基于信息融合技术的电弧炉终点预报方法的研究
|
摘要
铁合金冶炼行业是我国重要的基础产业部门之一,对我国社会经济发展起着至关重要的作用,在铸造、化学和有色金属冶炼中广泛应用。但是,我国的铁合金企业,与西方发达国家的同行相比,自动化程度低,能耗高,资源浪费严重,由此而造成企业生产成本高,缺乏市场竞争力。
在铁合金生产过程中,耗能设备主要是电弧炉。由于电弧炉炉温高、粉尘大、冶炼条件恶劣,电弧炉炉内的温度和熔池参数不宜直接连续检测。因此,选择先进的控制方法实现,对电弧炉冶炼锰硅合金的终点进行预报,能够有效的缩短电弧炉的冶炼周期,从而降低冶炼成本,并且提高生产效率。
针对冶炼锰硅合金的特点,本文以吉林某铁合金厂生产实际为背景,提出以多传感器信息融合技术为基础,对电弧炉冶炼锰硅合金的终点时刻进行预报,主要对以下几个方面的进行了研究：
(1)依托国家科技支撑计划项目,在对吉林某铁合金厂冶炼锰硅合金的终点判断调研的基础上,分析终点预报的内在特性；
(2)针对冶炼锰硅合金冶炼的实际环境和传统支持向量机的不足,提出对支持向量机进行多尺度分解,即多分辨率支持向量机；
(3)围绕LS-SVM算法,建立了LS-SVM分层结构模型,并详细描述了融合过程中LS-SVM的训练和测试两个重要环节,并对其测试结果进行MATLAB仿真；
(4)依托课题研究背景,将SVM模型应用于解决电弧炉冶炼锰硅合金的终点预报上,实验结果验证了本算法的有效性；
(5)针对实际生产控制需要,采用基于WEB系统开发的Java语言,结合Jsp、Javabean技术,实现实际问题的网络化。
本文为吉林铁合金冶炼锰硅合金终点判断提出了一条新的解决方案,打破传统的人工经验判断方式,对企业的节能降耗,具有广泛的实际应用前景。
Metallurgical industry is one of important basic industry sectors in China, which plays a key role on social economic development,it has been widely used in foundry industry, non-ferrous metals smelting and chemical industry. However, compared with peers in the western developed country, the ferroalloy enterprise of China has obvious drawbacks that include the low degree of automation, high energy consumption, high production cost and lack of market competitiveness.
The electric arc furnace (EAF) is the main dissipation energy equipment in manufacturing process of ferroalloy. Owing to the high temperature in electric arc furnace, excessive dust and abominable working condition, the composition and temperature can not be measured directly. So selecting an advanced control method to predicting the end point of smelting Mn-Si alloy process can reduce the loss of electric machine, consumption of resources and shorten melting time, thereby smelting costs are reduced and production rate is raised
This thesis is based on electric arc furnace of the ferroalloy subsidiary factory of Jilin steel group. The thesis includes the historic course, the actual research and the tendency of EAF ferroalloy smelting, which is based on the extensive knowledge of the craftwork of EAF and the development of ferroalloy smelting process of the prediction of end-point at home and on aboard, which all through study on a great deal of documents about prediction of end-point of the EAF.This thesis is devoted to theoretical research as follows:
(1) Among the country support science and technology project, On the basis of investigation to terminal judgment of smelting Mn-Si alloy in the ferroalloy subsidiary factory of Jilin steel group, internal quality of end-point prediction subjectes to analysis,which rely on the country to support science and technology projects.
(2) Considering the faults of the real condition of smelting Mn-Si alloy and support vector machine, it is multi-resolved on the basis of support vector machine, which is called multiresolution support vector machine.
(3) The LS-SVM layered model is established around the LS-SVM algorithm and describes two important links of LS-SVM training and testing in fusion process minutely. Then, by using MATLAB software the testing result is simulated.
(4) Relying on the research background, the model of the LS-SVM was applied to solve the problems of end-point prediction.
(5) Designed the Man-Machine Interface with the practical conditions,the interface is developed by Java which based on WEB system, comparing Jsp and Javabean technology
A new produce model different from tradition artificial experience judgment model is proposed.Meanwhile a new enery saving solution with a great application proposed for the prediction of end-point of the EAF.
