高速列车转向架蛇行失稳的MEEMD-LSSVM预测模型

摘要

To forecast hunting instability state of high-speed train bogie,a new forecasting model which com-bines modified ensemble empirical mode decomposition (MEEMD)and least squares support vector machine (LSSVM)was presented in this paper,it focus on the normal,transition and hunting instability states of bogie vibration signal.Firstly,the vibration signal will be decomposed by MEEMD.Then,the Hilbert transforma-tion (HT)will be used to analyze the time-frequency-energy features.Meanwhile,the energy feature of intrin-sic mode functions (IMFs)will be extracted to be used to train by LSSVM.Finally,hunting instability state will be forecasted by recognizing the transition state.The results show that forecast accuracy up to 93.33％, and the accuracy and calculation time are superior to ensemble empirical mode decomposition-support vector machine (EEMD-SVM)when the train at 350 km/h.The validity and rapidity of the forecasting model were proved.%为预测列车转向架蛇行失稳异常运动状态,提出一种改进的集总平均经验模态分解-最小二乘法支持向量机(MEEMD-LSSVM)的预测模型.以转向架正常、过渡、蛇行失稳3种状态下振动信号为研究对象,通过MEEMD对信号进行分解,再用Hilbert变换(HT)分析其时频能量特征,最后采用固有模态函数(IMF)的能量特征作为LSSVM的输入,通过识别过渡状态,预测列车蛇行失稳.试验表明,列车处于330～350 km/h之间时,预测准确率为93.33％,并且MEEMD-LSSVM方法准确率和计算耗时优于EEMD-SVM方法,证明该预测模型的有效性和快速性.