Department of Electrical Engineering and Computer Science, Texas A&M University-Kingsville, Kingsville, Texas, United States of America.
Department of Engineering Technology, Purdue University Northwest, Hammond, Indiana, United States of America.
PLoS One. 2022 Nov 15;17(11):e0277490. doi: 10.1371/journal.pone.0277490. eCollection 2022.
Local attacks in networked systems can often propagate and trigger cascading failures. Designing effective healing mechanisms to counter cascading failures is critical to enhance system resiliency. This work proposes a self-healing algorithm for networks undergoing load-based cascading failure. To advance understanding of the dynamics of networks with concurrent cascading failure and self-healing, a general discrete-time simulation framework is developed, and the resiliency is evaluated using two metrics, i.e., the system impact and the recovery time. This work further explores the effects of the multiple model parameters on the resiliency metrics. It is found that two parameters (reactivated node load parameter and node healing certainty level) span a phase plane for network dynamics where three regimes exist. To ensure full network recovery, the two parameters need to be moderate. This work lays the foundation for subsequent studies on optimization of model parameters to maximize resiliency, which will have implications to many real-world scenarios.
网络系统中的局部攻击往往会传播并引发级联故障。设计有效的修复机制来应对级联故障对于提高系统弹性至关重要。这项工作提出了一种针对基于负载的级联故障的网络自修复算法。为了深入了解具有并发级联故障和自修复的网络动态，开发了一个通用的离散时间仿真框架，并使用两个指标（系统影响和恢复时间）来评估弹性。本工作进一步探讨了多个模型参数对弹性指标的影响。结果发现，两个参数（重新激活节点负载参数和节点修复确定性水平）跨越了网络动态的相平面，其中存在三个区域。为了确保网络完全恢复，两个参数需要适中。这项工作为后续的研究奠定了基础，即优化模型参数以最大化弹性，这将对许多现实场景产生影响。
