Statistical Validation of a Rocking Numerical Model

Abstract

Validation of numerical models is a paramount task for the design of structures and introduction of new construction solutions in practice. Up to now, rocking structures have been considered unpredictable, hampering their use in practice. Using a recently proposed statistical validation method, this paper claims that rocking motion can be predicted. Herein, a simple 3D wobbling rocking model is statistically validated against multiple shake table tests. Both numerical and experimental seismic response are clustered in six bins, sorted according the excitation type and intensity, and their cumulative distribution functions and median response are compared. The paper shows that, even though the numerical model does not account for sliding and twisting of the columns, it still represents well the statistics of the horizontal displacement of the rocking system. Only in 1 out of the 6 cases the numerical model fails to predict well the statistics of the tested rocking system. Thus, as structures are usually designed for set ground of motions and not individual ground motions, the paper claims that the statistical validation is sufficient and appropriate for validation of rocking models.