基于神经网络的LF炉终点预报研究
|
摘要
当今世界先进炼钢生产工艺流程的一个重要环节是精炼炉炼钢,它是将电弧炉冶炼后的钢水,加入脱氧剂、合金料,进行脱氧、脱硫和均匀合金成分,炼制优质钢和特种钢的钢水二次冶炼工艺。其中,精炼期的终点预报是精炼炉优化控制的重要组成部分。如何精确控制钢液温度和终点成分,是一个迫切需要解决的课题。LF炉钢水温度以及终点成分的准确预测,对提高钢水质量、降低炼钢成本、合理组织生产、对操作人员选择最佳控制策略是很有帮助的。
本课题在查阅了大量国内外相关文献的基础上,对精炼炉终点预报方法作了深入的研究。在对各种预报方法了解的基础上,结合实际工艺和目前研究的实际条件决定采用改进的神经网络算法作为终点预报方法。它的基本思想是：基于对以往现场数据进行分析和掌握神经元网络模型,并结合炼钢的实际工艺特点,确定神经网络模型的各种参数,进而获得终点预报模型,再根据已经训练好的神经网络模型对现场采集来的数据进行预报。
在精炼炉炼钢的过程中,其中一个重要的控制目标是精炼炉炼钢终点温度和钢水成分。本课题首先应用统计模型进行终点预报,最后应用一种改进BP神经网络,通过建立钢水的终点温度、成分与各影响因素之间的模型,对炼钢终点进行预报。其预报结果说明了这个神经网络模型有较强的自学习能力,并且收敛特性比传统算法更好,预报结果具有较高的精度。
As an important part of the advanced steel making product process in the world, the ladle furnace(LF) steel making is a kind of secondary smelting process which adds deoxidant and alloy to the molten steel to deoxygenate, desulphurize and uniform the component, in order to smelt high quality and special type steel. Thereinto, endpoint prediction of ladle furnace steel making is important part of the ladle furnace steel control. It is a very importunate question how to control temperature and component of steel on endpoint accurately. It is helpful to operator choosing the most effective control policy. The accurate prediction of LF's temperature and component of steel in endpoint is very helpful to organising production rationally and improving quality of steel and reducing steel making's cost. It is valuable to operator choosing the most effective control policy too.
The thesis have made a particular study to endpoint prediction of ladle furnace based on referring to the large numbers of literature. Knowing different prediction method, the paper have neural network predictive control as the basal prediction method, which considering practical process and actual research condition. The main idea of neural network predictive control is calculating the parameter of the model, and then predict the other spot data according the confirmed model and the controller, based on the analysis of the past spot data and the mastery of the previous neural network model and considering practical characteristic of steel making process.
The very important control objects of the LF steel making process is that temperature and component of steel in endpoint. Using an improved BP neural network model, a mathematical model of endpoint temperature and components and correlative factors are developed to predict the endpoint of LF-steel making. The prediction results showed the algorithm has comparatively strong ability of self-studying. And this mathematic algorithm has better convergence character than conventional BP one. The precision of the results is comparatively high.
本课题在查阅了大量国内外相关文献的基础上,对精炼炉终点预报方法作了深入的研究。在对各种预报方法了解的基础上,结合实际工艺和目前研究的实际条件决定采用改进的神经网络算法作为终点预报方法。它的基本思想是：基于对以往现场数据进行分析和掌握神经元网络模型,并结合炼钢的实际工艺特点,确定神经网络模型的各种参数,进而获得终点预报模型,再根据已经训练好的神经网络模型对现场采集来的数据进行预报。
在精炼炉炼钢的过程中,其中一个重要的控制目标是精炼炉炼钢终点温度和钢水成分。本课题首先应用统计模型进行终点预报,最后应用一种改进BP神经网络,通过建立钢水的终点温度、成分与各影响因素之间的模型,对炼钢终点进行预报。其预报结果说明了这个神经网络模型有较强的自学习能力,并且收敛特性比传统算法更好,预报结果具有较高的精度。
As an important part of the advanced steel making product process in the world, the ladle furnace(LF) steel making is a kind of secondary smelting process which adds deoxidant and alloy to the molten steel to deoxygenate, desulphurize and uniform the component, in order to smelt high quality and special type steel. Thereinto, endpoint prediction of ladle furnace steel making is important part of the ladle furnace steel control. It is a very importunate question how to control temperature and component of steel on endpoint accurately. It is helpful to operator choosing the most effective control policy. The accurate prediction of LF's temperature and component of steel in endpoint is very helpful to organising production rationally and improving quality of steel and reducing steel making's cost. It is valuable to operator choosing the most effective control policy too.
