带有批间控制器的混合产品批次过程监控和故障诊断

Process monitoring and fault diagnosis of batch manufacturing is very important for its safe and efficient operation. However, with the application of the run-to-run controller and the increase of product categories, the traditional approaches meet new challenges. In this project, the mixed-product batch manufacturing process with run-to-run controller is dealt with, the influence of the run-to-run controller and product transition on the process and its data is considered, data feature of "pseudo normal" state and the process with multiple model, strong dynamics and batch inconsistency is extracted, the disturbance change and the faults on key/non-key variables are distinguished, new process monitoring and fault diagnosis approaches based on multivariate statistics, pattern recognition and machine learning methods are developed. It will provide a theoretical basis for solving practical problems in batch industrial process monitoring. At the same time, simulation and experimental platform using the enterprise data are constructed to verify the theoretical research results, which will lay foundation for the large-scale application of these results. This project is in the subject frontier,which comes from the actual production, and return to the engineering application,and has great theoretical significance and practical value.

批次制造工业的过程监控和故障诊断对于其安全高效的运行至关重要。然而随着批间控制器的使用和产品种类的增加，传统的过程监控和故障诊断方法遇到了新的挑战。本项目针对带有批间控制器的混合产品批次制造过程，考虑批间控制器和混合产品切换给过程及其数据带来的影响，提取"伪正常"状态和多模型、强动态性和批次不一致现象的数据特征，区分干扰变化和关键/非关键变量上的故障，在多元统计、模式识别和机器学习方法的基础上研究新的过程监控和故障诊断方法，为批次制造工业过程的工程实际问题的解决提供理论依据。同时，利用企业数据构建仿真和实验平台，对理论研究成果加以验证，为未来该方法的大规模运用奠定基础。本项目处于学科前沿，既来源于生产实际，又回归于工程应用，具有重大的理论意义和实用价值。

批次制造工业的过程监控和故障诊断对于其安全高效的运行至关重要。本项目针对带有控制器的制造过程，考虑了批间控制器以及PID、MPC等其他控制器对过程及其数据带来的影响，提取“伪正常”状态和多模态、强动态的数据特征，通过自适应Lasso等系统辨识方法建立了时间序列数据模型，设计了控制器性能指标、稳定性指标和模型质量指标来监控系统运行状态，进行故障检测和诊断；分析多个变量之间的相关性，采用Takagi–Sugeno模糊模型来处理未知的随机度量时延,在此基础上建立批次过程的批间补偿算法；采用了最小风险贝叶斯决策理论将误诊的风险和以前的故障诊断结果引入到当前的诊断过程中，提出了一种并行偏最小二乘的数据模型来分别检测与过程数据和质量数据有关的故障，提高了故障诊断效率。并且，相关成果在半导体制造业和和薄膜打印业中得到了应用，解决了实际工业监控问题。本项目处于学术前沿，既来源于生产实际，又回归于工程应用，具有重大的理论意义和实用价值。