煮糖结晶过程的最优控制
|
摘要
煮糖结晶过程是一非线性、慢时变过程。因其内部机理复杂，再加上其各变量间相互耦合，要想弄清其中某个变量与过饱和度(过饱和度直接影响结晶)间的关系都很困难，更不要说建立其机理模型了。近年来，随着人工智能、人工神经网络的发展，尤其是BP网络在非线性过程的模拟和控制中所显示出的优越性，使它们得到了广泛地应用。
本文采用BP网络对煮糖结晶过程进行建模，并对此网络模型进行了仿真，最后还提出了采用专家系统进行优化控制的简单方案。网络模型的仿真结果表明：该模型能较好地描述过程性能，且其记忆能力和泛化能力明显优于回归模型。
Cane Sugar's boiling and crystallisation is an non-linear and slowly time-varied process. Because its internal mechanism is complex, and its each variable is coupled with others, It is difficult to make the supersaturation (which affect Crystallisation directly)clear, saying nothing of establishing its mechanism model. In recent years, with the development of Artificial Intelligence and Artificial Neural Network(ANN) technology, the neural network especially BP Network is used successfully in the identification and control of the complicated process.
The main work of the thesis is to model and simulate the process of Cane Sugar's crystallisation . In the end, I have also put forward a simple scheme to optimize by Expert System. The simulation result shows that the model can describe the performance of process perfectly, and its prediction and generalization abilities are superior to the recurrent model.
本文采用BP网络对煮糖结晶过程进行建模，并对此网络模型进行了仿真，最后还提出了采用专家系统进行优化控制的简单方案。网络模型的仿真结果表明：该模型能较好地描述过程性能，且其记忆能力和泛化能力明显优于回归模型。
Cane Sugar's boiling and crystallisation is an non-linear and slowly time-varied process. Because its internal mechanism is complex, and its each variable is coupled with others, It is difficult to make the supersaturation (which affect Crystallisation directly)clear, saying nothing of establishing its mechanism model. In recent years, with the development of Artificial Intelligence and Artificial Neural Network(ANN) technology, the neural network especially BP Network is used successfully in the identification and control of the complicated process.
The main work of the thesis is to model and simulate the process of Cane Sugar's crystallisation . In the end, I have also put forward a simple scheme to optimize by Expert System. The simulation result shows that the model can describe the performance of process perfectly, and its prediction and generalization abilities are superior to the recurrent model.
引文
[1] J．Zupan．J．Gasteiger著，潘忠校、陈玲然译 《神经网络及其在化学中的应用》 中国科学技术大学出版社 2000年5月第一版
[2] 丛爽编著《面向Matlab工具箱的神经网络理论与应用》 中国科学技术大学出版社 1998年11月第一版
[3] 闻新、周露、王丹力、熊晓英编著 《Matlab神经网络应用设计》 科学出版社 2000年9月第一版
[4] 张立明编著 《人工神经网络的模型及其应用》 复旦大学出版社 1993年7月第一版
[5] 高俊斌编著 《Matlab5.0语言与程序设计》 华中理工大学出版社 1999年1月第一版
[6] 楼顺天、于卫、闫华梁编著 《Matlab程序设计语言》 西安电子科技大学出版社 1997年8月第一版
[7] 王华 等编著 《Matlab在电信工程中的应用》 中国水利水电出版社 2001年4月第一版
[8] 《甘蔗制糖工业手册》编写组 编 《甘蔗制糖工业手册》(上册) 轻工业出版社 1989年5月第一版
[9] 《甘蔗制糖工业手册》编写组 编 《甘蔗制糖工业手册》(下册) 轻工业出版社 1989年5月第一版
[10] 广东省糖纸工业公司 李尔煊编写 《制糖工艺技术学习班(澄清、蒸发)学习讲义》 广东省地方国营雷高糖厂翻印 1991年1月27日
[11] 广东省糖纸工业公司 李尔煊编写 《制糖工艺技术学习班(煮糖)学习讲义》 广东省地方国营雷高糖厂翻印 1991年1月27日
