面向物联制造的作业车间同步调度研究

In Internet of things (IoT)-enabled manufacturing systems, manufacturing resources and tasks are converted to intelligent agents that can make distributed decisions with real time captured data. Aiming to such great changes, this project proposes a new operational mode named as synchronized scheduling with the idea of leader agent being followed by subordinate agents. Three key research problems of synchronized scheduling in a job shop environment, i.e., synchronized scheduling modelling, system characteristics analysis, mechanism design are fully studied. We first investigate dynamic multi-attribute leader agent identification method. Based on identified leader agent, we propose a bi-level programming model for synchronized manufacturing. Then the characteristics of interactions among agents as well as the system dynamics are correspondingly analyzed in detail. Finally, we propose an incentive compatible mechanism to motivate subordinate agents to follow the leader agents. Besides, a cooperative game among agents are built and a core profit allocation scheme is presented. The proposed synchronized scheduling mode focuses on the decisions of leader agent rather than all agents, which simplifies the complexity of interactions and decisions of the whole system. The new mode is a great breakthrough of how to schedule numerous agents in IoT-enabled manufacturing systems, which can also be adopted for other intelligent manufacturing systems.

物联制造呈现出制造要素智能化、决策离散化、信息实时化等新特征，针对物联制造作业车间海量智能体调度问题，项目提出同步调度解决方案，提炼同步调度模型构建、特性分析、机制设计三个关键研究内容。提出动态多属性同步主体识别方法、构建面向主体的双层规划同步调度模型；揭示机器、工件等智能体交互行为，分析同步调度系统动力学特性；建立激励相容机制驱使智能体与同步主体进行自主同步，提出合作机制促使智能体进行相互合作。同步调度关注同步主体，使其他智能体与同步主体同步，简化了智能体交互和决策过程，降低了调度的复杂度，是解决物联制造海量智能体调度和控制的一种创新方法，为探索新型智能制造模式提供重要的理论依据。

针对物联制造系统调度问题，本项目研究了同步调度模型、机制和性能。提出了一个考虑生产和出货自私性的双层规划同步调度模型，以生产为主出货为从，设计了一个双层模拟退火算法，实现了生产与出货的同步运作。考虑异地资源运输成本和离散服务窗口，构建了一个考虑资源异地、离散的云服务调度混合整数规划模型，实现了异地云服务协同分配。提出一种基于VCG拍卖的智能车间主动调度机制，设计了两种基于知识驱动的估值函数，赋予工件智能体对排序的逻辑推理和估值能力，机器智能体根据最大化社会效益进行工件选择，实现智能车间高度自治的智能化运作。基于合作博弈思想，提出一种新型的云服务调度用户合作机制，让每个参与合作的用户都能受益，并且不愿脱离合作的联盟，实现制造资源的有效配置。提出了一种漂移瓶颈驱动的TOC启发式算法，通过识别动态瓶颈，确定动态的产品优先级进行资源分配，充分挖掘瓶颈的生产能力，非瓶颈跟瓶颈进行同步，实现产品组合配置优化。提出了一个数据驱动的产线生产能力评估方法，将复杂生产过程分解到工序，构建工序级工时预测模型，基于数据挖掘工序工时模型参数，实现产线生产能力智能评估。实验结果验证了所提方案的有效性，并对重要影响参数进行了敏感性分析。项目所提的同步调度模型、机制、性能分析方法能够有效提升生产效率、降低生产成本、提高交货准时率，为实现制造系统的自主、高效、智能运作提供理论和方法支撑。