Layers Heights Randomness Effect on Seismic Response of a Site in Algiers (Algeria)

This paper deals with the stochastic seismic analysis of a site in the city of Algiers (Algeria), which is composed by superposed layers over an elastic halfspace. Layers heights are assumed to be random variables with a lognormal distribution, which is suitable for strictly non-negative random variables. The analysis is carried out via Monte Carlo simulations coupled to the stiffness matrix method. A parametric study is conducted do derive the stochastic behavior of the extreme ground acceleration, its response spectrum and the transfer function. The soil height coefficient of variation is varied from 0 (corresponding to the homogeneous case) to 0.5. The input seismic acceleration corresponds to Boumerdes earthquake (Algeria, May 21st, Mw=6.5). Layers height variability causes an increase of the NS component of the peak ground acceleration and a decrease of the EW component. The soil profile height heterogeneity affects the form of the PGA but slightly its amplitude. This variability induces a slight increase of the soil profile fundamental frequency and an extension of the frequency content, indicating that the resonance phenomenon concerns a larger number of structures. The soils height randomness acts as the variation of the incident angle: decreasing the amplitude and shifting of the resonant frequencies. The maximum values of the confidence interval of the transfer function, for both mean and standard deviation, correspond to eigenfrequencies, which implies that the uncertainty is amplified by the resonance phenomenon.