High dimensional model representation for the probabilistic assessment of seismic pounding

Abstract

The study presented herein aims to analyse the seismic performance of a two-dimensional eight-storey non-ductile reinforced concrete frame against structural pounding with an adjacent three-storey stiff frame having different storey heights. The examined case of pounding refers to the extremely critical floor-to-column structural pounding for three different initial separation gaps between the said structures. Seismic vulnerability analysis is usually performed by way of developing fragility curves for a set of damage and intensity measures using a suitable fragility curve generation technique. For this study, damage measures are characterized by the percentage maximum inter-storey drifts of the taller, flexible frame while the peak ground accelerations of the ground motion data are used as the corresponding intensity measures. Displacement-based fragility curves were generated for 9 sampling points using the High Dimensional Model Representation (HDMR) technique and the results were compared with actual probabilistic data obtained using Monte-Carlo Simulations (MCS). The results of this study imply that the proposed use of HDMR provides excellent fragility curves for the estimation of pounding risks with a significant reduction in the number of simulations required, thereby reducing the computational cost by huge margins. Results also indicate that fragility curves for target separation distances can also be obtained using HDMR without performing additional simulations. This can further be used for the mitigation of pounding risks and for the reliability-based design of buildings for target separation distances and damage measures.