流程对象建模方法的研究与实现
|
摘要
现代流程工业的一个重要特点是向大型化和综合自动化方向发展。流程工业综合自动化通过集成过程控制,经营管理,计划调度和市场销售等技术手段,实现全局,局部各级优化,以最短的周期,最小的成本产生最大的经济效益。Babnde A. Ogunnaikeatu讲述了现代流程工业面临的五大问题:在线测量,严重的非线性,控制系统的建模和辨识,为仿真和执行机构的训练而进行的建模,过程检测和故障诊断。其中,一个重要问题就是弄清楚系统的运作机制,建立系统的描述模型,以便进行严格的科学管理和规划。
在流程工业中,水泥工业是国民经济发展的支柱产业之一,也是能耗较大的工业之一,其工艺特征也表现出明显的流程生产特性。与化工企业相比,水泥生产过程的不可控性,特别是对象的不可控因素因其物料的制备和煅烧过程的复杂性而非常难以分析。60年代末,随着电子计算机及其计算技术的发展,运用计算机来求解工程问题的最优方案,已逐渐成为解决实际设计问题的重要手段。它不仅大大提高了解决问题的效率和可靠性,而且解决了许多过去认为不可求解的实际问题。特别是近些年来,随着世界经济全球化发展的不断深入,各国流程工业更是以迅猛的速度增长,同时,流程工业生产过程具有高度的复杂性、强关联性、非线性、以及不确定性特点。生产规模的扩大势必进一步增加生产过程参数的复杂性以及参数之间的关联性。目前对流程工业的研究主要集中在过程优化控制上,而忽略了过程优化控制需要用到的参数的选择问题,参数选择的恰当与否对过程优化控制的效果有直接的影响。
基于以上对流程工业的分析,本课题以新型干法水泥生产线为研究对象,主要研究内容是把预分解炉、回转窑和篦冷机作为一个整体进行研究,建立其生产过程模型。其中预分解炉的主要负责水泥生料的分解,生料在分解炉的分解率可达到90%-95%;从分解炉出来的物料由窑尾进入窑体,回转窑一方面使未分解的生料进一步分解,另一方面使水泥生料煅烧为水泥熟料;水泥熟料由回转窑的窑头流向篦冷机的篦床后进行冷却,所以预分解炉、回转窑和篦冷机三个工艺过程是一个紧密连接,相互配合的整体,水泥质量的好坏不是取决于某一个生产工艺环节,而是依赖于三个过程的整体运行状况。因此本课题利用发展较为成熟的神经网络技术和数据挖掘技术,并且通过与经验丰富的水泥生产专家进行交流和沟通,抽取影响这个三个生产过程的主要参数,然后利用本文用到的柔性神经数模型对整个生产过程进行建模。本文在算法实现上的创新点:传统的FNT模型的进化代数是一个固定值,这是没有科学依据的,最优的模型可能在这个固定值之后出现,也可能在这个值之前就已经出现,这不仅浪费时间也使最优的模型随进化代数的增加而消失。本算法的改进之处是用平均误差率控制进化代数,当模型进化到最好的时候即误差率最小的时候则停止进化。利用此模型可以对流程工业的生产过程控制参数进行自动筛选,从而找到影响生产过程的重要参数,为流程工业的生产控制提供理论依据,使流程工业生产过程控制更加具有科学性和针对性,取得更好的优化控制效果。
在建模的过程中筛选出对水泥生产过程产生影响最大的因素,此建模的过程的结构优化是利用概率增强式程序进化(Probabilistic Incremental Program Evolution, PIPE)算法进行的;然后找出这些主要参数之间的关联,最后从它们的关联中找到最优的组合方案,即模型的优化,此优化过程是利用模拟退火(Simulation Annealing, SA)算法实现的。通过此建模过程不仅可以找到同一生产过程中参数之间的关系,也可以找出不同生产过程中参数间的比例关系。通过对整个生产过程进行建模和优化而达到稳定水泥生产过程,提高生料的分解率,降低用煤量的目的,把最终的研究方案利用DCS系统应用到实际的生产控制中。
One important feature of modern process industry is the advancement towards large-scale and integrated automation. Through the integration of process control, management, scheduling and marketing and other technical means, integrated Automation of Process Industry can achieve global and local optimization with the shortest cycle and minimum cost for the greatest economic benefit. Babnde A. Ogunnaikeatu discusses five problems facing modern process industry: online measurement, serious nonlinearity, modeling and recognition of the control system, modeling for the training of simulation and execution units, and process monitoring and fault diagnosis. Among them, an important issue is to clarify the operating mechanism of the system and to establish the system description model in order for strict scientific management and planning.
