高速列车监测数据处理及故障诊断

车体性能好坏直接影响列车的行车安全,文章利用安装在车体上的传感器所采集到的振动信号,选取合适的信号特征提取方法进行评估,达到列车故障早期预警的目的。试验数据表明,车体的振动信号具有非线性、非平稳的特点,先对振动信号提取小波包能量矩特征进行时频域分析,发现该特征提取方法可以直观地反映车辆横向和垂向振动情况。引入基于局部分析的拉普拉斯特征映射算法(LE),对故障工况的小波包能量矩熵特征所构造的高维特征向量空间进行降维,发现能够从垂向加速度信号识别出空气弹簧失气工况,从横向加速度信号识别出抗蛇行减振器故障和横向减振器故障。这与车辆动力学分析结果一致,同时也证实了流形学习方法对列车性能评估具有一定的作用。

Monitoring Data Processing and Fault Diagnosis for High-speed Train

The performance of the carbody affects its running safety directly.This paper uses the vibration signals collected by the sensors that installed on the carbody to select the appropriate signal feature extraction method for evaluation,so as to achieve the purpose of early warning of train faults.The test data shows that the vibration signal of carbody has the characteristics of nonlinear and non-stationary.The time-frequency domain analysis of the wavelet packet energy moment feature extracted from the vibration signal shows that the feature extraction method can intuitively reflect the lateral and vertical vibration of the vehicle.The Laplacian Eigenmap(LE)based on local analysis is introduced to reduce the dimension of the high-dimensional feature vector space constructed by the wavelet packet energy moment entropy feature of the fault condition,and it is found that the air spring air loss for can be identified from the vertical acceleration signal,the failure of the anti-hunting damper and the failure of the lateral damper are identified from the lateral acceleration signal.This is consistent with the result of vehicle dynamics analysis,and it also confirms that the manifold learning method can be used on train performance evaluation.