在铁合金生产过程中,耗能设备主要是电弧炉。由于电弧炉炉温高、粉尘大、冶炼条件恶劣,电弧炉炉内的温度和熔池参数不宜直接连续检测。因此,选择先进的控制方法实现,对电弧炉冶炼锰硅合金的终点进行预报,能够有效的缩短电弧炉的冶炼周期,从而降低冶炼成本,并且提高生产效率。
针对冶炼锰硅合金的特点,本文以吉林某铁合金厂生产实际为背景,提出以多传感器信息融合技术为基础,对电弧炉冶炼锰硅合金的终点时刻进行预报,主要对以下几个方面的进行了研究：
(1)依托国家科技支撑计划项目,在对吉林某铁合金厂冶炼锰硅合金的终点判断调研的基础上,分析终点预报的内在特性；
(2)针对冶炼锰硅合金冶炼的实际环境和传统支持向量机的不足,提出对支持向量机进行多尺度分解,即多分辨率支持向量机；
(3)围绕LS-SVM算法,建立了LS-SVM分层结构模型,并详细描述了融合过程中LS-SVM的训练和测试两个重要环节,并对其测试结果进行MATLAB仿真；
(4)依托课题研究背景,将SVM模型应用于解决电弧炉冶炼锰硅合金的终点预报上,实验结果验证了本算法的有效性；
(5)针对实际生产控制需要,采用基于WEB系统开发的Java语言,结合Jsp、Javabean技术,实现实际问题的网络化。
本文为吉林铁合金冶炼锰硅合金终点判断提出了一条新的解决方案,打破传统的人工经验判断方式,对企业的节能降耗,具有广泛的实际应用前景。
Metallurgical industry is one of important basic industry sectors in China, which plays a key role on social economic development,it has been widely used in foundry industry, non-ferrous metals smelting and chemical industry. However, compared with peers in the western developed country, the ferroalloy enterprise of China has obvious drawbacks that include the low degree of automation, high energy consumption, high production cost and lack of market competitiveness.
The electric arc furnace (EAF) is the main dissipation energy equipment in manufacturing process of ferroalloy. Owing to the high temperature in electric arc furnace, excessive dust and abominable working condition, the composition and temperature can not be measured directly. So selecting an advanced control method to predicting the end point of smelting Mn-Si alloy process can reduce the loss of electric machine, consumption of resources and shorten melting time, thereby smelting costs are reduced and production rate is raised
This thesis is based on electric arc furnace of the ferroalloy subsidiary factory of Jilin steel group. The thesis includes the historic course, the actual research and the tendency of EAF ferroalloy smelting, which is based on the extensive knowledge of the craftwork of EAF and the development of ferroalloy smelting process of the prediction of end-point at home and on aboard, which all through study on a great deal of documents about prediction of end-point of the EAF.This thesis is devoted to theoretical research as follows:
(1) Among the country support science and technology project, On the basis of investigation to terminal judgment of smelting Mn-Si alloy in the ferroalloy subsidiary factory of Jilin steel group, internal quality of end-point prediction subjectes to analysis,which rely on the country to support science and technology projects.
(2) Considering the faults of the real condition of smelting Mn-Si alloy and support vector machine, it is multi-resolved on the basis of support vector machine, which is called multiresolution support vector machine.
(3) The LS-SVM layered model is established around the LS-SVM algorithm and describes two important links of LS-SVM training and testing in fusion process minutely. Then, by using MATLAB software the testing result is simulated.
(4) Relying on the research background, the model of the LS-SVM was applied to solve the problems of end-point prediction.
(5) Designed the Man-Machine Interface with the practical conditions,the interface is developed by Java which based on WEB system, comparing Jsp and Javabean technology
A new produce model different from tradition artificial experience judgment model is proposed.Meanwhile a new enery saving solution with a great application proposed for the prediction of end-point of the EAF.
引文
[1]袁平,电弧炉冶炼过程先进控制方法的研究与应用[D],东北大学,2005。
[2]冯波,多传感器信息融合技术的研究[D],南京航空航天大学,2004.2。
[3]王雷,基于多源信息融合的驾驶员跟车行为研究[D],山东理工大学,2007.10。
[4]咸宝金,基于专家系统的数据融合技术及在机器人避障中的应用[D], 北方工业大学,2008。
[5]Amari S,Wu S.Improving support vector machine classifier by modifying kernel functions[M].Neural Networks,1999,12(9):783-789.
