Fault detection for chemical process based on memory-weighted kernel principal component analysis

doi:10.16265/j.cnki.issn1003-3033.2026.04.1571

Abstract

Abstract:

To enhance the detection ability of common faults in the chemical process and ensure the stable and reliable operation, a new fault detection method, termed M-WKPCA, was proposed. The method was based on historical fault data and KPCA, incorporating historical fault data through a weighting mechanism. Initially, kernel principal components(KPC) of normal data and historical fault data were calculated according to KPCA. KPC that can highlight the fault information were selected based on a comparison of the constructed indexes. A weighting matrix was then used to highlight the fault information, and a new statistic was constructed to establish an online fault detection model based on M-WKPCA method. Then, the M-WKPCA method was used to detect common faults in parallel. A parallel WKPCA fault detection strategy was proposed to achieve high precision detection of common faults. Finally, the proposed method was verified using simulated data from the Tennessee Eastman (TE) chemical process. The results show that the proposed method achieves an average accurate detection accuracy of 82.25%. This is much higher than that of the comparison methods, demonstrating its superiority in fault detection. At the same time, since fault information is incorporated during the selection of KPC, the detected fault data are significantly different from the normal data in terms of statistics.

Key words: memory-weighted kernel principal component analysis(M-WKPCA), kernel principal component analysis(KPCA), chemical process, kernel density estimation(KDE), fault detection

CLC Number:

X937

Yuan Shuai, Yang Chunxi, Zhang Xiufeng, Wang Xian, Li Gengen. Fault detection for chemical process based on memory-weighted kernel principal component analysis[J]. China Safety Science Journal, 2026, 36(4): 168-175.

Figures/Tables 5

Fig.1

Fig.2

Fig.3

Fig.4

Table 1

References 21

[1]	田宇. 基于优化动态核主元分析的复杂工业过程故障诊断研究[D]. 天津: 天津理工大学, 2020.
	Tian Yu. Research on fault diagnosis of complex industrial process based on optimized dynamic kernel principal analysis[D]. Tianjin: Tianjin University of Technology, 2020.
[2]	杨余, 杨鑫, 王英, 等. 基于mini-1D-CNN模型的TE过程故障诊断[J]. 中国安全科学学报, 2023, 33(2): 173-178. doi: 10.16265/j.cnki.issn1003-3033.2023.02.0017
	Yang Yu, Yang Xin, Wang Ying, et al. Research on TE process fault diagnosis based on mini-1D-CNN model[J]. China Safety Science Journal, 2023, 33(2):173-178. doi: 10.16265/j.cnki.issn1003-3033.2023.02.0017
[3]	王斌, 施祖建, 匡蕾. 基于改进KFDA和RW v-SVM的化工生产系统故障快速诊断[J]. 中国安全科学学报, 2013, 23(8): 84-89.
	Wang Bin, Shi Zujian, Kuang Lei. A fast method for diagnosing fault in chemical production system based on improved KFDA and RW v-SVM[J]. China Safety Science Journal, 2013, 23(8): 84-89.
[4]	OdendAAL H M, Jones T. Actuator fault detection and isolation: an optimised parity space approach[J]. Control Engineering Practice, 2014, 26: 222-232. doi: 10.1016/j.conengprac.2014.01.013
[5]	Wang Zun, Han Yongming, Li Chengfei, et al. Input-output networks considering graphlet-based analysis for production optimization: application in ethylene plants[J]. Journal of Cleaner Production, 2021, 278: 1-8.
[6]	张成, 高宪文, 李元. 基于k近邻主元得分差分的故障检测策略[J]. 自动化学报, 2020, 46(10): 2229-2238.
	Zhang Cheng, Gao Xianwen, Li Yuan. Fault detection strategy based on principal component score difference of k nearest neighbors[J]. Acta Automatica Sinica, 2020, 46(10): 2229-2238.
[7]	Ku Wenfu, Storer R H, Georgakis C. Disturbance detection and isolation by dynamic principal component analysis[J]. Chemometrics and Intelligent Laboratory Systems, 1995, 30(1): 179-196. doi: 10.1016/0169-7439(95)00076-3
[8]	Lee J, Yoo C K, Choi S W, et al. Nonlinear process monitoring using kernel principal component analysis[J]. Chemical Engineering Science, 2004, 59(1): 223-234. doi: 10.1016/j.ces.2003.09.012
[9]	Jong M, Qin S J, Lee I B. Fault detection and diagnosis based on modified independent component analysis[J]. Aiche Journal, 2006, 52(10): 3501-3514. doi: 10.1002/aic.v52:10
[10]	柴天佑. 工业过程控制系统研究现状与发展方向[J]. 中国科学:信息科学, 2016, 46(8): 1003-1015.
	Chai Tianyou. Industrial process control systems: research status and development direction[J]. Scientia Sinica: Informationis, 2016, 46(8): 1003-1015. doi: 10.1360/N112016-00062
[11]	Zhao Chunhui, Gao Furong. Fault-relevant principal component analysis (FPCA) method for multivariate statistical modeling and process monitoring[J]. Chemometrics and Intelligent Laboratory Systems, 2014, 133: 1-16. doi: 10.1016/j.chemolab.2014.01.009
[12]	Jiang Qingchao, Yan Xuefeng. Chemical processes monitoring based on weighted principal component analysis and its application[J]. Chemometrics and Intelligent Laboratory Systems, 2012, 119: 11-20. doi: 10.1016/j.chemolab.2012.09.002
[13]	Jiang Qingchao, Yan Xuefeng. Weighted kernel principal component analysis based on probability density estimation and moving window and its application in nonlinear chemical process monitoring[J]. Chemometrics and Intelligent Laboratory Systems, 2013, 127: 121-131. doi: 10.1016/j.chemolab.2013.06.013
[14]	Tao Yang, Shi Hongbo, Song Bing, et al. A novel dynamic weight principal component analysis method and hierarchical monitoring strategy for process fault detection and diagnosis[J]. IEEE Transactions on Industrial Electronics, 2019, 67(9): 7994-8004. doi: 10.1109/TIE.41
[15]	Chen Q, Wynne R J, Goulding P, et al. The application of principal component analysis and kernel density estimation to enhance process monitoring[J]. Control Engineering Practice, 2000, 8(5): 531-543. doi: 10.1016/S0967-0661(99)00191-4
[16]	Cheng Chunyuan, Hsu C, Chen Muchen. Adaptive kernel principal component analysis (KPCA) for monitoring small disturbances of nonlinear processes[J]. Industrial & Engineering Chemistry Research, 2010, 49(5): 2254-2262. doi: 10.1021/ie900521b
[17]	Guo Lingling, Wu Ping, Lou Siwei, et al. A multi-feature extraction technique based on principal component analysis for nonlinear dynamic process monitoring[J]. Journal of Process Control, 2020, 85: 159-172. doi: 10.1016/j.jprocont.2019.11.010
[18]	Bathelt a, Ricker N, Jelali M. Revision of the Tennessee Eastman process model[J]. IFAC-Papers on Line, 2015, 48(8): 309-314.
[19]	林扬, 何亚东, 袁壮, 等. 基于PCA-SVDD的化工过程异常工况检测[J]. 过程工程学报, 2022, 22(7): 970-978. doi: 10.12034/j.issn.1009-606X.221399
	Lin Yang, He Yadong, Yuan Zhuang, et al. Abnormal condition detection in chemical process based on PCA-SVDD[J]. The Chinese Journal of Process Engineering, 2022, 22(7): 970-978. doi: 10.12034/j.issn.1009-606X.221399
[20]	Wang Jin, He Q P. Multivariate statistical process monitoring based on statistics pattern analysis[J]. Industrial & Engineering Chemistry Research, 2010, 49(17): 7858-7869. doi: 10.1021/ie901911p
[21]	Samuel R T, Cao Yi. Nonlinear process fault detection and identification using kernel PCA and kernel density estimation[J]. Systems Science & Control Engineering, 2016, 4(1): 165-174.

