Numerical investigation of the effects of symmetric and eccentric earthquake-induced pounding on accelerations and response spectra for two-storey structures
Résumé
This paper examines the effect of symmetric and eccentric earthquake-induced pounding between adjacent two-storey structures on acceleration times histories and response spectra. The numerical investigation benefits from experimental data issued from an extensive shake table campaign involving two-storey steel-concrete composite structures and slab-to-slab contacts. It is shown that the high frequency branch of the floor response spectra is only related to the maximum impulse acceleration spike during time, characterized by its impulse value, contact duration and time-history shape. In order to numerically reproduce such characteristics, three-dimensional detailed models are set up by modeling the slab with shell elements or hexahedral elements. Previously developed explicit time integrator, the CD-Lagrange scheme, based on non-smooth contact dynamics approach, is employed for ensuring the contact conditions expressed in terms of velocity and impulse, without requiring any contact parameters such as contact stiffness, local damping and restitution coefficient. The eccentric pounding between slabs is simulated by introducing a small defect in the parallelism of the colliding slabs, as measured in experience. The results of eccentric and symmetric pounding, assuming elastic behavior for the steel and concrete materials, are compared with the experimental data in terms of acceleration time-histories and pseudo-acceleration response spectra. It is concluded that symmetric pounding predicts an acceleration spike with a more abrupt increase to the peak and a shorter contact duration in comparison to experimental data, leading to an overestimation of the high frequency branch of the response spectra. On the contrary, eccentric pounding achieves to reproduce the experimental acceleration time histories as well as response spectra up to 400 Hz. Finally, the robustness of the proposed numerical approach is underlined by changing the time step size of the CD-Lagrange scheme as well as by prospecting different relevant choices concerning the Rayleigh damping.