-
摘要: 煉鋼過程是極其復雜的工業場景，影響因素多且安全性要求極高，是當前深度學習尚未大規模應用的領域之一。在對深度學習的原理和類型進行梳理的基礎之上，結合國內外應用實例，總結了深度學習在煉鋼過程中的發展歷程與研究現狀。指出了深度學習在煉鋼過程中應用主要有特征提取簡單、泛化能力強、模型可塑性高的優勢，同時也面臨數據依賴性高、預處理難度大、生產安全性有待驗證的挑戰。提出了未來隨著高精度傳感器的應用、物聯網的普及、計算硬件的迭代、以及算法的創新，深度學習模型可以更加有效地應用于煉鋼的更多場景中，將推動冶金工業智能化發展。Abstract: The steel industry is an important embodiment of national productivity and contributes to the development of the national economy and defense construction as a material foundation. Recently, China’s crude steel production ranked first in the world and in 2020, it exceeded 1 billion tons for the first time, reaching 1.065 billion tons. However, the steel industry is also a major energy consumer and polluter. In the current national coordination to do a good job of “carbon peak” and “carbon-neutral” background, the traditional steelmaking process urgently needs to be transformed into intelligent and green. Recently, as an important branch of machine learning, with artificial neural networks as the basic architecture, deep learning, a nonlinear modeling algorithm that can extract features from data and realize knowledge learning, has been applied in various industrial fields. The steelmaking process is an extremely complex industrial scenario with many influencing factors and high-security requirements. It is also an area where deep learning has not been applied on a large scale yet. Accordingly, in this study, the principles and types of deep learning were compared, and the development history and research status of deep learning in the steelmaking process with domestic and foreign application examples were summarized. The application of deep learning to the steelmaking process mainly has the advantages of simple feature extraction, strong generalization ability, and high model plasticity, but it also faces the challenges of high data dependency, difficult preprocessing, and verification of production safety. In the future, with the application of high-precision sensors, popularization of the Internet of Things, iteration of computing hardware, and innovation of algorithms, deep learning models can be effectively applied to more scenarios in steelmaking, which will promote the intelligent development of the metallurgical industry.
-
Key words:
- steelmaking process /
- deep learning /
- neural network /
- application scenarios /
- data processing /
- nonlinear /
- intelligent
-
圖 1 人工神經網絡神經元結構及工作原理
Figure 1. Structure and working principle of artificial neural network neurons
圖 3 神經網絡分類. (a) 前饋神經網絡; (b) 反饋神經網絡; (c) 自組織神經網絡
Figure 3. Neural network classification: (a) feed-forward neural network; (b) feedback neural network; (c) self-organizing neural network
圖 4 深度學習模型在煉鋼過程中的部分應用
Figure 4. Partial applications of the deep learning model in the steelmaking process
圖 5 煉鋼過程信息采集、傳輸和運算的要求
Figure 5. Requirements for information collection, transmission and operation in steelmaking process
表 1 幾種深度學習主流方法特征對比
Table 1. Comparison of the features of several mainstream methods of deep learning
Deep learning methods Advantages Disadvantages BP (1) Strong nonlinear mapping capability; (2) High self-learning and self-adaptive capabilities; (3) Some fault tolerance (1) Slow convergence speed
