柔性材料多工位连续制造系统智能维护虚实一体建模方法的研究

Roll to roll (R2R) processing is a new manufacturing method for flexible thin film electronic materials, however the flexibility of material itself, the performance of manufacturing system is very easy to cause irregular deformation of material, so it is difficult to improve processing quality and efficiency. The project takes the system performance degradation analysis as the breakthrough point to study the method of improving manufacturing quality and extending the normal usage time of equipment, it includes: to model and simulate local loading and deformation for flexible material under the whole process, to extract influencing factors of controllable, uncontrollable disturbance variables; to study innovatively error prediction SOV method of multistage R2R continuous manufacturing process, and to solve error source of this process under the influence of multi factors; performance degradation prediction and health state evaluation methods will be studied based on adaptive fuzzy clustering analysis of manufacturing data, and qualitative classification problem of performance state will be solved; a knowledge discovery method of fuzzy rough sets is proposed that control rules can be summarized heuristically and updated incrementally, and a decision control model is constructed based on online parameter adjustment for multistage continuous manufacturing system. Finally A multi-level fusion and associated system control self decision adjustment mechanism will be formed with the core of "manufacturing quality - performance degradation - collaborative decision", and it will expand the application range of Intelligent Maintenance System Theory in multistage continuous manufacturing system control modeling.

卷对卷(R2R)加工是新型柔性薄膜电子材料的制造方式，但由于材料本身的柔软性，制造系统性能的变化极易造成材料不规则变形，加工质量和效率难于提高。项目以系统性能衰退分析为切入点研究提升制造质量、延长设备正常使用时间的方法，内容包括：全过程建模仿真柔性材料局部加载变形，提取可控、不可控干扰变量影响因素；创新性研究R2R多工位连续制造过程误差预测SOV方法，解决多因素相互关联影响下的多工位连续制造过程误差溯源；探索性研究基于制造数据模糊自适应聚类分析的性能衰退预测与健康状态评估方法，解决性能状态定性分类问题；提出了一种控制规则可启发式归纳与渐增更新的模糊粗糙集知识发现方法，构建面向多工位连续制造系统的在线参数可调整的决策控制模型，形成以“制造质量-性能衰退-协同决策”为核心的多层面融合与关联的系统控制自我决策调整机制，拓广智能维护系统理论在多工位连续制造系统控制建模中的应用范围。

项目从R2R制造系统性能衰退评估的角度来研究提升系统自我感知、预测、补偿与重构的智能化水平的模型和方法，目的是以最小的代价保证预期产品质量，又能延长设备正常使用的工作时间。围绕着柔性材料R2R多工位连续制造系统智能维护虚实一体建模问题，创新性地研究复杂因素影响下的柔性材料R2R制造过程误差预测SOV模型，并与质量控制图相结合，建立了加工误差预测与故障源定位方法，解决多因素相互关联影响下的多工位连续制造过程误差溯源、制造过程误差快速预测的问题；探索性研究基于模糊自适应聚类的性能衰退预测方法，研究通过隶属度函数的变化曲线来反映R2R加工系统性能衰退的趋势，提出了一种结合可能性c均值模糊聚类和TS-FNN的神经网络预测模型，有效地提高了R2R加工系统性能衰退与健康预测的稳定性和敏感性；创造性提出了一种基于粗糙集的柔性材料R2R加工控制决策生成方法，解决了现有柔性材料R2R加工中材料种类多、特性各异、变形影响因素复杂而造成R2R加工过程工艺控制难的问题。.项目构建了多工位连续制造系统智能维护的虚实一体模型，将制造系统决策优化控制问题的研究从“复杂控制模型预测”提升为“加工大数据知识发现”自我决策控制的层次，为多工位连续制造系统的智能化决策控制知识的自动沉淀奠定理论基础，并拓广智能维护系统理论在制造系统控制建模优化中的应用范围。