Rocket Force University of Engineering
The National Natural Science Foundation of China (General Program, Key Program)；The National Natural Science Foundation of Shaanxi Province (General Program)
独立成分分析(Independent component analysis, ICA)是一种多变量统计分析方法, 常被用于非高斯过程监测. 它能够有效利用信号的高阶统计信息（三阶以上）提取相互独立的独立成分, 在工业过程监测中得到了广泛的应用, 是当前国际过程监测领域的研究热点. 对此, 本文介绍了经典ICA模型、改进ICA模型及其在工业过程的过程监测技术. 首先, 对经典ICA模型进行了介绍, 在此基础上对经典ICA模型进行分类并指出其优缺点. 其次, 针对经典ICA模型存在的缺陷, 从ICA自身存在的问题、噪声和离群值三方面梳理了改进ICA模型的发展. 然后, 以工业过程为主要应用背景, ICA的过程监测技术从简单工业过程衍变至复杂工业过程. 面向工业过程运行数据的单一特性和混合特性, 综述了ICA及其扩展模型在工业过程监测中的研究现状. 最后, 探讨了该研究领域亟需解决的问题和未来的发展方向.
The independent component analysis (ICA) is a multivariate statistical analysis method, which is often used for non-gaussian process monitoring. It can effectively use the high-order statistical information (exceeding the third order) of the signal to extract independent components, which has been widely used in industrial process monitoring and is a research hotspot in the current international process monitoring field. As such, this paper introduces the classic ICA model, the improved ICA model and its process monitoring technology in industrial processes. Firstly, the classic ICA model is introduced, which is then classified and the advantages and disadvantages are pointed out. Secondly, in view of the shortcoming of the classic ICA model, the development of the improved ICA model is sorted out from three aspects, including ICA"s own problems, noise and outliers. Then, the industrial process is applied as the main application background, and the ICA process monitoring technology is evolved from simple to complex industrial processes. Faced with the single and mixed characteristics of operating data in the industrial process, the current research status of ICA and its extended models in industrial process monitoring are reviewed. Finally, the problems to be solved in this research field and the future development directions are discussed.