基于多支持向量机融合建模的直线电机结构设计研究
|
摘要
数控机床在对于提高我国的钢铁经济效率中占据了一个非常重要的地位,而直线电机又是数控机床中的一个非常重要的一部分。直线电机是一种新的研究方向,由于其进给速度范围宽、速度特性好、加速度大、定位精度高、结构简单、运动平稳、效率高、使用寿命长、安全可靠等优点,在数控机床进给系统中得到了推广应用。直线电机的结构优化是实现进给系统高速驱动的关键问题,对提高其推力品质具有重要理论和实践意义,直线电机也正是由于其结构的多变性,成为如今社会中各种领域所研究的对象。本文主要是围绕直线电机,对其结构进行研究建模,围绕该高效模型的参数进行各种优化,再对优化结果进行比较,深入研究电机的结构。本文主要研究内容如下：
一.研究多支持向量机融合建模方法
以单支持向量机建模方法为元件和基础,研究多支持向量机融合建模方法,对直线电机进行建模,建立出的高效模型具有单支持向量机模型所不具有的“多输入多输出”性能,同时也具有非常好的计算效率,是研究直线电机结构优化的重要理论前提及保障。
二.直线电机模型的实际验证
直线电机的有限元模型已经得到了实际验证,本文在此将多支持向量机融合方法得出的模型与有限元模型进行比较,通过推力与电流的比较,证明了该模型的正确性。
三.直线电机模型参数的优化
分别对直线电机模型的参数进行粒子群优化算法与差分进化算法进行优化,得出直线电机模型的最优参数,进一步验证了多支持向量机融合模型的正确性以及直线电机推力品质的提高。
四.提出退火差分进化算法
近年来所常用的粒子群优化算法与差分进化算法其实并非本文最理想的优化算法,本文在此提出一种新的优化算法,将模拟退火与差分进化算法进行融合得出一种新的优化算法,同时具有模拟退火算法的全局最优特性以及差分进化算法的高效计算特性,并且进一步得出了直线电机模型的更优参数值。
综上所述,本文采用多支持向量机融合方法对直线电机进行建模,分别进行实际验证,再对模型的参数进行优化,得出直线电机的最优参数。
Nc machine tools occupies a very important position to Enhance China's steel economic efficiency, and the linear motor is a very important part of nc machine. Linear motor is a kind of new research direction, because of its wide range, feeding speed, acceleration velocity characteristic, higher precision and simple structure, smooth movement, high efficiency, long service life, safe and reliable. It gets a widely application in the numerical control machine tool feeding system. Structure optimization design of linear motor is the key problems of realizing the feeding system and high-speed drive, it has the important theoretical and practical significance to improve its thrust the quality. Because of structure's polytrope, linear motor becomes the various field today. This article around the structure of the linear motor surrounded the efficient and parameters of the model optimization, and pay more attention of motel structure, which include several sections as following:
Firstly, study M-SVM method. We learned the M-SVM model for the Basic of the single SVM. Efficient model of linear motor also has a Multiple-input and multiple-output function, which single SVM model doesn't have, and computational efficiency. It is very useful to study the structure optimization of linear motor.
Secondly, the actual test of the linear motor M-SVM model. We have an actual test of linear motor FEM model. And will compare with FEM model for the thrust and current. And the result is that M-SVM model also can pass the actual test.
Thirdly, both the PSO and DE algorithms are used for linear motor model. The structure parameters of linear motor model approve that the rapid calculation model of linear motor is established and the model's accuracy together with efficiency is verified.
Fourthly, both the PSO and DE algorithms are not the best algorithms, and this article has another algorithm which is mixed by SA and DE algorithms. This algorithms has all of the two algorithms functions, global optimum and high efficiency. And it also produces a better structure value.
In short, this article uses M-SVM to have a model of linear motor, and then make an optimization of the model structure parameters.
