基于社团结构的化工过程风险演化路径深度挖掘

doi:10.11949/0438-1157.20221618

摘要/Abstract

摘要：

在化工生产过程中，由于物料的危险性和流程的复杂性，导致恶性事故频发。为了保证化工过程的安全平稳运行，需对其风险演化机制进行深入解析，以抑制次衍生灾害事故的传播、扩散及演变。然而，传统方法过于依赖专家经验或先验信息，风险评估结果不精准；以复杂网络为基础的社团结构虽可视为风险演化路径的高度抽象，但已有算法难以兼顾划分结果的合理性和准确率。为此，提出一种基于DSAE-Louvain社团结构的化工过程风险演化路径深度挖掘方法。首先对多源过程数据进行处理，构建风险演化网络模型，同时利用Dijkstra算法、跳数法等手段，获取相似度矩阵；进而引入深度稀疏自编码器（DSAE）与Louvain算法，经稀疏处理开展社团结构划分；最后，根据节点重要度排序追踪整个化工过程的风险薄弱节点和关键演化路径。以Tennessee Eastman （TE）过程为例，对比GN算法、Louvain算法和DSAE-GN方法，结果表明DSAE-Louvain方法能够提升社团结构划分的精细化、高效化程度，且所挖掘的风险演化路径更为符合实际生产工艺流程。

关键词: 化工过程, 风险演化, 路径挖掘, 过程系统, 神经网络, 安全

Abstract:

In the process of chemical production, due to the danger of materials and the complexity of the process, serious accidents occur frequently. It is necessary to reveal risk evolution mechanism to ensure the safe and stable operation of chemical processes, as well as to inhibit the spread, diffusion, or even evolution of secondary and derivative disasters. However, many traditional methods heavily rely on expert experience or prior information, which will cause the inaccuracy of risk assessment results. Although the community structure in complex network can be regarded as a highly abstract of risk evolution path, the existing algorithms cannot consider both of the rationality and accuracy division results. Therefore, a deep mining method of risk evolution path is proposed for chemical processes based on the DSAE-Louvain community structure. First, multi-source process data should be processed to establish a risk evolution network, and the Dijkstra and hop-count algorithms are further applied to obtain a similarity matrix. Then, the deep sparse autoencoder (DSAE) and Louvain algorithm are combined to divide the community structure using sparse analysis. Finally, the risk weak nodes and critical evolution paths in whole chemical processes are deeply mined according to the ranking of node importance. To illustrate its validity, the Tennessee Eastman (TE) process is selected as a test case. The results show that the proposed DSAE-Louvain method are more refined and high-efficiency in the community structure division by comparing with the GN algorithm, Louvain algorithm, and DSAE-GN method. Particularly, the mined risk evolution paths are more in accord with actual production processes.

Key words: chemical processes, risk evolution, path mining, process systems, neural networks, safety

中图分类号:

X 795

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.

Cheng YUN, Qianlin WANG, Feng CHEN, Xin ZHANG, Zhan DOU, Tingjun YAN. Deep-mining risk evolution path of chemical processes based on community structure[J]. CIESC Journal, 2023, 74(4): 1639-1650.

图/表 17

图1 深度稀疏自编码器（DSAE）结构

Fig.1 The structure of deep sparse autoencoder (DSAE)

图2 技术路线图

Fig.2 The flowchart of the proposed method

表1 某多源过程数据

Table 1 An example of multi-source process data

时刻	$a 1$	$a 2$	$a 3$	$a 4$
1	1.0	0.2	2.5	1.2
2	1.2	0.4	2.5	1.2
3	0.7	2.7	1.0	2.5
4	0.8	0.5	2.9	1.7
5	2.7	0.3	1.0	2.0
6	0.6	0.3	2.4	2.5
7	0.4	0.6	2.4	2.5

表1 某多源过程数据

Table 1 An example of multi-source process data

时刻	$a 1$	$a 2$	$a 3$	$a 4$
1	1.0	0.2	2.5	1.2
2	1.2	0.4	2.5	1.2
3	0.7	2.7	1.0	2.5
4	0.8	0.5	2.9	1.7
5	2.7	0.3	1.0	2.0
6	0.6	0.3	2.4	2.5
7	0.4	0.6	2.4	2.5