[12] 华东理工大学硕士学位论文 姚颖写 《挡板流化化床丁烯氧化脱氢过程的建模与优化》 1998年5月20日
[13] K.F. Miller 《Crystal growth and vacuum pan productivity》 INT. SUGAR. JNL., 1999, VOL. 101, NO. 1204 p216-218
[14] Michel Benne,Brigitte Grondin-Perez,Jean-Danlel Luk and Jean-Pierre Chabriat. Modelling cane sugar evaporation and crystallization (Part 1:A new approach using neural netwoks) INT. SUGAR JNL., 1999,VOL.101,NO. 1208 p418-422
[15] Michel Benne,Brigitte Grondin-Perez,Jean-Danlel Luk and Jean-Pierre Chabriat. Modelling of evaporation and crystallization (Part 2:A new approach using neural netwoks). INT.SUGAR JNL., 1999,VOL.101,NO.1208 p418-422
[16] A. D. Randolph & S. A. Ziebold: Continuous Sucrose Nuclearion, I. S. J., Vol 76(1974)
[17] E. W. Krause: "New Equipments and Processes in the Sugar Industry" Licht F. O. L. 1987 E5-E29
[18] Paturau:" Medium and Long Term Development of Cane Sugar Factories" Presentation, 1988 Expert Group Meeting in Preparation of First Consultation on Sugar Industry, UNIDO
[19]Chun Y. Chen and Babu Joseph, On-line optimization using a two phase approach: An application study. Ind. Eng. Chem. Res. Vol. 26,1924-1930,1987
[20]叶江祺、何伟然、李世俊编著 《热工仪表和控制设备的安装》 水利电力出版社 1983年7月第一版
[21]邵裕森编 《过程控制及仪表》 上海交通大学出版社 1986年9月第1版
[22]中国自动化控制系统总公司 编 《自控系统成套设备选型样本》 陕西科学技术出版社 1989年1月第1版
[23]广东省糖纸食品公司 编《糖膏煮炼与助晶》 轻工业出版社 1985年2月第1版
[24]无锡轻工业学院、华南工学院编著 《甘蔗制糖工艺学》 轻工业出版社 1982年1月第一版
[25]霍汗镇 主编 陈树功、陈世治、林乐新等编 (广东省制糖学会组织) 《现代制糖工业技术》 中国轻工业出版社 1992年9月第一版
[26]高大维、陈树功等：煮糖起晶制种新方法，《广东省制糖学会年会论文选集》，1990年
[27]Shi-shang Jang;Babu Joseph and Hiro Mukai.On-line optimization of constrained multivariable chemical processes.AIChE J.Vol.33,26-35,1987.
[28]A.Lucia and J.Xu,Nonconvex progress optimization .Comp & Chen. Engng Vol.20, 1375-1398,1996
[29]A.V. Ooyen, Improving the convergence of the back-propagation algorithm. Neural Networks Vol.5.465-571,1992
[30]B.Robitaille,B.Marcos and M.Veillette,Modified quasi-Newton methods for trainning neural network. Comp. Engng Sci. Vol.25,859-865
[31]B.Schenker and M.Agarwal, Cross-validated structure selection for neural networks. Comp & Chem. Engng Vol.20, No.2,175-186,1996
[32]C.D.Pschogios and L.H.Ungar, Direct and indirect model based control using artificial neural networks.Ind.Eng.Chem.Res.Vol.30,2564-2573,1991
[33]D.Melo and J.Friedly, On-line closed-loop identification of multivariable systems.Ind.Eng.Chem.Res. Vol.36,1-22,1982
[34]E.P. Gerald, C.S.C.Daniel, R.A.Stephen and E.B.William, Using Taguchi's method of experimental design to control errors in layered perceptrons. IEEE Trans. Neural Networks Vol.6,949-960,1995
[35]F.J.Solis and J.B.Wets, Minimization by random search techniques. Math. Oper. Res. Vol.6, 19-30, 1981.
[36]G.L.Bilbro and W. E. Snyder, Optimization of function with many minima. IEEE Trans. Syst. Man Cybernet. Vol.21,840-849, 1991.