一.研究多支持向量机融合建模方法
以单支持向量机建模方法为元件和基础,研究多支持向量机融合建模方法,对直线电机进行建模,建立出的高效模型具有单支持向量机模型所不具有的“多输入多输出”性能,同时也具有非常好的计算效率,是研究直线电机结构优化的重要理论前提及保障。
二.直线电机模型的实际验证
直线电机的有限元模型已经得到了实际验证,本文在此将多支持向量机融合方法得出的模型与有限元模型进行比较,通过推力与电流的比较,证明了该模型的正确性。
三.直线电机模型参数的优化
分别对直线电机模型的参数进行粒子群优化算法与差分进化算法进行优化,得出直线电机模型的最优参数,进一步验证了多支持向量机融合模型的正确性以及直线电机推力品质的提高。
四.提出退火差分进化算法
近年来所常用的粒子群优化算法与差分进化算法其实并非本文最理想的优化算法,本文在此提出一种新的优化算法,将模拟退火与差分进化算法进行融合得出一种新的优化算法,同时具有模拟退火算法的全局最优特性以及差分进化算法的高效计算特性,并且进一步得出了直线电机模型的更优参数值。
综上所述,本文采用多支持向量机融合方法对直线电机进行建模,分别进行实际验证,再对模型的参数进行优化,得出直线电机的最优参数。
Nc machine tools occupies a very important position to Enhance China's steel economic efficiency, and the linear motor is a very important part of nc machine. Linear motor is a kind of new research direction, because of its wide range, feeding speed, acceleration velocity characteristic, higher precision and simple structure, smooth movement, high efficiency, long service life, safe and reliable. It gets a widely application in the numerical control machine tool feeding system. Structure optimization design of linear motor is the key problems of realizing the feeding system and high-speed drive, it has the important theoretical and practical significance to improve its thrust the quality. Because of structure's polytrope, linear motor becomes the various field today. This article around the structure of the linear motor surrounded the efficient and parameters of the model optimization, and pay more attention of motel structure, which include several sections as following:
Firstly, study M-SVM method. We learned the M-SVM model for the Basic of the single SVM. Efficient model of linear motor also has a Multiple-input and multiple-output function, which single SVM model doesn't have, and computational efficiency. It is very useful to study the structure optimization of linear motor.
Secondly, the actual test of the linear motor M-SVM model. We have an actual test of linear motor FEM model. And will compare with FEM model for the thrust and current. And the result is that M-SVM model also can pass the actual test.
Thirdly, both the PSO and DE algorithms are used for linear motor model. The structure parameters of linear motor model approve that the rapid calculation model of linear motor is established and the model's accuracy together with efficiency is verified.
Fourthly, both the PSO and DE algorithms are not the best algorithms, and this article has another algorithm which is mixed by SA and DE algorithms. This algorithms has all of the two algorithms functions, global optimum and high efficiency. And it also produces a better structure value.
In short, this article uses M-SVM to have a model of linear motor, and then make an optimization of the model structure parameters.
引文
[1]丁锦红,高速数控机床概述[J],中国制造业信息化,2006(6),p：36-41.
[2]Baum, E.B. and Haussler, D.,1989. "What size net gives valid generalization?" Neural Computation [M], vol.1, p151-160.
[3]宋刚,基于粒子群和差分进化算法的新型混合算法在配电网重构中的应用[M].
[4]Vapnik V N. Estimation of Dependencies Based on Empirical Data[M]. Berlin: Spring-Verlag,1982.
[5]张学工统计学习理论的本质[J]北京：清华大学出版社,1999.
[6]Cherkassky V, Mulier F. Learning from Data:Concepts[J], Theory and Methods NY:John Viley & Sons,1997.
[7]张学工,关于统计学习理论与支持向量机[J],自动化学报,2000.
[8]边肇祺等,模式识别[M]。北京清华大学出版社,1998.
[9]王俊伟.粒子群优化算法的改进及应用[D].哈尔滨：东北大学,2006.1.
[10]Kim H K, Chong J K, Park KY, et al. Differential Evolution Strategy for Constrained Global Optimization and Application to Practical Engineering Problems[M]. IEEE Tram on Magnetics,2007,43(4):1565-1568.
[11]H-C Kim, "Construction support vector machine ensemble", Pattern Recognition, Vol.36, pp2757-2767,2003.
[12]Bing-Yu Sun and D.S.Huang, "Least squares support vector machine ensemble, "The 2004 International Joint Conference on Neural Networks (IJCNN2004), Budapest Hungary[J], July 25-29,2004.
[13]Bing-Yu Sun and De-Shuang Huang, "Support vector machine committee for classification," Lecture Notes in Computer Science 3173:pp.648-653, Springer-Verlag[J],2004.
