列车脱轨系数的精确预测方法研究

Research on Accurate Prediction Method of Train Derailment Coefficient

研究列车脱轨系数的精确预测问题。列车在运行过程中,脱轨系数过大会产生列车脱轨隐患,因此列车脱轨系数是评价列车运行安全的重要依据。针对传统脱轨系数预测方法成本高和预测精度低等问题,提出RB算法的NARX神经网络的脱轨系数预测方法。以实测的轨道不平顺为输入,脱轨系数为输出,分别建立BP神经网络和NARX神经网络两种预测模型,并分析对比了两种神经网络模型的预测性能。试验结果表明:与BP神经网络相比,基于BR算法的NARX神经网络可实现对脱轨系数的精确预测。

The accurate prediction of the train derailment coefficient was studied on the paper.The hidden risks can be caused by the over derailment coefficient on the train operation.As a result,the derailment coefficient is of great basis to evaluate the train operation safty.The problems of the high cost and the low accuracy exist in traditional prediction method of the derailment coefficient.A prediction method based on NARX neural network with BR training algorithm was discussed in the paper.The input was the normalized track irregularity,and the output was the derailment coefficient.BP and NARX neural networks were established to analyse their performaces of prediction.The test result shows that;compared with the BP neural network,the NARX neural network with BR training algorithm obtains higher predction accuracy of the derailment coefficient.