; (2) Easy to fall into local minimaCNN (1) Partial connection; (2) Value sharing
; (3) Hierarchical expression(1) Need to normalize the dataset; (2) No memory function; (3) Poor natural language processing skills WNN (1) Fast network convergence; (2) Avoid getting stuck in a local optimum; (3) High precision (1) Difficult to determine the nodes in the hidden layer; (2) No adaptive selection of functions SOM (1) Self-organization changes network parameters; (2) Only one neuron becomes the competition winner (1) Need to predetermine the number of neurons; (2) Randomly generate the initial value of the weight vector 
下載: 導出CSV
表 2 深度學習模型探索使用案例
Table 2. Deep learning model exploration use cases
No. Year Country Application companies Process Application 1 1990 USA Copperweld Steel Mill Electric arc furnace Process control 2 1990 USA North Star Steel Mill Electric arc furnace Process control 3 1991 Finland Rahhe Steel Mill Continuous casting Pourability forecast 4 1991 Japan Yawata Steel Mill Continuous casting Steel leakage forecast 5 1994 China Guangzhou Steel Mill Electric arc furnace Electrode control 6 1995 China Baoshan Steel Converters Dynamic model 7 1997 China Wuhan Steel Converters Endpoint control 8 2001 China Xingcheng Special Steel Electric arc furnace Temperature forecast
下載: 導出CSV
表 3 國內外煉鋼企業智能化發展布局
Table 3. Intelligent development layout of domestic and foreign steelmaking enterprises
No. Year Country Application companies Cooperation unit Project content 1 2017 Korea POSCO POSCO Technical Research Laboratories Deep learning projects 2 2017 USA Big River Steel Noodle AI Artificial intelligence platform 3 2018 China Baowu Steel Baidu Online Network Technology AI + steel quality inspection 4 2018 China Anshan Iron and Steel Kingsoft Corporation Limited Precision steel cloud platform 5 2018 China Xiangtan Iron and Steel Huawei Technologies Smart factory project 6 2018 India Tata Steel Tata Steel Digie-Shala Department Process optimization solutions 7 2019 China Jinnan Iron and Steel Alibaba Group Steel scrap AI grading system 8 2019 China Magang Holding Company Tencent Intelligent decision-making and control platform 9 2019 Japan Nippon Steel NS Solutions Corporation NS-DIG intelligent platform 10 2019 Germany Thyssenkrupp Microsoft “Alfred” artificial intelligence solution 11 2020 China Luli Group Ramon Science and Technology Remote intelligent grading system for steel scrap 12 2020 China Baowu Steel Shanghai Baosight Software Baowu ecotechnology platform
下載: 導出CSV
久色视频表 4 煉鋼企業不同應用要求的最佳解決方案
Table 4. Best solution for different applications required by the industry
No. Occurrence frequency Impact Application examples Solutions 1 High Serious Endpoint prediction and defect detection Building deep learning models 2 Low Serious Secondary oxidation of steel and continuous casting leakage Improvement from the process route and
operation system3 High Minor Small fluctuations in the amount of raw and auxiliary materials added Solving through lean management 4 Low Minor Temperature measurement on the gun failure and spare part overdue Solving through routine inspection
