Entropy‐based intensity measure selection for site‐specific probabilistic seismic risk assessment. (2nd September 2020)

• Record Type: Journal Article
• Authors: Du, Ao; Padgett, Jamie E.
• Is Part Of: Earthquake engineering and structural dynamics. Volume 50:Number 2(2021)
• Page Start: 560
• Page End: 579
• Publication Date: 2020-09-02
• Subjects: confident seismic risk estimates -- information theory -- multiresponse structural systems -- multivariate surrogate demand modeling -- uncertainty quantification and propagation; Structural dynamics -- Periodicals; Earthquake engineering -- Periodicals; 624.1762
• DOI: 10.1002/eqe.3346
• Languages: English
• ISSNs: 0098-8847

Abstract

Summary: Intensity measure (IM) selection is a critical step in probabilistic seismic risk assessment (PSRA). In many past studies, the efficiency of an IM, which quantifies its explanatory power within a probabilistic seismic demand model, has been a predominant IM evaluation criterion. However, because PSRA requires convolution of the conditional demand model and the probabilistic seismic hazard, IM selection solely based on efficiency ignores the influence of uncertainty contribution particularly from the ground motion prediction equations (GMPEs), which is also dependent on the IMs, and may lead to biased IM selection. In the present study, the concept of joint entropy from information theory is introduced to quantify the uncertainty of unconditional multivariate seismic demands for general multiresponse structural systems, offering a holistic consideration of different sources of uncertainties. We propose a novel entropy‐based IM evaluation criterion that can serve as the basis for IM selection and develop a practical framework to facilitate the implementation of the entropy‐based IM selection in site‐specific PSRA. Based on two case‐study highway bridges, the merit of the proposed IM selection methodology is demonstrated, and the influence of IM selection on demand entropy and loss estimation in PSRA is evaluated. From the results, the proposed entropy‐based IM selection manifests improved capability in considering different sources of uncertainties and is found to deliver much more consistent and confident seismic risk estimates.