基于ICA过程监控方法的研究

针对目前实际工业生产中变量不能严格服从高斯分布,且变量之间存有严重相关性的特点,进行了基于独立成分分析的过程监控方法研究。该方法不仅去除了变量之间的相关性,而且充分利用过程信息的高阶统计特性,建立过程信息的独立元模型。利用独立元模型对仿真实时数据进行故障检测研究,最后对离散系统多变量过程模型进行了仿真验证,仿真结果表明:该方法能快速准确的检测到运行中发生的异常,验证了该方法的有效性以及与PCA方法相比所存在的优越性。

Research of process monitoring based on independent component analysis

For the feature of non-Gaussian and severe correlation between process data in practical industrial process, a new fault monitoring method based on independent component analysis is presented. It not only removes the correlation between variables, but also makes full use of higher-order statistical characteristics and establishes ICA model of process data. Using the model to research the fault monitoring, the new method is applied to muhivariable process model in discrete system. The simulation results show that： the methods can quickly and accurately detect abnormal operation, verify the validity of the method, as well as in comparison with the superiority of PCA method