Bayesian network methodology for post-earthquake infrastructure risk management Armen Der Kiureghian, University of California, Berkeley, CA, USA Michelle Bensi, University of California, Berkeley, CA, USA Daniel Straub, University of California, Berkeley, CA, USA

ABSTRACT: In the immediate aftermath of an earthquake affecting infrastructure systems, decisions must be made regarding the deployment of emergency personnel and equipment, evacuation of people, inspection, closure or opening of facilities, and other actions to assure safety of people. Furthermore, soon after the earthquake event, selections must be made among alternative actions to restore functionality to vital infrastructure services. The key ingredient for such decision-making is information: information about the nature and characteristics of the hazard, about the states of the system and its components, and about the consequences of various decision alternatives. In the aftermath of an earthquake, the available information is usually incomplete, highly uncertain, and rapidly evolving in time. We present a Bayesian network methodology for information updating regarding an infrastructure subject to earthquake hazard. Given observed information about the earthquake and the infrastructure components, such as sensor measurements of the ground motion at selected points or observations of damage/no damage of the infrastructure components, we use Bayesian updating to assess the probabilistic state of the infrastructure both at the local (component) and global (system) levels. This analysis properly accounts for the spatial correlation structure of earthquake ground motions, which has been neglected in previous studies.