[37] H.Derrick,Nguyen and Bernard Widrow, Neural network for self-learning control system . IEEE Control Sys. Mag. Vol.4,18-23,1990．
[38] H. N. Robert, Theory of the Back-Propagation neural network. IJCNN, 1989
[39] M.A.Wolf, Numerical methods for unconstrained optimization. New York :Van Nostrand Reinhold,1978
[40] J.Saint-Donat, N. Bhat and T. J. McAvoy, Neural net based model predictive control. Int. J. Control Vol.54,1453-1468,1991d
[41] J. Savkovic-Stevanovic, Neural networks fo process analysis and optimization: modeling and applications. Comp & Chem. Engng Vol. 18,1149-1155,1994
[42] K. Juha and J. Jyrki, Generalization of principal compoent analysis, optimization problems and neural networks. Neural Networks Vol.8, 549-562,1995
[43] K. S. Narenda and K. Parthasarathy, Identification and control of dynamical system using neural networks. IEEE Trains. Neural Network Vol. 1,4-27,1990
[44] L. Leonard and MA.Kramer, Improvement of the backpropagation algorithm for training neural networks. Comp & Chem. Engng Vol.14,337-341,1990
[45] L. Megan and D. J. Cooper, A neural network strategy for disturbance pattern classifaction and adaptive multivariable control. Comp & Chem. Engng Vol. 19,171-186,1995
[46] M. J. Box, A new method of constrained optimization and a comparison with other methods. Computer J. Vol.8,42-52,1965
[47] N. Baba, Convergence of a random optimization method for constrained optimization problems. J. Optim. Theory Appl. Vol. 33,451-461,1981．
[48] N.Bhat and T. J. McAvoy, Use of neural nets for dynamic modeling and control of chemical process systems. Comp & Chem. Engng Vol.14,573-583,1990
[49] P. Tanartkit and L. T. Biegler, A nested , simutaneous approach for dynamic optimization problem. Comp & Chem. Engng Vol.20, 735-741,1996
[50] R. S. H. Mah and V. Chakravarthy, Pattern recognition using artificial neural Networks. Comp & Chem. Engng Vol.12, 555-562,1973
[51] Widrow B, et al. Neural Networks Application in Industry, Business and Science. Communcation of the ACM, 1994,3 7:93-105
[52] Special Issue on Application of Neural Network. Proc IEEE, 1996,84(10)
[53] SliMin F. An Expert Network for DNA Sequence Analysis, IEEE Intelligent Systems, Jan. /Feb. 1999,65-71
[54] Cybenko G . Approximation by Superpositions of Sigmoidal Function. Mathematics of Control, Signals and Systems, 1980, 2:303-314
[55] Hornik K, et al. Multilayer Feedforward Network are Universal Approximators. Neural Network, 1989,2:359-366
[56] Patrick P, et al.Minimisation Method for Training Feedforward Neural Networks. Neural Networks. 1994,7:1～11
[57] Yi S, et al. Global Optimization for Training. IEEE Computer, 1996, 3:45～54
[58] Van Rooij A J F, et al. Neural Networks Training Using Genetic Algorithm, World Scientific, 1996
[59] Looney C G. Advances in Feedforward Neural Netwoks, Denstifying Knowledge Acquiring Black Baxes.IEEE Trans Knowledge and Data Engeneering, 1996, 8:211～226
[60] Cohn D, et al. How Tight are the Vapnik-Chervoneko Bounds? Neural Computation, 1992,4:249～269
