The effect of the angle of seismic incidence when defining a statistical model for structural demand

Abstract

Probabilistic analyses often require making assumptions concerning the statistical distribution of the variables involved. In this context, the statistical distribution of two demand parameters is examined herein when threedimensional analysis of RC buildings is performed considering multiple angles of seismic incidence. Six buildings are subjected to multi-stripe analysis with bi-directional ground motion groups of various sizes applied along one to twelve angles of incidence and datasets of demand parameters are collected. Seven statistical models are then fitted to the datasets and their performance is examined based on quantile-quantile plots and the corresponding coefficient of determination. The results show that using more than one angle of seismic incidence doesn't affect the type of demand distribution, unless highly inelastic structural response is achieved. The generalized extreme distribution exhibits a marginally higher fitting performance when compared to that of the lognormal distribution. However, due to the larger variability of the coefficient of determination of the generalized extreme value distribution when small ground motion groups are used, the lognormal distribution is preferred to fit the demand data even when multiple angles of seismic incidence are used.