下載: 導出CSV
-
參考文獻
[1] Xing Y, Zhang W, Su W, et al. Research of ultra-low emission technologies of the iron and steel industry in China. Chin J Eng, 2021, 43(1): 1邢奕, 張文伯, 蘇偉, 等. 中國鋼鐵行業超低排放之路. 工程科學學報, 2021, 43(1):1 [2] National Bureau of Statistics of China. Statistical bulletin of the People's Republic of China on national economic and social development in 2020. China Stat, 2021(3): 8國家統計局. 中華人民共和國2020年國民經濟和社會發展統計公報. 中國統計, 2021(3):8 [3] Feng K. Development analysis of steel materials and industry. Met World, 2021(2): 7 doi: 10.3969/j.issn.1000-6826.2021.02.0002馮凱. 鋼鐵材料與鋼鐵工業未來發展分析. 金屬世界, 2021(2):7 doi: 10.3969/j.issn.1000-6826.2021.02.0002 [4] Zeng J. Thoughts on intellectualization improvement of iron and steel production process. Metall Ind Autom, 2019, 43(1): 13曾加慶. 關于鋼鐵流程智能化提升的思考. 冶金自動化, 2019, 43(1):13 [5] Schmidhuber J. Deep learning in neural networks: An overview. Neural Netw, 2015, 61: 85 doi: 10.1016/j.neunet.2014.09.003 [6] Bengio Y, Courville A, Vincent P. Representation learning: A review and new perspectives. IEEE Trans Pattern Anal Mach Intell, 2013, 35(8): 1798 doi: 10.1109/TPAMI.2013.50 [7] Collobert R. Deep learning for efficient discriminative parsing // 14th International Conference on Artificial Intelligence and Statistics. Ft. Lauderdale, 2011: 224 [8] Liu X, Tao F, Yu W B. A neural network enhanced system for learning nonlinear constitutive law and failure initiation criterion of composites using indirectly measurable data. Compos Struct, 2020, 252: 112658 doi: 10.1016/j.compstruct.2020.112658 [9] Rosenblatt F. The perceptron: A probabilistic model for information storage and organization in the brain. Psychol Rev, 1958, 65(6): 386 doi: 10.1037/h0042519 [10] Pei Y Y, Yang X B, Chuan J P, et al. Time series prediction of microseismic energy level based on feature extraction of onedimensional convolutional neural network. Chin J Eng, 2021, 43(7): 1003裴艷宇, 楊小彬, 傳金平, 等. 一維卷積神經網絡特征提取下微震能級時序預測. 工程科學學報, 2021, 43(7):1003 [11] Elman J L. Finding structure in time. Cogn Sci, 1990, 14(2): 179 doi: 10.1207/s15516709cog1402_1 [12] Mangiameli P, Chen S K, West D. A comparison of SOM neural network and hierarchical clustering methods. Eur J Oper Res, 1996, 93(2): 402 doi: 10.1016/0377-2217(96)00038-0 [13] Mcculloch W S, Pitts W. A logical calculus of the ideas immanent in nervous activity. Bull Math Biol, 1990, 52(1-2): 99 doi: 10.1016/S0092-8240(05)80006-0 [14] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors. Nature, 1986, 323(6088): 533 doi: 10.1038/323533a0 [15] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks. Science, 2006, 313(5786): 504 doi: 10.1126/science.1127647 [16] Bi X G. Application of artificial intelligence and expert systems in the steel industry. J Wuhan Univ Sci Tech, 1995, 18(2): 146畢學工. 人工智能和專家系統在鋼鐵工業中的應用. 武漢鋼鐵學院學報, 1995, 18(2):146 [17] Wang Y H. Steel Temperature-Soft-Measurement [Dissertation]. Tianjin: Tianjin University of Science and Technology, 2001王玉輝. 鋼水溫度軟測量[學位論文]. 天津: 天津科技大學, 2001 [18] Liu L. The full automatic control technique for converter blowing process. Metall Ind Autom, 1999, 23(4): 1劉瀏. 