[14]Zhi-Hua Zhou, "Enesmbling neural network:Many could be better than all", Artifical intelligence[J], Vol.137, pp236-263,2002.
[15]G.Ratsch, "Soft Margins for Adaboost", Machine Learning[J]," pp1-35,2000.
[16]B.Duerr, W.Haettich, and H. Tropf, "A combination of statistic and syntactical pattern recognition applied to classification of unconstrained handwritten numerals". Pattern Recognition[J],12(1):189-199,1980.
[17]P.Ahmed and C.Y Suen, "Computer Recognition of totally unconstrained handwritten ZIP codes"[J]. Pattern Recognition and Artificial Intelligence, 1(1):1-15,1987.
[18]叶云岳,直线电机原理与应用[M].北京,机械工业出版社,2000.
[19]王秉中,计算电磁学[M],北京,科学出版社,2002年7月第一版.
[20]Dong-Yeup Lee; Chun-Gil Jung; Kang-Jun Yoon; Gyu-Tak Kim;A study on the efficiency optimum design of a permanent magnet type linear synchronous motor [J], Magnetics, IEEE Transactions on, May.2005,1860-1863,Issue:5.
[21]赵吉文；孔凡让；孙丙宇,基于SVM和混沌优化的新型直线电机参数优化[J],中国机械工程,2005(20),p：1697-1701.
[22]Krivy, Ivan, Tvrdik, Josef, Krpec, Radek, Stochastic algorithms in nonlinear regression [J], Computational Statistics and Data Analysis, v 33,n 3, May,2000,p 277-290.
[23]Rahman, Mezbahur, Gokhale, D.V., Ullah, Aman, Note on combining parametric and non-parametric regression, Communications in Statistics[J]. Part B:Simulation and Computation, v 26,n 2,1997,p 519-529.
[24]Kim, Kwang Jae, Chen, Hsien-Ruey, Comparison of fuzzy and nonparametric linear regression[J], Computers & Operations Research,v 24,n 6,Jun,1997,p 505-519.
[25]Delicado,P., del Rio,M., Weighted nonparametric regression, Communications in Statistics Theory and Methods[J], v 26,n 12,Dec,1997, p 2983-2998.
[26]Kundu, Debasis, Murali, G., Model selection in linear regression[J], Computational Statistics & Data Analysis, v 22,n 5, Sep 30,1996,p 461-469.
[27]Camacho, F., Avolio, A.P., Lovell, N.H., A Linear Regression Model for Assessment of Interdependent Parameters of Central Pressure Determined from the Peripheral Pulse[J], Annual International Conference of the IEEE Engineering in Medicine and Biology Proceedings, v 3,2003,p 2826-2829.
[28]边肇祺,张学工等,编著,模式识别[M],北京,清华大学出版社,2000年1月.
[29]Bernhard E et al. A training algorithm for optimal margin classifier[J]. In:Proc of the 5th Annual ACM Workshop on Computational Learning Theory. Pittsburgh: ACM Press,1992.144-152.
[30]Cortes C, Vapnic V. Support-vector networks[J]. Machine Learning, 1995,20(3):273-397.
[31]Vapnic V N. The Nature of Statistical Learning Theory [J]. New York: Springer-Verlag,1995.
[32]赵吉文,新型直线电机支持向量机非线性建模研究[J],光学精密工程,2006(06),p：450-455.
[33]Storn R, Price K, Differential evolution——a simple and efficient adaptive scheme for global optimization over continuous spaces [J]. Berkeley:University of California,2006.
[34]Cherkassky, Vladimir; Ma, Yunquian, Practical selection of SVM parameters and noise estimation for SVM regression[J], Neural Networks Volume:17,Issue:1,January,2004, pp.113-126.
[35]赵吉文；刘永斌；孔凡让；陈军宁,核参数遗传选优的SVM在直线电机建模中的应用[J],系统仿真学报,2006(22),P：3547-3549+3553.
[36]Lopez C I L, Willigenburg L G, van Straten G. Efficient Differential Evolution Algorithms for Multimodal Optimal Control Problems [J]. Applied Soft Computing, 2003,3(2):97-122.
[37][19]崔逊学.多目标进化算法及其应用[M].北京：国防工业出版社,2006.
