Resilience Metrics and Outliers

Main content

Resilience Metrics and Outliers

The traditional approach to protect systems against external and internal disruptions is to design a system with limit-state resistance—that is able to withstand a high magnitude-low probability disruption—based on statistical methods. However, disruptions that are ‘beyond extreme value distributions’, which we refer to as ‘dragon kings’, cannot be predicted with traditional statistical approaches, putting complex infrastructure systems at risk.

Resilience is an emerging concept that enhances the ‘resistance’ view with ‘recovery’, ‘adaptation’ and ‘learning’ capabilities of the system in order to cope with increasing uncertainty. In fact, promising approaches to anticipate or even predict ‘beyond extreme value distribution’ events do exist.

The ‘Resilience Metrics and Outliers’ research module aims to:

  1. develop a metrics to characterise the resilience of single socio-technical systems and system-of-systems
  2. identify a network representation for critical infrastructure systems with node resilience and between-nodes interdependencies to estimate overall system resilience, and to integrate site- and technology-specific disruptions
  3. investigate the predictive strength of methods to detect ‘dragon kings’ ex-post and ex-ante
  4. evaluate strategies to cope with these ‘dragon kings’

The Resilience Metrics for an All-Hazard Approach submodule explores possible resilience metrics based on real-world data sets such as financial system and electric power systems to investigate how the response-recovery functionality of a set of system components may be quantified and aggregated at the systems level. This allows the study of how combinations of different disruption classes may be assessed at the systems level.

The Extreme Events in Industrial and Cyber Systems submodule investigates the effectiveness of different methods to detect ‘dragon king’ events by systematically studying the conditions and circumstances under which dysfunctions can cascade into ‘dragon king’ events. Researchers will model the interactions of endogenous components, pre-existing vulnerabilities, and the fragility of the human-industry complex in a dynamical framework to understand the dynamics and emergence of extreme events. Finally, we explore the possibility of suppressing ‘dragon kings’ through the introduction of occasional small perturbations.

Page URL:
Fri Jun 23 19:48:33 CEST 2017
© 2017 Eidgenössische Technische Hochschule Zürich