Among the process industry, cement industry is not only one of pillar industries of national economic development, but also one of larger energy consumption industries and its technological property obviously exhibits the characteristics of process production. In the end of the 60s, with the development of computer technology, the use of computers to solve optimal scheme of engineering problems has become an increasingly important means of solving practical design problems. It not only enhances the efficiency and reliability of the problem-solving greatly, but also resolves a lot of practical problems that could not be solved in the past. Especially in recent years, process industry is growing at a rapid pace with the development of globalization of world economy. At the same time, the production in process industry is high complex, strong associated, non-linear and indeterministic. The expansion of production scale will surely complicate the parameters of the process and enhance the association among them. At present, the research on process industry concentrates mainly on the control of the process optimization and neglects the problem of parameter selection required by process optimal control; however, proper selection of parameters will influence the effect of process optimal control directly.
Based on the above analysis of the process industry, this research is to treat pre-calciner, rotary kiln and grate cooler as a whole and to build its production process model. Decomposing furnace is mainly responsible for the decomposition of the raw material, and the decomposition rate could reach 90% -95%; then the decomposed material flows into rotary calciner, which takes on two tasks: one is to decompose the remaining raw material, and the other is to calcine the decomposed raw material; finally the calcined material is cooled in the grate bed. The three processes are closely linked with each other. Therefore, pre-calciner, rotary kiln and grate cooler are closely linked and mutually interacting. The quality of cement is not only dependent on one single process, but on the overall operation of the three courses. Using the sophisticated neural network technology and data mining techniques, and with the cooperation with experienced experts in cement production, we extract the parameters that influence the three production processes, and then use the flexible neural model to model the entire production process. The creative point of this work is as follows. The evolutional generation of traditional FNT model is fixed traditionally, yet the best model is not always formed in this generation, so to fix the evolutional generation is unreasonable. This improved algorithm uses mean error rate to control the number of evolutional generation instead of fixing it. This method provides theoretical basis for the production control of process industry so that the process control is more scientific and pertinent. On the other hand, it establishes a good basis for process optimal control of process industry. Using this method, we can not only obtain the best model and parameters, but enhanced the efficiency and accuracuy.
In the modeling process, the most important factor that impacts of cement production process is screened out, and this modeling process is based on Probabilistic Incremental Program Evolution (PIPE) algorithm. Then the link between these main parameters is found, and finally the best combination of programs from their association is found, namely, the optimization model. This optimization process uses the Simulation Annealing (SA) algorithm. Through this modeling process, not only the parameters relationship between the same productions processes are found, but the proportional relationship between parameters among a different manufacturing process is also found. Through the entire production process modeling and optimization, the purpose of stabilizing the cement production process and improving the rate of decomposition of raw materials and reducing the amount of coal consumption are reached. Finally, the ultimate research result will be applied to the practical production control using DCS system.
