基于高斯混合模型的MWPCA高炉异常监测算法

doi:10.11949/0438-1157.20201708

摘要/Abstract

摘要：

大型高炉是钢铁制造过程中的重要装备，由于高炉运行过程复杂，干扰因素繁多，经常会有异常炉况发生。为及时监测异常炉况、保证高炉顺行，本文利用高炉运行数据，开发了一种基于MWPCA和高斯混合模型的算法对高炉异常过程进行监测。由于高炉运行数据存在非高斯分布和时变的特点，利用高斯混合模型改进了传统PCA监测模型的T²统计量，使算法可以适应高炉数据的独特分布特征，并加入了滑窗机制，使算法具有实时更新的能力。随后，将算法应用在华南某大型钢铁集团的真实高炉数据上，检测了算法的有效性，并将其与基础算法进行了对比分析，证明了算法在高炉异常监测能力上有所提高。

关键词: 高炉, 过程系统, 主元分析, 高斯混合模型, 过程监测, 算法

Abstract:

Large-scale blast furnaces are important equipment in the steel manufacturing process. Due to the complex operation of the blast furnaces and the many interference factors, abnormal furnace conditions often occur. In order to monitor the abnormal furnace conditions in time and ensure the blast furnace is running smoothly, this paper develops an algorithm based on the principal component analysis and gaussian mixture model to monitor the abnormal process of the blast furnace. Due to the non-Gaussian distribution and time-varying characteristics of blast furnace operating data, the Gaussian mixture model is used to improve the T² statistics of the traditional PCA monitoring model, so that the algorithm can adapt to the unique distribution characteristics of blast furnace data. And the sliding window mechanism is added to give the algorithm the ability to update in real time. Subsequently, the algorithm was applied to the real blast furnace data of a large iron and steel group in South China. The effectiveness of the algorithm was tested and compared with the basic algorithm to prove the improvement of the algorithm's ability to monitor blast furnace anomalies.

Key words: blast furnace, process systems, principal component analysis, Gaussian mixture model, process monitoring, algorithm

中图分类号:

朱雄卓, 张瀚文, 杨春节. 基于高斯混合模型的MWPCA高炉异常监测算法[J]. 化工学报, 2021, 72(3): 1539-1548.

ZHU Xiongzhuo, ZHANG Hanwen, YANG Chunjie. MWPCA blast furnace anomaly monitoring algorithm based on Gaussian mixture model[J]. CIESC Journal, 2021, 72(3): 1539-1548.

图/表 13

图1 常见高炉故障监测方法

Fig.1 General blast furnace fault monitoring methods

表1 数据集的变量列表

Table 1 A list of variables for the dataset

序号	变量	单位	序号	变量	单位	序号	变量	单位
1	富氧率	％	13	冷风流量	10⁴ m³/h	25	实际风速	m/s
2	高炉渗透系数		14	炉腹煤气量	m³	26	冷风温度	℃
3	CO体积	％	15	理论燃烧温度	℃	27	顶温东北	℃
4	H₂体积	％	16	炉顶压力(1)	kPa	28	顶温西南	℃
5	CO₂体积	％	17	炉顶压力(2)	kPa	29	顶温西北	℃
6	标准风速	m/s	18	炉顶压力(3)	kPa	30	顶温东南	℃
7	富氧流量	m³/h	19	炉顶压力(4)	kPa	31	顶温下降管	℃
8	炉腹煤气指数		20	富氧压力	MPa	32	热风温度	℃
9	鼓风动能	kJ	21	冷风压力(1)	MPa	33	设定喷煤量	t/h
10	阻力系数		22	冷风压力(2)	MPa	34	本小时喷煤量	t
11	鼓风湿度		23	热风压力(1)	MPa	35	上小时喷煤量	t
12	全压差	MPa	24	热风压力(2)	MPa

图2 关键指标一天内变化情况

Fig.2 General blast furnace fault monitoring methods

图3 关键指标分布直方图

Fig.3 Distribution histogram of key index

表2 KS正态分布检验

Table 2 A list of variables for the dataset

变量	p	正态分布假设
热风压力	<0.001	拒绝
全压差	<0.001	拒绝
理论燃烧温度	<0.001	拒绝
鼓风动能	<0.001	拒绝

表3 两组数据均值与标准差

Table 3 Data mean and standard deviation of the two groups

变量	7月1日均值	7月20日均值	7月1日标准差	7月20日标准差
热风压力/MPa	0.393	0.389	0.006	0.017
全压差/MPa	165.974	161.558	6.433	7.636
理论燃烧温度/℃	2262.765	2277.597	8.345	23.872
鼓风动能/kJ	135.888	138.128	9.264	11.450

