基于多Agent与集成神经网络的制造网络信息调度模型的研究

制造网络的面向服务架构（SOA）应当满足敏捷制造的需求，在SOA中任何瓶颈，如网络资源的分配策略，会影响到制造网络的服务敏捷性。在分析制造网络多Agent结构的基础上，提出了一种制造信息的分类方法。利用制造应用模型，提取系统运行的状态空间，通过制造系统单元间的组合和不同部件网络输出的决策间的融合，集成神经网络输出各制造单元信息调度权重。实验结果表明，此方法符合制造网络设计的原则和依据，以制造应用为中心，最大限度满足制造应用的需求。

Research on Scheduling of Manufacturing Network' s Information Based on Multi-agent and Integrated Neural Networks

Service oriented architecture(SOA)of manufacturing network should meet the requirement of agile manufacture,and any bottleneck in SOA,such as the policy of network resource's distribution,can influence the agility of manufacturing network.A classifying approach of manufacturing information is proposed based on the multi-agent architecture for manufacturing networks.In this approach,the system state space is extracted by the model of manufacturing applications,and the final information scheduling weights for each manufacturing unit are given by the integrated neural networks through the combination of manufacturing units and the decision of different subnets.The experiment shows that the proposed approach accords the rule of designing a manufacturing network, centering on manufacturing applications and meeting their requirements furthest.