机车牵引特性综合测试系统的研究
|
摘要
机车牵引特性测试是在新建线路投入运营、客运列车提速、货运列车增加牵引吨位、司乘制度改革及机车交路变化等情况下所进行的一项基础性技术工作,为铁路部门经营决策、指定运输方案、挖潜增效提供科学依据。它的核心是能够实时计算得到机车运行过程中的牵引力和制动力。针对传统的牵引特性测试方法得到的牵引力与制动力误差较大的不足,提出了采用人工神经网络实时计算机车运行过程中牵引力与制动力的方法。
本文分析探讨了现有测试方法的不足,同时,通过人工神经网络技术的原理和特性的研究,以及其非线性逼近能力的分析,论证了神经网络技术应用于机车牵引特性测试上的可行性。
本文采用了误差反向传播网络(BP网络)建立了牵引力与制动力的计算模型,详细阐述了该网络的结构设计,讨论了BP网络隐含层节点数目的选择、数据的预处理等问题;并针对BP算法易陷入局部极小、收敛速度慢等不足,提出了改进措施。
最后,采用由成都铁路局提供的实验数据在MATLAB6.5环境下进行了仿真,实验结果表明,采用B P网络计算机车的牵引力和制动力,计算结果精度更高。
Under the circumstances of putting new railway lines into use or improving passenger train speed or increasing freight train tonnage or reforming driving system, locomotive tractive characteristic test has become a foundational technical job. It provides scientific data for railway administration to make management decision and transportations plan, to dig potential and increase profit. The real time calculation of tractive force and braking force is the core of the system. Aiming at the disadvantages of traditional calculations of locomotive tractive and braking force,the method of Artificial Neural Network model is proposed. The structure design of the neural network iselaborated.
The paper finds out the disadvantages of traditional method. At the meantime,by researching theory and characteristic of the Artificial Neural Network and analyzing its nonlinear approaching capacity,the writer probes into the feasibility of the application of Artificial Neural Network applying to the locomotive tractive characteristic test.
A calculation of tractive force and braking force model using Error Back-Propagation Training neural network(BP netwok) is proposed. The structure design of the neural network iselaborated.the problems such as choice about the node number of hidden layer,data pretreatment ,and so on are dicussed;in allusion to the shortages of BP arithmetic such as easy to fall into local smallness and slow constringency speed,the paper brought forward the amelioratin measures.
In the end, the simulation is carried out with Matlab 6.5 using the data provided by Chengdu Railway Bureau and these results are presented. It was shown that the accuracy of the result is increased.
本文分析探讨了现有测试方法的不足,同时,通过人工神经网络技术的原理和特性的研究,以及其非线性逼近能力的分析,论证了神经网络技术应用于机车牵引特性测试上的可行性。
本文采用了误差反向传播网络(BP网络)建立了牵引力与制动力的计算模型,详细阐述了该网络的结构设计,讨论了BP网络隐含层节点数目的选择、数据的预处理等问题;并针对BP算法易陷入局部极小、收敛速度慢等不足,提出了改进措施。
最后,采用由成都铁路局提供的实验数据在MATLAB6.5环境下进行了仿真,实验结果表明,采用B P网络计算机车的牵引力和制动力,计算结果精度更高。
Under the circumstances of putting new railway lines into use or improving passenger train speed or increasing freight train tonnage or reforming driving system, locomotive tractive characteristic test has become a foundational technical job. It provides scientific data for railway administration to make management decision and transportations plan, to dig potential and increase profit. The real time calculation of tractive force and braking force is the core of the system. Aiming at the disadvantages of traditional calculations of locomotive tractive and braking force,the method of Artificial Neural Network model is proposed. The structure design of the neural network iselaborated.
The paper finds out the disadvantages of traditional method. At the meantime,by researching theory and characteristic of the Artificial Neural Network and analyzing its nonlinear approaching capacity,the writer probes into the feasibility of the application of Artificial Neural Network applying to the locomotive tractive characteristic test.
A calculation of tractive force and braking force model using Error Back-Propagation Training neural network(BP netwok) is proposed. The structure design of the neural network iselaborated.the problems such as choice about the node number of hidden layer,data pretreatment ,and so on are dicussed;in allusion to the shortages of BP arithmetic such as easy to fall into local smallness and slow constringency speed,the paper brought forward the amelioratin measures.
In the end, the simulation is carried out with Matlab 6.5 using the data provided by Chengdu Railway Bureau and these results are presented. It was shown that the accuracy of the result is increased.
引文
[1]张中央.列车牵引计算.北京:中国铁道出版社,2004
[2]曾理.提速列车牵引计算软件自动计算功能研究.硕士学位论文.成都:西南交通大学,2004
[3]郑雪洋.动车组牵引性能试验方法的探讨,铁道机车车辆,2002.3:40-42
[4]孙中央.列车牵引计算规程实用教程.北京:中国铁道出版社,1999
[5]彭其渊.城市轨道交通列车牵引计算.成都:西南交通大学出版社,2005
[6]刘友梅.韶山3型4000系电力机车,北京:中国铁道出版社,2000
[7]凌鹤奇.电力机车牵引电机.北京:中国铁道出版社,1980
[8]西南交通大学电机系,牵引电机.中国铁道出版社,北京,1981。
[9]李占峰,机车牵引试验测量系统的研究.西南交通大学,硕士学位论文,2006
[10]中华人民共和国铁道部.列车牵引计算规程.北京:中国铁道出版社,1999
[11]蒋宗礼.人工神经网络导论.北京:高等教育出版社,2001.
