Abstract
SummaryA computationally efficient framework is presented for modification of stochastic ground motion models to establish compatibility with the seismic hazard for specific seismicity scenarios and a given structure/site. The modification pertains to the probabilistic predictive models that relate the parameters of the ground motion model to seismicity/site characteristics. These predictive models are defined through a mean prediction and an associated variance, and both these properties are modified in the proposed framework. For a given seismicity scenario, defined for example by the moment magnitude and source‐to‐site distance, the conditional hazard is described through the mean and the dispersion of some structure‐specific intensity measure(s). Therefore, for both the predictive models and the seismic hazard, a probabilistic description is considered, extending previous work of the authors that had examined description only through mean value characteristics. The proposed modification is defined as a bi‐objective optimization. The first objective corresponds to comparison for a chosen seismicity scenario between the target hazard and the predictions established through the stochastic ground motion model. The second objective corresponds to comparison of the modified predictive relationships to the pre‐existing ones that were developed considering regional data, and guarantees that the resultant ground motions will have features compatible with observed trends. The relative entropy is adopted to quantify both objectives, and a computational framework relying on kriging surrogate modeling is established for an efficient optimization. Computational discussions focus on the estimation of the various statistics of the stochastic ground motion model output needed for the entropy calculation.
Highlights
The modification of stochastic ground motion models to establish hazard compatibility for specific seismicity scenarios was discussed in this paper
The modification of the ground motion model was defined as an adjustment of the probabilistic predictive models/relationships that relate the parameters of the ground motion model to seismicity characteristics
The proposed modification was defined as a bi‐objective optimization with dual objective of minimizing the discrepancy between the hazard for a given structure/site and the predictions established through the stochastic ground motion model, while maintaining a small deviation from the original predictive relationships, assumed to facilitate similarity to observed regional trends
Summary
The relevance of techniques that model acceleration time‐series of seismic events has increased during the past decades due to the growing popularity of simulation‐based probabilistic seismic risk assessment[1,2,3] and performance‐based earthquake engineering.[4,5,6] Though the most popular methodology for performing this task is the selection of real (ie, recorded from past events) ground motions,[7,8,9,10] potentially scaled based on a target intensity measure (IM), an alternative philosophy is the use of simulated ground motions.[11,12] A specific modeling approach for the latter which has been steadily gaining increasing attention by the structural engineering community,[13,14,15] is the use of stochastic ground motion models.[16,17,18,19,20,21,22] These models are based on a parametric description of the spectral and temporal characteristics of the excitation, with synthetic time‐histories obtained by filtering a stochastic sequence through the resultant frequency and time domain modulating functions. The current study extends approach[28] to (1) match the prescribed conditional hazard (not mean IMs) for a specific site and structure (or range of structures) while (2) preserving desired trends and correlations in the physical characteristics of the resultant ground acceleration time‐series, including consideration of the variability of these characteristics This is again formulated as a bi‐objective optimization problem. As discussed in the introduction, the formulation of the predictive model for θ provides synthetic ground motions whose statistics (mean and dispersion) of output IMs do not necessarily match the intended hazard for specific structures and sites For accommodating such a match, a modification of the existing predictive model for θ is proposed for specific seismicity scenarios defined through z, with objective to get a suite of acceleration time‐series for that scenario whose (1) mean and dispersion match a target IM mean and dispersion vectors, while (2) maintaining similarity to the predictive relationships already established for the model.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