[38]Lampinen J. A bibliography of differential evolution algorithm [EB/OL]. http://www.lut.fi/～jlampine/debiblio.htm.2002-10-14.
[39]Lin Y C, Hwang K S, Wang F S. Co-evolutionary hybrid differential evolution for mixed-integer optimization problems[J]. Engineering Optimization, 2001;33(6):663-682.
[40]Cheng S, Hwang C. Optimal approximation of linear systems by a differential evolution algorithm[J]. IEEE Trans on Systems, Man and Cybernetics:A, 2001;31(6):698-707.
[41]Kim H K, Chong J K, Park KY, et al. Differential Evolution Strategy for Constrained Global Optimization and Application to Practical Engineering Problems[J]. IEEE Tram on Magnetics,2007,43(4):1565-1568.
[42]Omran M G H, Engelbrecht A P. Serf-Adaptive Differential Evolution Methods for Unsupervised Image Classification Prec of the IEEE Conference on Cybernetics and Intelligent Systems[J]. Bangkok, Thailand.2006:1-6.
[43]Zhang Renqian, Ding Jianxun. Non-Linear Optimal Control of Manufacturing System Based on Modified Differential Evolution Proc of the IMACS Multi-conference on Computational Engineering in Systems Applications[J]. Beijing, China,2006:1797-1803.
[44]Dhahri H. Alimi A M. The Modified Differential Evolution and the RBF(MDE —RBF)Neural Network for Time Series Prediction Proc of the International Joint Conference on Neural Networks [J]. Vancouver.USA,2006:2938-2943.
[45]Yang Shiwen, GanYB, Qing Anyong. Sideband Suppression in Time-Modulated Linear Arrays by the Differential Evolution Algorithm[J]. IEEE Tram on Antennas and Wireless Propagation Letters,2002,1(1):173-175.
[46]Masza A, Pastorino M, Randazzo A. Optimization of the Directivity of a Monopulse Antenna with a Subarray Weishting by a Hybrid Differential Evolution Method[J]. IEEE Tram on Antennas and Wireless Propagation LettelB,2006,5(1): 155-158.
[47]范瑜,金荣洪,耿军平,等.基于差分进化算法和遗传算法的混合优化算法及其在阵列天线方向图综合中的应用[J].电子学报,2004。32(12)：1997-2000.
[48]吴亮红,王耀南,袁小芳,等.自适应二次变异差分进化算法[J].控制与决策,2006,21(8)：898-902.
[49]Su C T, Lee C S. Network Reconfiguration of Distribution Systems Using Improved Mixed-Integer Hybrid Differential Evolution[J]. IEEE Tram Oil Power Delivery Review,2002,22(12):60-66.
[50]翟捷,王春峰,李光泉.基于差分进化方法的投资组合管理模型[J].天津大学学报：自然科学与工程技术版,2002,35(3)：304-308.
[51]万东.差分进化算法研究及其应用[J].科学技术与工程,2009.11.
[52]Metropolis N, Rosenbluth A, Rosenbluth M.Equation of state calculations by fast computing machines [J]. Journal of Chemical Physics,1953,56 (21):1087-1092.
[53]Kirkpatrick S, Gelatt Jr C D, Vecchi M P.Optimization by simulated annealing[J].Science,1983,220 (21):671-650.
[54]宋刚,基于粒子群和差分进化算法的新型混合算法在配电网重构中的应用[D].
[55]郭德龙,夏慧明,周永权,混合模拟退火-进化策略在非线性参数估计中的应用[J],数学的实践与认识,2010-11.
[2]Baum, E.B. and Haussler, D.,1989. "What size net gives valid generalization?" Neural Computation [M], vol.1, p151-160.
[3]宋刚,基于粒子群和差分进化算法的新型混合算法在配电网重构中的应用[M].
[4]Vapnik V N. Estimation of Dependencies Based on Empirical Data[M]. Berlin: Spring-Verlag,1982.
[5]张学工统计学习理论的本质[J]北京：清华大学出版社,1999.
[6]Cherkassky V, Mulier F. Learning from Data:Concepts[J], Theory and Methods NY:John Viley & Sons,1997.
[7]张学工,关于统计学习理论与支持向量机[J],自动化学报,2000.
[8]边肇祺等,模式识别[M]。北京清华大学出版社,1998.
[9]王俊伟.粒子群优化算法的改进及应用[D].哈尔滨：东北大学,2006.1.
