Contrastive learning based on method for identifying operating conditions of ethylene cracking furnace

doi:10.11949/0438-1157.20241291

Abstract

Abstract:

Ethylene cracking furnace is the core device of ethylene production, and the complex high-temperature cracking reaction of hydrocarbon raw materials occurs in the cracking furnace. It is very important to timely identify the changes in the operating conditions of the cracking furnace for the safe and efficient operation of the equipment. A large number of process data generated during the operation of ethylene cracking furnace are usually multi-variable and high-dimensional, which increases the complexity of data processing and analysis. How to timely detect the change of ethylene cracking furnace operating conditions based on process data has become an urgent problem to be solved. Referring to the excellent performance of contrastive learning algorithm in image classification, this paper proposes a method of cracking furnace operating condition recognition based on contrastive learning. Firstly, the industrial data of ethylene cracking furnace were normalized, and the data were dynamically extracted using time windows of different lengths and converted into grayscale images. According to the information in the image, the image is data-enhanced and then input into the encoder to obtain the global semantics, category, content invariance and other features of the image. These features are applied to calculate the loss function of contrastive learning, and the classification of gray images is realized by minimizing the contrastive loss function. Through the method in this paper, the changes of working conditions can be quickly found according to the process data. Its classification accuracy is significantly improved compared with the self-supervised contrastive learning for universal time series representation learning (TimesURL) method, which can effectively realize the identification of ethylene cracking furnace working conditions.

Key words: ethylene cracking furnace, safety, unsupervised learning, contrastive learning, algorithm, neural networks, data visualization, working condition identification

摘要：

乙烯裂解炉是乙烯生产的核心装置，烃类原料在裂解炉中发生复杂的高温裂解反应，及时识别裂解炉运行工况变化对设备安全高效运行非常重要。裂解炉运行过程中产生大量的过程数据，这些数据通常具有多变量、高维度特性，增加了数据处理和分析的复杂性，如何基于过程数据及时检测乙烯裂解炉工况变化成为亟需解决的问题。借鉴对比学习算法在图片分类中的优秀性能，提出一类基于对比学习的裂解炉运行工况识别方法。首先，将乙烯裂解炉工业数据经归一化后，使用不同长度的时间窗动态提取数据，将其转化为灰度图片。根据图片中的信息，将图片进行数据增强后输入编码器，得到图片的全局语义、类别、内容不变性等特征。将这些特征应用于计算对比学习的损失函数，通过最小化对比损失函数，实现对灰度图片的分类。通过本文方法，可以根据过程数据快速发现工况变化，其分类准确度较通用时间序列表示学习的自监督对比学习（self-supervised contrastive learning for universal time series representation learning，TimesURL）方法有明显提升，可有效实现乙烯裂解炉工况识别。

关键词: 乙烯裂解炉, 安全, 无监督学习, 对比学习, 算法, 神经网络, 数据图像化, 工况识别

CLC Number:

TP 274

Yulun WU, Zhenlei WANG, Xin WANG. Contrastive learning based on method for identifying operating conditions of ethylene cracking furnace[J]. CIESC Journal, 2025, 76(6): 2733-2742.

吴与伦, 王振雷, 王昕. 基于对比学习的乙烯裂解炉运行工况识别方法[J]. 化工学报, 2025, 76(6): 2733-2742.