在流程工业中,水泥工业是国民经济发展的支柱产业之一,也是能耗较大的工业之一,其工艺特征也表现出明显的流程生产特性。与化工企业相比,水泥生产过程的不可控性,特别是对象的不可控因素因其物料的制备和煅烧过程的复杂性而非常难以分析。60年代末,随着电子计算机及其计算技术的发展,运用计算机来求解工程问题的最优方案,已逐渐成为解决实际设计问题的重要手段。它不仅大大提高了解决问题的效率和可靠性,而且解决了许多过去认为不可求解的实际问题。特别是近些年来,随着世界经济全球化发展的不断深入,各国流程工业更是以迅猛的速度增长,同时,流程工业生产过程具有高度的复杂性、强关联性、非线性、以及不确定性特点。生产规模的扩大势必进一步增加生产过程参数的复杂性以及参数之间的关联性。目前对流程工业的研究主要集中在过程优化控制上,而忽略了过程优化控制需要用到的参数的选择问题,参数选择的恰当与否对过程优化控制的效果有直接的影响。
基于以上对流程工业的分析,本课题以新型干法水泥生产线为研究对象,主要研究内容是把预分解炉、回转窑和篦冷机作为一个整体进行研究,建立其生产过程模型。其中预分解炉的主要负责水泥生料的分解,生料在分解炉的分解率可达到90%-95%;从分解炉出来的物料由窑尾进入窑体,回转窑一方面使未分解的生料进一步分解,另一方面使水泥生料煅烧为水泥熟料;水泥熟料由回转窑的窑头流向篦冷机的篦床后进行冷却,所以预分解炉、回转窑和篦冷机三个工艺过程是一个紧密连接,相互配合的整体,水泥质量的好坏不是取决于某一个生产工艺环节,而是依赖于三个过程的整体运行状况。因此本课题利用发展较为成熟的神经网络技术和数据挖掘技术,并且通过与经验丰富的水泥生产专家进行交流和沟通,抽取影响这个三个生产过程的主要参数,然后利用本文用到的柔性神经数模型对整个生产过程进行建模。本文在算法实现上的创新点:传统的FNT模型的进化代数是一个固定值,这是没有科学依据的,最优的模型可能在这个固定值之后出现,也可能在这个值之前就已经出现,这不仅浪费时间也使最优的模型随进化代数的增加而消失。本算法的改进之处是用平均误差率控制进化代数,当模型进化到最好的时候即误差率最小的时候则停止进化。利用此模型可以对流程工业的生产过程控制参数进行自动筛选,从而找到影响生产过程的重要参数,为流程工业的生产控制提供理论依据,使流程工业生产过程控制更加具有科学性和针对性,取得更好的优化控制效果。
在建模的过程中筛选出对水泥生产过程产生影响最大的因素,此建模的过程的结构优化是利用概率增强式程序进化(Probabilistic Incremental Program Evolution, PIPE)算法进行的;然后找出这些主要参数之间的关联,最后从它们的关联中找到最优的组合方案,即模型的优化,此优化过程是利用模拟退火(Simulation Annealing, SA)算法实现的。通过此建模过程不仅可以找到同一生产过程中参数之间的关系,也可以找出不同生产过程中参数间的比例关系。通过对整个生产过程进行建模和优化而达到稳定水泥生产过程,提高生料的分解率,降低用煤量的目的,把最终的研究方案利用DCS系统应用到实际的生产控制中。
One important feature of modern process industry is the advancement towards large-scale and integrated automation. Through the integration of process control, management, scheduling and marketing and other technical means, integrated Automation of Process Industry can achieve global and local optimization with the shortest cycle and minimum cost for the greatest economic benefit. Babnde A. Ogunnaikeatu discusses five problems facing modern process industry: online measurement, serious nonlinearity, modeling and recognition of the control system, modeling for the training of simulation and execution units, and process monitoring and fault diagnosis. Among them, an important issue is to clarify the operating mechanism of the system and to establish the system description model in order for strict scientific management and planning.
Among the process industry, cement industry is not only one of pillar industries of national economic development, but also one of larger energy consumption industries and its technological property obviously exhibits the characteristics of process production. In the end of the 60s, with the development of computer technology, the use of computers to solve optimal scheme of engineering problems has become an increasingly important means of solving practical design problems. It not only enhances the efficiency and reliability of the problem-solving greatly, but also resolves a lot of practical problems that could not be solved in the past. Especially in recent years, process industry is growing at a rapid pace with the development of globalization of world economy. At the same time, the production in process industry is high complex, strong associated, non-linear and indeterministic. The expansion of production scale will surely complicate the parameters of the process and enhance the association among them. At present, the research on process industry concentrates mainly on the control of the process optimization and neglects the problem of parameter selection required by process optimal control; however, proper selection of parameters will influence the effect of process optimal control directly.
