High-speed rails with determined length and load run for long periods at almost uniform speeds along fixed routes, constituting a new stable and repeatable artificial seismic source. Studies have demonstrated the wide bands and discrete spectra of high-speed rail seismic signals. Exploring the abundant information contained in massive high-speed rail seismic signals has great application value in the safety monitoring of high-speed rail operation and subgrade. However, given the complex environment around the rail network system, field data contain not only high-speed rail seismic waves but also ambient noise and the noise generated by various human activities. The foundation and key to effectively using high-speed rail seismic signals is to extract them from field data. In this paper, we propose an adaptive variational mode decomposition (VMD)-based separation algorithm for high-speed rail seismic signals. The optimization algorithm is introduced to VMD, and sample entropy and energy difference are used to construct the fitness function for the optimal adjustment of the mode number and penalty factor. Furthermore, time–frequency analysis is performed on the extracted high-speed rail signals and field data using the synchrosqueezed wavelet transform (SSWT). After verifying the processing of simulated signals, the proposed method is applied to field data. Results show that the algorithm can effectively extract high-speed rail seismic signals and eliminate other ambient noises, providing a basis for the imaging and inversion of high-speed rail seismic waves.

确定长度和载荷的高铁沿固定路线以几乎匀速的速度长时间运行，构成了一种新的稳定且可重复的人工震源。研究证明了高铁地震信号的宽带和离散频谱。挖掘海量高铁地震信号中蕴含的丰富信息，在高铁运营及路基安全监测中具有巨大的应用价值。然而，由于轨道交通网络系统周围环境复杂，现场数据中不仅包含高铁地震波，还包含环境噪声以及各种人类活动产生的噪声。有效利用高铁地震信号的基础和关键是从现场数据中提取高铁地震信号。在本文中，我们提出了一种基于自适应变分模态分解（VMD）的高铁地震信号分离算法。将优化算法引入VMD，利用样本熵和能量差构建适应度函数，对模式数和惩罚因子进行优化调整。此外，使用同步压缩小波变换（SSWT）对提取的高铁信号和现场数据进行时频分析。在验证了模拟信号的处理效果后，将所提出的方法应用于现场数据。结果表明，该算法能够有效提取高铁地震信号并消除其他环境噪声，为高铁地震波成像和反演提供基础。