轉爐全自動吹煉技術. 冶金自動化, 1999, 23(4):1 [19] Ding R, Liu L. Artificial intelligence static control model in converter steelmaking. Iron &Steel, 1997, 32(1): 22丁容, 劉瀏. 轉爐煉鋼過程人工智能靜態控制模型. 鋼鐵, 1997, 32(1):22 [20] Sun S Y, Li S P, Wang J R, et al. Intelligent control method for the secondary cooling of continuous casting. J Univ Sci Technol Beijing, 1997, 19(2): 188孫韶元, 李世平, 王俊然, 等. 連鑄二冷控制的智能化方法. 北京科技大學學報, 1997, 19(2):188 [21] Yu Z X, Liu L C, XiaoW B, et al. Development of computer controlled LD blowing process at NO. 3 steel plant, WISCO. Iron Steel, 2004, 39(8): 58余志祥, 劉路長, 肖文斌, 等. 武鋼三煉鋼計算機煉鋼技術的新進展. 鋼鐵, 2004, 39(8): 58 [22] Wang Q K, Hu R F, Lei J Y, et al. Development and application of dynamic controling system for sub-lance used in NIPPON steel corporation. Iron Steel, 1985, 20(1): 51王慶奎, 胡瑞富, 雷家源, 等. 新日鐵副槍動態控制系統的發展和應用. 鋼鐵, 1985, 20(1):51 [23] Hua C J, Wang M, Zhang M Y, et al. Effect of submerged entry nozzle wall surface morphologies on boundary layer structure and alumina inclusions transport. Chin J Eng, 2021, 43(7): 925華承健, 王敏, 張孟昀, 等. 浸入式水口內壁特征對邊界層流場結構和氧化鋁夾雜物運動行為的影響. 工程科學學報, 2021, 43(7):925 [24] Zhou C G, Hu J Z, Jiang C M, et al. Prediction model of phosphorus content in dephosphorization converter end point based on BP neural network algorithm. Steelmaking, 2021, 37(2): 10周朝剛, 胡錦榛, 蔣朝敏, 等. 基于BP神經網絡算法的脫磷轉爐終點磷含量預報模型. 煉鋼, 2021, 37(2):10 [25] Li C R, Zhao H W, Xie X, et al. Prediction of end-point phosphorus content for BOF based on LM BP neural network. Iron Steel, 2011, 46(4): 23李長榮, 趙浩文, 謝祥, 等. 基于L-M算法BP神經網絡的轉爐煉鋼終點磷含量預報. 鋼鐵, 2011, 46(4):23 [26] He F, Zhang L Y. Prediction model of end-point phosphorus content in BOF steelmaking process based on PCA and BP neural network. J Process Control, 2018, 66: 51 doi: 10.1016/j.jprocont.2018.03.005 [27] Gao F, Bao Y P, Wang M, et al. Prediction model of end-point phosphorus content of converter based on FA-ELM. Iron Steel, 2020, 55(12): 24高放, 包燕平, 王敏, 等. 基于FA-ELM的轉爐終點磷含量預測模型. 鋼鐵, 2020, 55(12):24 [28] Xuan M T, Li J J, Wang N, et al. Endpoint prediction of basic oxygen furnace steelmaking based on FOA-GRNN model. J Mater Metall, 2019, 18(1): 31鉉明濤, 李嬌嬌, 王楠, 等. 基于FOA-GRNN模型的轉爐煉鋼終點預報. 材料與冶金學報, 2019, 18(1):31 [29] Qi Z Y, Gao K, Zhao B F, et al. Research on application of RBF neural network in endpoint prediction of converter steelmaking. Wirel Internet Technol, 2017(4): 106 doi: 10.3969/j.issn.1672-6944.2017.04.049祁子怡, 高坤, 趙寶芳, 等. 基于RBF神經網絡在轉爐煉鋼終點預報中的應用研究. 無線互聯科技, 2017(4):106 doi: 10.3969/j.issn.1672-6944.2017.04.049 [30] Zhang H Y, Zhou Q L, Yuan Z X, et al. RBF neural network base on affinity propagation clustering and its application in BOF steelmaking. J Iron Steel Res, 2014, 26(1): 22張輝宜, 周奇龍, 袁志祥, 等. 基于AP聚類的RBF神經網絡研究及其在轉爐煉鋼中的應用. 鋼鐵研究學報, 2014, 26(1):22 [31] Wang Z, Chang J, Ju Q P, et al. Prediction model of end-point manganese content for BOF steelmaking process. ISIJ Int, 2012, 52(9): 1585 doi: 10.2355/isijinternational.52.1585 [32] Liu Z M, Zhan D P, Ge Q Z, et al. Prediction model of mass fraction of endpoint carbon of electric furnace based on BP neural network. Ind Heat, 2018, 47(4): 28 doi: 10.3969/j.issn.1002-1639.2018.04.008劉志明, 戰東平, 葛啟楨, 等. 