An enhanced sequential ground motion selection for risk assessment using a Bayesian updating approach

In the context of probabilistic prediction of seismic demand parameters, this research presents a novel method for sequential ground motion selection based on the Bayesian updating approach. Unlike traditional methods of ground motion selection, which rely on spectral shape or intensity measures, the ground motions in this approach are chosen directly through a probabilistic evaluation of the structural responses. In this method, the probability density function produced by a limited number of nonlinear analyses closely resembles the curve produced by a large number of nonlinear analyses. To accomplish this, a large number of linear analysis results obtained in a short period of time are utilized to form the initial belief in a Bayesian model. Subsequently, in a two-way interaction, the results of fresh nonlinear analysis can be introduced sequentially to improve the predictive power of the model while concurrently selecting new ground motions. For this purpose, three case study buildings, including an 8-, 12-, and 20-story structure, were studied at three hazard levels. The efficiency of the proposed approach was investigated for sequential ground motion selection, and its effect on the probability density function of the peak floor acceleration and maximum story drift were studied. The proposed method is highly independent of the hazard level or the height of the building, and it may be used in almost every situation with significant improvement. This method of selecting ground motions is fairly efficient in terms of computing. The results show a significant improvement, especially at higher demand values that are more important in risk assessment. Significant improvements in peak floor acceleration results are also observed, which are notoriously difficult to evaluate.