Improved prediction model for BOF end-point manganese content based on IPSO-RELM method
-
摘要: 分析了影響轉爐冶煉終點鋼水中錳含量的因素, 針對基于BP神經網絡算法的轉爐冶煉終點錳含量預測模型存在的收斂速度慢, 預測精度低等問題, 提出了一種基于極限學習機(ELM) 算法建模的新思路, 并引入正則化以及改進粒子群優化算法(IPSO), 建立了基于改進粒子群算法優化的正則化極限學習機(IPSO-RELM) 的轉爐終點錳含量預測模型; 應用國內某煉鋼廠轉爐實際生產數據對模型進行訓練和驗證, 并與基于BP、ELM和RELM算法的三類模型進行比較.結果表明, 采用IPSO-RELM方法構建的模型, 錳含量預測誤差在±0. 025%范圍內的命中率達到94%, 均方誤差為2. 18×10-8, 擬合優度R2為0. 72, 上述三項指標均顯著優于其他三類模型, 此外, 該模型還具有良好的泛化能力, 對于轉爐實際冶煉過程具有一定的指導意義.Abstract: The basic oxygen furnace (BOF) steelmaking process, as the predominant steelmaking method used around the world, involves very complex physical and chemical phenomena such as multi-component reactions, multi-phase fluid dynamics, and high temperature. The main task of the BOF process is tailoring the temperature and melt components to meet the requirements of high-quality steel production. With the development of intelligent steelmaking, the prediction of the end-point manganese content is an extremely important task for the BOF process, and improving the level of control regarding the end-point of BOF steelmaking can reduce production costs and enhance efficiency. In this paper, the mechanism of the BOF steelmaking process and the factors influencing the endpoint manganese content were analyzed. The control variables for predicting the end-point manganese content were also determined. To solve the problems of slow convergence, weak generalization ability, and low prediction accuracy in the prediction model established for the BP neural network, a new modeling concept based on an extreme learning machine (ELM) algorithm was proposed. By introducing regularization and improved particle swarm optimization (IPSO), a prediction model for the end-point manganese content in a converter based on improved particle swarm optimization and a regularized ELM (IPSO-RELM) was established. The paper then trained and verified the performance of these models with actual production data. A comparison of the performance of the proposed model with those of the prediction model of the BP neural network, the ELM model, and the RELM model reveals that the IPSO-RELM prediction model has the highest prediction accuracy and the best generalization performance. The hit ratio of the IPSO-RELM prediction model is 94%when the predictive errors of the model are within 0. 025%, the mean square error is 2. 18 × 10-8, and the fitting degree is 0. 72. Relative to the above three models, the IPSO-RELM prediction model may provide a more accurate prediction of the end-point manganese content and thus serves as a good reference point for actual production.
-
圖 1 典型的單隱層前饋神經網絡模型
Figure 1. Typical single hidden layer feedforward neural network model
圖 3 錳含量預測值與實測值的比較. (a) BP; (b) ELM; (c) RELM; (d) IPSO--RELM
Figure 3. Comparison of predicted and observed end-point manganese contents: (a) BP; (b) ELM; (c) RELM; (d) IPSO--RELM
圖 4 錳含量預測誤差的分布. (a) BP; (b) ELM; (c) RELM; (d) IPSO--RELM
Figure 4. Distribution of prediction errors of end-point manganese content: (a) BP; (b) ELM; (c) RELM; (d) IPSO-RELM
表 1 轉爐終點錳含量影響因素的皮爾遜相關系數
Table 1. Pearson correlation coefficients of process parameters in the prediction of end-point manganese content for BOF
變量 Pearson
相關系數變量 Pearson
相關系數化渣劑加入量 -0.545** 鐵水S質量分數 -0.231** 廢鋼裝入量 0.376** 鐵水溫度 0.231** 鐵水Mn質量分數 0.371** 鐵水P質量分數 0.217** 石灰加入量 0.288** 氧耗量 0.201** 輕燒白云石加入量 0.256** 鐵水加入量 0.193** 注：** 表示變量與終點錳含量在0.01水平上顯著相關；*表示變量與終點錳含量在0.05水平上顯著相關. 