[12]王奕淳,基于人工神经网络的期货预测.吉林大学,硕士学位论文,2007
[13]李贞珍,神经网络在城市交通流预测模型中应用的研究.硕士学位论文.上海:上海交通大学,2007
[14]张铃,张钹,人工神经网络技术及其应用[M],浙江:浙江科学技术出版社,1997
[15]沈世镒.神经网络系统理论及其应用[M],北京:高等教育出版社,2001
[16]Neural Network Toolbox.Math Works,2007
[17]孟维伟 基于神经网络的交通量预测技术研究.硕士学位论文.南京:南京理工大学,2006
[18]阔丹锋.人工神经网络在化工物性定量预测中的应用研究.博士学位论文.北京化工大学.1992
[19]赵振宇,徐用惫.模糊理论和神经网络的基础与应用[M].北京、南宁:清华大学出版社、广西科学技术出版社,1992.70-88
[20]廖文辉.供应链环境下的汽车零配件供应商综合评价体系的研究.浙江大学,硕士学位论文.2006
[21]张玉朵,唐钢二炼铁厂3号高炉铁水硅含量神经网络预报模型.河北理工 学院,硕士学位论文,2003
[22]刘典政,冯晓云,等.机车牵引力的在线实时计算[J].铁道机车车辆,2005(12)
[23]高大启.有教师的线性基本函数前向三层神经网络结构研究[J].计算机学报,1998,21(1):80-85.
[24]Rumelhart D E.Learning Representations by Backpropagating Errors[J].Nature,1986,323(9):533-536.
[25]Johnson E M,Dowla F U,Goodman D M.Backpropagating Learning for Multi-layer Feed-forward Neural Networks Using the Conjugate Gradient Method[J].Int J Neural System,1991,2(4):291-302.
[26]Friedrich B K,Barmann F.A Learning Algorithm for Multilayered Neural Networks Based on Linear Least Square Problems[J].Neural Networks,1993,4(6):127-602.
[27]刘典政,冯晓云.基于LM神经网络模型的机车牵引力和制动力的计算.机车电传动,2007(3)
[28]蔡清谊.基于人工神经网络的洪水预报与灾度评价研究.合肥工业大学,硕士学位论文,2008
[29]李贞珍,神经网络在城市交通流预测模型中应用的研究.硕士学位论文.上海:上海交通大学,2007
[30]杨伟 等.BP网络权值初始值与收敛性问题研究.电力系统及其自动化学报,2002,14(1)
[31]刘刚.一种综合改进的BP神经网络及其实现[J].武汉理工大学学报,2002,24(10):57-60.
[32]徐志钮等,结合N-W方法的L-M算法在变压器故障诊断中的应用。华北电力大学学报,2005.7.32-7。
[33]张向东等.基于MATLAB的ANN方法在桩基岩石强度预测中的应用.地下空间与工程学报,2005,1(7)
[34]周开利,康耀红.神经网络模型及其MATLAB仿真程序设计[M].北京:清华大学出版社,2004.
[35]飞思科技产品研发中心.神经网络理论与MATLAB7实现[M].北京:电子工业出版社,2005.
[36]徐海云,涂雄苓.海量数据导入与导出的有效方法.数理统计与管理,2007.26(4)
[37]罗成汉.基于神经网络工具箱的网络实现.计算机仿真,2004,21(5)
[38]白润才,李建刚,郭旭颖.基于图形用户界面的BP神经网络建模与仿真.系统仿真学报,2005,17(11)
[39]Hagan M T,Menhaj M B.Training feedforward networks with the marquardt algorithm[J].IEEE Trans Neural Net,1994,5(6):989-993.
[40]Hagan M T,Howard B Demuth.神经网络设计[M].北京,机械工业出版社,2002.
[41]赵弘,等.基于Levenberg_Marquardt算法的神经网络监督控制[J].西安交通大学学报,2002(5)
[42]R.C.Eberhart and R.W.Dobbins,Neural Network Pc tools,Academic Press,1990,pp93-155.
[43]NguygoD,WidrowB.Improving the learning speed of 2-layer Neural Networks by choosing initial values of the adaptive weights[C].SanDiego,USA:international joint Conference on Neural network,1990,17-21.
[44]ShimodariaH.A Weight Value Initialization Method for Improved Learning Performance of the Back Propagation Algorithm in Neural Networks[J].Proc.of the 6th International Conference on Tools with Artificial Intelligence,1994,2:854.
