滑动电接触摩擦力的BP与RBF人工神经网络建模

BP and RBF Atificial Neural Network Modeling on Friction of Sliding Electrical Contact

中文关键词:  弓网系统  滑动电接触  摩擦力  人工神经网络  建模

英文关键词:pantograph catenary system  sliding electrical contact  friction  artificial neural network  modeling

中文摘要:

      在弓网系统中，受电弓滑板与接触网导线之间的摩擦力受机械、电气以及它们之间的交互作用影响。摩擦力与运行速度、接触电流以及受电弓滑板和接触网导线之间的压力载荷有着密切关系。通过对浸金属碳销（滑板）与铜盘（导线）的载流摩擦实验，得出在不同载流、速度以及载荷条件下的摩擦力特性规律。因摩擦力较难实现数学建模，故采用神经网络中应用最广的BP及RBF算法分别建立了以摩擦力作为输出，以接触压力、滑动速度和接触电流为输入的预测模型，并通过Matlab进行仿真与测试。结果表明：两种算法建立的预测模型均有较高的准确度及良好的泛化能力，为进一步波动载荷下弓网滑动电接触摩擦力预测建模的应用提供参考。

英文摘要:

      The friction between the pantograph slide and contact wire in pantograph catenary system is basically attributed to machinery,electricity and interaction between them,and it is closely related to the running speed,the contact current,and the pressure between the pantograph slide and the contact wire.The results of friction are obtained under various current,velocity and pressure through metal impregnated carbon rubbing against the copper wire.Due to the inconvenience of mathematical modeling,back propagation(BP) and radial basis function(RBF) most widely used in neural networks were used to establish the predictive model where the friction as output，contact pressure,sliding velocity and contact current as the import.The Matlab results show that the prediction models established by the two algorithms have high accuracy and good generalization ability,and provide reference for further application of predictive modeling of pantograph sliding electrical contact friction under dynamic load.