Figures/Tables 13

References 38

[1]	Deng J L, Chen G R. Research on online fault diagnosis model based on IoT in process industry[J]. Advanced Materials Research, 2013, 765/766/767: 2089-2092.
[2]	刘宇航. 基于主成分分析的故障监测方法及其应用研究[D]. 上海: 华东理工大学, 2012.
	Liu Y H. Research on fault monitoring method based on principal component analysis and its application[D]. Shanghai: East China University of Science and Technology, 2012.
[3]	常亚娜, 武锦涛, 代玉强. 数据分析与BP神经网络相结合的乙烯装置智能故障诊断系统[J]. 石油炼制与化工, 2023, 54(6): 97-104.
	Chang Y N, Wu J T, Dai Y Q. Research on intelligent fault diagnosis system of ethylene plant based on data analysis and bp neural network[J]. Petroleum Processing and Petrochemicals, 2023, 54(6): 97-104.
[4]	吴家葆, 曾国辉, 张振华, 等. 基于K-means分层聚类的TCN-GRU和LSTM动态组合光伏短期功率预测[J]. 可再生能源, 2023, 41(8): 1015-1022.
	Wu J B, Zeng G H, Zhang Z H, et al. Dynamic combination of TCN-GRU and LSTM photovoltaic short-term power prediction based on K-means hierarchical clustering[J]. Renewable Energy Resources, 2023, 41(8): 1015-1022.
[5]	刘星宇, 宋韶旭, 黄向东, 等. Apache TsFile中的短时间序列分组压缩及合并方法[J]. 软件学报, 2025, 36(3): 941-961.
	Liu X Y, Song S X, Huang X D, et al. Short time series group compression and merging method in Apache TsFile[J]. Journal of Software, 2025, 36(3): 941-961.
[6]	程翔. 基于机器学习的滚动轴承故障诊断研究[D]. 淮南: 安徽理工大学, 2024.
	Cheng X. Research on fault diagnosis of rolling bearing based on machine learning[D]. Huainan: Anhui University of Science & Technology, 2024.
[7]	徐娇. 基于时间序列的高精度外测数据预处理方法研究[D]. 西安: 西安理工大学, 2022.
	Xu J. Research on high precision external measurement data preprocessing method based on time series[D]. Xi'an: Xi'an University of Technology, 2022.
[8]	李娟, 赵友刚, 于洋, 等. 含大时滞和噪声的网络化控制系统的最优故障诊断[J]. 自动化学报, 2012, 38(5): 858-864.
	Li J, Zhao Y G, Yu Y, et al. Optimal fault diagnosis for networked control systems with large time-delays and noises[J]. Acta Automatica Sinica, 2012, 38(5): 858-864.
[9]	赵凯阳. 基于改进主元分析的微小故障诊断方法研究[D]. 西安: 西安工业大学, 2024.
	Zhao K Y. Research on micro-fault diagnosis method based on improved principal component analysis[D]. Xi'an: Xi'an Technological University, 2024.
[10]	刘晰宇. 乙烯精馏塔预测控制技术及其应用研究[D]. 大庆: 东北石油大学, 2023.
	Liu X Y. Predictive control technology of ethylene distillation column and its application[D]. Daqing: Northeast Petroleum University, 2023.
[11]	陈德坤. 大型乙烯裂解炉烧焦过程中COT随机分布系统建模与控制研究[D]. 北京: 北京化工大学, 2021.
	Chen D K. Study on modeling and control of COT random distribution system during coking of large ethylene cracking furnace[D]. Beijing: Beijing University of Chemical Technology, 2021.
[12]	于静, 金秀章. 基于RBF神经网络的PFC-PID主汽温串级预测控制[J]. 华北电力大学学报(自然科学版), 2020, 47(6): 91-98.
	Yu J, Jin X Z. PFC-PID main steam temperature cascade predictive control based on RBF neural network[J]. Journal of North China Electric Power University (Natural Science Edition), 2020, 47(6): 91-98.
[13]	韩先尧, 王宁, 申青蓉, 等. 基于累计T ²贡献值的乙烯裂解炉故障诊断分析[J]. 控制工程, 2021, 28(10): 1917-1922.
	Han X Y, Wang N, Shen Q R, et al. Fault diagnosis analysis of ethylene cracking furnace based on accumulated T ² contribution value[J]. Control Engineering of China, 2021, 28(10): 1917-1922.
[14]	Li G, Ba Z G, Zhang H Z. Expert system of fault diagnosis for flight simulator based on fault tree analysis[J]. Advanced Materials Research, 2014, 1055: 371-374.
[15]	李长青, 张德龙, 董泽. 主汽温系统线性自抗扰优化控制[J]. 计算机仿真, 2018, 35(8): 81-86.
	Li C Q, Zhang D L, Dong Z. Optimized linear active disturbance rejection controller design for main steam temperature systems[J]. Computer Simulation, 2018, 35(8): 81-86.
[16]	Lu Y S, Wang F L, Jia M X, et al. Centrifugal compressor fault diagnosis based on qualitative simulation and thermal parameters[J]. Mechanical Systems and Signal Processing, 2016, 81: 259-273.
[17]	任佳, 师琪, 陈晓, 等. 数据驱动的化工过程故障诊断方法综述[C]//第35届中国过程控制会议论文集. 2024.
	Ren J, Shi Q, Chen X, et al. Review on data-driven based fault detection and diagnosis approaches for chemical industrial process[C]//Proceedings of the 35th China Process Control Conference. 2024.
[18]	崔潇艺. 基于深度学习的化工过程故障诊断算法研究[D]. 长春: 长春工业大学, 2024.
	Cui X Y. Research on fault diagnosis algorithm of chemical process based on deep learning[D]. Changchun: Changchun University of Technology, 2024.
