ABSTRACT Monitoring the condition of suspension systems is significant to ensure the safe operation of modern railway vehicles. For this purpose, an online modal identification scheme, denoted as Correlation Subset based Stochastic Subspace Identification (CoS-SSI) is proposed in this paper to monitor the suspension conditions. Because of the widespread of the dynamic contact status between wheel and track, especially under faulty suspension cases, the vibration responses measured online exhibit high nonstationarity and nonlinearity. To take into account these characteristics of signals, the input correlation signals for SSI are clustered into several successive subsets according to their magnitudes, on which SSI is implemented one by one. In this way it yields a magnitude adaptive SSI for more reliable and accurate identification. Experimental studies were conducted on a 1/5th scaled roller rig system to verify the effectiveness of the proposed method for suspension monitoring. The experimental results show that the CoS-SSI outperform the conventional SSI in that it produces more reliable and realistic identification for the nonlinear system. Furthermore, the effectiveness of the CoS-SSI was verified experimentally with two faulty suspension faults induced into the system.

中文翻译：

---

基于相关信号子集的随机子空间识别用于铁路车辆悬架系统在线识别

摘要 监测悬架系统的状态对于保证现代铁路车辆的安全运行具有重要意义。为此，本文提出了一种在线模态识别方案，称为基于相关子集的随机子空间识别（CoS-SSI），以监测悬架条件。由于轮轨之间的动态接触状态普遍存在，特别是在故障悬挂情况下，在线测量的振动响应表现出高度的非平稳性和非线性。考虑到信号的这些特性，SSI 的输入相关信号根据它们的大小被聚类成几个连续的子集，在这些子集上一一实现 SSI。通过这种方式，它产生幅度自适应 SSI，以实现更可靠和准确的识别。在 1/5 比例的滚轮钻机系统上进行了实验研究，以验证所提出的悬挂监测方法的有效性。实验结果表明，CoS-SSI 优于传统的 SSI，因为它对非线性系统产生了更可靠和现实的识别。此外，CoS-SSI 的有效性通过系统中引入的两个故障悬挂故障进行了实验验证。