Abstract: When tackling supply chain optimization for large manufacturing companies,the conflict among economics,consumer loyalty,and operational efficiency is a significant problem.Focusing on a five-level supply chain network with suppliers level,manufacturers level,transit warehouses level,retailers level and consumers level,a collaborative supply chain operation model based on an information platform was built by taking the total cost,consumer satisfaction and cash conversion cycle of the supply chain network as the core multiple optimization targets.The Non-dominated Sorting Genetic Algorithms II with Dynamic Crowding distance (NSGA-II-DC) was designed through modifying the NSGA-II and incorporating the hybrid coding and the dynamic crowding distance environment selection strategy.The theoretical function test results showed that NSGA-II-DC significantly outperformed the five classical multi-objective evolutionary algorithms in terms of convergence and diversity.Moreover,by verifying a five-level supply chain network with multiple levels and products,the results showed that the proposed model was effectively optimized by NSGA-II-DC.Finally,an optimal solution of the supply chain network model was obtained by Analytic Hierarchy Process (AHP),which may provide a strong theoretical decision basis for decision-makers.

Key words: supply chain network design, multi-objective optimization, dynamic crowding distance, non-dominated sorting genetic algorithm Ⅱ

摘要： 针对大型制造企业在供应链优化过程中存在的经济性、消费者忠诚性和运营效率冲突问题,以供应商、制造商、中转仓库、零售商和消费者在内的五级供应链网络为研究对象,选取供应链网络的总成本、消费者满意度和现金转换周期等多个目标建立优化模型,并提出一种基于信息平台的供应链协同运作模型。结合实例,通过改进第二代非支配排序遗传算法(NSGA-Ⅱ),加入混合编码和动态拥挤距离环境选择策略,设计了改进算法NSGA-Ⅱ-DC。理论函数测试结果表明,NSGA-Ⅱ-DC在收敛性和多样性上明显优于5款经典的多目标进化算法。通过在一个具有多产品、多阶段的五级供应链网络模型上验证表明,NSGA-Ⅱ-DC能够对所提模型进行有效优化。通过层次分析法得到供应链网模型的最优方案,可为决策者提供较强的理论决策依据。

关键词: 供应链网络设计, 多目标优化, 动态拥挤距离, 第二代非支配排序遗传算法