Wheel flat, as a common defect of railway vehicles, can cause large impact forces on both the vehicle and infrastructure components and thus seriously hinder the vehicle running stability and safety. Considering the complex track irregularities and variable-speed conditions, the vehicle vibration responses often contain strong interference signal components and exhibit time-varying frequency contents, which poses severe challenges to wheel flat detection. In this paper, a vehicle-track coupled model considering wheel flats under variable-speed conditions is employed to calculate and analyze the vehicle vibration accelerations at first. Then, according to the vibration characteristics, a two-level adaptive chirp mode decomposition (ACMD) approach is developed for the wheel flat detection. Specifically, in Level 1 of the approach, the ACMD is integrated with a Gini index-based mode selection and regrouping scheme to separate different fault signal modes under strong interferences. Based on the separated signal modes, the high-resolution ACMD-based time-frequency analysis method is applied to accurately extracting the time-varying fault characteristic frequencies and thus achieving the fault detection in Level 2. Both the dynamics simulation and experiment results indicate that the proposed approach can accurately detect wheel flats of small sizes under strong interferences for a variable-speed railway vehicle.

车轮漏气是铁路车辆的常见缺陷，会在车辆和基础设施部件上产生很大的冲击力，从而严重妨碍车辆的行驶稳定性和安全性。考虑到复杂的轨道不规则性和变速条件，车辆的振动响应通常包含强烈的干扰信号分量，并表现出随时间变化的频率内容，这对轮和平整度检测提出了严峻挑战。在本文中，首先考虑了在变速条件下轮辋的车辆-轨道耦合模型来计算和分析车辆的振动加速度。然后，根据振动特性，开发了一种两级自适应线性调频模式分解（ACMD）方法，用于轮毂检测。具体来说，在该方法的第1级中，ACMD与基于基尼索引的模式选择和重组方案集成在一起，以在强烈干扰下分离不同的故障信号模式。基于分离的信号模式，基于高分辨率ACMD的时频分析方法可准确地提取时变故障特征频率，从而实现2级故障检测。动力学仿真和实验结果均表明：对于变速铁路车辆，该方法可以在强干扰下准确地检测出小尺寸的车轮平面。