Proposed Bayesian Network Framework to Model Multisite Seismic Hazard with Existing Probabilistic Seismic Hazard Analysis Results: Fid-Bau Portal

Proposed Bayesian Network Framework to Model Multisite Seismic Hazard with Existing Probabilistic Seismic Hazard Analysis Results

Gibson, Emily M. / Bensi, Michelle T.

Abstract

Seismic events concurrently affect multiple facilities within an infrastructure system or portfolio distributed across a geographic region. When calculating the multisite seismic hazard to support system-level risk assessments, various sources of hazard correlation need to be considered. A Bayesian network (BN) framework is proposed herein, which transparently characterizes the dependencies between seismic hazard input parameters while leveraging the outcomes of an existing probabilistic seismic hazard assessment (PSHA). Specifically, one of the major benefits of the framework is that it utilizes mean hazard curves and deaggregation from an existing PSHA, thereby reducing the computational resources required to estimate the multisite seismic hazard. Derivations of the mathematical formulations required to manipulate the PSHA data are provided along with the configuration of the proposed framework. An example problem is then presented to validate the BN against Monte Carlo simulations.

In engineering practice, seismic risk is most often assessed for a single facility or infrastructure component. However, when an earthquake occurs, it will impact an entire region, and multiple structures will be impacted at the same time. Some of these structures are part of a larger system, such as a river with multiple dams, and system failure may create larger consequences than if a single structure failed. Available methods for assessing the seismic hazard at multiple sites often require significant computing resources. This research proposes a new framework to assess this hazard that is more transparent than existing methods and requires less effort to develop and compute. This framework would be especially useful in the central and eastern US where it can be difficult to justify the cost of expensive seismic analyses because the seismic risk is not as apparent to most stakeholders. A simpler, cheaper framework would allow people in this region to better understand the seismic risk posed to their infrastructure systems.