[61] Holden S B. On the Practical Applicability of VC Dimension Bounds. Neural Computation, 1995, 7:1265～1288
[62] Wei-Tsih Lee, et al. On an Asympototically Optimal Adaptive Classifier Design. IEEE Trans, 1993, PAMI-15:312～318
[63] Cohn A M, et al. On the Geometry of Feedward NN Weight Space. In:Proc 2nd IEEE Conference on ANN, London:IEEE Press,1991,1～4
[64] Hole A. Vapnik-Chervonenko Genneralization Bounds for Real Valued NN. Neural Computation, 1996, 8:1277-1299
[65] Guyon I, et al. What size Test Set Gives Good Error Rate Estimates? IEEE Trans PAMI,1998,(20):52～64
[66] Twomey J M, et al. Bias and Variance of Validation Methods for Function Approximation Neural Networks Chder Conditions of Sparse Data. IEEE Trans SMC PARTC, Q998, 28:417～430
[67] Devoroye L.Automatic Pattern Recognition:A Study of the Probability of Error. IEEE Trans PAMI, 1988, 28:417～430
[68] Bax E. Validation of Average Error Rate Over Classifiers Pattern Recongnition Letters, 1998, 19:127～132
[69] Battit R. First and Second Order Methods for learning: Between Steepest Descent and Newton's Method. Neural Computation, 1992, 4:141～166
[70] 阎平凡 最小描述长度与多层前馈网络设计中的一些问题．模式识别与人工智能，1993，6：143～148
[71] 阎平凡、张长水编著 《人工神经网络与模拟进化计算》 清华大学出版社 2000年11月第1版
[72] 陈维均、许斯欣全书主编，许斯欣、陈维均、林福兰编著 《甘蔗制糖原理与技术(第四分册)——蔗糖结晶与成糖》 中国轻工业出版社 2001年1月第1版
[73] 王俊普主编 《智能控制》 中国科学技术大学出版社 1996年9月第1版
[2] 丛爽编著《面向Matlab工具箱的神经网络理论与应用》 中国科学技术大学出版社 1998年11月第一版
[3] 闻新、周露、王丹力、熊晓英编著 《Matlab神经网络应用设计》 科学出版社 2000年9月第一版
[4] 张立明编著 《人工神经网络的模型及其应用》 复旦大学出版社 1993年7月第一版
[5] 高俊斌编著 《Matlab5.0语言与程序设计》 华中理工大学出版社 1999年1月第一版
[6] 楼顺天、于卫、闫华梁编著 《Matlab程序设计语言》 西安电子科技大学出版社 1997年8月第一版
[7] 王华 等编著 《Matlab在电信工程中的应用》 中国水利水电出版社 2001年4月第一版
[8] 《甘蔗制糖工业手册》编写组 编 《甘蔗制糖工业手册》(上册) 轻工业出版社 1989年5月第一版
[9] 《甘蔗制糖工业手册》编写组 编 《甘蔗制糖工业手册》(下册) 轻工业出版社 1989年5月第一版
[10] 广东省糖纸工业公司 李尔煊编写 《制糖工艺技术学习班(澄清、蒸发)学习讲义》 广东省地方国营雷高糖厂翻印 1991年1月27日
[11] 广东省糖纸工业公司 李尔煊编写 《制糖工艺技术学习班(煮糖)学习讲义》 广东省地方国营雷高糖厂翻印 1991年1月27日
[12] 华东理工大学硕士学位论文 姚颖写 《挡板流化化床丁烯氧化脱氢过程的建模与优化》 1998年5月20日
[13] K.F. Miller 《Crystal growth and vacuum pan productivity》 INT. SUGAR. JNL., 1999, VOL. 101, NO. 1204 p216-218
[14] Michel Benne,Brigitte Grondin-Perez,Jean-Danlel Luk and Jean-Pierre Chabriat. Modelling cane sugar evaporation and crystallization (Part 1:A new approach using neural netwoks) INT. SUGAR JNL., 1999,VOL.101,NO. 1208 p418-422
[15] Michel Benne,Brigitte Grondin-Perez,Jean-Danlel Luk and Jean-Pierre Chabriat. Modelling of evaporation and crystallization (Part 2:A new approach using neural netwoks). INT.SUGAR JNL., 1999,VOL.101,NO.1208 p418-422
[16] A. D. Randolph & S. A. Ziebold: Continuous Sucrose Nuclearion, I. S. J., Vol 76(1974)
[17] E. W. Krause: "New Equipments and Processes in the Sugar Industry" Licht F. O. L. 1987 E5-E29
[18] Paturau:" Medium and Long Term Development of Cane Sugar Factories" Presentation, 1988 Expert Group Meeting in Preparation of First Consultation on Sugar Industry, UNIDO
[19]Chun Y. Chen and Babu Joseph, On-line optimization using a two phase approach: An application study. Ind. Eng. Chem. Res. Vol. 26,1924-1930,1987