[10]Kim H K, Chong J K, Park KY, et al. Differential Evolution Strategy for Constrained Global Optimization and Application to Practical Engineering Problems[M]. IEEE Tram on Magnetics,2007,43(4):1565-1568.
[11]H-C Kim, "Construction support vector machine ensemble", Pattern Recognition, Vol.36, pp2757-2767,2003.
[12]Bing-Yu Sun and D.S.Huang, "Least squares support vector machine ensemble, "The 2004 International Joint Conference on Neural Networks (IJCNN2004), Budapest Hungary[J], July 25-29,2004.
[13]Bing-Yu Sun and De-Shuang Huang, "Support vector machine committee for classification," Lecture Notes in Computer Science 3173:pp.648-653, Springer-Verlag[J],2004.
[14]Zhi-Hua Zhou, "Enesmbling neural network:Many could be better than all", Artifical intelligence[J], Vol.137, pp236-263,2002.
[15]G.Ratsch, "Soft Margins for Adaboost", Machine Learning[J]," pp1-35,2000.
[16]B.Duerr, W.Haettich, and H. Tropf, "A combination of statistic and syntactical pattern recognition applied to classification of unconstrained handwritten numerals". Pattern Recognition[J],12(1):189-199,1980.
[17]P.Ahmed and C.Y Suen, "Computer Recognition of totally unconstrained handwritten ZIP codes"[J]. Pattern Recognition and Artificial Intelligence, 1(1):1-15,1987.
[18]叶云岳,直线电机原理与应用[M].北京,机械工业出版社,2000.
[19]王秉中,计算电磁学[M],北京,科学出版社,2002年7月第一版.
[20]Dong-Yeup Lee; Chun-Gil Jung; Kang-Jun Yoon; Gyu-Tak Kim;A study on the efficiency optimum design of a permanent magnet type linear synchronous motor [J], Magnetics, IEEE Transactions on, May.2005,1860-1863,Issue:5.
[21]赵吉文；孔凡让；孙丙宇,基于SVM和混沌优化的新型直线电机参数优化[J],中国机械工程,2005(20),p：1697-1701.
[22]Krivy, Ivan, Tvrdik, Josef, Krpec, Radek, Stochastic algorithms in nonlinear regression [J], Computational Statistics and Data Analysis, v 33,n 3, May,2000,p 277-290.
[23]Rahman, Mezbahur, Gokhale, D.V., Ullah, Aman, Note on combining parametric and non-parametric regression, Communications in Statistics[J]. Part B:Simulation and Computation, v 26,n 2,1997,p 519-529.
[24]Kim, Kwang Jae, Chen, Hsien-Ruey, Comparison of fuzzy and nonparametric linear regression[J], Computers & Operations Research,v 24,n 6,Jun,1997,p 505-519.
[25]Delicado,P., del Rio,M., Weighted nonparametric regression, Communications in Statistics Theory and Methods[J], v 26,n 12,Dec,1997, p 2983-2998.
[26]Kundu, Debasis, Murali, G., Model selection in linear regression[J], Computational Statistics & Data Analysis, v 22,n 5, Sep 30,1996,p 461-469.
[27]Camacho, F., Avolio, A.P., Lovell, N.H., A Linear Regression Model for Assessment of Interdependent Parameters of Central Pressure Determined from the Peripheral Pulse[J], Annual International Conference of the IEEE Engineering in Medicine and Biology Proceedings, v 3,2003,p 2826-2829.
[28]边肇祺,张学工等,编著,模式识别[M],北京,清华大学出版社,2000年1月.
[29]Bernhard E et al. A training algorithm for optimal margin classifier[J]. In:Proc of the 5th Annual ACM Workshop on Computational Learning Theory. Pittsburgh: ACM Press,1992.144-152.
[30]Cortes C, Vapnic V. Support-vector networks[J]. Machine Learning, 1995,20(3):273-397.
[31]Vapnic V N. The Nature of Statistical Learning Theory [J]. New York: Springer-Verlag,1995.
[32]赵吉文,新型直线电机支持向量机非线性建模研究[J],光学精密工程,2006(06),p：450-455.
[33]Storn R, Price K, Differential evolution——a simple and efficient adaptive scheme for global optimization over continuous spaces [J]. Berkeley:University of California,2006.
