The feasibility of automated railway track segment characterisation, as part of a broader track condition based maintenance procedure, is explored via railway–vehicle–based random vibration signals and Statistical Time Series (STS) methods. In particular, three methods within a Multiple Model (MM) framework, which are founded on data-driven stochastic parametric models for the representation of the partial vehicle-rail dynamics, are employed. The random vibration signals are obtained from two sensors, which are mounted on the axlebox and the bogie frame, respectively, of an Athens Metro railway vehicle moving under three different speeds (60, 70 or 80 km/h). The performance assessment of the methods is based on two distinct track characterisation problems. The first corresponds to the characterisation of a specific track segment after 6 months of continuous use, while the second to the comparison between two nominally identical track segments which were installed with a 2 month time interval and are used by different numbers and types of trains. The results indicate deterioration of the track segment in the first characterisation problem and differences between the two nominally identical segments in the second. The superiority of the employed methods over a state-of-the-art method is also demonstrated via proper comparisons.

作为更广泛的基于轨道状况的维护程序的一部分，自动化铁路轨道段表征的可行性通过基于铁路车辆的随机振动信号和统计时间序列 (STS) 方法进行了探索。特别是，采用了多模型 (MM) 框架内的三种方法，这些方法基于数​​据驱动的随机参数模型，用于表示部分车辆-轨道动力学。随机振动信号来自两个传感器，它们分别安装在雅典地铁铁路车辆在三种不同速度（60、70 或 80 公里/小时）下行驶的轴箱和转向架上。这些方法的性能评估基于两个不同的轨道特征问题。第一个对应于连续使用 6 个月后特定轨道段的特征，而第二个对应于两个名义上相同的轨道段之间的比较，它们以 2 个月的时间间隔安装并由不同数量和类型的火车使用。结果表明，第一个表征问题中轨道段的恶化以及第二个表征问题中两个名义上相同的段之间的差异。通过适当的比较也证明了所采用的方法优于最先进的方法。结果表明，第一个表征问题中轨道段的恶化以及第二个表征问题中两个名义上相同的段之间的差异。通过适当的比较也证明了所采用的方法优于最先进的方法。结果表明，第一个表征问题中轨道段的恶化以及第二个表征问题中两个名义上相同的段之间的差异。通过适当的比较也证明了所采用的方法优于最先进的方法。