Supply chain resilience under stochastic disruptions

Abstract

The relevance of this research is determined by the growing vulnerability of supply chain networks to stochastic disruptions caused by local failures, capacity losses, and clustered damage. The objective of the study is to develop and validate a compact graph-flow model and a reproducible computational framework for assessing structural and functional resilience under uncertainty. The proposed approach represents a supply chain as a directed capacitated graph and defines network performance through normalized maximum flow from supply sources to demand nodes. The methodology combines stochastic damage generation, including node failures, edge degradation, cluster disruptions, and irreversible losses, with a baseline discrete recovery process and Monte Carlo simulation. Resilience is evaluated by a time-dependent performance trajectory, an integral resilience index, robustness, final recovery level, median resilience, interquartile range, and a lower-tail conditional risk metric. Computational experiments compare hierarchical, hub, redundant, and dense network archetypes across different disruption types, redundancy levels, network sizes, metric parameters, and damage irreversibility settings. The results show that dense topology provides the highest typical and lower-tail resilience, whereas hub topology may remain effective in median scenarios but demonstrates weaker behavior in adverse cases. Redundancy is found to be especially important for reducing tail risks rather than only improving typical performance. The analysis also shows that cluster disruptions are not automatically more destructive than random node failures when the actual damage dose is normalized, while irreversibility sharply reduces recovery completeness and lower-tail resilience. The scientific novelty of the study lies in integrating flow-based functionality, stochastic disruption modeling, recovery dynamics, and risk-oriented metrics into a single benchmark procedure. The practical value of the framework is its applicability as a basis for comparing supply chain network structures and for developing future prioritized, decentralized, and multi-agent recovery algorithms.