Robustness of complex supply chain networks to targeted attacks

Abstract

In this thesis, we study the robustness of complex supply chain systems from a network science perspective. Through the simulation of targeted attacks to nodes and edges using different hierarchical measures from network science to select the most relevant components, we evaluate the extent to which local centrality measures can estimate the relevance of a node in maintaining the connectivity and the efficient communication across the network. We perform the experiments on two real-world supply chain data sets, and on an ensemble of networks generated from network growth models that share simple topological properties with the real-world networks. It is found that all models produce more robust networks than the data sets of choice. In addition, the removal of high average neighbor degree nodes seems to have little impact on the connectivity of the network, and a highly varying impact on the efficiency of the network. Finally, robustness against targeted node and edge removal is found to be more associated to the number of nodes and links in the network than to more complex network measures such as the degree distribution.