颗粒机智能控制模型的研究
|
摘要
在计算机屏上操作管理的自动化生产在化工、医药行业已得到了较广泛的应用,而饲料生产管理还处于相对落后的状态。经过“九五”、“十五”技术攻关实现了生产过程控制自动化及计算机生产管理,使饲料厂的自动化水平大大提高。但是,计算机的引入主要还是解决生产自动化问题,对于改善产品质量、产品品种还没有发挥应有的作用。从大多数饲料厂的现状来看,由于目前大量信息在生产现场只是作记录,来不及分析和处理,往往是出现问题后再调查、再分析和处理,无法把事故控制在萌芽状态,给企业造成不必要的损失和浪费。影响饲料产品质量的因素较多及其不确定性,传统的建模方法很难对不确定的非线性对象进行建模,而径向基函数(Radial Basic Function,RBF)神经网络可实现系统输入输出的非线性映射,可用于建立饲料制粒系统智能控制数学模型。
本文在研究颗粒饲料加工工艺的基础上,根据人工神经网络理论,提出采用径向基函数(Radial Basic Function,RBF)神经网络进行建模,模型结构为3层。在实验及分析的基础上,选定调质温度、调质水分、喂料速度、蛋白质含量做为网络的输入,以颗粒饲料硬度衡量饲料的质量,并做为网络的输出。
RBF网络的模型的实现需要掌握计算机编程语言及较高的编程能力,这在一定程度上不利于神经网络技术的推广和使用。MATLAB软件提供了一个现成的神经网络工具箱(Neural Network Toolbox,简称NNT)为解决这个难题提供了便利条件。本文在简要介绍了RBF神经网络基本原理及其算法的基础上,详细介绍了利用MATLAB神经网络工具箱进行RBF网络模型建立、训练、仿真的编程方法,并建立饲料制粒系统智能控制数学模型。
Production process control and automation on industrial control computer has been widely used in chemical and medical industry, on the other hand, it received rare attention by feed plant. Currently, it has been greatly improved in feed plant with the development of computer-aid production process control and automation system during the ninth and tenth five-year technique promotion plans. However, computer aided production control system is mainly focused on the production automation, yet leaves the problem how to improve product quality intact. From actualities of most feed plant, lots derived information are not dealt with through analyzing just recording, which brings value loss and waste of feed plant because problems and troubles cannot be handled in time. On the other hand , there are many factors with uncertainty influencing the quality of feed pellet, and traditional approachs do not facitating modeling the nonlinear relationships with uncertainty, therefore. Radial Basic Function (RBF) neural network is adopted to establish the mathematic model for feed drilling intelligent control system for its characteristic of approximating the relationships between input variables to output variables to real-word relationships in any precisions.
Based on the study of product feed pellet production process, a Radial Basic Function (RBF) neural network is proposed to build the drilling model, which has three lays. Based on experiment design and factor analysis, conditioning temperature, conditioning coisture, feeding speed and protein content are adopted as input of network while hardness is employed as the quality of feed pellet, as well as the output of neural network.
Modeling and realizing of RBF network requires outstanding programming capability, which greatly prevent the neural network technique from spreading and application. However, Matlab software provides an effective toolkit called Neural Network Toolbox (NNT) to solve this problem. In this paper, mathematic model for feed drilling intelligent control system is established based on the introduction of RBF neural network's principal and algorithm, along with RBF neural network model building, training, simulating and programming.
本文在研究颗粒饲料加工工艺的基础上,根据人工神经网络理论,提出采用径向基函数(Radial Basic Function,RBF)神经网络进行建模,模型结构为3层。在实验及分析的基础上,选定调质温度、调质水分、喂料速度、蛋白质含量做为网络的输入,以颗粒饲料硬度衡量饲料的质量,并做为网络的输出。
RBF网络的模型的实现需要掌握计算机编程语言及较高的编程能力,这在一定程度上不利于神经网络技术的推广和使用。MATLAB软件提供了一个现成的神经网络工具箱(Neural Network Toolbox,简称NNT)为解决这个难题提供了便利条件。本文在简要介绍了RBF神经网络基本原理及其算法的基础上,详细介绍了利用MATLAB神经网络工具箱进行RBF网络模型建立、训练、仿真的编程方法,并建立饲料制粒系统智能控制数学模型。
Production process control and automation on industrial control computer has been widely used in chemical and medical industry, on the other hand, it received rare attention by feed plant. Currently, it has been greatly improved in feed plant with the development of computer-aid production process control and automation system during the ninth and tenth five-year technique promotion plans. However, computer aided production control system is mainly focused on the production automation, yet leaves the problem how to improve product quality intact. From actualities of most feed plant, lots derived information are not dealt with through analyzing just recording, which brings value loss and waste of feed plant because problems and troubles cannot be handled in time. On the other hand , there are many factors with uncertainty influencing the quality of feed pellet, and traditional approachs do not facitating modeling the nonlinear relationships with uncertainty, therefore. Radial Basic Function (RBF) neural network is adopted to establish the mathematic model for feed drilling intelligent control system for its characteristic of approximating the relationships between input variables to output variables to real-word relationships in any precisions.