[20]叶江祺、何伟然、李世俊编著 《热工仪表和控制设备的安装》 水利电力出版社 1983年7月第一版
[21]邵裕森编 《过程控制及仪表》 上海交通大学出版社 1986年9月第1版
[22]中国自动化控制系统总公司 编 《自控系统成套设备选型样本》 陕西科学技术出版社 1989年1月第1版
[23]广东省糖纸食品公司 编《糖膏煮炼与助晶》 轻工业出版社 1985年2月第1版
[24]无锡轻工业学院、华南工学院编著 《甘蔗制糖工艺学》 轻工业出版社 1982年1月第一版
[25]霍汗镇 主编 陈树功、陈世治、林乐新等编 (广东省制糖学会组织) 《现代制糖工业技术》 中国轻工业出版社 1992年9月第一版
[26]高大维、陈树功等：煮糖起晶制种新方法，《广东省制糖学会年会论文选集》，1990年
[27]Shi-shang Jang;Babu Joseph and Hiro Mukai.On-line optimization of constrained multivariable chemical processes.AIChE J.Vol.33,26-35,1987.
[28]A.Lucia and J.Xu,Nonconvex progress optimization .Comp & Chen. Engng Vol.20, 1375-1398,1996
[29]A.V. Ooyen, Improving the convergence of the back-propagation algorithm. Neural Networks Vol.5.465-571,1992
[30]B.Robitaille,B.Marcos and M.Veillette,Modified quasi-Newton methods for trainning neural network. Comp. Engng Sci. Vol.25,859-865
[31]B.Schenker and M.Agarwal, Cross-validated structure selection for neural networks. Comp & Chem. Engng Vol.20, No.2,175-186,1996
[32]C.D.Pschogios and L.H.Ungar, Direct and indirect model based control using artificial neural networks.Ind.Eng.Chem.Res.Vol.30,2564-2573,1991
[33]D.Melo and J.Friedly, On-line closed-loop identification of multivariable systems.Ind.Eng.Chem.Res. Vol.36,1-22,1982
[34]E.P. Gerald, C.S.C.Daniel, R.A.Stephen and E.B.William, Using Taguchi's method of experimental design to control errors in layered perceptrons. IEEE Trans. Neural Networks Vol.6,949-960,1995
[35]F.J.Solis and J.B.Wets, Minimization by random search techniques. Math. Oper. Res. Vol.6, 19-30, 1981.
[36]G.L.Bilbro and W. E. Snyder, Optimization of function with many minima. IEEE Trans. Syst. Man Cybernet. Vol.21,840-849, 1991.
[37] H.Derrick,Nguyen and Bernard Widrow, Neural network for self-learning control system . IEEE Control Sys. Mag. Vol.4,18-23,1990．
[38] H. N. Robert, Theory of the Back-Propagation neural network. IJCNN, 1989
[39] M.A.Wolf, Numerical methods for unconstrained optimization. New York :Van Nostrand Reinhold,1978
[40] J.Saint-Donat, N. Bhat and T. J. McAvoy, Neural net based model predictive control. Int. J. Control Vol.54,1453-1468,1991d
[41] J. Savkovic-Stevanovic, Neural networks fo process analysis and optimization: modeling and applications. Comp & Chem. Engng Vol. 18,1149-1155,1994
[42] K. Juha and J. Jyrki, Generalization of principal compoent analysis, optimization problems and neural networks. Neural Networks Vol.8, 549-562,1995
[43] K. S. Narenda and K. Parthasarathy, Identification and control of dynamical system using neural networks. IEEE Trains. Neural Network Vol. 1,4-27,1990
[44] L. Leonard and MA.Kramer, Improvement of the backpropagation algorithm for training neural networks. Comp & Chem. Engng Vol.14,337-341,1990
[45] L. Megan and D. J. Cooper, A neural network strategy for disturbance pattern classifaction and adaptive multivariable control. Comp & Chem. Engng Vol. 19,171-186,1995
[46] M. J. Box, A new method of constrained optimization and a comparison with other methods. Computer J. Vol.8,42-52,1965