图3 某风险演化网络

Fig.3 An example of risk evolution network

表2 最短路径计算过程

Table 2 An example of calculation process for the shortest paths

复杂网络	邻接矩阵	最短路径结果
	$0101000101010100$	$0011001100010010$

表2 最短路径计算过程

Table 2 An example of calculation process for the shortest paths

复杂网络	邻接矩阵	最短路径结果
	$0101000101010100$	$0011001100010010$

图4 TE过程的工艺流程图

Fig.4 The P&ID of the TE process

表3 TE过程的测量变量（22个过程变量）

Table 3 Measured variables of the TE process （22 process variables）

变量	说明	变量	说明
XMEAS(1)	A进料流量值	XMEAS(12)	产品分离器液位值
XMEAS(2)	D进料流量值	XMEAS(13)	产品分离器压力值
XMEAS(3)	E进料流量值	XMEAS(14)	产品分离器下部出料值
XMEAS(4)	A和C进料流量值	XMEAS(15)	汽提塔液位值
XMEAS(5)	循环流量值	XMEAS(16)	汽提塔压力值
XMEAS(6)	反应器进料流量值	XMEAS(17)	汽提塔下部出料值
XMEAS(7)	反应器压力值	XMEAS(18)	汽提塔温度值
XMEAS(8)	反应器液位值	XMEAS(19)	汽提塔蒸汽流量值
XMEAS(9)	反应器温度值	XMEAS(20)	压缩机功率值
XMEAS(10)	放空流量值	XMEAS(21)	反应器冷却水出口温度值
XMEAS(11)	产品分离器温度值	XMEAS(22)	冷凝器冷却水出口温度值

表4 TE过程的邻接矩阵

Table 4 Adjacency matrix of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	0.571		1	0
XMEAS(10)	0	0		0.800	0		0.800	0
$⋮$			···			···
XMEAS(21)	0	0		1	0.571		0.857	0
XMEAS(22)	0	0		0	0		0	0

表4 TE过程的邻接矩阵

Table 4 Adjacency matrix of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	0.571		1	0
XMEAS(10)	0	0		0.800	0		0.800	0
$⋮$			···			···
XMEAS(21)	0	0		1	0.571		0.857	0
XMEAS(22)	0	0		0	0		0	0

表5 TE过程的最短路径结果

Table 5 Results from the shortest paths of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	1		1	0
XMEAS(10)	0	0		1	0		1	0
$⋮$			···			···
XMEAS(21)	0	0		1	1		0	0
XMEAS(22)	0	0		0	0		0	0

表5 TE过程的最短路径结果

Table 5 Results from the shortest paths of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	1		1	0
XMEAS(10)	0	0		1	0		1	0
$⋮$			···			···
XMEAS(21)	0	0		1	1		0	0
XMEAS(22)	0	0		0	0		0	0

表6 TE过程的相似度矩阵

Table 6 Similarity matrix of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	0		1	0
XMEAS(10)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(21)	0	0		1	0		0	0
XMEAS(22)	0	0		0	0		0	0

表6 TE过程的相似度矩阵

Table 6 Similarity matrix of the TE process

Item	XMEAS(1)	XMEAS(2)	···	XMEAS(9)	XMEAS(10)	···	XMEAS(21)	XMEAS(22)
XMEAS(1)	0	0		0	0		0	0
XMEAS(2)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(9)	0	0		0	0		1	0
XMEAS(10)	0	0		0	0		0	0
$⋮$			···			···
XMEAS(21)	0	0		1	0		0	0
XMEAS(22)	0	0		0	0		0	0

表7 DSAE模型的最优参数

Table 7 Optimal parameters of the DSAE model

参数	最优值
各层节点数	256-64-16-64-256
训练批次	4
迭代次数	150
惩罚因子β	1×10^-6
损失	0.691

图5 TE过程的社团结构（DSAE-Louvain）

Fig.5 Community structures of the TE process (DSAE-Louvain)