下載: 導出CSV
表 2 IPSO-RELM模型的基本參數
Table 2. Fundamental parameters of IPSO-RELM model
參數名稱 設置值 參數名稱 設置值 輸入層節點 10 輸出層節點 1 隱含層數 1 種群規模，P 20 隱含層節點數 20 迭代次數，N 50 學習因子，c1 2.8 最大權重系數，ωmax 1.2 學習因子，c2 1.2 最小權重系數，ωmin 0.4
下載: 導出CSV
中文字幕在线观看表 3 4種算法性能比較
Table 3. Comparison of performances of four kinds of algorithms
模型 均方誤差，MSE/10-8 R2 誤差±0.025%命中率 BP 3.49 0.49 0.78 ELM 2.71 0.65 0.84 RELM 2.64 0.66 0.88 IPSO--RELM 2.18 0.72 0.94
下載: 導出CSV
-
參考文獻
[1] Takawa T, Katayama K, Hoteiya M, et al. Mathematical model of end point control for the top and bottom blowing process in BOF. Trans ISIJ, 1987, 27(12): 951 doi: 10.2355/isijinternational1966.27.951 [2] He F, He D F, Xu A J, et al. Hybridmodel of molten steel temperature prediction based on ladle heat status and artificial neural network. J Iron Steel Res Int, 2014, 21(2): 181 doi: 10.1016/S1006-706X(14)60028-5 [3] Wang H B, Cai J, Feng K. Predicting theendpoint phosphorus content of molten steel in BOF by two-stage hybrid method. J Iron Steel Res Int, 2014, 21(Suppl 1): 65 http://www.sciencedirect.com/science/article/pii/S1006706X14601230 [4] Cox I J, Lewis R W, Ransing R S, et al. Application of neural computing in basic oxygen steelmaking. J Mater Process Technol, 2002, 120(1-3): 310 doi: 10.1016/S0924-0136(01)01136-0 [5] Zhang G Y, Wan X F, Lin D, et al. Carbon content and temperature variation of bath based on exhaust gas analysis. J Iron Steel Res, 2006, 18(11): 56 doi: 10.3321/j.issn:1001-0963.2006.11.014張貴玉, 萬雪峰, 林東, 等. 基于爐氣分析的熔池碳含量及溫度變化研究. 鋼鐵研究學報, 2006, 18(11): 56 doi: 10.3321/j.issn:1001-0963.2006.11.014 [6] Takawa T, Sato M, Okada T, et al. Development of automatic blowing technique in BOF based on a mathematical model. Tetsu-to-Hagane, 1988, 74(4): 664高輪武志, 佐藤光信, 岡田剛, 等. 數式モデルによる転爐自動吹錬技術の開発. 鉄と鋼, 1988, 74(4): 664 [7] Yang L H, Liu L, He P. [Mn]ep prediction model for melt in oxygen converter. Steelmaking, 2003, 19(1): 10 doi: 10.3969/j.issn.1002-1043.2003.01.002楊立紅, 劉瀏, 何平. 轉爐冶煉終點錳成分的預報模型. 煉鋼, 2003, 19(1): 10 doi: 10.3969/j.issn.1002-1043.2003.01.002 [8] Liu K, Liu L, He P. Endpoint phosphorus and manganese content control model based on sublance technique and optimization of dephosphorization process. Iron Steel, 2008, 43(7): 32 doi: 10.3321/j.issn:0449-749X.2008.07.007劉錕, 劉瀏, 何平. 基于副槍的轉爐終點磷錳控制模型與脫磷優化. 鋼鐵, 2008, 43(7): 32 doi: 10.3321/j.issn:0449-749X.2008.07.007 [9] Wang Z, Chang J, Ju Q P, et al. Predictionmodel of end-point manganese content for BOF steelmaking process. ISIJ Int, 2012, 52(9): 1585 doi: 10.2355/isijinternational.52.1585 [10] Huang G B, Zhu Q Y, Siew C K. Extreme learning machine: Theory and applications. Neurocomputing, 2006, 70(1-3): 489 doi: 10.1016/j.neucom.2005.12.126 [11] Zhao Q Q, Hou B L. Parameter identification of a shell transfer arm using FDA and optimized ELM. Chin J Eng, 2017, 39(4): 611 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201704017.htm趙搶搶, 侯保林. 