[19]	Bachman P, Hjelm R D, Buchwalter W. Learning representations by maximizing mutual information across views[EB/OL]. .
[20]	吕巧莉. 基于自监督对比学习的图像分类算法研究[D]. 保定: 河北大学, 2024.
	Lyu Q L. Research on image classification algorithm based on self-supervised contrastive learning[D]. Baoding: Hebei University, 2024.
[21]	Tian Y L, Krishnan D, Isola P. Contrastive multiview coding[M]//Vedaldi A, Bischof H, Brox T, et al. Lecture Notes in Computer Science. Cham: Springer International Publishing, 2020: 776-794.
[22]	李希, 刘喜平, 李旺才, 等. 对比学习研究综述[J]. 小型微型计算机系统, 2023, 44(4): 787-797.
	Li X, Liu X P, Li W C, et al. Survey on contrastive learning research[J]. Journal of Chinese Computer Systems, 2023, 44(4): 787-797.
[23]	张重生, 陈杰, 李岐龙, 等. 深度对比学习综述[J]. 自动化学报, 2023, 49(1): 15-39.
	Zhang C S, Chen J, Li Q L, et al. Deep contrastive learning: a survey[J]. Acta Automatica Sinica, 2023, 49(1): 15-39.
[24]	He K M, Fan H Q, Wu Y X, et al. Momentum contrast for unsupervised visual representation learning[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA: IEEE, 2020: 9726-9735.
[25]	Hadsell R, Chopra S, LeCun Y. Dimensionality reduction by learning an invariant mapping[C]//2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06). New York, NY, USA: IEEE, 2006: 1735-1742.
[26]	Wu Z R, Xiong Y J, Yu S X, et al. Unsupervised feature learning via non-parametric instance discrimination[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 3733-3742.
[27]	Dosovitskiy A, Springenberg J T, Riedmiller M, et al. Discriminative unsupervised feature learning with convolutional neural networks[J/OL]. .
[28]	Gutmann M, Hyvärinen A. Noise-contrastive estimation: a new estimation principle for unnormalized statistical models[C]//Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics. 2010: 297-304.
[29]	Oord A, Li Y, Vinyals O. Representation learning with contrastive predictive coding[J/OL]. .
[30]	Pathak D, Krähenbühl P, Donahue J, et al. Context encoders: feature learning by inpainting[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA: IEEE, 2016: 2536-2544.
[31]	李三仟. 基于多视觉特征和卷积神经网络先验信息的图像复原[D]. 南昌: 南昌大学, 2019.
	Li S Q. Image restoration based on multi-visual features and prior information of convolutional neural network[D]. Nanchang: Nanchang University, 2019.
[32]	Zhang R, Isola P, Efros A A. Colorful image colorization[M]//Leibe B, Matas J, Sebe N, et al. Lecture Notes in Computer Science. Cham: Springer International Publishing, 2016: 649-666.
[33]	Tao C X, Wang H H, Zhu X Z, et al. Exploring the equivalence of Siamese self-supervised learning via a unified gradient framework[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA: IEEE, 2022: 14411-14420.
[34]	Chen T, Kornblith S, Norouzi M, et al. A simple framework for contrastive learning of visual representations[C]//International Conference on Machine Learning. PMLR, 2020: 1597-1607.
[35]	Meybodi Z H, Mohammadi A, Abouei J, et al. CoPo: self-supervised contrastive learning for popularity prediction in MEC networks[C]//2023 24th International Conference on Digital Signal Processing (DSP). Rhodes (Rodos), Greece: IEEE, 2023: 1-5.
[36]	Nguyen T, Pham T X, Zhang C N, et al. DimCL: dimensional contrastive learning for improving self-supervised learning[J]. IEEE Access, 2023, 11: 21534-21545.
[37]	Li Y F, Hu P, Liu Z T, et al. Contrastive clustering[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(10): 8547-8555.
[38]	Liu J X, Chen S C. TimesURL: self-supervised contrastive learning for universal time series representation learning[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(12): 13918-13926.

参数	本文	TimesURL
模型架构	ResNet-50	TCN
学习率	0.03	0.001
批量大小	256	16
动量系数	0.999	无

参数	本文	TimesURL
模型架构	ResNet-50	TCN
学习率	0.03	0.001
批量大小	256	16
动量系数	0.999	无

灰度图大小	本文方法的准确率/%	本文方法的测试准确率/%	TimesURL 准确率/%
5×13	71.2	59.4	55.1
10×13	85.8	71.6
20×13	90.3	79.1
50×13	92.1	81.7

灰度图大小	本文方法的准确率/%	本文方法的测试准确率/%	TimesURL 准确率/%
5×13	71.2	59.4	55.1
10×13	85.8	71.6
20×13	90.3	79.1
50×13	92.1	81.7

燃料燃烧效率范围	对应热效率等级	对应热效率含义
(2.0, 3.02)	1	低
(3.02, 3.04)	2	较低
(3.04, 3.06)	3	中等
(3.06, 3.08)	4	较高
(3.08, 3.10)	5	高