Selection and scaling of the positively correlated ground motion suite for seismic performance assessment

Abstract

Various ground motion selection frameworks reported in the prior art recommend the construction of target spectra taking into account the underlying correlation or covariance matrix. A number of target spectra in the order of thousands are generally required to capture the true correlation structure. In contrast, a suite of about 10–20 ground motions is generally considered in seismic performance assessment owing to the inherent computational cost and paucity of recorded ground motions in the available database. This, in turn, leads to the failure probability of an engineering demand parameter (EDP) sensitive to the selection of a number of ground motion records in the suite and the resulting failure probability could be potentially under- or over-estimated. One possible way out in such a case, without increasing the number of ground motion records/spectra in a suite, is to construct the target spectra using positive correlation and in other words, using the fractile spectra. This is aimed to achieve an upper limit of the failure probability given the uncertainty of not capturing the true correlation with a small number of records. A framework for the selection and scaling of the positively correlated ground motion suite is proposed. Novelty involved in the proposed framework is demonstrated through an example of a ground motion selection exercise followed by the seismic performance assessment of a structure. The resulting various EDP hazards are compared with a widely adopted framework in the prior art to illustrate the possible impact of selecting a positively correlated ground motion suite.