函數型數據分析與優化極限學習機結合的彈藥傳輸機械臂參數辨識. 工程科學學報, 2017, 39(4): 611 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201704017.htm [12] Deng W Y, Zheng Q H, Chen L. Regularizedextreme learning machine//IEEE Symposium on Computational Intelligence & Data Mining. Nashville, 2009: 389 [13] Xiong W T, Yu B J, Sun L. Improved particle swarm optimization of rolling schedule on 420 mm 5-stand tandem cold strip mill. J Iron Steel Res, 2014, 26(11): 25 https://www.cnki.com.cn/Article/CJFDTOTAL-IRON201411006.htm熊文濤, 禹寶軍, 孫林. 改進粒子群算法對1420mm五機架冷連軋機軋制規程的優化. 鋼鐵研究學報, 2014, 26(11): 25 https://www.cnki.com.cn/Article/CJFDTOTAL-IRON201411006.htm [14] Kennedy J, Eberhart R. Particle swarm optimization//Proceedings of ICNN' 95-International Conference on Neural Networks. Perth, 1995: 1942 [15] Li Q, Xu Y M, Zhang D Z, et al. Global path planning method for mobile robots based on the particle swarm algorithm. J Univ Sci Technol Beijing, 2010, 32(3): 397 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201003022.htm李擎, 徐銀梅, 張德政, 等. 基于粒子群算法的移動機器人全局路徑規劃策略. 北京科技大學學報, 2010, 32(3): 397 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201003022.htm [16] Tao H L. Study on Forecast of Railway Traffic Volume Based on Hybrid Intelligent Algorithm[Dissertation]. Lanzhou: Lanzhou Jiaotong University, 2012陶海龍. 基于混合智能算法的鐵路運量預測研究[學位論文]. 蘭州: 蘭州交通大學, 2012 [17] Bueno-Crespo A, García-Laencina P J, Sancho-Gómez J L. Neural architecture design based on extreme learning machine. Neural Networks, 2013, 48: 19 doi: 10.1016/j.neunet.2013.06.010 [18] Chen H Z, Yang J P, Lu X C, et al. Quality prediction of the continuous casting bloom based on the extreme learning machine. Chin J Eng, 2018, 40(7): 815 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201807007.htm陳恒志, 楊建平, 盧新春, 等. 基于極限學習機(ELM)的連鑄坯質量預測. 工程科學學報, 2018, 40(7): 815 https://www.cnki.com.cn/Article/CJFDTOTAL-BJKD201807007.htm [19] Na W B, Su Z W, Ji Y F. Research ofsingle well production prediction based on improved extreme learning machine. Appl Mech Mater, 2013, 333-335: 1296 doi: 10.4028/www.scientific.net/AMM.333-335.1296 [20] Martínez-Martínez J M, Escandell-Montero P, Soria-Olivas E, et al. Regularized extreme learning machine for regression problems. Neurocomputing, 2011, 74(17): 3716 doi: 10.1016/j.neucom.2011.06.013 [21] Hornik K, Stinchcombe M, White H. Multilayer feedforward networks are universal approximators. Neural Networks, 1989, 2(5): 359 doi: 10.1016/0893-6080(89)90020-8 [22] Du B, Lin Y. Development andapplication of an universal automatic modeling tools. Comput Technol Autom, 2003, 22(2): 103 doi: 10.3969/j.issn.1003-6199.2003.02.030杜斌, 林云. 通用智能自動建模軟件開發與應用. 計算技術與自動化, 2003, 22(2): 103 doi: 10.3969/j.issn.1003-6199.2003.02.030 -
點擊查看大圖
圖(5) / 表(3)
計量
- 文章訪問數: 1158
- HTML全文瀏覽量: 455
- PDF下載量: 39
- 被引次數: 0