基于BP神經網絡的電爐終點碳質量分數預報模型. 工業加熱, 2018, 47(4):28 doi: 10.3969/j.issn.1002-1639.2018.04.008 [33] Ma R. Study on Intelligent Control Technique for the Electric Arc Furnace Steelmaking [Dissertation]. Xi’an: Northwestern Polytechnical University, 2006馬戎. 智能控制技術在煉鋼電弧爐中的應用研究[學位論文]. 西安: 西北工業大學, 2006 [34] Liu K, Liu L, He P, et al. Application of increment artificial neural network model to prediction of end-point carbon, phosphorus and temperature for an 100 t EAF steelmaking. Special Steel, 2004, 25(3): 40 doi: 10.3969/j.issn.1003-8620.2004.03.013劉錕, 劉瀏, 何平, 等. 增量神經網絡模型預報100t電弧爐終點碳、磷和溫度的應用. 特殊鋼, 2004, 25(3):40 doi: 10.3969/j.issn.1003-8620.2004.03.013 [35] Li Q, Cao G. Forecasting model for the molten steel temperature in refining furnace based on artificial neural network and expert system. Heavy Mach, 2010(6): 22 doi: 10.3969/j.issn.1001-196X.2010.06.007李強, 曹剛. 基于人工神經網絡和專家系統的精煉過程鋼水溫度預測模型. 重型機械, 2010(6):22 doi: 10.3969/j.issn.1001-196X.2010.06.007 [36] Wu Y. Study on Temperature Forecast and Control Model of RH Vacuum Refining [Dissertation]. Wuhan: Wuhan University of Science and Technology, 2014吳揚. RH真空精煉溫度預報與控制模型的研究[學位論文]. 武漢: 武漢科技大學, 2014 [37] He D F, He F, Xu A J, et al. On-line liquid steel temperature control for the steelmaking-continuous casting process. J Univ Sci Technol Beijing, 2014, 36 (Suppl 1): 200賀東風, 何飛, 徐安軍, 等. 煉鋼連鑄流程在線鋼水溫度控制. 北京科技大學學報, 2014, 36(增刊1): 200 [38] Fu G Q, Liu Q, Wang Z, et al. Grey box model for predicting the LF end-point temperature of molten steel. J Univ Sci Technol Beijing, 2013, 35(7): 948付國慶, 劉青, 汪宙, 等. LF精煉終點鋼水溫度灰箱預報模型. 北京科技大學學報, 2013, 35(7):948 [39] Feng C S, Xiao B Q, He F. A presetting model of molten steel temperature based on BP neural network. Res Iron Steel, 2012, 40(3): 30馮春松, 肖步慶, 何飛. 基于BP神經網絡的鋼水溫度預定模型. 鋼鐵研究, 2012, 40(3):30 [40] Fu J, Tao B S, Chen C Y, et al. Research on oxygen model for BOF based on BP neural network. Metall Ind Autom, 2014, 38(4): 11 doi: 10.3969/j.issn.1000-7059.2014.04.003付佳, 陶百生, 陳春雨, 等. 基于BP神經網絡的轉爐供氧模型研究. 冶金自動化, 2014, 38(4):11 doi: 10.3969/j.issn.1000-7059.2014.04.003 [41] Ai X L, Wang Y S, Tang W M. Prediction of oxyen blow rate in BP neural network based converter refining. Steelmaking, 2013, 29(2): 34艾曉禮, 王玉生, 唐文明. 基于BP神經網絡的轉爐煉鋼吹氧量預測. 煉鋼, 2013, 29(2):34 [42] Li A L, Zhao D Z, Guo Z B, et al. Prediction of converter oxygen consumption in improved deep belief network. China Meas Test, 2020, 46(6): 1 doi: 10.11857/j.issn.1674-5124.2019080078李愛蓮, 趙多禎, 郭志斌, 等. 改進深度信念網絡的轉爐耗氧量預測. 中國測試, 2020, 46(6):1 doi: 10.11857/j.issn.1674-5124.2019080078 [43] Zhang Z Y, Sun Y G. Prediction of oxygen amount in converter based on grey Elman neural network. Comput Appl Softw, 2018, 35(11): 103張子陽, 孫彥廣. 基于灰色Elman神經網絡轉爐吹氧量的預測. 計算機應用與軟件, 2018, 35(11):103 [44] Yang Z Y, Ren X J. Mathematic model of neural network with powder consumption optimization for hot metal pretreatment. Res Iron Steel, 2011, 39(3): 16楊志勇, 任小佳. 鐵水預處理粉劑用量優化的神經網絡模型. 鋼鐵研究, 2011, 39(3):16 [45] Zhang H, Chen F Y, Wang Y H. End point optimized control for BOF steel-making process based on the characteristic of subsidiary material’s movement. J Iron Steel Res, 2013, 25(1): 5張華, 陳鳳銀, 王艷紅. 