摘要
针对工业过程微小故障以及工业过程数据普遍存在的非线性问题,提出基于近邻指标累积和(Cumulative Sum of Neighbor Statistic,CUSUM-NS)的故障检测方法。使用互信息主元分析(Mutual Information Principal Component Analysis,MIPCA)对训练数据降维,提取互信息主元构成新样本空间,对降维后的新样本空间计算k近邻(k-Nearest-Neighbor,kNN)距离平方和指标,充分提取过程数据的非线性特征。利用累积和(CUSUM)方法对近邻距离平方和进行累积,捕捉过程数据的微小变化。通过TE(Tennessee Eastman)过程进行仿真,结果验证了所提方法有效性。
Aiming at the small faults and the common data non-linear problems of industrial process, a fault detection method based oncumulative sum of neighbor statistic(CUSUM-NS) is proposed. Mutual information principal component analysis(MIPCA) is used to reduce the dimension of training data, and the principal components based on mutual information are extracted to construct a new sample space. For the new sample space after dimensionality reduction, the nonlinear features of the process data can be fully extracted through the distance square sum statistics of k nearest neighbors. Cumulative summation(CUSUM) method is used to accumulate the sum of squares of neighboring distances to capture the small changes in process data. The simulation is conducted through the Tennessee Eastman(TE) process, and the results verify the effectiveness of the proposed method.
作者
郭小萍
高嘉俊
郭建斌
李元
Guo Xiaoping;Gao Jiajun;Guo Jianbin;Li Yuan(Information Engineering School,Shenyang University of Chemical Technology,Shenyang 110142,China)
出处
《系统仿真学报》
CAS
CSCD
北大核心
2021年第4期792-800,共9页
Journal of System Simulation
基金
国家自然科学基金重大项目(61490701)
国家自然科学基金(61673279)。
关键词
互信息主元分析
K近邻
累积和
微小故障
PCA based on mutual information
k-Nearest Neighbor
cumulative sum
small faults
