单元式混流装配生产作业计划的基因调控方法研究

Cellular mixed model assembly, a customized production mode that fits for multi-product and small-batch production, is important trend for enterprises at assembly stage. Its production planning has significant impacts on production efficiency and order deliveries for diversified needs of customers. This project investigates gene-based production control method for mixed model assembly cells. Against the complexity, large scale and multiple objectives of production optimization problem for mixed model assembly cells, the mapping relationships that represent decision variables, constraints and objectives with genes, gene regulations and gene expression levels are investigated to construct the multiple-level genetic regulatory network based production optimization problem model, of which different levels are connected by relationships between orders, products and tasks, as well as relationships between resources, cells and durations. The evolution of genetic regulatory network is restricted by production constraints to ensure the feasibility of production plans in gene expression procedure, and critical process quantities in the state transition space of genetic regulatory network is investigated to reveal the regulatory mechanism in production planning. The gene activation mechanism that takes critical process quantities as main control parameters is designed to develop multi-level production control policies for order assigning, production cell balancing and product sequencing, and the coordination of policy parameters is investigated to realize integrated optimization of multiple objectives in production planning. The proposed method is to be applied to case studies including antiaircraft missile assembly shops to validate its effectiveness. This research will provide important theoretical basis for the improvement of production efficiency and order deliveries in mixed model assembly cells.

单元式混流装配是一种面向多品种小批量的客户化定制生产模式，是制造企业在装配阶段的重要发展趋势。合理的生产作业计划能够提高单元式混流装配车间生产效率，保证订单准时交付，满足客户多样化需求，具有十分重要的工程意义。本项目研究单元式混流装配生产作业计划的基因调控方法。针对生产作业优化问题的复杂大规模和多目标特点，研究以生产事件为基因节点、约束条件为基因调控关系、优化目标为基因表达水平的多维映射关系，根据单元式混流装配车间的订单-产品-任务关系与资源-能力-时间关系，构建基于多层次基因调控网络的生产作业优化问题模型；通过研究约束条件限制下生产作业计划对应的基因调控网络演化规律，保证基因表达过程中所形成生产作业计划的可行性，并提取优化目标在基因表达过程中产生的关键过程量，揭示生产作业计划的调控作用机理；通过研究以关键过程量为控制对象的基因状态激活机制，设计调控关系参数驱动下订单分配、单元平衡与产品排序的多重控制策略，以调控参数协同优化下的基因状态激活机制，实现生产作业计划的多目标综合优化。最后在防空导弹装配车间等工程对象中进行应用验证。本项目的研究，对提升单元式混流装配生产效率与多品种小批量订单的准时交付能力，具有重要理论意义和实际应用价值。

针对多品种小批量客户化定制生产模式下的单元式混流装配生产作业计划优化需求，本项目根据生产作业计划包含的订单分配、资源配置与产品排序等内容，构建了三层基因调控网络模型，利用二进制变量表示的基因节点、微分方程形式的基因调控关系与常微分方程形式的基因表达水平，分别描述了生产作业计划优化问题的生产事件、约束条件与优化目标；根据生产作业计划在约束条件作用下的生产性能动态变化特点，提出了基于基因表达过程的性能指标演化规律分析方法，通过构建生产作业计划可行解对应的基因调控网络状态转移空间，并计算订单准时交付、单元负载均衡和物料消耗平顺化等性能指标对应的基因表达水平，分析了基因调控关系对各层基因表达水平的影响情况，发现了生产作业计划的有效控制策略集合；根据生产作业计划对各项性能指标的综合优化需求，设计了控制策略强化学习驱动的调控优化算法，在集成有效控制策略的基因调控关系中提取生产参数集合，设计了面向基因调控网络状态转移空间的生产参数调控动作网络与生产性能指标评价网络，利用近端策略优化算法规划调控动作与奖励函数，以合理的基因调控关系驱动基因表达过程，生成了实现订单准时交付与单元负载均衡的单元式混流装配生产作业计划优化解。以某航天制造企业总装车间为背景，开发了单元式混流装配生产作业计划优化原型系统，并进行了应用验证。本项目研究，对提升单元式混流装配生产效率与多品种小批量订单的准时交付能力，具有重要理论意义和实际应用价值。