[47] N. Baba, Convergence of a random optimization method for constrained optimization problems. J. Optim. Theory Appl. Vol. 33,451-461,1981．
[48] N.Bhat and T. J. McAvoy, Use of neural nets for dynamic modeling and control of chemical process systems. Comp & Chem. Engng Vol.14,573-583,1990
[49] P. Tanartkit and L. T. Biegler, A nested , simutaneous approach for dynamic optimization problem. Comp & Chem. Engng Vol.20, 735-741,1996
[50] R. S. H. Mah and V. Chakravarthy, Pattern recognition using artificial neural Networks. Comp & Chem. Engng Vol.12, 555-562,1973
[51] Widrow B, et al. Neural Networks Application in Industry, Business and Science. Communcation of the ACM, 1994,3 7:93-105
[52] Special Issue on Application of Neural Network. Proc IEEE, 1996,84(10)
[53] SliMin F. An Expert Network for DNA Sequence Analysis, IEEE Intelligent Systems, Jan. /Feb. 1999,65-71
[54] Cybenko G . Approximation by Superpositions of Sigmoidal Function. Mathematics of Control, Signals and Systems, 1980, 2:303-314
[55] Hornik K, et al. Multilayer Feedforward Network are Universal Approximators. Neural Network, 1989,2:359-366
[56] Patrick P, et al.Minimisation Method for Training Feedforward Neural Networks. Neural Networks. 1994,7:1～11
[57] Yi S, et al. Global Optimization for Training. IEEE Computer, 1996, 3:45～54
[58] Van Rooij A J F, et al. Neural Networks Training Using Genetic Algorithm, World Scientific, 1996
[59] Looney C G. Advances in Feedforward Neural Netwoks, Denstifying Knowledge Acquiring Black Baxes.IEEE Trans Knowledge and Data Engeneering, 1996, 8:211～226
[60] Cohn D, et al. How Tight are the Vapnik-Chervoneko Bounds? Neural Computation, 1992,4:249～269
[61] Holden S B. On the Practical Applicability of VC Dimension Bounds. Neural Computation, 1995, 7:1265～1288
[62] Wei-Tsih Lee, et al. On an Asympototically Optimal Adaptive Classifier Design. IEEE Trans, 1993, PAMI-15:312～318
[63] Cohn A M, et al. On the Geometry of Feedward NN Weight Space. In:Proc 2nd IEEE Conference on ANN, London:IEEE Press,1991,1～4
[64] Hole A. Vapnik-Chervonenko Genneralization Bounds for Real Valued NN. Neural Computation, 1996, 8:1277-1299
[65] Guyon I, et al. What size Test Set Gives Good Error Rate Estimates? IEEE Trans PAMI,1998,(20):52～64
[66] Twomey J M, et al. Bias and Variance of Validation Methods for Function Approximation Neural Networks Chder Conditions of Sparse Data. IEEE Trans SMC PARTC, Q998, 28:417～430
[67] Devoroye L.Automatic Pattern Recognition:A Study of the Probability of Error. IEEE Trans PAMI, 1988, 28:417～430
[68] Bax E. Validation of Average Error Rate Over Classifiers Pattern Recongnition Letters, 1998, 19:127～132
[69] Battit R. First and Second Order Methods for learning: Between Steepest Descent and Newton's Method. Neural Computation, 1992, 4:141～166
[70] 阎平凡 最小描述长度与多层前馈网络设计中的一些问题．模式识别与人工智能，1993，6：143～148
[71] 阎平凡、张长水编著 《人工神经网络与模拟进化计算》 清华大学出版社 2000年11月第1版
[72] 陈维均、许斯欣全书主编，许斯欣、陈维均、林福兰编著 《甘蔗制糖原理与技术(第四分册)——蔗糖结晶与成糖》 中国轻工业出版社 2001年1月第1版
[73] 王俊普主编 《智能控制》 中国科学技术大学出版社 1996年9月第1版