当前位置: 首页>博士论文>资源详情
多级供应链复杂系统建模仿真与演化分析研究
中文摘要

随着工业4.0的推进及经济全球化发展,制造业供应链愈加呈现出地域范围广、企业数量庞大、相互依赖性强、动态性强、复杂化等特点。探索供应链演化规律和鲁棒性能,对于行业和企业的管理具有重大的实践价值和理论意义。 主要研究内容和创新点如下: 1.提出了基于节点度与竞争优势系数连接偏好的、局域世界选择连接的、哑铃型多级供应链的刻画规则和演化算法,建立供需网络及其子网、综合网络及其子网的演化模型,分析了网络统计特性的演化规律,揭示了全局网络和子网之间、供需网络和综合网络之间统计特性的相关性。结果表明:相同演化条件下,度分布幂指数大小与网络平均度正相关、与供应链级数负相关;综合网络较大的平均度使得网络的平均距离更短,效率更高,合作因子更大,有利于改善网络性能。 2.以上述模型为基础,提出了基于核心级的攻击策略和基于全网的攻击策略,设计了具有连接关系的同类相似节点级联失效算法,从多个维度设计评价了供应链的动静态鲁棒性能,揭示了度分布和网络级数与供应链鲁棒性的关系。结果表明:相同演化条件下,供应链级数越少,幂指数越小,静态鲁棒性越强;较全网攻击相比,供应链在核心级攻击策略下表现出较强的静态鲁棒性;同类节点的级联失效会降低供应链的动态鲁棒性能。 3.采用CAS理论,基于ABM/DES的混合建模方法,提出并建立了以多个制造商为核心的、以价格和广告为竞争策略的多级供应链刺激反应演化模型。仿真发现:采取低价策略的企业较早进入供应链系统,在供应链形成期具有明显的竞争优势,但后劲不足。采取高质量策略的企业虽然融入系统迟,但凭借质量优势持续拓展市场,具有后发优势。特定条件下,制造商的价格策略起主导作用,广告策略影响较小。 4.基于CN/CAS/GIS,构建了考虑产品效用和距离连接偏好的、局域世界连接的、多级供应链演化模型及其适应度指标,发现了供应链点权、单位权、度分布之间的关系,阐明了适应性供应链的鲁棒性能。结果表明:考虑点权的现实供应链鲁棒性较强;供应链的核心级在面对蓄意攻击和随机攻击时,适应度较强且差异小;同级节点的有限连接改善了供需供应链的网络性能和静态鲁棒性,也可以规避同类节点过度连接造成的级联失效风险。 5.基于CN/CAS/GIS模型,提出了基于PSO的多Agent学习机制,分析了社会学习能力和自我学习能力对供应链的影响,揭示了不同学习能力对供应链演化和鲁棒性的影响规律。结果表明:适当的社会学习能力有利于提高供应链的利润和产品效用;加强自我学习能力,供应链上产品更加多样化,消费者对商品的选择范围更广;具有学习能力的供应链有利于改善网络的鲁棒性能。 关键词:供应链:复杂网络;复杂适应系统;鲁棒性;粒子群学习算法 国家门然科学基金研究计划资助项目(项目编号11072192)

英文摘要

With the advancement of industrie 4.0 and economic globalization, the manufacturing supply chain has become more and more characterized by a wide range of areas, a large number of enterprises, strong interdependence, dynamic and complicated. Exploring the evolution law of supply chain has great practical value and theoretical significance for management of industry and enterprises. The main contents and innovations of this paper are as follows: 1.The description rules and evolutionary algorithm of dumbbell-type multi-level supply chain complex network were proposed based on node degree and competitive coefficient and local world selection connection. The evolution model of supply and demand network (SDN) and its subnets, integrated network (IN ) and its subnets were established. The evolution rule of multi-level supply chain statistical features was analyzed, and the relationship between subnet and global network was revealed, and the relationship between SDN and IN was revealed too. The results show that the power index of degree distribution is positively correlated with the average degree and negatively correlated with level number of supply chain under the same evolution condition, bigger average degree makes IN has shorter average distance and higher efficiency and bigger cooperative factor compared with SDN, which is beneficial to improve network performance. 2.Based on the above model, the global network attack strategy and core level attack strategy were proposed, and the cascading failure algorithm of similar nodes with connection was designed, and the robustness of the complex supply chain was evaluated from multiple dimensions. The relationship between degree distribution and static robustness, and between network level and static robustness of supply chain were revealed. The results show that the fewer level and the smaller power index of supply chain, the stronger static robustness under the same evolution condition, the supply chain display stronger static robustness under core level attack strategy compared with global network attack strategy, and the cascading failure robustness of comprehensive network is greatly reduced. 3.Based on theory of complex adaptive system (CAS) and ABM/DES hybrid modeling method, the multi-level supply chain stimulus-response model was proposed and established, which used multiple manufacturers as the core and considerd pricing and advertising strategy. Simulation shows that the manufacturer with low pricing strategy enter supply chain system firstly, and has obvious competitive advantages in the formation, but has insufficient stamina and has a smaller market share eventually. However the manufacturer with quality strategy enter the supply chain lately, but its market share is increase with time, and takes a large proportion market share eventually, and has advantage of backwardness. Under certain conditions, manufacturer's pricing strategy plays a leading role, and advertising strategy has less influence. 4.Based on CN/CAS/GIS, the multi-level supply chain model and its adaptability index were put forward considered product utility and distance utility of node and local world connection. The relationship between point weight distribution and unit weight distribution and degree distribution power index were revealed, and the robustness of supply chain was clarified. The results show that the realistic supply chain considering point weight has strong robustness, and supply chain adaptability is strong and has little difference when core level faced with deliberate attack and random attack. Limited connections of similar nodes not only improve the performance and static robustness, but also avoid cascading failure caused by the excessive connection of similar nodes. 5.Based on CN/CAS/GIS evolution model, the multi-agent learning mechanism based on PSO learning algorithm was proposed, and the impact of social learning ability and the self-learning ability on supply chain was analyzed. The influence of different learning ability on the evolution and robustness of the supply chain was revealed. The results show that appropriate social learning ability is beneficial to improve the profit and product effectiveness of supply chain. The product of supply chain are more diversified, and consumer's choice are wider if emphasize self-learning ability. The supply chain with learning ability is propitious to improve the robustness of supply chain. Keywords: Supply chain; Complex network; Complex adaptive system; Robustness; PSO The study is sponsored by the Funds of National Natural Science Foundation of China (No. 11072192 ).

作者相关
主题相关
看过该书的人还在看哪些书