Applying Computational Network Analysis Methods for Critical Infrastructure Robustness: Integrating Centrality Metrics and Model Simplification for Efficient Failure Simulations

This research work presents computational network analysis methods to assess the robustness of critical infrastructure under failure scenarios while improving simulation efficiency. We propose modeling infrastructure systems as unweighted and undirected graphs to evaluate robustness through percolation-style node removal failure simulations. Removal of nodes are prioritized by topologically-focused network metrics, degree and betweenness centrality, with and without recalculating the rankings, and compared to random selection. To reduce computational burden, we introduce a model simplification that removes sequential degree-two nodes while preserving overall topology. We demonstrate the methods on the Seattle road network ( 23,303 nodes, 34,243 edges) and a simplified version with 9 % fewer nodes and 6 % fewer edges. Results show that targeted removals degrade network connectivity more effectively than random failures. Dynamic update of the betweenness centrality metric over the percolation process is the most disruptive scenario which highlights the value of adaptive prioritization of assets in the evolving network during failures. The simplified model yields an average 16 % reduction in computation time with minimal impact on collapse behavior, indicating that omitting degree-two nodes can improve practicality without sacrificing accuracy. The study provides transferable methods for identifying critical assets and conducting efficient robustness analysis of large-scale infrastructure networks with more than twenty thousand nodes.

Keywords: Infrastructure Resilience, Network Analysis, Model Simplification, Failure Simulation, Centrality Metrics.