[6]王志强,邓玉花,浅谈如何提高锰硅合金中锰的回收率[J],甘肃科技纵横,2010,39(3),38-39.
[7]Scholkopf B.,Smola A.,Williamson R.C.etal.NewSupport Vector Algorithms[M].Neural Computation,2000,12(5):1207-1245.
[8]Scholkopf B.,J.C.Plat,J.Shawe-Taylor etal.Estimating the Support of a High dimensional Distribution[M]. Neural Computation,2001,13(7):1443-1471.
[9]Lee Y.J.,Mangasarian O.L..RSVM:Reduced Support Vector Machines[J].Proceeding of the First SIAM International Conference on Data Mining 2001.
[10]Chew H G, Bonger Robert E, Lim C C.Dual No-support Vector Machine with Error Rate and Training Size Biasing[J].Proeeedings of 26 th IEEE ICASSP2001, salt LakeCity, USA,2001:1269-1272.
[11]Suykens J., Vandewalle J..Least Square Support Vector Machine Classifiers[J].Neural Processing Letters,1999,9(3):293-300.
[12]Tax D.,Duin R..Data Domain Description by Support Vector[J].Proceedings of ESANN.1999,251-256.
[13]王志勇,郭创新,曹一家,改进范例推理在短期负荷预测中的应用[J]。电力系统自动化,2005,29(12)：33-36.
[14]穆向阳,张太镒,周亚同,尺度核函数在最小二乘支持向量机信号逼近中的应用[J]。西安交通大学学报,2008,12：1464-1467.
[15]张素超,朱子宗,秦玉廷,王勇,高家城,锰硅合金生产节能研究[J],过程工程学报,2009,S1：142-145.
[16]陈永义,计算机学习的SVM方法与应用软件平台CMSVM[J],上海大学学报,2004,10：56-58.
[17]CHAPELLE O,VAPNIK V. Model selection for sup-port vector machines [M] Advances in Neural Information Processing Systems. Cambridge, MA, USA:MIT Press,2000:349-355.
[18]罗林开,支持向量机的核选择[D],厦门大学,2007.
[19]OPFER R. Tight frame expansions of multiscale reproducing kernels in Sobolev spaces[J]. Applied and Computational Harmonic Analysis,2006,20(1):357-374.
[20]郭晓妮,基于改进的SVM交通信息融合算法及应用研究[D],北京交通大学,2009.
[21]周雪梅,基于多尺度估计理论的组合导航系统研究[D],哈尔滨工程大学,2006.
[22]赵娜乐,丁雷,耿彦斌,陈旭海.基于SVM的数据多元ITS数据融合方法初探[J].交通运输系统工程信息,2007,4(7)：32-38.
[23]Koch K R,Yang Y.Rubust Kalman filter for rank deficient observation model[J] Journal of Geodesy,1998,72:436-441.
[24]Box QTiao G.Beyesian Inference in Statistical Analysis[M].Reading,MA:Addison Wesley,1973.
[25]H.F.Durrant-whyte.Sensor models and multisensor integration.Int[J].Robot.Res.1988, 7(6):97-113.
[26]马素丽,噪声环境下说话人识别技术研究[D],电子科技大学,2008.
[27]吴冷冬,基于支持向量机电话语音情感识别方法的研究与实现[D],北京大学,2009.
[28]刘洁,基于支持向量机的网络入侵检测系统研究[D],中南大学,2008
[29]李忠伟,支持向量机学习算法研究[D],哈尔滨工程大学,2006
[30]曾志明,网站开发技术的比较研究,电脑知识与技术[J],2010,6(5),1075-1078.
[31]文成林,潘泉,张洪才,戴冠中.多传感器多模型动态系统多尺度融合估计算法,自动化学报.2000,26(SB)：66-70.
[32]袁彦芹,基于支持向量机的大规模文本分类研究与设计[D],山东师范大学2007.
