串并联多工位制造系统误差传播分析与多变量过程质量控制

Quality control of production process is one of the critical problems in the operational optimization of manufacturing system. Product quality risk will increase greatly when the production process is out of control. In the proposal, we focus on the process quality control problem of the serial-parallel multi-station manufacturing system (SP-MMS) by applying novel methods on the basis of up-to-date achievements such as Stream of Variation (SoV), variation propagation analysis and Multivariate Statistical Process Control (MSPC). Firstly, the SoV method is extended from the serial manufacturing processes to the serial-parallel manufacturing processes. By combining the advantages of engineering knowledge-driven SoV method and data-driven MSPC method, a SoV/MSPC integrated "physical-statistical" variation propagation model is set up, which can reveal the basic rules of interaction, propagation and accumulation of variations in all production stages. Secondly, by analyzing the inter-dependency and auto-correlation of quality characteristics in depth, an economic-statistical control chart design model is proposed, and then the out-of-control patterns of control chart are automatically recognized and the variation sources are identified through the variation propagation analysis. Thirdly, the impact of "equipment/fixture/tool" failures on the quality characteristics are quantitatively analyzed to obtain the joint optimization model and the economic maintenance model by integrating the manufacturing system reliability analysis and quality control strategies. Such models can facilitate the diagnosis and predictive control of abnormal process fluctuation. Lastly, based on the variation propagation model, the process quality continuous improvement method are further discussed through the production line layout optimization, the quality-oriented assessment of process routes, and the joint optimization of quality and productivity performance measures.

过程质量控制是制造系统运行优化的核心问题，过程一旦失控将大大增加质量风险。本项目在误差流（SoV）、误差传播分析、多变量统计过程控制（MSPC）等最新成果的基础上，通过创新手段对串并联多工位制造系统的过程质量控制问题进行研究：首先将SoV方法从串联推广到串并联过程，融合工程驱动SoV和数据驱动MSPC的优点，建立SoV/MSPC集成的"物理-统计"误差传播模型，揭示阶段误差相互影响、传播、累计的规律；然后，深入分析质量数据相关性和自相关性，提出经济和统计综合优化的控制图设计方法，基于误差传播分析实现过程失控自动判别和误差源定位；再后，定量分析"设备/工装/刀具"失效对质量特性的影响，建立制造系统可靠性和质量控制集成优化模型及经济性维修策略，实现过程异常波动的诊断和预防控制；最后，基于误差传播模型，通过生产线布局优化、面向质量的工艺路线评估、质量和生产性能指标集成优化等途径实现过程持续改进。

过程质量控制是制造系统运行优化的核心问题，过程一旦失控将大大增加质量风险。本项目在误差传播分析、统计过程控制、维修决策等最新成果的基础上，通过创新手段对多工位制造系统的过程质量控制问题进行了研究。.1）分析了多阶段装配制造过程的误差传播规律，建立了多工序DP-SDT偏差综合分析理论的框架，并通过公差优化降低制造质量控制的难度。提出了制造系统多工序加工误差预测及控制理论。建立了制造系统运行效率和质量成本的综合优化模型。.2）以生产系统的质量成本最小为优化目标，以统计特性较优为约束条件，进行控制图的优化设计。针对多工位构成的串联制造系统，建立了多属性控制图的经济与统计指标综合优化设计方法。建立了小样本环境下的CUSUM控制图设计与分析方法。运行神经网络等机器学习方法，对失控模式进行自动辨识，对故障进行诊断分析。.3）研究了统计过程控制和维修决策的集成优化模型。针对过程质量均值偏移随时间变化的情况，提出了过程单一失效机制下的SPC与维修决策的集成优化模型。针对由生产过程与生产设备组成的系统，提出了混合失效机制下基于延迟监控策略的SPC与维修决策的集成优化模型。针对过程质量均值偏移为随机分布的情况，建立了经济-统计最优的指数加权移动平均（EWMA）控制图与预防性维修集成优化问题。针对过程失控与设备故障混合失效机制，建立了混合失效机制下多变量EWMA控制图与维修决策集成优化模型。.4）开发了基于偏差传播理论的装配公差分析原型系统、SPC与维修决策集成优化原型系统，在东方电机公司燃气轮机定子铁芯叠片自动装配项目、深圳某手机代工厂、三一重工装配车间进行了应用验证。.本项目共发表论文19篇，其中SCI论文7篇，EI论文12篇；申请发明专利4项。