The thesis have made a particular study to endpoint prediction of ladle furnace based on referring to the large numbers of literature. Knowing different prediction method, the paper have neural network predictive control as the basal prediction method, which considering practical process and actual research condition. The main idea of neural network predictive control is calculating the parameter of the model, and then predict the other spot data according the confirmed model and the controller, based on the analysis of the past spot data and the mastery of the previous neural network model and considering practical characteristic of steel making process.
The very important control objects of the LF steel making process is that temperature and component of steel in endpoint. Using an improved BP neural network model, a mathematical model of endpoint temperature and components and correlative factors are developed to predict the endpoint of LF-steel making. The prediction results showed the algorithm has comparatively strong ability of self-studying. And this mathematic algorithm has better convergence character than conventional BP one. The precision of the results is comparatively high.
引文
1.徐丽娜.神经网络控制[M],北京：电子工业出版社,2003,60～61.
2.张代远.神经网络新理论与方法[M],北京：清华大学出版社,2006,49～50.
3.肖静,周传平,郑冬喜.浅谈BP神经网络预测模型[J],九江大学学报,2007,33(2)：57～60.
4.艾立群.人工神经网络在钢铁工业中的应用[J],钢铁研究学报,1997,9(4)：60-62.
5.严六明等.运用化学键参数与神经网络模型在氧化物相图分析中的应用[J],科学通讯,1994(4)：36.
6.蔡煜东等.Kohonen自组织网络对金属间化合物结构的分析[J],金属学报,1995,31(6)：B280.
7.杨尚宝等.神经网络高炉专家系统[J],北京科技大学学报,1995,17(6)：524.
8.张文,张利,郭永新.基于人工神经元网络的预想事故电压预测[J],山东工业大学学报,2000,30(1)：42～46.
9.魏虹.基于人工神经网络温度场物性参数模型研究[J],内蒙古工业大学学报,2000,19(3)：183～186.
10.高宪文.电弧炉炼钢过程建模与智能优化控制[M],沈阳：东北大学出版社,1999.4,38～55.
11. Shi-feng Zhang, Shao-de Zhang, Li Kun, Zheng Xiao. Model Reference Adaptive Control based on Neural Network for Electrode System in Electric Arc Furnace [J], IEEE Trans. On Neural Networks,2006,5(1):49～55.
12.刘德明.电弧炉炼钢新技术[C],钢铁冶炼新技术讲座,2005,130(5)：53～58.
13.吕铭,付博,孟宪俭等.精炼炉工艺[J],莱钢科技,2007,127(1)：10～13.
14.张鉴.炉外精炼的理论与实践[M],北京：冶金工业出版社,1993,514～539.
15. Benoit Boulet, Gino Lalli, Mark Ajersch. Modeling and Control of an Electric Arc Furnace[C], IEEE Proceedings of the American Control Conference, Denver, Coiorado 2003,6,3060～3064.
16.郑淑胜,唐立冬,张茂存。济钢120tLF精炼炉的工艺优化[J],山东冶金,2006,28(3)：26～28.
17.杜勇,彭家清,姬健营.100t转炉LF精炼工艺的生产实践[J],中国冶金,2006,16(8)：17～20.
18.董锋斌,蒋军,王智忠等.钢包精炼炉智能底吹氩控制系统[J],机床与液压,2006, 8：195～197.
19.刘浏.中国炼钢技术的发展、创新与展望[J],炼钢,2007,23(2)：1～6.
20.黄华,蔡继明.现代电炉炼钢技术发展趋势[J],特钢技术,2006,12(49)：58～62.
21.李纯义,杨桂星.直流电弧炉冶炼及底电极温度分布试验[J],电弧炉—炉外精炼技术,第3辑：21.
22. Yan Xue, Ke Liu. Analysis of variable-sampling networked control system based on neural network [C], roceeding of the 2007 International Conference on Wavelet Analysis and Pattern Recognition, Beijing, China,2-4 Nov.2007,772～777.
23. M.Tossavainen, F. Engstrom, Q.Yang, N.Menad, M.Lidstrom Larsson, B.Bjorknian. Characteristics of steel slag under different cooling conditions[J], Waste Management, 2007,10(27):1335～1344.
24. Koo Y S, Kang T, Shin YK, etal. Thermal cycle model of ladle for steel temperature con-trol in melt shop and its application[C], Steelmaking Conference Proceeding, Chicago, Illinois; USA,1989(72):415～421.