[33]师帅,基于数据融合技术的烟化炉冶炼终点判断研究[D],昆明理工大学,2005
[34]朱慧,庄进,邓益,支持向量机算法及其应用的研究[J],北京电力专科学校学报,2010,29(3),134-135.
[35]邓开楠,基于专家系统的电弧炉炉况判断[D],东北大学,2005.
[36]J.Platt.Using Analytic QP and Sparseness to Speed Training of Support Vector Maehines[J].Advances in Neural Information Processing Systems,MITPress, Cambridge,MA,1999,11:557-563.
[37]聂捷楠,EMIS(教学管理信息系统)的研究与开发[D],西南交通大学,2010.
[38]楼新恒,水电站技术数据管理系统的研究与开发[D],中国农业大学,2006.
[39]贾建华,王军峰,冯冬青.人工神经网络在多传感器信息融合中的应用研究[J].微计算机信息,2006,7：192—194.
[40]柳晓菁,易建强,赵冬斌,王伟,基于最小二乘支持向量机的自适应逆扰动消除控制系统[J],西安交通大学学报,2008,12：1466-1469.
[41]Gen Junbao, Huang Shuhong. A method of rotating maehinery fault diagnosis based on the close degree of information entropy[J].International journal of Plant Engineering and Management,2006,11(3):137-143.
[42]严怀成,黄心汉,王敏.多传感器数据融合技术及其应用[J],传感器技术,2005,24(10),1-4.
[43]张雨,温熙森.设备故障信息融合问题的思考[J].长沙交通学院学报,1999,15：22-29.
[44]张永,基于模糊支持向量机的多类分类算法研究[D],大连理工大学,2008.
[45]王安娜、田慧欣,基于信息融合算法的LF炉钢水温度预测[J],钢铁研究学报,17(6)：P71～74,2005.
[46]何祥林,机器人中的多传感器信息融合技术[D],2006.
[47]刘海霞,基于多传感器行为融合基础上的AGV导航研究[D],合肥工业大学,2006.
[48]胡文波,HLA等位基因频率数据库的构建[D],中山大学,2009.
[2]冯波,多传感器信息融合技术的研究[D],南京航空航天大学,2004.2。
[3]王雷,基于多源信息融合的驾驶员跟车行为研究[D],山东理工大学,2007.10。
[4]咸宝金,基于专家系统的数据融合技术及在机器人避障中的应用[D], 北方工业大学,2008。
[5]Amari S,Wu S.Improving support vector machine classifier by modifying kernel functions[M].Neural Networks,1999,12(9):783-789.
[6]王志强,邓玉花,浅谈如何提高锰硅合金中锰的回收率[J],甘肃科技纵横,2010,39(3),38-39.
[7]Scholkopf B.,Smola A.,Williamson R.C.etal.NewSupport Vector Algorithms[M].Neural Computation,2000,12(5):1207-1245.
[8]Scholkopf B.,J.C.Plat,J.Shawe-Taylor etal.Estimating the Support of a High dimensional Distribution[M]. Neural Computation,2001,13(7):1443-1471.
[9]Lee Y.J.,Mangasarian O.L..RSVM:Reduced Support Vector Machines[J].Proceeding of the First SIAM International Conference on Data Mining 2001.
[10]Chew H G, Bonger Robert E, Lim C C.Dual No-support Vector Machine with Error Rate and Training Size Biasing[J].Proeeedings of 26 th IEEE ICASSP2001, salt LakeCity, USA,2001:1269-1272.
[11]Suykens J., Vandewalle J..Least Square Support Vector Machine Classifiers[J].Neural Processing Letters,1999,9(3):293-300.
[12]Tax D.,Duin R..Data Domain Description by Support Vector[J].Proceedings of ESANN.1999,251-256.
[13]王志勇,郭创新,曹一家,改进范例推理在短期负荷预测中的应用[J]。电力系统自动化,2005,29(12)：33-36.
[14]穆向阳,张太镒,周亚同,尺度核函数在最小二乘支持向量机信号逼近中的应用[J]。西安交通大学学报,2008,12：1464-1467.
[15]张素超,朱子宗,秦玉廷,王勇,高家城,锰硅合金生产节能研究[J],过程工程学报,2009,S1：142-145.
