中文摘要：

为解决列车高速运行时,出现的蛇行失稳故障难以被准确识别的问题,提出一种基于集合经验模态分解 （ensemble empirical mode decomposition, EEMD）摘特征和最小二乘法支持向量机 （ least squares support vector machine, LSSVM）的高速列车蛇行异常运动状态的诊断方法.首先通过EEMD对高速列车蛇行故障振动信号进行分 解,再提取固有模态函数（intrinsic mode function, IMF）分量的样本熵特征、香农熵特征和能量熵特征,最后分别用 LSSVM进行训练和识别.试验结果表明：高速列车在330耀350km/h的运行速度下, EEMD熵特征一LSSVM方法能准确 识别高速列车蛇行失稳状态,并且LSSVM的输入特征为能量熵特征时,识别效果优于样本熵特征和香农熵特征,识 别率达到95%.

英文摘要：

To address the issue of hunting instability for high-speed train, a new method which combines ensemble empirical mode decomposition （EEMD）, entropy features and least squares support vector machine （LSSVM） was presented in this paper to diagnose hunting abnormal motion state of high-speed train. Firstly, the vibration signal was decomposed by EEMD. Then, sample entropy features, Shannon entropy features and energy entropy features of IMFs were extracted. Lastly, the features were trained and recognized by LSSVM respectively. The test results show that the method of EEMD entropy features -LSSVM can accurately recognize the instability state of hunting motion when the speed of train is up to 330-350km/ h. Furthermore, what can be learned from the experiment is that as an input feature, the energy entropy recognition effect will be superior to Shannon entropy and sample entropy, up to 95%.