Based on the above analysis of the process industry, this research is to treat pre-calciner, rotary kiln and grate cooler as a whole and to build its production process model. Decomposing furnace is mainly responsible for the decomposition of the raw material, and the decomposition rate could reach 90% -95%; then the decomposed material flows into rotary calciner, which takes on two tasks: one is to decompose the remaining raw material, and the other is to calcine the decomposed raw material; finally the calcined material is cooled in the grate bed. The three processes are closely linked with each other. Therefore, pre-calciner, rotary kiln and grate cooler are closely linked and mutually interacting. The quality of cement is not only dependent on one single process, but on the overall operation of the three courses. Using the sophisticated neural network technology and data mining techniques, and with the cooperation with experienced experts in cement production, we extract the parameters that influence the three production processes, and then use the flexible neural model to model the entire production process. The creative point of this work is as follows. The evolutional generation of traditional FNT model is fixed traditionally, yet the best model is not always formed in this generation, so to fix the evolutional generation is unreasonable. This improved algorithm uses mean error rate to control the number of evolutional generation instead of fixing it. This method provides theoretical basis for the production control of process industry so that the process control is more scientific and pertinent. On the other hand, it establishes a good basis for process optimal control of process industry. Using this method, we can not only obtain the best model and parameters, but enhanced the efficiency and accuracuy.
In the modeling process, the most important factor that impacts of cement production process is screened out, and this modeling process is based on Probabilistic Incremental Program Evolution (PIPE) algorithm. Then the link between these main parameters is found, and finally the best combination of programs from their association is found, namely, the optimization model. This optimization process uses the Simulation Annealing (SA) algorithm. Through this modeling process, not only the parameters relationship between the same productions processes are found, but the proportional relationship between parameters among a different manufacturing process is also found. Through the entire production process modeling and optimization, the purpose of stabilizing the cement production process and improving the rate of decomposition of raw materials and reducing the amount of coal consumption are reached. Finally, the ultimate research result will be applied to the practical production control using DCS system.
引文
[1]舒云星.水泥烧成系统故障诊断与质量预测支持向量机方法的研究[博士学位论文].武汉:武汉理工大学,2008.
[2]王海东.水泥生产过程多AGENT控制系统的研究[博士学位论文].湖南:中南大学,2004.
[3]王成光.流程工业大型实时数据库理论、技术与应用[博士学位论文].浙江:浙江大学,2003.
[4]刘传文.流程工业企业综合自动化系统若干关键技术与应用研究[博士学位论文].浙江:浙江大学,2004.
[5]张英.基于支持向量机的过程工业数据挖掘技术研究[博士学位论文].浙江:浙江大学,2005.
[6]耿志强.流程工业粒度数据挖掘技术研究与应用[博士学位论文].北京:北京化工大学,2005.
[7]庄春来.水泥行业2005年回顾2006年展望.中国水泥,2006,2:8-11.
[8]李剑锋.分解炉温度自动控制系统.水泥技术,2005,(05).
[9]邹健.智能预测控制及其应用研究[博士学位论文].浙江:浙江大学,2002.
[10] http://www.dcement.com/Article/200901/71385.html.
[11] Richard J.Roiger,Michael W.Geatz,,翁敬农.数据挖掘教程(第一版).北京:清华大学出版社,2003.
[12]连志明.分解炉结皮原因分析.水泥技术,2002,(06).
[13]杨志强.模糊控制在水泥回转窑生产过程中的应用.石油化工自动化,2000,4:27-28.
[14]高玉琦.水泥回转窑的计算机控制.自动化学报,1991,17(2):166-173.