图4 GMM模型迭代流程图

Fig.4 Iteration flow chart of GMM model

图5 样本与T2统计量覆盖范围

Fig.5 Samples and the coverage of T2 statistics

图6 样本和GMM-T2统计量覆盖范围

Fig.6 Samples and the coverage of GMM-T2 statistics

图7 高斯性检验的监测结果

Fig.7 Monitoring results of Gaussian test

图8 加入滑窗前的监测结果

Fig.8 Monitoring results of moving windows were not added

图9 加入滑窗后的监测结果

Fig.9 Monitoring results after moving windows were added

表4 不同算法对比

Table 4 Comparison of different algorithms

统计量	误报率	故障报警持续时间	故障的报警起始时刻
$T 2$ 统计量	5.74%	186个时刻	1859时刻
MAX-T²统计量	1.14%	173个时刻	1861时刻
主元凸包算法	1.14%~5.74%	173~186个时刻	1859~1861时刻
两阶段PCA算法	3.40%	117个时刻	1896时刻
KPCA算法	4.33%	184个时刻	1860时刻
GMM-T²统计量	0.24%	189个时刻	1861时刻

表4 不同算法对比

Table 4 Comparison of different algorithms

统计量	误报率	故障报警持续时间	故障的报警起始时刻
$T 2$ 统计量	5.74%	186个时刻	1859时刻
MAX-T²统计量	1.14%	173个时刻	1861时刻
主元凸包算法	1.14%~5.74%	173~186个时刻	1859~1861时刻
两阶段PCA算法	3.40%	117个时刻	1896时刻
KPCA算法	4.33%	184个时刻	1860时刻
GMM-T²统计量	0.24%	189个时刻	1861时刻

参考文献 36

1	杨天钧. 中国高炉炼铁技术的进展[J]. 中国冶金, 2004, (6): 1-7.
	Yang T J. Technology process of blast furnace iron making in China[J]. China Metallurgy, 2004, (6): 1-7.
2	由文泉. 实用高炉炼铁技术[M]. 北京: 冶金工业出版社, 2002.
	You W Q. Applied Blast Furnace Ironmaking Technology[M]. Beijing: Metallurgical Industry Press, 2002.
3	王维兴. 中国高炉炼铁技术进展[J]. 钢铁, 2005, (10): 11-15.
	Wang W X. Technical progress of BF ironmaking in China[J]. Iron & Steel, 2005, (10): 11-15.
4	王维兴, 黄洁. 中国高炉炼铁技术发展评述[J]. 钢铁, 2007, 42(3): 1-4.
	Wang W X, Huang J. Review of technological development of BF ironmaking in China[J]. Iron & Steel, 2007, 42(3): 1-4.
5	Zhang T, Wang W, Ye H, et al. Fault detection for ironmaking process based on stacked denoising autoencoders[C]// 2016 American Control Conference (ACC). American Automatic Control Council (AACC), 2016.
6	周渝生, 钱晖, 张友平, 等. 现有主要炼铁工艺的优缺点和研发方向[J]. 钢铁, 2009, 44(2): 1-10.
	Zhou Y S, Qian H, Zhang Y P, et al. Advantages and disadvantages and its research direction of present ironmaking processes[J]. Iron & Steel, 2009, 44(2): 1-10.
7	安汝峤, 杨春节, 潘怡君. 基于加权图方法的高炉过程故障检测[J]. 高校化学工程学报, 2020, 34(2): 495-502.
	An R Q, Yang C J, Pan Y J. Fault detection based on the weight graph method in a blast furnace process[J]. Journal of Chemical Engineering of Chinese Universities, 2020, 34(2): 495-502.
8	An R, Yang C, Pan Y. Unsupervised change point detection using a weight graph method for process monitoring[J]. Industrial & Engineering Chemistry Research, 2019, 58(4): 1624-1634.
9	Wang L, Yang C J, Sun Y X, et al. Effective variable selection and moving window HMM-based approach for iron-making process monitoring[J]. Journal of Process Control, 2018, 68: 86-95.
10	Pan Y, Yang C, An R, et al. Robust principal component pursuit for fault detection in a blast furnace process[J]. Industrial & Engineering Chemistry Research, 2018, 57(1): 283-291.
11	Wang L, Yang C, Sun Y. Multimode process monitoring approach based on moving window hidden Markov model[J]. Industrial & Engineering Chemistry Research, 2018, 57(1): 292-301.
12	Pan Y, Yang C, An R, et al. Fault detection with improved principal component pursuit method[J]. Chemometrics and Intelligent Laboratory Systems, 2016, 157: 111-119.
13	Li R F, Wang X Z. Dimension reduction of process dynamic trends using independent component analysis[J]. Computers & Chemical Engineering, 2002, 26(3): 467-473.
14	Kano M, Tanaka S, Hasebe S, et al. Monitoring independent components for fault detection[J]. AIChE Journal, 2003, 49(4): 969-976.
15	Kano M, Tanaka S, Hasebe S, et al. Combined multivariate statistical process control[J]. IFAC Proceedings Volumes, 2004, 37(1): 281-286.
16	Ge Z, Song Z. Process monitoring based on independent component analysis- principal component analysis (ICA- PCA) and similarity factors[J]. Industrial & Engineering Chemistry Research, 2007, 46(7): 2054-2063.
17	Thissen U, Swierenga H, de Weijer A P, et al. Multivariate statistical process control using mixture modelling[J]. Journal of Chemometrics, 2005, 19(1): 23-31.
18	Yu J, Qin S J. Multimode process monitoring with Bayesian inference-based finite Gaussian mixture models[J]. AIChE Journal, 2008, 54(7): 1811-1829.
19	Chen T, Zhang R. On-line multivariate statistical monitoring of batch processes using Gaussian mixture model[J]. Computers & Chemical Engineering, 2010, 34(4): 500-507.
20	Yamamoto T, Sawada T, Shinohara K, et al. Blast furnace operational system with the application of advanced go/stop system at mizushima works[C]//IISC. The Sixth International Iron and Steel Congress. 1990: 364-371.
21	李芳. 高炉异常炉况预报专家系统研究[D]. 重庆: 重庆大学, 2007.
	Li F. The research of expert system for blast furnace exceptional situation forecast[D]. Chongqing: Chongqing University, 2007.
22	Tian H, Wang A. A novel fault diagnosis system for blast furnace based on support vector machine ensemble[J]. ISIJ International, 2010, 50(5): 738-742.
23	Liu L M, Wang A N, Sha M, et al. Multi-class classification methods of cost-conscious LS-SVM for fault diagnosis of blast furnace[J]. Journal of Iron and Steel Research, International, 2011, (10): 17-33.
24	赵明. 基于神经网络的高炉炉况诊断与预报研究[D]. 沈阳: 东北大学, 2010.
	Zhao M. Study on diagnose and forest of blast furnace condition based on neural network[D]. Shenyang: Northeastern University, 2010.
25	Vanhatalo E. Multivariate process monitoring of an experimental blast furnace[J]. Quality & Reliability Engineering International, 2010, 26(5): 495-508.
26	Zhang T, Ye H, Wang W, et al. Fault diagnosis for blast furnace ironmaking process based on two-stage principal component analysis[J]. ISIJ International, 2014, 54(10): 2334-2341.
27	Zhang T, Ye H, Wang W. Application of PCA based process monitoring method to ironmaking process[C]//2015 Chinese Automation Congress (CAC). Wuhan, 2015: 893-898.
28	Zhou B, Ye H, Zhang H, et al. Process monitoring of iron-making process in a blast furnace with PCA-based methods[J]. Control Engineering Practice, 2016, 47: 1-14.
29	孙梦园. 基于改进ICA算法的高炉故障诊断方法[D]. 杭州: 浙江大学, 2017.
	Sun M Y. Blast furnace fault diagnosis method based on improved ICA[D]. Hangzhou: Zhejiang University, 2017.
30	Shang J, Chen M, Ji H, et al. Dominant trend based logistic regression for fault diagnosis in nonstationary processes[J]. Control Engineering Practice, 2017, 66: 156-168.
31	Shang J, Chen M, Zhang H, et al. Increment-based recursive transformed component statistical analysis for monitoring blast furnace iron-making processes: an index-switching scheme[J]. Control Engineering Practice, 2018, 77: 190-200.
32	Pearson K. LIII. On lines and planes of closest fit to systems of points in space[J]. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 1901, 2(11): 559-572.
33	Hotelling H. Analysis of a complex of statistical variables into principal components[J]. Journal of Educational Psychology, 1933, 24(6): 417-441.
34	Jackson J E. Quality control methods for several related variables[J]. Technometrics, 1959, 1(4): 359-377.
35	Zheng J, Wen Q, Song Z. Recursive Gaussian mixture models for adaptive process monitoring[J]. Industrial & Engineering Chemistry Research, 2019, 58(16): 6551-6561.
36	Xie X, Shi H. Dynamic multimode process modeling and monitoring using adaptive Gaussian mixture models[J]. Industrial & Engineering Chemistry Research, 2012, 51(15): 5497-5505.

