Experimental and numerical analyses of variability in the responses of imperfect slender free rigid blocks under random dynamic excitations

Due to the well-known sensitivity of the behaviors of free structures under seismic excitations, the question of the aptitude of a numerical model to accurately represent them arise. To contribute to the answer to this question, this article presents experiments which were carried out on the shaking table of CEA/Saclay in France, on three rigid blocks with geometrical defects, inevitably due to the manufacturing process, subjected to 100 realizations of a random process. These tests were analyzed using specifically-developed indicators, and compared with the results yielded by two numerical models, one with a symmetrical geometry and the other with a non-symmetrical geometry, calibrated to reproduce out-of-plane behavior identified through release tests. Counter-intuitively, this article shows that a numerical model can predict motion over a longer period than an experiment performed on a supposedly identical block. From a statistical point of view, despite experimental uncertainties this article shows a good agreement between numerical and experimental results. Finally, a numerical study, performed using artificial seismic signals, showed that the assumption of perfect geometry can lead to an underestimation of the risk of overturning. Moreover, it is showed that a symmetrical model with a realistic slenderness correction can provide an overestimation of this risk under 1D excitation, but not in 2D.