基于輔料資源運行特性的煉鋼終點優化控制. 鋼鐵研究學報, 2013, 25(1):5 [46] Ou Q L, Wu X Z, Ou D X. PSO-BP-PID control of ladle furnace proportioning system. Control Eng China, 2013, 20(5): 825 doi: 10.3969/j.issn.1671-7848.2013.05.009歐青立, 吳興中, 歐達賢. 鋼包爐配料PSO-BP-PID控制研究. 控制工程, 2013, 20(5):825 doi: 10.3969/j.issn.1671-7848.2013.05.009 [47] Zhao Q, Chen Y R, Wang Y, et al. Light intensity and image information used in steelmaking end-point control. Chin J Sci Instrum, 2005, 26(8): 575趙琦, 陳延如, 王昀, 等. 光強與圖像信息在轉爐煉鋼終點判斷中的應用. 儀器儀表學報, 2005, 26(8):575 [48] Ma H T, Wang S S, Wu L B, et al. AOD furnace splash soft-sensor in the smelting process based on improved BP neural network // Proceedings of the Society of Photo-optical Instrumentation Engineers. Changchun, 2017(1060): 739 [49] Pang S Y, Wang S Y, Jia H S. Recognition of converter flame state based on ResNet neural network. Metall Ind Autom, 2021, 45(1): 34龐殊楊, 王姝洋, 賈鴻盛. 基于殘差神經網絡實現轉爐火焰狀態識別. 冶金自動化, 2021, 45(1):34 [50] Li C, Liu H. Carbon content prediction of converter steelmaking end-point based on improved MTBCD flame image feature extraction. Computer Integrated Manufacturing Systems, https://kns.cnki.net/kcms/detail/11.5946.TP.20210428.1806.020.html李超, 劉輝. 改進MTBCD火焰圖像特征提取的轉爐煉鋼終點碳含量預測. 計算機集成制造系統,https://kns.cnki.net/kcms/detail/11.5946.TP.20210428.1806.020.html [51] Mao X X, Liu Z, Ren J R, et al. Slab surface defect detection system based on deep learning. Ind Control Comput, 2019, 32(3): 66毛欣翔, 劉志, 任靜茹, 等. 基于深度學習的連鑄板坯表面缺陷檢測系統. 工業控制計算機, 2019, 32(3):66 [52] Konovalenko I, Maruschak P, Brezinová J, et al. Steel surface defect classification using deep residual neural network. Metals, 2020, 10(6): 846 doi: 10.3390/met10060846 [53] An B, Yan B, Liu Y J. The online quality evaluation of continuous casting billet based on BP and kohonen neural network. J North Univ China Nat Sci, 2016, 37(3): 268安波, 閆彬, 劉永姜. 基于BP和Kohonen神經網絡結合的鑄坯在線質量評估. 中北大學學報(自然科學版), 2016, 37(3):268 [54] Han Z. Research on Billet Quality Analysis Algorithm Based on Neural Networks [Dissertation]. Dalian: Dalian University of Technology, 2017韓舟. 基于神經元網絡技術的鑄坯質量分析算法研究[學位論文]. 大連: 大連理工大學, 2017 [55] Fan J D, Wang W Y, Rong Y C, et al. Application of RBF neural network to prediction of breakout in continuous casting process. J Shanghai Univ (Nat Sci Ed) , 2001, 7(5): 391范建東, 王唯一, 榮亦誠, 等. RBF神經網絡應用于連鑄漏鋼預報. 上海大學學報(自然科學版), 2001, 7(5):391 [56] Yang Q, Peng L. Quantum wavelet neural networks and its application in breakout prediction. Comput Eng Appl, 2008, 44(15): 242 doi: 10.3778/j.issn.1002-8331.2008.15.075楊琴, 彭力. 量子小波神經網絡及其在漏鋼預報中的應用. 計算機工程與應用, 2008, 44(15):242 doi: 10.3778/j.issn.1002-8331.2008.15.075 [57] Zhang B G, Zhang R Z, Wang G, et al. Breakout prediction for continuous casting using genetic algorithm-based back propagation neural network model. Int J Model Identif Control, 2012, 16(3): 199 doi: 10.1504/IJMIC.2012.047727 [58] Li Y, Wang Z, Ao Z G, et al. Optimization for breakout prediction system of BP neural network. Control Decis, 2010, 25(3): 453厲英, 王正, 敖志廣, 等. BP神經網絡漏鋼預測系統優化. 控制與決策, 2010, 25(3):453 -
點擊查看大圖
圖(5) / 表(4)
計量
- 文章訪問數: 1125
- HTML全文瀏覽量: 405
- PDF下載量: 170
- 被引次數: 0
