基于广义互熵主元分析的故障检测方法

Fault Detection Method Based on Generalized Correntropy Principal Component Analysis

针对实际化工过程会受到不同程度非高斯扰动影响的问题,提出一种基于广义互熵主元分析的故障检测方法。从重构误差的角度出发,考虑过程的非高斯性,采用广义互熵准则构建PCA模型,利用核密度估计法确定故障检测指标的控制限。将所提出的方法用于田纳森-伊斯曼过程进行故障检测,并与基于传统PCA的故障检测方法和基于核PCA的故障检测方法进行对比。由田纳森-伊斯曼过程21种故障检测结果可知,本文所提出的广义互熵PCA在处理非高斯系统的故障检测方面表现出良好的性能,即有较低的误报率和漏报率。

In this paper,a fault detection method based on generalized correntropy principal component analysis(PCA)is proposed to deal with the problem that actual chemical processes are affected by non-gaussian disturbances to varying degrees.First,traditional PCA algorithm is briefly reviewed.Then,from the perspective of reconstruction error,by considering the non-gaussian nature of process,the mean square error is not enough to describe its random characteristics.In this paper,the generalized correntropy criterion is adopted to construct the PCA model.Then,the kernel density estimation method is used to determine the control limit of the fault detection index.Finally,the proposed method is applied to the fault detection of Tenenson-Eastman process,and compared with the traditional PCA-based fault detection method and the kernel PCA-based fault detection method.It can be seen from the 21 fault detection results of Tenenson-Eastman process that the generalized correntropy PCA proposed in this paper has a good performance in dealing with the fault detection of non-gaussian systems,that is,it has a low false alarm rate and a low missing alarm rate.