中空玻璃生产过程中多离散问题的耦合分析与优化方法

The process flow of insulating glass involves in four optimization problems, say，the batching problem of orders, the cutting problem of pieces, the loading problem of tempering furnace and the planning problem of tasks. These four problems influence each other and form a complex coupling optimum problems. Solving these problems lonely will impair not only the balance of the production organization but also decrease the production efficiency. This project firstly analyses the structure of the coupling optimization problem from the system-level view and encloses the existed combinatorial explosion and iterative divergence problems. Then, it suggests an optimization method that mixes the compute decoupling mode and the strategy decoupling mode, and integrates a number of algorithms such as clustering, packing and dispatching to solve the coupling optimization problem. The innovation points of such project are as follows: first, it proposes a new strategy by which problem-depended gaps are used to reduce recursive times when traversing the domain of a discrete variable. Based on the strategy and extreme pairs with utilization ratio and loading ratio, it develops a new hybrid decoupling method to reduce the computing load and improve the convergence of the suggested optimization method. Second, it proposes a macro-micro combinational optimization mechanism to zip the solve space based on the group method. Last, it develops a new packing algorithm, in which grouping policy is employed, to speed up the solving process. The project can directly reduce the cost of raw materials and energy consumption, improve the production efficiency. The research will strike a new topic on the operation mechanism of a manufacturing system— run a manufacturing system at an optimal state by constructing the special coupling optimization problem in it and solving it.

中空玻璃生产过程中蕴含着订单组批、下料优化、钢化炉装载、作业计划四类相互耦合的离散问题，孤立地求解只会导致生产组织失衡、效率降低。项目从系统的角度深入分析了其间的耦合关系和耦合结构，揭示了求解过程中存在的“维数灾”与“迭代发散”现象，提出了计算与策略混合的解耦思路和“解耦计算+单元优化算法”的解耦方法，自主研发与集成聚类、排样及调度算法来有效求解这一问题。创新性在于：提出了新型的混合式解耦方法，利用排样率与装载率的自然维度间隙，建立极值对机制，有效降低耦合所带来组合计算负荷及迭代发散风险；提出了新颖的宏微复合优化机制，依托宏微二级组化及反馈机制，约简计算空间；提出了新式的构造式排样优化算法，发挥组件的隐性启发作用，加快求解速度。项目的研究可为中空玻璃生产过程的组织优化提供理论支撑，带来原料成本与能耗降低、产能释放等经济效益，也有助于推动以耦合优化问题求解为核心的制造系统优态运行机理的研究。

本项目针对中空玻璃生产过程中的“组批-下料-装载-计划”四类耦合优化问题展开研究，采用“计算性解耦与策略性解耦”相混合的思路，基于组化技术的排样优化算法，形成了以“解耦计算+单元算法”为基本框架的求解方法。项目搭建了数字孪生快速定制设计平台，通过硬件与软件仿真集成的方式，实现整体设计方案和整体优化方法的迭代优化。创新性提出了基于数字孪生的产线变型设计方法，解决了“产线重构失速”的核心难题。自主研发了数字孪生系统平台，平台具有三维建模、半实物仿真、多视图同步（数字孪生）、虚实互控、性能分析等功能模块，平台对产线变型设计的支撑功能优于国际主流仿真平台；有力支撑了产线设计集成模型从串行化向并行化的转变；形成了柔性产线变型设计与优化技术标志性成果，成功应用于中空玻璃生产、电子制造等领域。项目在JMS、IJPR、EJOR、IJCIM、RCIM等国际知名期刊发表SCI论文11篇，其中ESI高被引论文2篇；申请美国专利4项，其中授权2件；PCT专利检索6件，授权中国发明专利12件，软件著作权5件。