基于Copula函数的高速列车信号联合特征提取

Extracting joint feature of signals of high-speed train using Copula

针对高速列车在车体不同位置的通道测得的振动信号,提出一种基于Copula函数的通道间信号联合特征提取方法。该方法使用泛化高斯分布对各通道信号的边缘分布进行拟合,并使用Gaussian Copula函数构建信号间的联合分布函数。提取边缘分布参数以及联合分布函数的参数作为特征。通过对某型高速列车转向架正常、抗蛇行减振器失效、空气弹簧失效和横向减振器失效4种典型工况的振动信号进行分析和特征提取,并采用支持向量机进行识别,平均识别率超过97%,表明该特征提取方法的有效性。

Aiming at mechanical vibration signals of different channels of high-speed train bogie, an approach using Copula function to extract joint feature of signals among different channels was proposed in the paper. The marginal distribution functions of the signals were fitted by generalized Gaussian distribution. The joint distribution function of the marginal distribution functions was computed by Gaussian Copula function. The coefficients of marginal distribution functions and joint density function were extracted as the features. Vibration signals of a high-speed train bogie were obtained under four typical working conditions including normal condition, yaw damper fault, air spring fault, and lateral damper fault. The features were extracted using the proposed approach, and the working conditions were classified with the support vector machine. The average recognition rate was above 97%, which verified the effectiveness of the proposed feature extraction method.