Abnormal cascading on complex networks

Abstract

In the study of cascading failures on complex networks, a key issue is to define capacities of edges and nodes as realistically as possible. This leads to the consideration of intrinsic edge capacity associated with laws governing flows on networks, which goes beyond the existing definitions of capacity based on the initial load as quantified by the betweenness centrality. Limited edge capacity (or bandwidth) and high flux or attack can trigger cascading processes, which we find as characteristically different from those reported in the literature. In particular, there can be an abnormal parameter regime where incrementally augmenting the edge capacity can counterintuitively increase the severeness of the cascading process. Another striking finding is that heterogeneous flow distribution tends to suppress the cascading process, in contrast to the current understanding that heterogeneity can make the network more vulnerable to cascading. We provide numerical computations and analysis to substantiate these findings.