Designing Robust Integrated Forward/Reverse Logistics Networks Against Malicious Attacks

The logistics of modern manufacturing is becoming increasingly complex and networked,forming logistics networks.With recent industrial upgrades,the concepts of green manufacturing,remanufacturing,recycle economics and so on have been put forward one after another,and the traditional forward logistics networks are increasingly developing towards the integrated forward/reverse(or closed-loop)logistics networks.However,with the background that political confrontations and local armed conflicts are intensifying,malicious attacks,such as terrorist attacks,economic sanctions and technological blockades,are frequent,and the globalization of logistics is increasingly challenged and transferred into regionalization,the ultimate goal of this study is to design and build robust integrated forward/reverse logistics networks against malicious attacks on the basis of the existing forward logistics networks.The constraint is not to add new facilities or change the connection relationship between the original forward network nodes,but to transform a certain number of the forward logistics chains into integrated forward/reverse logistics chains.To this end,based on multi-partite networks,a network model,which is in line with the structural characteristics of the actual logistics systems,and the corresponding robustness indices are established.Then,distributed evolutionary algorithms are utilized to optimize the robustness of integrated forward/reverse networks against malicious attacks.The concrete research contents and conclusions of this dissertation are as follows.(1)Establishing a network model which can directly reflect the hierarchy of logistics networks and node heterogeneityIn view of the problem that the traditional network model is only represented by a single matrix,which is difficult to directly reflect the hierarchy of logistics networks and node heterogeneity,a group of adjacency matrices is utilized to represent the logistics networks based on the multi-partite network.In this way,the proposed logistics network model can also directly reflect the upstream and downstream relationship between nodes and express the complete logistics chain through matrix operations.Consequently,the structural characteristics of each echelon are analyzed based on the adjacency matrices of the logistics networks.The results show that community structure and nestedness structure are the common structural characteristics in logistics networks.Besides,for different echelons of the same logistics network,there may exist great differences in structural characteristics.Therefore,it is necessary to analyze the network structure at the echelon level compared with the traditional structural analysis of the whole network.(2)Establishing the robustness indices that can reflect the functional cascade failure between upstream and downstream nodesAs the traditional indices are proposed to measure the robustness of logistics networks,the cascading failure caused by the functional dependence between upstream and downstream nodes is not considered.In this dissertation,two kinds of robustness indices are established based on the proposed logistics network model,which can reflect the functional cascading failure on logistics.Then,under the different defined types of attacks,the robustness analysis of the logistics networks shows that the logistic networks are very vulnerable to malicious attacks when the importance of all node in the network are uniformly measured,but they are highly resistant to random attacks.Furthermore,the robustness analysis is carried out when the important nodes are protected.It is found that for different types of attacks,the relationship between the increased proportion of protected nodes and the effect of robustness improvement is very different,which provides a frame of reference for balancing the cost of node protection and the degree of robustness improvement.Besides,the above analysis is accompanied by the comparison of indices,and the analysis results show that the robustness indices proposed in this dissertation are more stringent compared with the traditional indices.(3)Distributed evolutionary algorithms are proposed for optimization design to obtain robust integrated forward/reverse logistics networks against malicious attacksIn order to solve the problem of how to construct the most robust integrated logistics networks against malicious attacks based on the original forward logistics networks.At first,the construction method is defined to transform the whole forward logistics chain into an integrated forward/reverse logistics chain.To take the complete logistics chains as the basic unit when a logistics network is being built,that is,to ensure that all the selected routes can be utilized to form complete reverse logistics chains,based on the multi-layer logistics network model in this dissertation,two different reverse route selection methods are proposed.For each of them,the corresponding optimization model and algorithm for the robustness of integrated network are proposed.The two optimization algorithms are called distributed evolutionary algorithm based on the island model and distributed nested evolutionary algorithm based on the master-slave model respectively.In terms of the innovation of the algorithms,for the former one,the two parent individuals of the crossover operation have different breakpoints to ensure that the number of ones in the offspring individuals remains unchanged.For the latter one,the individual of the algorithm is divided into the main code and subcode,and the operators of encoding,decoding,crossover and mutation should be constructed respectively.The population of sub code is initialized based on the main code individual.Furthermore,the best fitness value of the subcode population after evolution is taken as the fitness value of the corresponding main-code individual.Finally,the simulation results show that the proposed algorithms can significantly improve the robustness of the constructed integrated logistics networks against malicious attacks.Meanwhile,the robustness against random attacks is not destroyed among the optimization.