A Comparison of Network-Based Metrics of Behavioral Degradation in Complex Engineered Systems

Abstract

It has been assumed, but not yet tested, that the topological disintegration of networks is relatable to degradations in complex engineered system behavior and that extant network metrics are capable of capturing these degradations. This paper tests three commonly used network metrics used to quantify the topological robustness of networks for their ability to characterize the degree of failure in engineered systems: average shortest path length, network diameter, and a robustness coefficient. A behavioral network of a complex engineered system is subjected to “attack” to simulate potential failures to the system. Average shortest path length and the robustness coefficient showed topological disintegration patterns which differed between nominal and failed cases, regardless of failure implementation location. The network diameter metric is not sufficiently dependent on local cluster topology to show changes in topology with edge removal failure strategies. The results show that topological metrics from the field of complex networks are applicable to complex engineered systems when they account for both local and global topological changes.