表8 各社团结构所对应的风险演化网络节点

Table 8 The node corresponding to every community structure in risk evolution network

序号	网络节点
社团Ⅰ	XMEAS(4)—A和C组分进料流量、XMEAS(10)—放空流量、XMEAS(16)—汽提塔压力、XMEAS(22)—冷凝器冷却水出口温度
社团Ⅱ	XMEAS(2)—D组分进料流量、XMEAS(3)—E组分进料流量、XMEAS(7)—反应器压力、XMEAS(17)—汽提塔下部出料、XMEAS(18)—汽提塔温度
社团Ⅲ	XMEAS(5)—循环流量、XMEAS(6)—反应器进料流量、XMEAS(8)—反应器液位、XMEAS(12)—产品分离器液位、XMEAS(14)—产品分离器下部出料、XMEAS(19)—汽提塔蒸汽流量、XMEAS(20)—压缩机功率
社团Ⅳ	XMEAS(1)—A组分进料流量、XMEAS(9)—反应器温度、XMEAS(11)—产品分离器温度、XMEAS(13)—产品分离器压力、XMEAS(15)—汽提塔液位、XMEAS(21)—反应器冷却水出口温度

表9 TE过程的节点重要度排序

Table 9 Node importance ranking of the TE process

序号	重要度值	序号	重要度值
5	4.538	18	2.443
15	4.538	8	2.153
21	4.538	10	1.827
17	3.921	22	1.601
20	3.921	4	1.553
1	3.591	2	0
13	3.591	3	0
7	2.890	9	0
11	2.890	12	0
14	2.890	16	0
6	2.786	19	0

图6 TE过程的风险演化路径（DSAE-Louvain）

Fig.6 Risk evolution paths of the TE process (DSAE-Louvain)

图7 不同模型的模块度Q值对比

Fig.7 Comparison of the modularity Q in different models

图8 TE过程的社团结构（DSAE-GN）

Fig.8 Community structures of the TE process (DSAE-GN)