Based on the study of product feed pellet production process, a Radial Basic Function (RBF) neural network is proposed to build the drilling model, which has three lays. Based on experiment design and factor analysis, conditioning temperature, conditioning coisture, feeding speed and protein content are adopted as input of network while hardness is employed as the quality of feed pellet, as well as the output of neural network.
Modeling and realizing of RBF network requires outstanding programming capability, which greatly prevent the neural network technique from spreading and application. However, Matlab software provides an effective toolkit called Neural Network Toolbox (NNT) to solve this problem. In this paper, mathematic model for feed drilling intelligent control system is established based on the introduction of RBF neural network's principal and algorithm, along with RBF neural network model building, training, simulating and programming.
引文
[1]邓君明.张曦,饲料加工工艺的缓新研究进展。畜牧与兽医,2002,8:35-37
[2]焦李成.神经网络系统理论,西安:西安电子科技大学出版社,1990
[3]胡守仁土编神经网络导论。长沙:国防科技大学出版社,1992
[4]张德富,殷正坤。人工神经网络的发展及其哲理。科学技术与辩证法,2000,8:17-20
[5]徐丽娜,神经网络控制。哈尔滨:哈尔滨工业大学出版社,1999
[6]焦李成,神经网络系统理论。西安:西安电子科技大学出版社1992,5-7,34-3
[7]周继成,人工神经网络-第六代计算机的实现北京:科学普及出版社。1993
[8]李德发,范石军。饲料工业手册。北京:中国农业大学出版社,2002
[9]曹康.现代饲料加工技术。上海:上海科学技术文献出版社,2003
[10]盛骤,谢式千,潘承毅.概率论与数理统计.北京:高等教育出版社,1989:241-278.
[11]北京市饲料公司,北京市饲料科学技术研究所。饲料工业基础知识.北京:北京出版社,1989 561-562
[12]江志炜等,蛋白质加工技术,北京:化学工业出版社,2003
[13]王红英,邓志刚,于庆龙等。高蛋白饲料加工质量影响因素分析。饲料工业,2004,2:8-9.
[14]闻新,周露,王丹力.MATLAB神经网络应用设计.北京:科学出版社,2003:258-244
[15]闻新,周露,李翔等。MATLAB神经网络仿真与应用.北京:科学出社.,2003:258-243
[16]盛骤,谢式干,潘承毅,概率论与数理统计。北京:高等教育出版社,1989:241-278.
[17]北京市饲料公司,北京市饲料科学技术研究所。饲料工业基础知识,北京:北京出版社,1989 561-562
[18]张志涌,掌握和精通MATLAB。北京:航空肮天大学出版社,1997
[19]高宁,基于神经网络的农作物虫情预测预报及其MATLAB实现。[硕十学位论文]合肥::安徽农业大学,2003
[20]于海华,人工神经网络与煤发热量的测量。[硕十学位论文],哈尔滨:黑龙江大学,2003
[21]邱万里,关于饲料制粒工艺中要素的控制。粮食与饲料工业,2003,3,51-52
[22]黄其春,影响颗粒饲料品质的因素。粮食与饲料工业,2003,3,17-19
[23]姜伟忠,饲料制粒的目标。粮食与饲料工业,2000,10,22-24
[24]任春英等,影响压制颗粒饲料产量的因素。现代化农业2004(6):22-23
[25]David A.Fairfield.Pelleting for Profit,Feed and Feeding Digest。2003(6)
[26]全国饲料工作办公室,中国饲料工业协会,饲料行业HACCP管理技术指南。北京:中国农业出版社,2003
[27]邓志刚,李军国,于庆龙,颗粒饲料质量安全保证工艺技术研究。粮食与饲料工业,2005,2:28-29.
