Efficient Seismic Risk Analysis: Disaggregation-based Weighting of Non-linear Dynamic Analysis Results

Ground motion (GM) record selection represents one of the main issues in assessing the seismic response of a structure through numerical dynamic analysis. In the framework of the probabilistic performance-based assessment it is useful to introduce an intermediate parameter (scalar or vector), known as the ground motion intensity measure IM, to relate the GM record characteristics to the structural performance. This approach often assumes that structural response depends only upon the chosen IM, which is believed to carry the largest information with respect to other properties of the GM. This condition is termed “sufficiency” and must be carefully verified since, if it is not met, then the probability distribution for the damage measure will not only depend upon IM, but also upon the (other) properties of the records selected for analysis. In principle, careful record selection is not essential if the IM is demonstrated to be sufficient with respect to GM characteristics. However, establishing the sufficiency of an IM is not a trivial task and may be structure and/or response measure dependent. In cases where the IM is not sufficient, the GM record selection may be guided by the disaggregation of the seismic hazard for the site of interest. The results of seismic hazard disaggregation can be used to assign relative weights to a given GM record based on the likelihood that a GM having its corresponding magnitude, distance and deviation from the GM prediction model (epsilon) actually occurs. The weighted GMs can be used in order to make probability-based seismic assessments using non-linear dynamic analysis procedure both for a wide and a limited range of GM intensities. In this paper, the implications of using the weighted GM records are investigated in terms of seismic risk, which is represented herein by the mean annual frequency of exceeding the critical structural component demand to capacity ratio, in an existing reinforced concrete structure using both the peak ground acceleration and the first-mode spectral acceleration as IMs. It is demonstrated that the annual frequencies based on weighted records are comparable to those obtained using vector-valued intensity measures, yet requiring less computational effort