摘要
尽管轮轨力测量的测力轮对技术已相对成熟,轮轨作用点位置的测量却一直很困难。作用点位置的在线连续测量对脱轨机理的研究、机车车辆性能的研究有十分重要的意义。在常规测力轮对的基础上,增加一个电桥感应作用点位置的变化。采用神经网络拟合轮轨作用力位置变化与电桥输出间复杂的非线性映射关系,用不同作用点位置下各种横、垂向力的组合来训练神经网络,从而达到由电桥输出值得到作用点位置的目的。实验结果表明,网络不仅训练精度好,而且预测能力也令人满意。
Although techniques to measure rail-wheel loads have been developed in recent years,it is difficult to detect rail-wheel contact points.Online detecting of contact points is very important to study on derailment mechanism and vehicle performances.On the basis of a conventional instrumented wheelset,an electrical bridge to induce changes of contact points is added.A method based on neural network is studied to calculate contact points.The neural network is trained to learn the non-linear relationship between the rail-wheel contact point and the bridge outputs,along with combinations of various horizontal and vertical loads at different contact points.Then,the locations of contact points can be calculated from the outputs of the electrical bridge.Experimental results show that the neural network has good training accuracy and satisfactory predicting ability.
出处
《振动与冲击》
EI
CSCD
北大核心
2007年第5期90-92,共3页
Journal of Vibration and Shock
基金
牵引动力国家重点实验实开放课题(TPL0607)资助
关键词
测力轮对
轮轨力接触点
连续测量
神经网络
instrumented wheelset,wheel/rail contact point,continuous measurement,neural network