[15]姚伯威.水泥回转窑的一种Fuzzy控制实现.电子科技大学学报,1995,24(6):610-613.
[16]袁南儿.水泥回转窑操作的实时专家系统.化工自动化及仪表,1997,24(4):15-18.
[17]张学东.水泥回转窑实时专家控制系统的研制.中国矿业,2000,9(3):82-84.
[18]马翠红.基于专家知识的水泥回转窑模糊控制器.自动化仪表,2003,24(6):37-39.
[19]张建民等.一种动态神经网络在水泥回转窑温度控制中的应用.吉林化工学院学报,2001,18(1):28-30.
[20]于宏亮.面向综合自动化的水泥生产自动控制与过程优化研究[博士论文].沈阳:东北大学,2007.
[21]袁铸钢.水泥预分解窑模糊模型的建立.山东建材学院学报,1998,12(2):118-125.
[22]孙频东.回转窑模糊控制系统.信息技术,2002,(10):9-11.
[23]王庆陶.水泥回转窑的模糊控制模型.工科数学,2001,17(1):12-15.
[24]张宏斌.BP神经网络在水泥窑控制建模中的应用.计算机工程与应用,2002,14:235-238.
[25]刘晓琳.基于神经网络的分解炉温度控制过程建模.新世纪水泥导报,2004,1:39-42.
[26]邹健.模糊预测函数控制在水泥回转窑分解炉温控系统中的应用研究.硅酸盐学报,2001,29(4):318-321.
[27]姚维.模糊控制技术在水泥回转窑分解炉温度控制中的应用.测控技术,2000,19(9):36-40.
[28]姚维.水泥回转窑生产过程的模糊控制.化工自动化及仪表,2000,27(2):15-18.
[29]方千山.水泥工业模糊控制的研究与应用.水泥技术,1998,4:18-21.
[30] Chiang, D. etc. A design methodology of constriant-based fuzzy logic controller. IFSA World Congress and 20th NAFIPS international Conference,2001,(6):1259-1264.
[31] Correcher etc.. Failure diagnosis of a cement kiln using expert systems. 28th annual conference of the IEEE Industrial Electronics Society (IECON-2002),2002,(3): 1881-1886.
[32]王益民.我国建材工业自动化现状与发展.北京:中国建材工业出版社,2000.
[33]孙卫星,邝伯诚.浅谈我过水泥工业电气控制与自动化现状.中国建材科技,1998,06:13-18.
[34]殷晓辉.我厂2000T/D和4000T/D生产线控制系统介绍.水泥工程,1997,05:37-40.
[35]官军海.我公司4000T/D RSP窑生产线自控系统.水泥工程,1998,06:32-33.
[36]许荣康,王建民,江清如.我国水泥生产自动化技术与装备的应用现状.中国建材,1999,05:1-5.
[37]陈文伟,黄金才,赵新昱.数据挖掘技术.北京:北京工业大学出版社,2002.
[38]朱明.数据挖掘.北京:北京电子科技大学出版社,2002.
[39]丁永生.计算智能(第一版).北京:科学出版社,2004.
[40] Chen yuehui,Yang bo,Dong jiwen et al. Time-series Forecasting Using Flexible Neural Tree Model.Information Science,2005, 174: 219-235.
[41] R.P. Salustowicz, J. Schmidhuber. Probabilistic Incremental Program Evolution. Evol. Comput, 1997,2 (5):123–141.
[42] Yuehui Chen, S. Kawaji. System Identification and Control using Probabilistic Incremental Program Evolution Algorithm. Journal of Robotics and Mechatronics,2000,12(6):657-681.
[43] Yuehui Chen, S. Kawaji. Evolving ANNs by hybrid approaches of PIPE and Random Search Algorithm for System Identification. Proceedings of the 3rd Asian ControlConference, 2000:2206-2211.
[44]吴进波.免疫算法和模拟退火算法求解TSP问题的研究[硕士学位论文].武汉:武汉理工大学计算机科学与技术学院,2007.