25. Anon. Intelligent arc furnace operation at Birmingham steel [J], Steel Times International, 1996,20(1):20～21.
26. Olika B, Bjokman B. Prediction of steel temperature in ladle through time temperature simulation [J], Scandinavian Journal of Metallurgy,1993,22(4):213～218.
27. Castillejos A H, Acosta F A G, Betancourt A S, etal. On-line modeling for temperature control of ladles and steel during continuous thin slab casting[J], Iron &Steel Maker,1997, 24(7):53～63.
28. Amlan D, Mavoori H, Prem K K, etal. Adaptive Neural Net(ANN)Models for Desulphu-rization of Hot Metal and Steel[J], Steel Research,1994,65(11):466～471.
29.彭小奇,胡志坤,梅炽.炼铜转炉吹炼终点的神经网络和自适应残差补偿组合预测模型[J],控制理论与应用,2002,19(1)：149～151.
30. http://www.prosimu.net/simu-8.htm
31.李亮,姜周华,王文忠等.应用神经网络技术预报VD炉终点钢水温度[J],钢铁研究学报,2003,15(3)：56～59.
32.谢书明,陶钧,柴天佑.基于神经网络的转炉炼钢终点控制[J],控制理论与应用,2003,20(6)：903～907.
33.杨遴杰,陈伟庆,于平.LF-VD-CC钢液温度预报[J],钢铁,2000,35(1)：13～16.
34. Zhang Chun-xia, Wang Bao-jun, Zhou Shi-guang,etal. Hybrid Neural Network Model for RH Vacuum Refining Process Control [J], Iron and Steel Research International,2004, 11(1):12～16.
35.刘錕,刘浏,何平.增量模型预报电弧炉终点碳含量及温度的研究[J],冶金自动化, 2007,1：5～8.
36.杨虎,钟波等.应用数理统计[M],清华大学出版社,2006,217.
37.廖晓峰,李传东.神经网络研究发展趋势[J],国际学术动态,2006,5：43～46.
38.周开利,康耀红.神经网络模型及其MATLAB仿真程序设计[M],北京：清华大学出版社,2006,8-9.
39.李素杰,申东日,陈义俊等.一种基于神经网络辨识的预测方法[J],甘肃科学学报,2006,18(2)：66～69.
40. Jin-Quan Huang, F.L.Lewis. Neural-network predictive for nonlinear dynamic systems with time-delay[J], IEEE Tanasactions on Neural Networks March 2003,14(2).
41.蒋绍飞,张春丽,钟善桐.BP网络模型的改进方法探讨[J],哈尔滨建筑大学学报,2000,33(5)：57～60.
42. D.L.Yu, J.B.Gomm. Implementation of neural network control to a multivariable chemical reactor[J], Control Engineering Practice 2003,11,1315～1323.
43.飞思科技产品研发中心.神经网络理论与Matlab7实现[M],北京：电子工业出版社,2005：257～297.
2.张代远.神经网络新理论与方法[M],北京：清华大学出版社,2006,49～50.
3.肖静,周传平,郑冬喜.浅谈BP神经网络预测模型[J],九江大学学报,2007,33(2)：57～60.
4.艾立群.人工神经网络在钢铁工业中的应用[J],钢铁研究学报,1997,9(4)：60-62.
5.严六明等.运用化学键参数与神经网络模型在氧化物相图分析中的应用[J],科学通讯,1994(4)：36.
6.蔡煜东等.Kohonen自组织网络对金属间化合物结构的分析[J],金属学报,1995,31(6)：B280.
7.杨尚宝等.神经网络高炉专家系统[J],北京科技大学学报,1995,17(6)：524.
8.张文,张利,郭永新.基于人工神经元网络的预想事故电压预测[J],山东工业大学学报,2000,30(1)：42～46.
9.魏虹.基于人工神经网络温度场物性参数模型研究[J],内蒙古工业大学学报,2000,19(3)：183～186.
10.高宪文.电弧炉炼钢过程建模与智能优化控制[M],沈阳：东北大学出版社,1999.4,38～55.
11. Shi-feng Zhang, Shao-de Zhang, Li Kun, Zheng Xiao. Model Reference Adaptive Control based on Neural Network for Electrode System in Electric Arc Furnace [J], IEEE Trans. On Neural Networks,2006,5(1):49～55.
12.刘德明.电弧炉炼钢新技术[C],钢铁冶炼新技术讲座,2005,130(5)：53～58.
13.吕铭,付博,孟宪俭等.精炼炉工艺[J],莱钢科技,2007,127(1)：10～13.
14.张鉴.炉外精炼的理论与实践[M],北京：冶金工业出版社,1993,514～539.
