基于递推规范变量残差和核主元分析的微小故障检测

微小故障由于故障征兆不明显从而很难在故障发生早期对其进行检测. 针对该问题, 本文提出了一种基于递推规范变量残差和核主元分析(RCVD–KPCA)的微小故障检测方法. 首先构造规范变量残差, 从中提取数据的线性特征. 利用指数加权滑动平均法对规范变量残差进行递推滤波处理, 提高规范变量残差对微小故障的敏感程度;然后使用KPCA提取规范变量残差中的非线性主成分作为非线性特征, 根据提取的特征提出了两个新的故障检测统计量; 此外, 利用核密度估计确定故障检测统计量的控制限. 由于同时提取了过程数据的线性和非线性特征, 有效地提高了非线性动态过程中微小故障的可检测性. 以闭环连续搅拌釜式反应器过程为例进行了仿真分析, 仿真结果表明本文所提方法具有较好的故障检测性能.

Incipient fault detection based on recursive canonical variate dissimilarity and kernel principal component analysis

Incipient faults are difficult to detect in the early stage because the symptoms are not obvious. In response to this problem, this paper proposes an incipient fault detection method based on recursive canonical variate dissimilarity and kernel principal component analysis (RCVD-KPCA). First, the canonical variate dissimilarity is constructed, and the linear features are extracted from the canonical variate dissimilarity. The exponentially weighted moving average method is used to recurse and filter the canonical variate dissimilarity to improve the sensitivity of canonical variate dissimilarity to incipient faults. Then, KPCA is used to extract the nonlinear principal components in the canonical variate dissimilarity as nonlinear features. According to the extracted features, two new fault detection statistics are proposed. Furthermore, the kernel density estimation is used to determine the control limits of statistics. The method extracts the linear and nonlinear features of the process data at the same time, the detectability of incipient fault in the nonlinear dynamic process is improved. The process of a closed-loop continuous stirred tank reactor is taken as an example for simulation analysis, and the simulation results show that the proposed method has good performance in incipient fault detection.