methodologies for obtaining spectrum-compatible ground motion time histories, to be used in transient dynamic structural analysis, are available in literature. In particular, Abrahamson et al (2010) proposed to modify natural ground motion by time domain wavelets so as to perfectly match a target spectrum. This methodology is available through RSPMatch software. Another widely used methodology, based on the so-called Random Vibration Theory (RVT), is due to Vanmarcke. It is a stochastic model where a spectrum-compatible PSD is identified by using the relation between the standard deviation and the distribution of the maxima of a stationary Gaussian stochastic process. In order to improve the matching, the spectral content of the sample of time histories is subsequently adjusted in the frequency domain. This approach has been distributed via SMQKE software, Gasparini&Vanmarcke 1976. POWERSPEC, a software based on the work of Preumont 1985 used in the nuclear industry, as well as Code_Aster, propose a similar approach for the simulation of spectrum compatible ground motion. Recently, Baker and co-workers (2011) have proposed a new methodology for ground motion selection and possibly scaling that allows to obtain sets of spectrum compatible ground motion exhibiting realistic spectral shape. This approach is of particular interest when the target is a conditional mean spectrum, Baker (2011). Indeed, the notion of conditional mean spectrum (CMS) has been introduced in the framework of performance based safety assessment methods in order to provide a spectral shape compatible with specific event scenarios obtained by disaggregation from the Uniform 1 Expert research engineer, Institute of Mechanical Science and Industrial Applications, IMSIA UMR EDF-CNRS-CEA-ENSTA , Clamart, France, [email protected]