Probabilistic blind identification of site effects from ground surface signals

Abstract

A new blind identification technique is devised within a probabilistic framework to identify site effects from ground surface accelerograms. Ground response analysis is an essential part of most seismic assessment and ground motion characterization efforts and is fraught with inherent complexities due to epistemic as well as aleatory uncertainties. Whereas an extensive collection of analytical and numerical techniques is available to analyze ground responses, their accuracy depends on a priori knowledge of site properties and the availability of appropriate input motions. In the probabilistic blind identification method proposed herein, the site response is identified using only the ground surface accelerations recorded at multiple stations, under the assumption that the unknown input motion is similar for all these stations. The few competing site response identification methods rely on the strategic selection of a reference station and the present approach obviates this limitation and additionally yields the uncertainty of the identified results.