高速列车具有若干时变激励源，传统的时频分析方法只能对观测的混合振动的总体强度分布、时频域结构加以分析，不能分离出与各振源对应的信号分量从而明晰振源状态与故障特征。盲源分离是一种可行的分析方法，但由于高速列车振动信号具有时变振源数目、时变信号长度、受车速调制的变频非平稳等特征，传统的盲源分离方法不适用。为了提高高速列车非平稳信号的盲源分离效果，基于自适应滤波理论提出全局最优信噪比盲源分离新方法，并对其可分离性的判别依据进行论证。新方法的有效性经仿真计算和实测数据分析得到验证。研究表明：新方法对高速列车时变非平稳信号的盲源分离效果优于传统的基于非线性函数的盲源分离方法和基于高阶累积量的盲源分离方法。

There are a lot of time-varying drive sources in high speed train. The traditional time-frequency analysis methods could analysis the magnitude and the spectrum characteristics of the compound vibration, but couldn’t separate the source signals to know their properties and failure distribution. The vibration signal of the high speed train is nonstationary random signal modulated by velocity, and the number of sources as well as the length of signals are time-varying, the traditional blind source separation methods couldn’t deal with the difficult problem. A new blind source separation algorithm called globally optimal signal-to-noise ratio algorithm based on the adaptive filtering theory is proposed. The separability of the proposed method is deduced. The simulation and test analysis results show that the proposed method is effective, and obtains more satisfactory separation quality than the classical blind source separation methods based on nonlinearity function and high-order cumulant in nonstationary signal analysis of high speed train.