参考文献 38

25	Lima-Mendez G, Faust K, Henry N, et al. Ocean plankton. Determinants of community structure in the global plankton interactome[J]. Science, 2015, 348(6237): 1262073.
26	Newman M E J. Detecting community structure in networks[J]. The European Physical Journal B-Condensed Matter, 2004, 38(2): 321-330.
27	高庆一, 李牧. 基于GN算法的重叠社区识别方法[J]. 华中科技大学学报(自然科学版), 2015, 43(9): 13-18.
	Gao Q Y, Li M. Method of identifying overlapping communities based on GN algorithm[J]. Journal of Huazhong University of Science and Technology (Natural Science Edition), 2015, 43(9): 13-18.
28	Zhang Z Q, Pu P, Han D D, et al. Self-adaptive Louvain algorithm: fast and stable community detection algorithm based on the principle of small probability event[J]. Physica A: Statistical Mechanics and its Applications, 2018, 506: 975-986.
29	尚敬文, 王朝坤, 辛欣, 等. 基于深度稀疏自动编码器的社区发现算法[J]. 软件学报, 2017, 28(3): 648-662.
	Shang J W, Wang C K, Xin X, et al. Community detection algorithm based on deep sparse autoencoder[J]. Journal of Software, 2017, 28(3): 648-662.
30	李晓佳, 张鹏, 狄增如, 等. 复杂网络中的社团结构[J]. 复杂系统与复杂性科学, 2008, 5(3): 19-42.
	Li X J, Zhang P, Di Z R, et al. Community structure in complex networks[J]. Complex Systems and Complexity Science, 2008, 5(3): 19-42.
31	来杰, 王晓丹, 向前, 等. 自编码器及其应用综述[J]. 通信学报, 2021, 42(9): 218-230.
	Lai J, Wang X D, Xiang Q, et al. Review on autoencoder and its application[J]. Journal on Communications, 2021, 42(9): 218-230.
32	Bayer F M, Kozakevicius A J, Cintra R J. An iterative wavelet threshold for signal denoising[J]. Signal Processing, 2019, 162: 10-20.
1	Chen C, Reniers G. Chemical industry in China: the current status, safety problems, and pathways for future sustainable development[J]. Safety Science, 2020, 128: 104741.
2	张圣柱, 韩玉鑫, 曹旭, 等. 我国危险化学品产业转移及安全风险分析[J]. 中国安全生产科学技术, 2022, 18(11): 46-52.
	Zhang S Z, Han Y X, Cao X, et al. Analysis on industrial transfer of hazardous chemicals and its safety risk in China[J]. Journal of Safety Science and Technology, 2022, 18(11): 46-52.
3	王德亮, 周志茂, 林梦蕾, 等. 中国炼油转型化工现状及发展约束因素的思考[J]. 化工进展, 2021, 40(10): 5854-5860.
	Wang D L, Zhou Z M, Lin M L, et al. Status and thinking of development constraints of refining to chemical transformation in China[J]. Chemical Industry and Engineering Progress, 2021, 40(10): 5854-5860.
4	王子宗, 索寒生, 赵学良. 数字孪生智能乙烯工厂研究与构建[J]. 化工学报, 2023, 74(3): 1175-1186.
	Wang Z Z, Suo H S, Zhao X L. Research and construction of digital twin intelligent ethylene plant[J]. CIESC Journal, 2023, 74(3): 1175-1186.
5	王倩琳, 田文慧, 张东胜, 等. 基于FRAM的化工装置事故情景推演研究[J]. 过程工程学报, 2022, 22(6): 782-791.
	Wang Q L, Tian W H, Zhang D S, et al. Scenario deduction on chemical plant accidents using FRAM[J]. The Chinese Journal of Process Engineering, 2022, 22(6): 782-791.
6	刘庆龙, 曲秋影, 赵东风, 等. 基于多源异构数据融合的化工安全风险动态量化评估方法[J]. 化工学报, 2021, 72(3): 1769-1777.
	Liu Q L, Qu Q Y, Zhao D F, et al. Dynamic quantitative assessment method of chemical safety risk based on multi-source heterogeneous data fusion[J]. CIESC Journal, 2021, 72(3): 1769-1777.
7	Motalifu M, Tian Y, Liu Y, et al. Chemical process safety education in China: an overview and the way forward[J]. Safety Science, 2022, 148: 105643.
8	Wang H Z, Chen B Z, He X R, et al. SDG-based HAZOP analysis of operating mistakes for PVC process[J]. Process Safety and Environmental Protection, 2009, 87(1): 40-46.
9	Luo T Y, Wu C, Duan L X. Fishbone diagram and risk matrix analysis method and its application in safety assessment of natural gas spherical tank[J]. Journal of Cleaner Production, 2018, 174: 296-304.
10	Taleb-Berrouane M, Khan F, Hawboldt K. Corrosion risk assessment using adaptive bow-tie (ABT) analysis[J]. Reliability Engineering & System Safety, 2021, 214: 107731.
11	Song M C, Lind M, Yang J, et al. Integrative decision support for accident emergency response by combining MFM and Go-Flow[J]. Process Safety and Environmental Protection, 2021, 155: 131-144.
12	Li X H, Liu Y Z, Abbassi R, et al. A Copula-Bayesian approach for risk assessment of decommissioning operation of aging subsea pipelines[J]. Process Safety and Environmental Protection, 2022, 167: 412-422.