[45] Xiuyun Cui, Yuehui Chen, Kuifan Shi and Bo Yang. Application of artificial neural networks-genetic algorithms to prediction of cement strength,Journal of Shandong Institute of Building Materials,1998,(3):275-277.
[46]刘照莲.流程对象建模方法的研究与实现[硕士学位论文].济南:济南大学信息科学与工程学院,2009.
[47]张维明,邓苏,罗雪山等.信息系统建模技术与应用(第一版).北京:电子工业出版社,1997.
[48]蔡锁.数学建模原理与方法.北京:海洋出版社,2003.
[49]袁震东.数学建模方法(第一版).上海:华东师范大学出版社,2003.
[50] Yuehui Chen, Bo Yanga, Ajith Abrahama. Flexible neural trees ensemble for stock index modeling. Neurocomputing,2007, (70):697–703.
[51] Yuehui Chen, Lizhi Peng, Ajith Abraham. Gene Expression Profiling Using Flexible Neural Trees. Lecture Notes on Computer Science,2006, (4224),1121–1128.
[52] Yuehui Chen, Ajith Abraham, Bo Yang. Feature selection and classification using ?exible neural tree. NeuroComputing,2006,(70):305-313.
[53] Yuehui Chen,Lizhi Peng,Ajith Abraham. Exchange Rate Forecasting Using Flexible Neural Trees. Lecture Notes on Computer Science,2006,(3973):518–523.
[54] Yuehui Chen, Bo Yang, Jiwen Dong. Evolving Flexible Neural Networks Using Ant Programming and PSO Algorithm. Computer Science,2004,(3173):211-216.
[55] Yuehui Chen, Ajith Abraham, Bo Yang. Hybrid Flexible Neural-Tree-Based Intrusion Detection Systems. INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS,2007,(22):337–352.
[56]国治利.生产工艺概论.武汉:武汉理工大学出版社,2005.
[57]李坚利,周惠群.水泥生产工艺(第2版).武汉:武汉理工大学出版社,2008.
[58]陈亚明.基于神经网络模拟预分解窑煅烧的研究[硕士学位论文] .武汉:武汉理工大学材料学院,2003.
[59]李秀改,候媛彬.基于神经网络BP算法的模糊自适应控制器的研究与实现.电气传动自动化.2000,(04).
[60]李玲俐.异构数据整合技术在钢铁企业信息系统中的应用研究[硕士学位论文].武汉:武汉科技大学,2005.
[61] John K.Holt, Soon M.Chung. Mining association rules using inverted hashing and pruning. Information Processing Letters,83(2002):211-220.
[62]邓全亮,邹云涛,邹益仁.流程工业MES中的数据集成研究及应用.计算机工程,2006,32(19):258-260.
[63]阎威武,朱宏栋,邵惠鹤.基于最小二乘支持向量机的软测量建模.系统仿真学报,2003,15 (10):1494-1496.
[64]乔龙德.中国水泥工业发展的新特点与战略目标.新世纪水泥导报,2003,(5):5-7.
[65] Shouning Qu, Aifang Fu, Zhanglian Liu etal. The Improvement and Application of Structure Optimization Algorithm Based on Flexible Neural Trees. In: 2009 International Conference on Information Technology and Computer Science,2009.
[2]王海东.水泥生产过程多AGENT控制系统的研究[博士学位论文].湖南:中南大学,2004.
[3]王成光.流程工业大型实时数据库理论、技术与应用[博士学位论文].浙江:浙江大学,2003.
[4]刘传文.流程工业企业综合自动化系统若干关键技术与应用研究[博士学位论文].浙江:浙江大学,2004.
[5]张英.基于支持向量机的过程工业数据挖掘技术研究[博士学位论文].浙江:浙江大学,2005.
[6]耿志强.流程工业粒度数据挖掘技术研究与应用[博士学位论文].北京:北京化工大学,2005.
[7]庄春来.水泥行业2005年回顾2006年展望.中国水泥,2006,2:8-11.