[2]曾理.提速列车牵引计算软件自动计算功能研究.硕士学位论文.成都:西南交通大学,2004
[3]郑雪洋.动车组牵引性能试验方法的探讨,铁道机车车辆,2002.3:40-42
[4]孙中央.列车牵引计算规程实用教程.北京:中国铁道出版社,1999
[5]彭其渊.城市轨道交通列车牵引计算.成都:西南交通大学出版社,2005
[6]刘友梅.韶山3型4000系电力机车,北京:中国铁道出版社,2000
[7]凌鹤奇.电力机车牵引电机.北京:中国铁道出版社,1980
[8]西南交通大学电机系,牵引电机.中国铁道出版社,北京,1981。
[9]李占峰,机车牵引试验测量系统的研究.西南交通大学,硕士学位论文,2006
[10]中华人民共和国铁道部.列车牵引计算规程.北京:中国铁道出版社,1999
[11]蒋宗礼.人工神经网络导论.北京:高等教育出版社,2001.
[12]王奕淳,基于人工神经网络的期货预测.吉林大学,硕士学位论文,2007
[13]李贞珍,神经网络在城市交通流预测模型中应用的研究.硕士学位论文.上海:上海交通大学,2007
[14]张铃,张钹,人工神经网络技术及其应用[M],浙江:浙江科学技术出版社,1997
[15]沈世镒.神经网络系统理论及其应用[M],北京:高等教育出版社,2001
[16]Neural Network Toolbox.Math Works,2007
[17]孟维伟 基于神经网络的交通量预测技术研究.硕士学位论文.南京:南京理工大学,2006
[18]阔丹锋.人工神经网络在化工物性定量预测中的应用研究.博士学位论文.北京化工大学.1992
[19]赵振宇,徐用惫.模糊理论和神经网络的基础与应用[M].北京、南宁:清华大学出版社、广西科学技术出版社,1992.70-88
[20]廖文辉.供应链环境下的汽车零配件供应商综合评价体系的研究.浙江大学,硕士学位论文.2006
[21]张玉朵,唐钢二炼铁厂3号高炉铁水硅含量神经网络预报模型.河北理工 学院,硕士学位论文,2003
[22]刘典政,冯晓云,等.机车牵引力的在线实时计算[J].铁道机车车辆,2005(12)
[23]高大启.有教师的线性基本函数前向三层神经网络结构研究[J].计算机学报,1998,21(1):80-85.
[24]Rumelhart D E.Learning Representations by Backpropagating Errors[J].Nature,1986,323(9):533-536.
[25]Johnson E M,Dowla F U,Goodman D M.Backpropagating Learning for Multi-layer Feed-forward Neural Networks Using the Conjugate Gradient Method[J].Int J Neural System,1991,2(4):291-302.
[26]Friedrich B K,Barmann F.A Learning Algorithm for Multilayered Neural Networks Based on Linear Least Square Problems[J].Neural Networks,1993,4(6):127-602.
[27]刘典政,冯晓云.基于LM神经网络模型的机车牵引力和制动力的计算.机车电传动,2007(3)
[28]蔡清谊.基于人工神经网络的洪水预报与灾度评价研究.合肥工业大学,硕士学位论文,2008
[29]李贞珍,神经网络在城市交通流预测模型中应用的研究.硕士学位论文.上海:上海交通大学,2007
[30]杨伟 等.BP网络权值初始值与收敛性问题研究.电力系统及其自动化学报,2002,14(1)
[31]刘刚.一种综合改进的BP神经网络及其实现[J].武汉理工大学学报,2002,24(10):57-60.
[32]徐志钮等,结合N-W方法的L-M算法在变压器故障诊断中的应用。华北电力大学学报,2005.7.32-7。
[33]张向东等.基于MATLAB的ANN方法在桩基岩石强度预测中的应用.地下空间与工程学报,2005,1(7)
[34]周开利,康耀红.神经网络模型及其MATLAB仿真程序设计[M].北京:清华大学出版社,2004.
[35]飞思科技产品研发中心.神经网络理论与MATLAB7实现[M].北京:电子工业出版社,2005.
[36]徐海云,涂雄苓.海量数据导入与导出的有效方法.数理统计与管理,2007.26(4)
[37]罗成汉.基于神经网络工具箱的网络实现.计算机仿真,2004,21(5)
[38]白润才,李建刚,郭旭颖.基于图形用户界面的BP神经网络建模与仿真.系统仿真学报,2005,17(11)
[39]Hagan M T,Menhaj M B.Training feedforward networks with the marquardt algorithm[J].IEEE Trans Neural Net,1994,5(6):989-993.
[40]Hagan M T,Howard B Demuth.神经网络设计[M].北京,机械工业出版社,2002.
[41]赵弘,等.基于Levenberg_Marquardt算法的神经网络监督控制[J].西安交通大学学报,2002(5)
[42]R.C.Eberhart and R.W.Dobbins,Neural Network Pc tools,Academic Press,1990,pp93-155.
[43]NguygoD,WidrowB.Improving the learning speed of 2-layer Neural Networks by choosing initial values of the adaptive weights[C].SanDiego,USA:international joint Conference on Neural network,1990,17-21.
[44]ShimodariaH.A Weight Value Initialization Method for Improved Learning Performance of the Back Propagation Algorithm in Neural Networks[J].Proc.of the 6th International Conference on Tools with Artificial Intelligence,1994,2:854.