Efficient Approximation of the Survival Signature for Large Networks

authored by: Jasper Behrensdorf, Sebastian Brandt, Matteo Broggi, Michael Beer
Abstract: The reliability analysis of complex networks, e.g. water supply networks, transportation networks or electrical distribution networks, is of key importance to the resilience of communities. The concept of survival signature provides a novel basis for analyzing complex networks efficiently. The survival signature outperforms traditional analyses techniques, in particular, when estimating the reliability of networks. Its most unique feature is the separation of the network structure from its probabilistic properties, opening pathways for the consideration of, for instance, general dependencies, common cause failures, or vaguely specified probabilities. However, the numerical effort to calculate the survival signature is still prohibitive for large systems. While the issue of numerical efficiency can be addressed well with analytical approaches such as the use of binary decision diagrams, these approaches are limited by the number of components and types. In this paper we propose an approximation of the survival signature using a combination of graph theory and Monte Carlo simulation. By application of graph theory, we are able to predetermine certain fractions of the survival signature without explicitly evaluating it. The remaining fraction is then analyzed with Monte Carlo simulation in a targeted manner, circumventing high-effort-low-contribution calculations. The developed approach excels, in particular, in cases with a large number of different component types. Using an example we highlight the significant reduction in computational effort required to accurately determine the survival signature.
Organisation(s): Institute for Risk and Reliability
External Organisation(s): University of Liverpool
International Joint Research Center for Engineering Reliability and Stochastic Mechanics
Type: Conference contribution
Pages: 661-666
No. of pages: 6
Publication date: 2018
Publication status: Published
Peer reviewed: Yes
Electronic version(s): https://doi.org/10.3850/978-981-11-2726-7_crr14 (Access: Closed)