[8]李剑锋.分解炉温度自动控制系统.水泥技术,2005,(05).
[9]邹健.智能预测控制及其应用研究[博士学位论文].浙江:浙江大学,2002.
[10] http://www.dcement.com/Article/200901/71385.html.
[11] Richard J.Roiger,Michael W.Geatz,,翁敬农.数据挖掘教程(第一版).北京:清华大学出版社,2003.
[12]连志明.分解炉结皮原因分析.水泥技术,2002,(06).
[13]杨志强.模糊控制在水泥回转窑生产过程中的应用.石油化工自动化,2000,4:27-28.
[14]高玉琦.水泥回转窑的计算机控制.自动化学报,1991,17(2):166-173.
[15]姚伯威.水泥回转窑的一种Fuzzy控制实现.电子科技大学学报,1995,24(6):610-613.
[16]袁南儿.水泥回转窑操作的实时专家系统.化工自动化及仪表,1997,24(4):15-18.
[17]张学东.水泥回转窑实时专家控制系统的研制.中国矿业,2000,9(3):82-84.
[18]马翠红.基于专家知识的水泥回转窑模糊控制器.自动化仪表,2003,24(6):37-39.
[19]张建民等.一种动态神经网络在水泥回转窑温度控制中的应用.吉林化工学院学报,2001,18(1):28-30.
[20]于宏亮.面向综合自动化的水泥生产自动控制与过程优化研究[博士论文].沈阳:东北大学,2007.
[21]袁铸钢.水泥预分解窑模糊模型的建立.山东建材学院学报,1998,12(2):118-125.
[22]孙频东.回转窑模糊控制系统.信息技术,2002,(10):9-11.
[23]王庆陶.水泥回转窑的模糊控制模型.工科数学,2001,17(1):12-15.
[24]张宏斌.BP神经网络在水泥窑控制建模中的应用.计算机工程与应用,2002,14:235-238.
[25]刘晓琳.基于神经网络的分解炉温度控制过程建模.新世纪水泥导报,2004,1:39-42.
[26]邹健.模糊预测函数控制在水泥回转窑分解炉温控系统中的应用研究.硅酸盐学报,2001,29(4):318-321.
[27]姚维.模糊控制技术在水泥回转窑分解炉温度控制中的应用.测控技术,2000,19(9):36-40.
[28]姚维.水泥回转窑生产过程的模糊控制.化工自动化及仪表,2000,27(2):15-18.
[29]方千山.水泥工业模糊控制的研究与应用.水泥技术,1998,4:18-21.
[30] Chiang, D. etc. A design methodology of constriant-based fuzzy logic controller. IFSA World Congress and 20th NAFIPS international Conference,2001,(6):1259-1264.
[31] Correcher etc.. Failure diagnosis of a cement kiln using expert systems. 28th annual conference of the IEEE Industrial Electronics Society (IECON-2002),2002,(3): 1881-1886.
[32]王益民.我国建材工业自动化现状与发展.北京:中国建材工业出版社,2000.
[33]孙卫星,邝伯诚.浅谈我过水泥工业电气控制与自动化现状.中国建材科技,1998,06:13-18.
[34]殷晓辉.我厂2000T/D和4000T/D生产线控制系统介绍.水泥工程,1997,05:37-40.
[35]官军海.我公司4000T/D RSP窑生产线自控系统.水泥工程,1998,06:32-33.
[36]许荣康,王建民,江清如.我国水泥生产自动化技术与装备的应用现状.中国建材,1999,05:1-5.
[37]陈文伟,黄金才,赵新昱.数据挖掘技术.北京:北京工业大学出版社,2002.
[38]朱明.数据挖掘.北京:北京电子科技大学出版社,2002.
[39]丁永生.计算智能(第一版).北京:科学出版社,2004.
[40] Chen yuehui,Yang bo,Dong jiwen et al. Time-series Forecasting Using Flexible Neural Tree Model.Information Science,2005, 174: 219-235.
