大规模定制下虚拟单元节能调度的在线调控模型与规则研究

The manufacturing cellular in the mass customization products is relatively fixed, which may cause the production scheduling is very complicated currently. The increase of the environmental pressure and individual demand makes it urgent to integrate manufacturing cellular and production scheduling, so as to promote effective energy usage rate and resource sharing rate. However, the confliction of both the fixed manufacturing cellular and the various lot-size production, and the real-time change during production process, makes dynamical integration difficult.. Aiming to the manufacturing cellular mode under mass customization, this project takes the energy consumption calculation and the product process as the key points, and gradually achieves the online regulation with energy-saving, high-efficiency and stability in virtual cellular scheduling. The detail information is as follows. ① The energy consumption equation is explored based on the sliding window method. Then the characteristics of problem are analyzed successfully by establishing the linear two-layer hierarchical optimization model for mass customization. ② Based on the implicit experience and knowledge with the potential value of production data, the scenario classification method with Bayesian network and the rules discovering technique with meta-heuristic algorithm are designed as so to establish the scenarios and rules database. ③ And on this basis, a novel “correlation + regulation” dynamical operation optimization method is proposed so as to generate the superior scheme of online regulation quickly and efficiently. This will respond the production process timely and accurately, and keep it energy-saving, high-efficiency and stability. . The project will provide an effective way for improving the production management of the mass customization products, benefit the sustainable development of the manufacturing industry, and promote the practical application of the theoretical achievement.

大规模定制产品生产单元固化与调度过程复杂。随环境管制增强与个性需求膨胀，亟需集成虚拟单元与调度以提升能耗有效利用率和资源共享率。然而，单元固化与动态批次冲突，生产环境实时变化，致使难以动态集成。. 本项目围绕大规模定制单元加工模式，以能耗计算、产品工艺为切入点，实现虚拟单元调度的在线低耗、高效与平稳调控。具体实现方法：①探索基于滑动窗口的能耗计算方法、及面向大规模定制的线性双层递阶结构优化模型，实现问题特征分析；②深度探寻数据价值，挖掘问题特征下的隐性经验和知识，设计基于贝叶斯网络的情景分类和基于元启发式算法的规则发现技术，实现离线情景与规则库构建；③在此基础上，提出“关联+调控”动态运作优化方法，快速高效生成优越的在线调控方案，迅速精准实现生产运行的节能、高效、有序。. 研究成果为大规模定制产品生产管理提供一种新途径，有利于制造业的可持续发展，促进研究理论的实际应用。

大规模定制产品生产单元固化与调度过程复杂。随环境管制增强与个性需求膨胀，亟需集成虚拟单元与调度以提升能耗有效利用率和资源共享率。然而，单元固化与动态批次冲突，生产环境实时变化，致使难以动态集成。. 项目组围绕大规模定制单元加工模式，以能耗计算、产品工艺为切入点，实现虚拟单元调度的在线低耗、高效与平稳调控。研究成果包括：（1）在大量基础数据调研的基础上，设计了设备与AGV能耗计算方法，建立了大规模定制制造虚拟车间能耗特性框架；（2）设计了能耗、高效、平顺等特征的指标体系，构建了2D/3D单元划分模型、多目标单元调度模型、以及虚拟单元调度双层递阶多目标优化模型；（3）提出了大规模制造车间的局部搜索算法与群智能优化算法，实现了2D/3D单元划分问题、多目标单元调度问题、以及虚拟单元调度问题求解，为大规模定制下2D/3D模块化单元产线高效重构提供理论支撑。（4）设计了具有自学习、无监督的基因表达式编程算法，构建了兼容问题特征和节能期望的离线规则库；随着历史数据的不断完善，离线规则库也及时更新，为在线调控多情景规则匹配提供有效支撑；（5）提出了基于多智能体的深度强化学习算法实时框架，为快速优越生成调控方案，迅速精准响应实时动态变化提供保证。. 项目组围绕项目研究目标，在《Journal of Cleaner Production》、《Computers & Operations Research》、《Swarm and Evolutionary Computation》、《中国机械工程》、《计算机集成制造系统》等重要刊物发表论文28篇，其中SCI/EI收录17篇，申请专利5项，获得软件著作权2项，培养了硕士研究生9名，博士研究生2名。