Resilience analysis and recovery strategy for interdependent automated container port networks under cascading failures

The automated container port logistics system enhances operational efficiency by integrating information networks with physical equipment and control systems. However, this integration also introduces vulnerabilities that undermine system reliability and resilience. To address the interdependence between the information and physical layers, this study proposes a resilience evaluation model for interdependent port logistics networks. The model incorporates an enhanced load-capacity mechanism and a node buffering strategy, defining five node states and corresponding transition rules to simulate cascading failures from deliberate failures (e.g., cyberattacks) and random failures. A sensitivity analysis evaluates how load tolerance (α), redundancy capacity (β), adjustable parameter (γ), and overload threshold (OT) influence system resilience. Additionally, a sequential node recovery strategy is applied using Graph Convolutional Networks (GCN) and the Asynchronous Advantage Actor-Critic (A3C) algorithm. This strategy is compared with baseline methods based on node degree, betweenness, capacity-link preference, PageRank, and random selection. Simulation results reveal that targeted failures are more destructive, with network collapse occurring when random failures exceed 38%. Parameter variations significantly affect resilience, with functional resilience closely tied to load tolerance and redundancy. Strengthening critical-node redundancy enhances resilience. The proposed GCN-A3C strategy outperforms existing methods in recovery speed and efficiency. This research offers a theoretical foundation for resilience modeling and recovery decision-making in automated port logistics networks.