[28]徐毅峰,深化自动控制技术在保证饲料质量中的应用。粮食与食品工业,2004,2,41-44
[29]何玉彬、李新忠,神经网络控制技术及其应用,北京:科学出版社。2000
[30]张乃尧,阎平凡,神经网络与模糊控制。清华大学出版社,1998
[31]胡钧望等,食品化学。北京:科学出版社,1992
[32]Dach,-J.Niedbala,-G,Przybyl,-J.Application of the neural networks in agriculture.Inzynieria-Rolnicza(Poland).2001,1:57-62
[33]Hourigan-JA,Zhou-W,Tuck-DL,et al:Artificial neural networks in the dairy industy.Austrlian-Journal-of-Dairy-Technology. 1994,2:110-124
[34]孙虎章.自动控制原理。北京:中央广播电视大学出版社,1994
[35]Katsuhiko Ogata著,卢伯英,于海勋译。现代控制工程。北京:电子工业出版社,2003
[36]SonW C,Tse C K.Derivation of fussy controller forDC-DC converters using neural networks.Journal of Electric and Electronics,1996,2:85-90
[37]Lo J Y,Baker JA,Kornguth PJ,et all,Predicting breast cancer invasion with artificial neural net works on the basis of mammographic featured Radiology,1997,203::159-163
[38]胡上序,程翼宇,人工神经元计算导论。北京:科学出版社,1994
[39]胡守仁,人工神经网络应用技术。北京:国防科技大学出版社,1993
[40]周继成,人工神经网络——第六代计算机的实现。北京:科学普及出版社。1993
[41]王汝南,章明,第六代计算机——人工神经网络计算机。北京:科学出版社,1992
[2]焦李成.神经网络系统理论,西安:西安电子科技大学出版社,1990
[3]胡守仁土编神经网络导论。长沙:国防科技大学出版社,1992
[4]张德富,殷正坤。人工神经网络的发展及其哲理。科学技术与辩证法,2000,8:17-20
[5]徐丽娜,神经网络控制。哈尔滨:哈尔滨工业大学出版社,1999
[6]焦李成,神经网络系统理论。西安:西安电子科技大学出版社1992,5-7,34-3
[7]周继成,人工神经网络-第六代计算机的实现北京:科学普及出版社。1993
[8]李德发,范石军。饲料工业手册。北京:中国农业大学出版社,2002
[9]曹康.现代饲料加工技术。上海:上海科学技术文献出版社,2003
[10]盛骤,谢式千,潘承毅.概率论与数理统计.北京:高等教育出版社,1989:241-278.
[11]北京市饲料公司,北京市饲料科学技术研究所。饲料工业基础知识.北京:北京出版社,1989 561-562
[12]江志炜等,蛋白质加工技术,北京:化学工业出版社,2003
[13]王红英,邓志刚,于庆龙等。高蛋白饲料加工质量影响因素分析。饲料工业,2004,2:8-9.
[14]闻新,周露,王丹力.MATLAB神经网络应用设计.北京:科学出版社,2003:258-244
[15]闻新,周露,李翔等。MATLAB神经网络仿真与应用.北京:科学出社.,2003:258-243
[16]盛骤,谢式干,潘承毅,概率论与数理统计。北京:高等教育出版社,1989:241-278.
[17]北京市饲料公司,北京市饲料科学技术研究所。饲料工业基础知识,北京:北京出版社,1989 561-562
[18]张志涌,掌握和精通MATLAB。北京:航空肮天大学出版社,1997
[19]高宁,基于神经网络的农作物虫情预测预报及其MATLAB实现。[硕十学位论文]合肥::安徽农业大学,2003
[20]于海华,人工神经网络与煤发热量的测量。[硕十学位论文],哈尔滨:黑龙江大学,2003
[21]邱万里,关于饲料制粒工艺中要素的控制。粮食与饲料工业,2003,3,51-52
[22]黄其春,影响颗粒饲料品质的因素。粮食与饲料工业,2003,3,17-19
[23]姜伟忠,饲料制粒的目标。粮食与饲料工业,2000,10,22-24
[24]任春英等,影响压制颗粒饲料产量的因素。现代化农业2004(6):22-23
[25]David A.Fairfield.Pelleting for Profit,Feed and Feeding Digest。2003(6)
[26]全国饲料工作办公室,中国饲料工业协会,饲料行业HACCP管理技术指南。北京:中国农业出版社,2003
[27]邓志刚,李军国,于庆龙,颗粒饲料质量安全保证工艺技术研究。粮食与饲料工业,2005,2:28-29.
[28]徐毅峰,深化自动控制技术在保证饲料质量中的应用。粮食与食品工业,2004,2,41-44
[29]何玉彬、李新忠,神经网络控制技术及其应用,北京:科学出版社。2000
[30]张乃尧,阎平凡,神经网络与模糊控制。清华大学出版社,1998
[31]胡钧望等,食品化学。北京:科学出版社,1992
[32]Dach,-J.Niedbala,-G,Przybyl,-J.Application of the neural networks in agriculture.Inzynieria-Rolnicza(Poland).2001,1:57-62
[33]Hourigan-JA,Zhou-W,Tuck-DL,et al:Artificial neural networks in the dairy industy.Austrlian-Journal-of-Dairy-Technology. 1994,2:110-124
[34]孙虎章.自动控制原理。北京:中央广播电视大学出版社,1994
[35]Katsuhiko Ogata著,卢伯英,于海勋译。现代控制工程。北京:电子工业出版社,2003
[36]SonW C,Tse C K.Derivation of fussy controller forDC-DC converters using neural networks.Journal of Electric and Electronics,1996,2:85-90
[37]Lo J Y,Baker JA,Kornguth PJ,et all,Predicting breast cancer invasion with artificial neural net works on the basis of mammographic featured Radiology,1997,203::159-163
[38]胡上序,程翼宇,人工神经元计算导论。北京:科学出版社,1994
[39]胡守仁,人工神经网络应用技术。北京:国防科技大学出版社,1993
[40]周继成,人工神经网络——第六代计算机的实现。北京:科学普及出版社。1993
[41]王汝南,章明,第六代计算机——人工神经网络计算机。北京:科学出版社,1992