基于多尺度数据融合的智能制造系统状态监测与故障诊断方法

Manufacturing has become increasingly hyper-connected and accessible with the advancement of the industrial Internet of Things and Cyber-physical Systems, and this also provides unprecedented opportunities for process condition monitoring and diagnosis in smart manufacturing systems. However, due to the large volume, high dimensionality, nonstandard data acquirements, heterogeneity, and uncertainties of the engineering data from both of the system and device levels, it is still challenging to develop condition monitoring and diagnostic tools by fusing these data. The objective of this project is to develop new models and associated algorithms for smart manufacturing systems by fusing all-level data sources, in order to establish novel in-situ condition monitoring and accurate diagnostic approaches that are built upon an innovative data fusion framework. The engineering data fusion model and associated parameter estimation algorithms will be thoroughly investigated by incorporating data analytics as well as the domain knowledge. It is hopeful that this research could provide a solid theoretical basis and key technologies for a real-time health index and faulty tool diagnostic method for smart manufacturing systems.

信息物理系统和工业物联网技术的快速发展促进了智能制造系统各层面数据的采集和实时交互，为系统状态监测和故障诊断提供了前所未有的契机。然而，由于制造系统层面间和层面内的数据存在容量大、维度高、采集标准难以统一、数据结构差异性明显、数据噪声和不确定性原因繁多等特点，现有的监测与诊断技术尚缺乏能够有效融合制造系统全局数据的方法。本课题以智能制造系统中设备层和系统层的数据为研究对象，针对制造系统“难以全面融合各层级数据”这一突出问题，拟建立一套能够实时监测制造系统状态并精准诊断设备故障的方法论。课题充分考虑实际制造过程中存在的各种数据不确定性和条件约束，利用工程背景知识从传感数据到系统执行数据逐层建立数据融合模型，拟解决工程数据融合建模、参数估计、实时状态监测和故障诊断等具体问题。该研究将最终为解决智能制造系统及装备的故障诊断理论奠定科学基础，同时也为制造系统实时状态监测提供理论依据与关键技术。

本项目以先进制造中的信息物理系统为研究对象，充分考虑当前制造在多尺度环境下的维度高、异构程度复杂、信噪比低等特点，重点探索了多源异构传感数据在工业系统生产运营不同层面的数据融合方法，提升了制造系统在不同层级和数据尺度下的融合与建模方法水平，为在实际工况下的制造系统过程监测与诊断奠定了良好基础。在传感数据层融合方面，研究团队攻克了数据预处理和融合建模统一的数据处理方法，满足数据随机缺失条件下的数据融合方法；在设备间数据融合层面，研究团队攻克了多工序数据关联融合方法，并依此开发了相应的全局过程监测与诊断技术。研究方法在半导体制造领域和特钢制造领域得到了良好的验证，研究成果在在工业工程领域、质量与可靠性工程领域等国际知名期刊发表了多篇高水平论文。