故障序号	PCA	KPCA	DWPCA	M-WKPCA	P-WKPCA
1	0.992 7	0.993 8	0.994 8	0.996 9	0.999 0
2	0.985 4	0.984 4	0.977 0	0.986 5	0.987 5
3	0.212 5	0.245 8	0.299 6	0.249 0	0.249 0
4	0.760 4	0.852 1	0.991 6	0.997 9	0.997 9
5	0.396 9	0.431 3	0.990 6	0.996 9	0.999 0
6	0.994 8	0.994 8	0.996 9	0.999 0	1
7	0.996 9	0.994 8	0.764 0	1	1
8	0.980 2	0.977 1	0.802 7	0.989 6	0.991 7
9	0.206 3	0.214 8	0.190 0	0.191 7	0.230 2
10	0.530 2	0.581 3	0.793 3	0.868 8	0.868 8
11	0.672 9	0.721 9	0.833 0	0.812 5	0.829 2
12	0.986 5	0.980 2	0.986 4	0.981 3	0.995 8
13	0.954 2	0.955 2	0.838 2	0.962 5	0.962 5
14	0.996 9	0.994 8	0.999 0	0.999 0	0.999 0
15	0.226 0	0.266 7	0.295 4	0.255 2	0.309 4
16	0.386 5	0.455 2	0.788 1	0.909 4	0.909 4
17	0.865 6	0.880 2	0.913 3	0.966 7	0.969 8
18	0.911 5	0.910 4	0.913 3	0.919 8	0.924 0
19	0.307 3	0.388 5	0.181 6	0.750 0	0.766 7
20	0.587 5	0.641 7	0.762 0	0.815 6	0.816 7
21	0.526 0	0.531 3	0.422 8	0.625 0	0.685 4
平均值	0.689 4	0.714 1	0.749 2	0.822 5	0.832 9