[34]Cherkassky, Vladimir; Ma, Yunquian, Practical selection of SVM parameters and noise estimation for SVM regression[J], Neural Networks Volume:17,Issue:1,January,2004, pp.113-126.
[35]赵吉文；刘永斌；孔凡让；陈军宁,核参数遗传选优的SVM在直线电机建模中的应用[J],系统仿真学报,2006(22),P：3547-3549+3553.
[36]Lopez C I L, Willigenburg L G, van Straten G. Efficient Differential Evolution Algorithms for Multimodal Optimal Control Problems [J]. Applied Soft Computing, 2003,3(2):97-122.
[37][19]崔逊学.多目标进化算法及其应用[M].北京：国防工业出版社,2006.
[38]Lampinen J. A bibliography of differential evolution algorithm [EB/OL]. http://www.lut.fi/～jlampine/debiblio.htm.2002-10-14.
[39]Lin Y C, Hwang K S, Wang F S. Co-evolutionary hybrid differential evolution for mixed-integer optimization problems[J]. Engineering Optimization, 2001;33(6):663-682.
[40]Cheng S, Hwang C. Optimal approximation of linear systems by a differential evolution algorithm[J]. IEEE Trans on Systems, Man and Cybernetics:A, 2001;31(6):698-707.
[41]Kim H K, Chong J K, Park KY, et al. Differential Evolution Strategy for Constrained Global Optimization and Application to Practical Engineering Problems[J]. IEEE Tram on Magnetics,2007,43(4):1565-1568.
[42]Omran M G H, Engelbrecht A P. Serf-Adaptive Differential Evolution Methods for Unsupervised Image Classification Prec of the IEEE Conference on Cybernetics and Intelligent Systems[J]. Bangkok, Thailand.2006:1-6.
[43]Zhang Renqian, Ding Jianxun. Non-Linear Optimal Control of Manufacturing System Based on Modified Differential Evolution Proc of the IMACS Multi-conference on Computational Engineering in Systems Applications[J]. Beijing, China,2006:1797-1803.
[44]Dhahri H. Alimi A M. The Modified Differential Evolution and the RBF(MDE —RBF)Neural Network for Time Series Prediction Proc of the International Joint Conference on Neural Networks [J]. Vancouver.USA,2006:2938-2943.
[45]Yang Shiwen, GanYB, Qing Anyong. Sideband Suppression in Time-Modulated Linear Arrays by the Differential Evolution Algorithm[J]. IEEE Tram on Antennas and Wireless Propagation Letters,2002,1(1):173-175.
[46]Masza A, Pastorino M, Randazzo A. Optimization of the Directivity of a Monopulse Antenna with a Subarray Weishting by a Hybrid Differential Evolution Method[J]. IEEE Tram on Antennas and Wireless Propagation LettelB,2006,5(1): 155-158.
[47]范瑜,金荣洪,耿军平,等.基于差分进化算法和遗传算法的混合优化算法及其在阵列天线方向图综合中的应用[J].电子学报,2004。32(12)：1997-2000.
[48]吴亮红,王耀南,袁小芳,等.自适应二次变异差分进化算法[J].控制与决策,2006,21(8)：898-902.
[49]Su C T, Lee C S. Network Reconfiguration of Distribution Systems Using Improved Mixed-Integer Hybrid Differential Evolution[J]. IEEE Tram Oil Power Delivery Review,2002,22(12):60-66.
[50]翟捷,王春峰,李光泉.基于差分进化方法的投资组合管理模型[J].天津大学学报：自然科学与工程技术版,2002,35(3)：304-308.
[51]万东.差分进化算法研究及其应用[J].科学技术与工程,2009.11.
[52]Metropolis N, Rosenbluth A, Rosenbluth M.Equation of state calculations by fast computing machines [J]. Journal of Chemical Physics,1953,56 (21):1087-1092.
[53]Kirkpatrick S, Gelatt Jr C D, Vecchi M P.Optimization by simulated annealing[J].Science,1983,220 (21):671-650.
[54]宋刚,基于粒子群和差分进化算法的新型混合算法在配电网重构中的应用[D].
[55]郭德龙,夏慧明,周永权,混合模拟退火-进化策略在非线性参数估计中的应用[J],数学的实践与认识,2010-11.