New Emulation-Based Approach for Probabilistic Seismic Demand

Abstract eng:
An advanced statistical emulation-based approach to compute the probabilistic seismic demand is introduced here. The emulation approach, which is a version of kriging, uses a mean function as a first approximation of the mean conditional distribution of Engineering Demand Parameter given Intensity Measure (EDP|IM) and then models the approximation errors as a Gaussian Process (GP). The main advantage of the kriging emulator is its flexibility, as it does not impose a fixed mathematical form on the EDP|IM relationship as in other approaches (i.e. standard cloud method). In this study, a casestudy building representing the Special-Code vulnerability class, is analyzed at a high level of fidelity, namely nonlinear dynamic analysis. For the evaluation of the emulator, two different scenarios are considered, each corresponding to a different “assumed reality” represented by an artificially generated IM:EDP relationship derived from the sample of real analysis data. These “assumed realities” are used as a reference point to assess the performance of the emulation-based approach, and to compare against the outcomes of the standard cloud analysis. A number of input configurations are tested, utilizing two sampling processes (i.e. random and stratified sampling) and varying the number of training inputs. The outcomes of the current work show that the proposed statistical emulation-based approach, when calibrated with high fidelity analysis data and combined with the advanced IM INp , outperforms the standard cloud method in terms of coverage probability and average length, while insignificant differences are obtained in the mean squared error estimations (MSE). The improved performance of emulator over the cloud method is maintained in both “assumed realities” tested showing the capability of the former approach to better estimate the EDP|IM relationship for the cases that does necessarily follow a favorable pattern (e.g. power-law).

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket