基于稀疏过滤特征学习的化工过程故障检测方法

doi:10.11949/0438-1157.20190894

化工学报 ›› 2019, Vol. 70 ›› Issue (12): 4698-4709.DOI: 10.11949/0438-1157.20190894

基于稀疏过滤特征学习的化工过程故障检测方法

江升(),旷天亮,李秀喜()

华南理工大学化学与化工学院，广东广州 510640

收稿日期:2019-08-11 修回日期:2019-08-19 出版日期:2019-12-05 发布日期:2019-12-05
通讯作者: 李秀喜
作者简介:江升（1994—），男，硕士研究生，1339625174@qq.com
基金资助:
（1）假设对于训练集X,其中<inline-formula><mml:math fxid="FX_MAT_ID80005EA6" id="M1" xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msup><mml:mrow><mml:mi mathvariant="bold-italic">X</mml:mi></mml:mrow><mml:mrow><mml:mi mathvariant="bold-italic">i</mml:mi></mml:mrow></mml:msup><mml:mtext>∈</mml:mtext><mml:mi>m</mml:mi><mml:mtext>×</mml:mtext><mml:mn mathvariant="normal">1</mml:mn></mml:math></inline-formula>是第i个样本,m为样本的变量数,训练集中一共有n个样本,那么,首先初始化权重矩阵W,假设输入的特征个数为L个,把训练样本集变换为初步特征矩阵?,通过以下非线性激活函数公式进行特征变换

A chemical process fault detection method based on sparse filtering feature learning

Sheng JIANG(),Tianliang KUANG,Xiuxi LI()

School of Chemistry and Chemical Engineering, South China University of Technology, Guangzhou 510640，Guangdong，China

Received:2019-08-11 Revised:2019-08-19 Online:2019-12-05 Published:2019-12-05
Contact: Xiuxi LI

摘要/Abstract

摘要：

过程安全一直以来是化学工业中尤为重要的问题之一，故障检测与诊断（FDD）作为化工异常工况管理最有力的工具之一，给过程安全提供了保障。随着深度学习的发展，很多智能学习算法已经被提出，然而这些算法却很少被应用到FDD中来。提出了一种基于稀疏过滤和逻辑回归（SFLR）算法的化工过程故障检测新方法。采用TE过程和环己烷无催化氧化制环己酮过程对提出的方法进行了验证，结果表明，所提出的方法均具有较高的诊断精度，案例研究表明提出的方法可以及时有效地诊断出故障。

关键词: 故障检测与诊断, 安全, 稀疏过滤, 逻辑回归, 神经网络, 过程系统

Abstract:

Process safety has always been one of the most important issues in the chemical industry, fault detection and diagnosis(FDD) is one of the most powerful tools for chemical process abnormal events management, which provides guarantee for process safety. With the development of deep learning, many intelligent learning algorithms have been proposed, but these algorithms are rarely applied to FDD. In this paper, a novel chemical process fault diagnosis method is proposed based on sparse filtering and logistic regression (SFLR) .The main idea of this method is to divide chemical process raw data into training data and test data firstly, then standardized and whitening preprocessing, and then train sparse filtering(SF) model with three layers of neural networks, and use the SF model for unsupervised feature learning. Finally, logistic regression(LR) model with supervised learning is used to classify the chemical process health conditions using learned features. The proposed method was verified by the TE process and the non-catalytic oxidation of cyclohexane to cyclohexanone. The results show that the proposed method has high diagnostic accuracy, and the case study shows that the proposed method can diagnose faults in a timely and effective manner.

Key words: fault detection and diagnosis, safety, sparse filtering, logistic regression, neural networks, process systems

中图分类号:

TQ 021.8

江升, 旷天亮, 李秀喜. 基于稀疏过滤特征学习的化工过程故障检测方法[J]. 化工学报, 2019, 70(12): 4698-4709.

Sheng JIANG, Tianliang KUANG, Xiuxi LI. A chemical process fault detection method based on sparse filtering feature learning[J]. CIESC Journal, 2019, 70(12): 4698-4709.

图/表 15

图1 稀疏过滤模型的三层神经网络结构图

Fig.1 Three-layer neural network structure diagram of sparse filtering model

图2 基于SFLR故障检测新方法的流程

Fig.2 Flow chart based on new method of SFLR fault detection

图3 田纳西-伊斯曼（TE）化工过程工艺流程图

Fig.3 Flow chart of Tennessee-Eastman (TE) chemical process

表1 TE过程的20个预先设定的故障

Table 1 20 pre-set faults in TE process

故障序号	过程变化	干扰类型
1	A/C物料进料比例扰动，B成分恒定	阶跃变化
2	B组分扰动，A/C比例恒定	阶跃变化
3	组分D进料温度扰动	阶跃变化
4	反应器冷却水入口温度	阶跃变化
5	反应器冷却水入口温度	阶跃变化
6	A组分泄漏	阶跃变化
7	组分C压力下降扰动	阶跃变化
8	A、B、C进料成分	随机变化
9	组分D进料温度扰动	随机变化
10	组分C进料温度扰动	随机变化
11	反应器冷却水入口温度	随机变化
12	反应器冷却水入口温度	随机变化
13	反应器动力性能	缓慢漂移
14	反应器冷却水调节阀	堵塞
15	反应器冷却水调节阀	堵塞
16	未知	未知
17	未知	未知
18	未知	未知
19	未知	未知
20	未知	未知

图4 故障检出率和时间与特征数量的变化关系

Fig.4 Relationship between fault detection rate and time and feature quantity

图5 误报率和时间与特征数量的变化关系

Fig.5 Relationship between false positive rate and time and feature quantity

表2 TE过程不同方法的故障检出率

Table 2 Failure detection rates for different methods of TE process/%

Fault	PCA		改进的ICA		KPCA		提出的方法(SFLR)
Fault	T²	SPE	T²	SPE	T²	SPE	提出的方法(SFLR)
1	99	100	100	100	100	100	99.75
2	98	96	98	98	98	98	98.375
3	2	1	1	1	4	8	9.875
4	6	99	65	96	9	100	99.875
5	23	20	24	23	25	27	31.375
6	99	100	100	100	99	100	100
7	42	100	100	100	100	100	100
8	97	89	97	98	95	97	95.25
9	1	1	1	2	4	4	9.75
10	30	18	70	67	43	51	90.5
11	22	72	43	66	24	81	68.875
12	97	90	98	97	97	98	98
13	93	95	95	94	94	95	95
14	81	100	100	100	79	100	99.875
15	1	2	1	2	5	7	12.5
16	13	16	76	73	30	52	56.625
17	74	93	87	94	74	95	92.125
18	89	90	90	90	90	90	90.25
19	0	29	26	29	3	49	24.625
20	32	45	70	66	41	55	77

图6 各故障监测结果

Fig.6 Monitoring result of each fault

图7 测试数据未学习和已学习前三个主元分析比较

Fig.7 Comparison of first three principal component analysis of test data unlearned and learned

图8 环己烷无催化氧化分解工段流程示意图

Fig.8 Schematic diagram of process of non-catalytic oxidation decomposition of cyclohexane

图9 故障检出率和时间与特征数量的变化关系

Fig.9 Relationship between fault detection rate and time and feature quantity

图10 误报率和时间与特征数量的变化关系

Fig.10 Relationship between false positive rate and time and feature quantity

表3 环己烷无催化氧化过程故障检测训练和测试结果

Table 3 Fault detection training and test results of cyclohexane non-catalytic oxidation process /%

测试FDR	训练FDR	测试FAR	训练FAR
80.08	77.92	5.50	4.62

图11 基于SFLR方法的环己酮生产过程监测结果

Fig.11 Monitoring results of cyclohexanone production process based on SFLR method

图12 测试数据前三个主元分析比较

Fig.12 Comparison of the first three principal components of test data

参考文献 31

1	Shu Y, Ming L, Cheng F, et al. Abnormal situation management: challenges and opportunities in the big data era[J]. Computers & Chemical Engineering, 2016, 91: 104-113.
2	Ge Z, Song Z, Gao F. Review of recent research on data-based process monitoring[J]. Industrial & Engineering Chemistry Research, 2013, 52(10): 3543-3562.
3	Venkatasubramanian V, Rengaswamy R, Yin K, et al. A review of process fault detection and diagnosis(Ⅰ): Quantitative model-based methods[J]. Computers & Chemical Engineering, 2003, 27(3): 293-311.
4	Venkatasubramanian V, Rengaswamy R, Kavuri S N, et al. A review of process fault detection and diagnosis(Ⅱ): Qualitative models and search strategies[J]. Computers & Chemical Engineering, 2003, 27(3): 313-326.
5	Venkatasubramanian V, Rengaswamy R, Kavuri S N, et al. A review of process fault detection and diagnosis(Ⅲ): Process history based methods[J]. Computers & Chemical Engineering, 2003, 27(3): 327-346.
6	Isermann R. Model-based fault-detection and diagnosis-status and applications[J]. Annual Reviews in Control, 2005, 29(1): 71-85.
7	Frank P M, Ding X. Survey of robust residual generation and evaluation methods in observer-based fault detection systems[J]. Journal of Process Control, 1997, 7(6): 403-424.
8	Chen J, Liao C. Dynamic process fault monitoring based on neural network and PCA[J]. Journal of Process Control, 2002, 12(2): 277-289.
9	Lau C, Ghosh K, Hussain M, et al. Fault detection of the Tennessee Eastman process with multi-scale PCA and ANFIS[J]. Chemometrics and Intelligent Laboratory Systems, 2013, 120(2): 1-14.
10	Lu N, Gao F, Yang Y, et al. PCA-based modeling and on-line monitoring strategy for uneven-length batch process [J]. Industrial & Engineering Chemistry Research, 2004, 43(13): 3343-3352.
11	Lu N, Gao F, Wang F. Sub-PCA modeling and on-line monitoring strategy for batch process[J]. AIChE Journal, 2004, 50(1): 255-259.
12	Albazzaz H, Wang X Z. Statistical process control charts for batch operations based on independent component analysis[J]. Industrial and Engineering Chemistry Research, 2004, 43(21): 6731-6741.
13	Lee J, Yoo C, Lee I. Statistical process monitoring with independent component analysis[J]. Journal of Process Control, 2004, 14(5): 467-485.
14	Yu J, Chen J, Rashid M M. Multiway independent component analysis mixture model and mutual information based fault detection and diagnosis approach of multiphase batch process[J]. AIChE Journal, 2013, 59(8): 2761-2779.
15	Zhang Y, Ma C. Fault diagnosis of nonlinear process using multiscale KPCA and multiscale KPLS[J]. Chemical Engineering Science, 2011, 66(1): 64-72.
16	Zhang Y, Zhou H, Qin S J, et al. Decentralized fault diagnosis of large-scale processes using multiblock kernel partial least squares[J]. IEEE Transactions on Industrial Informatics, 2010, 6(1): 3-10.
17	Rad M A A, Tazdanpanah M J. Designing supervised local neural network classifiers based on EM clustering for fault diagnosis of Tennessee Eastman Process[J]. Chemometrics and Intelligent Laboratory Systems, 2015, 146: 149-157.
18	Zhang J. Improved on-line process fault diagnosis through information fusion in multiple neural networks[J]. Computers & Engineering, 2006, 30(3): 558-571.
19	Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.
20	郭丽丽, 丁世飞. 深度学习研究进展[J]. 计算机科学, 2015, 42(5): 28-33.
	Guo L L, Ding S F. Research progress on deep learning[J]. Computer Science, 2015, 42(5): 28-33.
21	Bengio Y, Lamblin P, Dan P, et al. Greedy layer-wise training of deep networks[C]//International Conference on Neural Information Processing Systems, 2006: 153-160.
22	Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2014, 18(7): 1527-1554.
23	LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436.
24	Ngiam J, Pang W K, Chen Z, et al. Sparse filtering[C]//International Conference on Neural Information Processing Systems. Granada, Spain, 2011: 1125-1133.
25	Lei Y, Jia F, Lin J, et al. An intelligent fault diagnosis method using unsupervised feature learning towards mechanical big data[J]. IEEE Transactions on Industrial Electronics, 2016, 63(5): 3137-3147.
26	Bielza C, Robles V, Larrañaga P. Regularized logistic regression without a penalty term: an application to cancer classification with microarray data[J]. Expert Systems with Applications, 2011, 38(5): 5110-5118.
27	Oommen T, Baise L G, Vogel R M. Sampling bias and class imbalance in maximum-likelihood logistic regression[J]. Mathematical Geosciences, 2011, 43(1): 99-120.
28	Downs J J, Vogel E F. A plant-wide industrial process control problem[J]. Computers and Chemical Engineering, 1993, 17(3): 245-255.
29	Shen Y, Ding S X, Haghani A, et al. A comparison study of basic data-driven fault diagnosis and process monitoring methods on the benchmark Tennessee Eastman Process[J]. Journal of Process Control, 2012, 22(9): 1567-1581.
30	Zhang Y. Enhanced statistical analysis of nonlinear processes using KPCA, KICA and SVM[J]. Chemical Engineering Science, 2009, 64(5): 801-811.
31	Qin S J. Process data analytics in the era of big data[J]. AIChE Journal, 2014, 60(9): 3092-3100.

[1]	温凯杰, 郭力, 夏诏杰, 陈建华. 一种耦合CFD与深度学习的气固快速模拟方法[J]. 化工学报, 2023, 74(9): 3775-3785.
[2]	陈哲文, 魏俊杰, 张玉明. 超临界水煤气化耦合SOFC发电系统集成及其能量转化机制[J]. 化工学报, 2023, 74(9): 3888-3902.
[3]	李珍宝, 李超, 王虎, 王绍瑞, 黎泉. MPP抑制铝镁合金粉尘爆炸微观机理研究[J]. 化工学报, 2023, 74(8): 3608-3614.
[4]	尹刚, 李伊惠, 何飞, 曹文琦, 王民, 颜非亚, 向禹, 卢剑, 罗斌, 卢润廷. 基于KPCA和SVM的铝电解槽漏槽事故预警方法[J]. 化工学报, 2023, 74(8): 3419-3428.
[5]	郑玉圆, 葛志伟, 韩翔宇, 王亮, 陈海生. 中高温钙基材料热化学储热的研究进展与展望[J]. 化工学报, 2023, 74(8): 3171-3192.
[6]	诸程瑛, 王振雷. 基于改进深度强化学习的乙烯裂解炉操作优化[J]. 化工学报, 2023, 74(8): 3429-3437.
[7]	闫琳琦, 王振雷. 基于STA-BiLSTM-LightGBM组合模型的多步预测软测量建模[J]. 化工学报, 2023, 74(8): 3407-3418.
[8]	徐野, 黄文君, 米俊芃, 申川川, 金建祥. 多源信息融合的离心式压缩机喘振诊断方法[J]. 化工学报, 2023, 74(7): 2979-2987.
[9]	李贵贤, 曹阿波, 孟文亮, 王东亮, 杨勇, 周怀荣. 耦合固体氧化物电解槽的CO₂制甲醇过程设计与评价研究[J]. 化工学报, 2023, 74(7): 2999-3009.
[10]	刘晓洋, 喻健良, 侯玉洁, 闫兴清, 张振华, 吕先舒. 螺旋微通道对掺氢甲烷爆轰传播的影响[J]. 化工学报, 2023, 74(7): 3139-3148.
[11]	邵远哲, 赵忠盖, 刘飞. 基于共同趋势模型的非平稳过程质量相关故障检测方法[J]. 化工学报, 2023, 74(6): 2522-2537.
[12]	高学金, 姚玉卓, 韩华云, 齐咏生. 基于注意力动态卷积自编码器的发酵过程故障监测[J]. 化工学报, 2023, 74(6): 2503-2521.
[13]	黄磊, 孔令学, 白进, 李怀柱, 郭振兴, 白宗庆, 李平, 李文. 油页岩添加对准东高钠煤灰熔融行为影响的研究[J]. 化工学报, 2023, 74(5): 2123-2135.
[14]	肖忠良, 尹碧露, 宋刘斌, 匡尹杰, 赵亭亭, 刘成, 袁荣耀. 废旧锂离子电池回收工艺研究进展及其安全风险分析[J]. 化工学报, 2023, 74(4): 1446-1456.
[15]	贠程, 王倩琳, 陈锋, 张鑫, 窦站, 颜廷俊. 基于社团结构的化工过程风险演化路径深度挖掘[J]. 化工学报, 2023, 74(4): 1639-1650.

基于稀疏过滤特征学习的化工过程故障检测方法

A chemical process fault detection method based on sparse filtering feature learning

RichHTML

PDF (PC)

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 31

相关文章 15

编辑推荐

Metrics

本文评价