Fault detection method based on variable sub-region PCA

doi:10.11949/j.issn.0438-1157.20160217

Abstract

Abstract:

Aiming at the problem that traditional principal component analysis (PCA) method can't highlight the local variable information in industrial process monitoring, this paper proposes a variable sub-region PCA (VSR-PCA) fault detection method. First, PCA is used to decompose original data space into principal component subspace (PCS) and residual subspace (RS), and mutual information between variables and PCS is calculated to measure their correlation which is utilized to obtain the variable sub-regions. Then, local T² statistics and local SPE statistics are calculated in each variable sub-region. Bayesian inference is applied to integrate information in every sub-region to construct global statistics which are able to emphasize the local variable information while preserving the whole process information. Simulation results on the continuous stirred tank reactor (CSTR) system show that VSR-PCA method has better process monitoring performance.

Key words: fault detection, principal component analysis, process systems, dynamic simulation, variable sub-region, Bayesian inference

摘要：

针对工业过程监控中传统主元分析（PCA）方法没有突出局部变量信息的问题，提出一种基于变量子域PCA（variable sub-region PCA，VSR-PCA）的故障检测方法。首先使用PCA将原始数据空间分解成主元子空间（principal component subspace，PCS）和残差子空间（residual subspace，RS），计算变量与PCS的互信息来度量两者的相关性并以此划分变量子域。然后在变量子域中计算局部T²统计量和局部SPE统计量，并通过贝叶斯推理整合所有子域的信息构造全局统计量，使得在利用所有过程信息的同时挖掘局部变量信息。在连续搅拌反应釜系统上的仿真结果表明，VSR-PCA方法具有更好的过程监控性能。

关键词: 故障检测, 主元分析, 过程系统, 动态仿真, 变量子域, 贝叶斯推理

CLC Number:

TP277

WANG Lei, DENG Xiaogang, XU Ying, ZHONG Na. Fault detection method based on variable sub-region PCA[J]. CIESC Journal, 2016, 67(10): 4300-4308.

王磊, 邓晓刚, 徐莹, 钟娜. 基于变量子域PCA的故障检测方法[J]. 化工学报, 2016, 67(10): 4300-4308.

References 22

[1]	GE Z Q, SONG Z H, GAO F R. Review of recent research on data-based process monitoring[J]. Industrial & Engineering Chemistry Research, 2013, 52(10):3543-3562.
[2]	解翔, 侍洪波. 多稳态化工过程的全局监控策略[J]. 化工学报, 2012, 63(7):2156-2162. XIE X, SHI H B. Global monitoring strategy for multimode chemical processes[J]. CIESC Journal, 2012, 63(7):2156-2162.
[3]	钟娜, 邓晓刚, 徐莹. 基于LECA的多工况过程故障检测方法[J]. 化工学报, 2015, 66(12):4929-4940. ZHONG N, DENG X G, XU Y. Fault detection method based on LECA for multimode process[J]. CIESC Journal, 2015, 66(12):4929-4940.
[4]	KIM D, LEE I B. Process monitoring based on probabilistic PCA[J]. Chemometrics & Intelligent Laboratory Systems, 2003, 67(2):109-123.
[5]	许仙珍, 谢磊, 王树青. 基于PCA混合模型的多工况过程监控[J]. 化工学报, 2011, 62(3):743-752. XU X Z, XIE L, WANG S Q. Multi-mode process monitoring method based on PCA mixture model[J]. CIESC Journal, 2011, 62(3):743-752.
[6]	夏陆岳, 潘海天, 周猛飞, 等. 基于改进多尺度主元分析的丙烯聚合过程监测与故障诊断[J]. 化工学报, 2011, 62(8):2312-2317. XIA L Y, PAN H T, ZHOU M F, et al. Process monitoring and fault diagnosis of propylene polymerization based on improved multiscale principal component analysis[J]. CIESC Journal, 2011, 62(8):2312-2317.
[7]	JIANG Q C, YAN X F. Just-in-time reorganized PCA integrated with SVDD for chemical process monitoring[J]. AIChE Journal, 2014, 60(3):949-965.
[8]	DENG X G, TIAN X M, CHEN S. Modified kernel principal component analysis based on local structure analysis and its application to nonlinear process fault diagnosis[J]. Chemometrics & Intelligent Laboratory Systems, 2013, 127(16):195-209.
[9]	WISE B M, RICKER N L, VELTKAMP D F, et al. A theoretical basis for the use of principal component models for monitoring multivariate processes[J]. Process Control & Quality, 1990, 1(1):41-51.
[10]	KU W, STORER R H, GEORGAKIS C. Disturbance detection and isolation by dynamic principal component analysis[J]. Chemometrics & Intelligent Laboratory Systems, 1995, 30(1):179-196.
[11]	WANG X, KRUGER U, IRWIN G W, et al. Nonlinear PCA with the local approach for diesel engine fault detection and diagnosis[J]. IEEE Transactions on Control Systems Technology, 2008, 16(1):122-129.
[12]	YU J B. Local and global principal component analysis for process monitoring[J]. Journal of Process Control, 2012, 22(7):1358-1373.
[13]	CHERRY G A, QIN S J. Multiblock principal component analysis based on a combined index for semiconductor fault detection and diagnosis[J]. IEEE Transactions on Semiconductor Manufacturing, 2006, 19(2):159-172.
[14]	GE Z Q, ZHANG M G, SONG Z H. Nonlinear process monitoring based on linear subspace and Bayesian inference[J]. Journal of Process Control, 2010, 20(5):676-688.
[15]	GE Z Q, SONG Z H. Distributed PCA model for plant-wide process monitoring[J]. Industrial & Engineering Chemistry Research, 2013, 52(5):1947-1957.
[16]	WANG B, JIANG Q C, YAN X F. Fault detection and identification using a Kullback-Leibler divergence based multi-block principal component analysis and Bayesian inference[J]. Korean Journal of Chemical Engineering, 2014, 31(6):930-943.
[17]	LI Y, XIE Z, ZHOU D H. Fault detection and isolation based on abnormal sub-regions using the improved PCA[J]. Journal of Chemical Engineering of Japan, 2004, 37(4):514-522.
[18]	WANG J Y, WANG Y, ZHAO S G, et al. Maximum mutual information regularized classification[J]. Engineering Applications of Artificial Intelligence, 2015, 37(37):1-8.
[19]	KRASKOV A, STÖGBAUER H, GRASSBERGER P. Estimating mutual information[J]. Physical Review E:Statistical Nonlinear & Soft Matter Physics, 2004, 69(6):279-307.
[20]	SILVERMAN B W. Density estimation for statistics and data analysis[J]. Journal of the American Statistical Association, 1988, 83(401):600-620.
[21]	DENG X G, TIAN X M. Sparse kernel locality preserving projection and its application in nonlinear process fault detection[J]. Chinese Journal of Chemical Engineering, 2013, 21(2):163-170.
[22]	DENG X G, TIAN X M. Multimode process fault detection using local neighborhood similarity analysis[J]. Chinese Journal of Chemical Engineering, 2014, 22(11):1260-1267.