[16]陈永义,计算机学习的SVM方法与应用软件平台CMSVM[J],上海大学学报,2004,10：56-58.
[17]CHAPELLE O,VAPNIK V. Model selection for sup-port vector machines [M] Advances in Neural Information Processing Systems. Cambridge, MA, USA:MIT Press,2000:349-355.
[18]罗林开,支持向量机的核选择[D],厦门大学,2007.
[19]OPFER R. Tight frame expansions of multiscale reproducing kernels in Sobolev spaces[J]. Applied and Computational Harmonic Analysis,2006,20(1):357-374.
[20]郭晓妮,基于改进的SVM交通信息融合算法及应用研究[D],北京交通大学,2009.
[21]周雪梅,基于多尺度估计理论的组合导航系统研究[D],哈尔滨工程大学,2006.
[22]赵娜乐,丁雷,耿彦斌,陈旭海.基于SVM的数据多元ITS数据融合方法初探[J].交通运输系统工程信息,2007,4(7)：32-38.
[23]Koch K R,Yang Y.Rubust Kalman filter for rank deficient observation model[J] Journal of Geodesy,1998,72:436-441.
[24]Box QTiao G.Beyesian Inference in Statistical Analysis[M].Reading,MA:Addison Wesley,1973.
[25]H.F.Durrant-whyte.Sensor models and multisensor integration.Int[J].Robot.Res.1988, 7(6):97-113.
[26]马素丽,噪声环境下说话人识别技术研究[D],电子科技大学,2008.
[27]吴冷冬,基于支持向量机电话语音情感识别方法的研究与实现[D],北京大学,2009.
[28]刘洁,基于支持向量机的网络入侵检测系统研究[D],中南大学,2008
[29]李忠伟,支持向量机学习算法研究[D],哈尔滨工程大学,2006
[30]曾志明,网站开发技术的比较研究,电脑知识与技术[J],2010,6(5),1075-1078.
[31]文成林,潘泉,张洪才,戴冠中.多传感器多模型动态系统多尺度融合估计算法,自动化学报.2000,26(SB)：66-70.
[32]袁彦芹,基于支持向量机的大规模文本分类研究与设计[D],山东师范大学2007.
[33]师帅,基于数据融合技术的烟化炉冶炼终点判断研究[D],昆明理工大学,2005
[34]朱慧,庄进,邓益,支持向量机算法及其应用的研究[J],北京电力专科学校学报,2010,29(3),134-135.
[35]邓开楠,基于专家系统的电弧炉炉况判断[D],东北大学,2005.
[36]J.Platt.Using Analytic QP and Sparseness to Speed Training of Support Vector Maehines[J].Advances in Neural Information Processing Systems,MITPress, Cambridge,MA,1999,11:557-563.
[37]聂捷楠,EMIS(教学管理信息系统)的研究与开发[D],西南交通大学,2010.
[38]楼新恒,水电站技术数据管理系统的研究与开发[D],中国农业大学,2006.
[39]贾建华,王军峰,冯冬青.人工神经网络在多传感器信息融合中的应用研究[J].微计算机信息,2006,7：192—194.
[40]柳晓菁,易建强,赵冬斌,王伟,基于最小二乘支持向量机的自适应逆扰动消除控制系统[J],西安交通大学学报,2008,12：1466-1469.
[41]Gen Junbao, Huang Shuhong. A method of rotating maehinery fault diagnosis based on the close degree of information entropy[J].International journal of Plant Engineering and Management,2006,11(3):137-143.
[42]严怀成,黄心汉,王敏.多传感器数据融合技术及其应用[J],传感器技术,2005,24(10),1-4.
[43]张雨,温熙森.设备故障信息融合问题的思考[J].长沙交通学院学报,1999,15：22-29.
[44]张永,基于模糊支持向量机的多类分类算法研究[D],大连理工大学,2008.
[45]王安娜、田慧欣,基于信息融合算法的LF炉钢水温度预测[J],钢铁研究学报,17(6)：P71～74,2005.
[46]何祥林,机器人中的多传感器信息融合技术[D],2006.
[47]刘海霞,基于多传感器行为融合基础上的AGV导航研究[D],合肥工业大学,2006.
[48]胡文波,HLA等位基因频率数据库的构建[D],中山大学,2009.