关联供应链网络级联失效机理及鲁棒性研究

摘要

To analyze the robustness of complex interdependent supply chain networks suffering disruption events, the cascading failure mechanism of both intra-network and inter-network is studied.Without loss of generality, the undirected information layer network and directed physical layer network comprising the interdependent supply chain network are generated via the stochastic rule.Then, the network characteristics, i.e.,node load and node capacity, are expressed by parameters α, β, σ, etc.The redistribution strategies of failure loads corresponding to the condition of existing edge flow constraints are presented.Through giant component functions, the valid nodes still with function could be judged when the redistribution is finished.Furthermore, the time-varied state equations are constructed based on one to one interdependent relationship between the cyber-layer network and physical-layer network, which can depict the dynamic propagation throughout the interdependent supply chain network.Finally, numerical simulations including two cases: single node removal and multi-node removal, are given and the robustness of interdependent supply chain networks with different parameters is compared.In particular, multi-node removals are classified in three ways, i.e., degree ascending, degree descending, and random degree.The simulation results show that the multi-node removal in the ascending degree way leads to the worst robustness performance among the three removal ways, if β=0.5, 1, 1.5, 2.Meanwhile, the first phase transition of the cascading failure for interdependent supply chain networks, after multi-node removal, is more obvious than for isolate layer networks, i.e., a small fraction of nodes removal will result in the crash of the whole interdependent supply chain network.%为了分析复杂关联供应链网络在遭遇干扰事件时的鲁棒性,研究关联网络层内和层间级联失效机理.通过随机规则生成供应链无向信息层网络和有向物理层网络,描述并分析由α、β、σ等参数表征的节点负荷、容量等网络结构特性.针对存在边流量约束的情况,提出相应的失效负荷分流策略;通过构建极大簇函数,判断关联供应链网络中经分流策略后仍具运作功能的有效节点,并依据供应链信息层网络和物理层网络间一对一匹配的关联关系,构建时变失效迭代状态方程,从而有效描述关联网络间的动态失效传递.最后,在不同参数控制下,对初始单个节点故障和多个节点故障两种情况下的关联供应链网络综合鲁棒性进行数值仿真分析.按节点度从大到小、从小到大、随机3种方式去除多个节点,结果表明在β=0.5, 1, 1.5, 2 4种参数下,按度从小到大方式去除多节点,其引发的关联网络级联失效规模较之其它两种方式大;同时发现关联供应链网络在多节点去除情况下,其级联失效较之单层网络具有明显的一级相变特性,即少量的节点移除就会导致整个关联供应链网络崩溃.