15. Benoit Boulet, Gino Lalli, Mark Ajersch. Modeling and Control of an Electric Arc Furnace[C], IEEE Proceedings of the American Control Conference, Denver, Coiorado 2003,6,3060～3064.
16.郑淑胜,唐立冬,张茂存。济钢120tLF精炼炉的工艺优化[J],山东冶金,2006,28(3)：26～28.
17.杜勇,彭家清,姬健营.100t转炉LF精炼工艺的生产实践[J],中国冶金,2006,16(8)：17～20.
18.董锋斌,蒋军,王智忠等.钢包精炼炉智能底吹氩控制系统[J],机床与液压,2006, 8：195～197.
19.刘浏.中国炼钢技术的发展、创新与展望[J],炼钢,2007,23(2)：1～6.
20.黄华,蔡继明.现代电炉炼钢技术发展趋势[J],特钢技术,2006,12(49)：58～62.
21.李纯义,杨桂星.直流电弧炉冶炼及底电极温度分布试验[J],电弧炉—炉外精炼技术,第3辑：21.
22. Yan Xue, Ke Liu. Analysis of variable-sampling networked control system based on neural network [C], roceeding of the 2007 International Conference on Wavelet Analysis and Pattern Recognition, Beijing, China,2-4 Nov.2007,772～777.
23. M.Tossavainen, F. Engstrom, Q.Yang, N.Menad, M.Lidstrom Larsson, B.Bjorknian. Characteristics of steel slag under different cooling conditions[J], Waste Management, 2007,10(27):1335～1344.
24. Koo Y S, Kang T, Shin YK, etal. Thermal cycle model of ladle for steel temperature con-trol in melt shop and its application[C], Steelmaking Conference Proceeding, Chicago, Illinois; USA,1989(72):415～421.
25. Anon. Intelligent arc furnace operation at Birmingham steel [J], Steel Times International, 1996,20(1):20～21.
26. Olika B, Bjokman B. Prediction of steel temperature in ladle through time temperature simulation [J], Scandinavian Journal of Metallurgy,1993,22(4):213～218.
27. Castillejos A H, Acosta F A G, Betancourt A S, etal. On-line modeling for temperature control of ladles and steel during continuous thin slab casting[J], Iron &Steel Maker,1997, 24(7):53～63.
28. Amlan D, Mavoori H, Prem K K, etal. Adaptive Neural Net(ANN)Models for Desulphu-rization of Hot Metal and Steel[J], Steel Research,1994,65(11):466～471.
29.彭小奇,胡志坤,梅炽.炼铜转炉吹炼终点的神经网络和自适应残差补偿组合预测模型[J],控制理论与应用,2002,19(1)：149～151.
30. http://www.prosimu.net/simu-8.htm
31.李亮,姜周华,王文忠等.应用神经网络技术预报VD炉终点钢水温度[J],钢铁研究学报,2003,15(3)：56～59.
32.谢书明,陶钧,柴天佑.基于神经网络的转炉炼钢终点控制[J],控制理论与应用,2003,20(6)：903～907.
33.杨遴杰,陈伟庆,于平.LF-VD-CC钢液温度预报[J],钢铁,2000,35(1)：13～16.
34. Zhang Chun-xia, Wang Bao-jun, Zhou Shi-guang,etal. Hybrid Neural Network Model for RH Vacuum Refining Process Control [J], Iron and Steel Research International,2004, 11(1):12～16.
35.刘錕,刘浏,何平.增量模型预报电弧炉终点碳含量及温度的研究[J],冶金自动化, 2007,1：5～8.
36.杨虎,钟波等.应用数理统计[M],清华大学出版社,2006,217.
37.廖晓峰,李传东.神经网络研究发展趋势[J],国际学术动态,2006,5：43～46.
38.周开利,康耀红.神经网络模型及其MATLAB仿真程序设计[M],北京：清华大学出版社,2006,8-9.
39.李素杰,申东日,陈义俊等.一种基于神经网络辨识的预测方法[J],甘肃科学学报,2006,18(2)：66～69.
40. Jin-Quan Huang, F.L.Lewis. Neural-network predictive for nonlinear dynamic systems with time-delay[J], IEEE Tanasactions on Neural Networks March 2003,14(2).
41.蒋绍飞,张春丽,钟善桐.BP网络模型的改进方法探讨[J],哈尔滨建筑大学学报,2000,33(5)：57～60.
42. D.L.Yu, J.B.Gomm. Implementation of neural network control to a multivariable chemical reactor[J], Control Engineering Practice 2003,11,1315～1323.
43.飞思科技产品研发中心.神经网络理论与Matlab7实现[M],北京：电子工业出版社,2005：257～297.