[1]	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.
[2]	曹跃, 余冲, 李智, 杨明磊. 工业数据驱动的加氢裂化装置多工况切换过渡状态检测[J]. 化工学报, 2023, 74(9): 3841-3854.
[3]	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.
[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]	王光, 单发顺, 钱禹丞, 焦建芳. 基于集成学习传递熵的化工过程微小故障检测方法[J]. 化工学报, 2023, 74(7): 2967-2978.
[7]	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.
[8]	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.
[9]	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.
[10]	周继鹏, 何文军, 李涛. 异形催化剂上乙烯催化氧化失活动力学反应工程计算[J]. 化工学报, 2023, 74(6): 2416-2426.
[11]	郑书闽, 郭鹏程, 颜建国, 王帅, 李文博, 周淇. 微小通道内过冷流动沸腾阻力特性实验及预测研究[J]. 化工学报, 2023, 74(4): 1549-1560.
[12]	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.
[13]	宋冰, 郑城风, 侍洪波, 陶阳, 谭帅. 基于VAE-OCCA的质量相关故障检测方法研究[J]. 化工学报, 2023, 74(4): 1630-1638.
[14]	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.
[15]	顾学荣, 刘硕士, 杨思宇. 基于并行EGO和代理模型辅助的多参数优化方法研究[J]. 化工学报, 2023, 74(3): 1205-1215.