13	李沛洁, 杨博, 李宏光. 基于关联规则的条件状态模糊Petri网及其在故障诊断中的应用[J]. 化工学报, 2018, 69(8): 3517-3527.
	Li P J, Yang B, Li H G. Association rules based conditional state fuzzy Petri nets with applications in fault diagnosis[J]. CIESC Journal, 2018, 69(8): 3517-3527.
14	Bauer M, Thornhill N F. A practical method for identifying the propagation path of plant-wide disturbances[J]. Journal of Process Control, 2008, 18(7/8): 707-719.
15	Amin M T. An integrated methodology for fault detection, root cause diagnosis, and propagation pathway analysis in chemical process systems[J]. Cleaner Engineering and Technology, 2021, 4: 100187.
16	苏婉君, 周迎, 周诚. 地铁深基坑施工风险时空演化及控制[J]. 土木工程与管理学报, 2017, 34(6): 133-140.
	Su W J, Zhou Y, Zhou C. Risk evolution and control in metro foundation constructions[J]. Journal of Civil Engineering and Management, 2017, 34(6): 133-140.
17	罗文慧, 蔡凤田, 吴初娜, 等. 基于文本挖掘的道路运输安全风险源辨识模型[J]. 西南交通大学学报, 2021, 56(1): 147-152.
	Luo W H, Cai F T, Wu C N, et al. Text-mining based risk source identification model for transportation safety[J]. Journal of Southwest Jiaotong University, 2021, 56(1): 147-152.
18	Xu K K, Hu J Q, Zhang L B, et al. A risk factor tracing method for LNG receiving terminals based on GAT and a bidirectional LSTM network[J]. Process Safety and Environmental Protection, 2023, 170: 694-708.
19	Watts D J, Strogatz S H. Collective dynamics of ‘small-world’ networks[J]. Nature, 1998, 393(6684): 440-442.
20	Barabasi A L, Albert R. Emergence of scaling in random networks[J]. Science, 1999, 286(5439): 509-512.
21	汪小帆, 李翔, 陈关荣. 复杂网络理论及其应用[M]. 北京: 清华大学出版社, 2006.
	Wang X F, Li X, Chen G R. Theory and Application of Complex Networks[M]. Beijing: Tsinghua University Press, 2006.
22	Abedi A, Gaudard L, Romerio F. Review of major approaches to analyze vulnerability in power system[J]. Reliability Engineering & System Safety, 2019, 183: 153-172.
23	Cats O, Koppenol G J, Warnier M. Robustness assessment of link capacity reduction for complex networks: application for public transport systems[J]. Reliability Engineering & System Safety, 2017, 167: 544-553.
24	Wei D J, Zhang X G, Mahadevan S. Measuring the vulnerability of community structure in complex networks[J]. Reliability Engineering & System Safety, 2018, 174: 41-52.
33	Yu Y, Zhu D, Wang J D, et al. Abnormal data detection for multivariate alarm systems based on correlation directions[J]. Journal of Loss Prevention in the Process Industries, 2017, 45: 43-55.
34	Koramati S, Bandhu Majumdar B, Pani A, et al. A registry-based investigation of road traffic fatality risk factors using police data: a case study of Hyderabad, India[J]. Safety Science, 2022, 153: 105805.
35	Wang J Y, Yu X Y, Zong R W, et al. Evacuation route optimization under real-time toxic gas dispersion through CFD simulation and Dijkstra algorithm[J]. Journal of Loss Prevention in the Process Industries, 2022, 76: 104733.
36	张军祥, 李书琴, 刘斌. 基于平滑l₁范数的深度稀疏自动编码器社区识别算法[J]. 计算机应用研究, 2020, 37(4): 1063-1068.
	Zhang J X, Li S Q, Liu B. Sparse autoencoder community recognition algorithm based on smoothed l₁ norm[J]. Application Research of Computers, 2020, 37(4): 1063-1068.
37	陈雨, 韩永明, 王尊, 等. 基于数据复杂网络理论的系统故障检测方法[J]. 化工学报, 2014, 65(11): 4503-4508.
	Chen Y, Han Y M, Wang Z, et al. System fault detection based on data-driven and complex networks theory[J]. CIESC Journal, 2014, 65(11): 4503-4508.
38	Liu N, Wang J, Sun S L, et al. Optimized principal component analysis and multi-state Bayesian network integrated method for chemical process monitoring and variable state prediction[J]. Chemical Engineering Journal, 2022, 430: 132617.

[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]	邵伟明, 韩文学, 宋伟, 杨勇, 陈灿, 赵东亚. 基于分布式贝叶斯隐马尔可夫回归的动态软测量建模方法[J]. 化工学报, 2023, 74(6): 2495-2502.
[12]	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.
[13]	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.
[14]	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.
[15]	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.