[41] R.P. Salustowicz, J. Schmidhuber. Probabilistic Incremental Program Evolution. Evol. Comput, 1997,2 (5):123–141.
[42] Yuehui Chen, S. Kawaji. System Identification and Control using Probabilistic Incremental Program Evolution Algorithm. Journal of Robotics and Mechatronics,2000,12(6):657-681.
[43] Yuehui Chen, S. Kawaji. Evolving ANNs by hybrid approaches of PIPE and Random Search Algorithm for System Identification. Proceedings of the 3rd Asian ControlConference, 2000:2206-2211.
[44]吴进波.免疫算法和模拟退火算法求解TSP问题的研究[硕士学位论文].武汉:武汉理工大学计算机科学与技术学院,2007.
[45] Xiuyun Cui, Yuehui Chen, Kuifan Shi and Bo Yang. Application of artificial neural networks-genetic algorithms to prediction of cement strength,Journal of Shandong Institute of Building Materials,1998,(3):275-277.
[46]刘照莲.流程对象建模方法的研究与实现[硕士学位论文].济南:济南大学信息科学与工程学院,2009.
[47]张维明,邓苏,罗雪山等.信息系统建模技术与应用(第一版).北京:电子工业出版社,1997.
[48]蔡锁.数学建模原理与方法.北京:海洋出版社,2003.
[49]袁震东.数学建模方法(第一版).上海:华东师范大学出版社,2003.
[50] Yuehui Chen, Bo Yanga, Ajith Abrahama. Flexible neural trees ensemble for stock index modeling. Neurocomputing,2007, (70):697–703.
[51] Yuehui Chen, Lizhi Peng, Ajith Abraham. Gene Expression Profiling Using Flexible Neural Trees. Lecture Notes on Computer Science,2006, (4224),1121–1128.
[52] Yuehui Chen, Ajith Abraham, Bo Yang. Feature selection and classification using ?exible neural tree. NeuroComputing,2006,(70):305-313.
[53] Yuehui Chen,Lizhi Peng,Ajith Abraham. Exchange Rate Forecasting Using Flexible Neural Trees. Lecture Notes on Computer Science,2006,(3973):518–523.
[54] Yuehui Chen, Bo Yang, Jiwen Dong. Evolving Flexible Neural Networks Using Ant Programming and PSO Algorithm. Computer Science,2004,(3173):211-216.
[55] Yuehui Chen, Ajith Abraham, Bo Yang. Hybrid Flexible Neural-Tree-Based Intrusion Detection Systems. INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS,2007,(22):337–352.
[56]国治利.生产工艺概论.武汉:武汉理工大学出版社,2005.
[57]李坚利,周惠群.水泥生产工艺(第2版).武汉:武汉理工大学出版社,2008.
[58]陈亚明.基于神经网络模拟预分解窑煅烧的研究[硕士学位论文] .武汉:武汉理工大学材料学院,2003.
[59]李秀改,候媛彬.基于神经网络BP算法的模糊自适应控制器的研究与实现.电气传动自动化.2000,(04).
[60]李玲俐.异构数据整合技术在钢铁企业信息系统中的应用研究[硕士学位论文].武汉:武汉科技大学,2005.
[61] John K.Holt, Soon M.Chung. Mining association rules using inverted hashing and pruning. Information Processing Letters,83(2002):211-220.
[62]邓全亮,邹云涛,邹益仁.流程工业MES中的数据集成研究及应用.计算机工程,2006,32(19):258-260.
[63]阎威武,朱宏栋,邵惠鹤.基于最小二乘支持向量机的软测量建模.系统仿真学报,2003,15 (10):1494-1496.
[64]乔龙德.中国水泥工业发展的新特点与战略目标.新世纪水泥导报,2003,(5):5-7.
[65] Shouning Qu, Aifang Fu, Zhanglian Liu etal. The Improvement and Application of Structure Optimization Algorithm Based on Flexible Neural Trees. In: 2009 International Conference on Information Technology and Computer Science,2009.