Abstract Diagnosis and prognostics of rolling element bearings have been widely studied in recent years, but very few researches were dealing with high-speed train wheel set bearings (HSTWSB). Most prognostics and health management (PHM) models are generally based on obtaining the remaining useful life (RUL) of concerned bearings. Since it is difficult to quantify and to monitor bearing status from vibration signal and there is no clear definition what is the end of bearing service life, determine RUL is not realistic in industrial practice. In order to achieve reliable fault diagnosis and prognosis for HSTWSB, it is of great importance and necessity to conduct a thorough research under realistic or close to reality operation conditions. Therefore, in this paper two types of techniques, i.e. vibration and acoustic emission, have been particularly studied. Different from many previous PHM studies which seek seeking bearing’s RUL by establishing physics model or artificial neural network model, a new hybrid model based on extendable useful life (EUL) under continuous monitoring and bearing status classification is proposed. Statistical properties of typical time domain features extracted from vibration and acoustic emission are studied. Correlations of these parameters with bearing status are reviewed and feasible parameters are evaluated for bearing status quantification. By driving an electric multiple unit (EMU) speed up to 350 km/h, a test device close to real running environment was introduced. A batch of bearings with different level of nature defects instead of artifacts were particularly selected as database samples of this paper. Test procedure was designed to allow fault diagnosis to be verified under low, medium and high speeds and the corresponding database and knowledgebase of bearing status assessment are established. Defect geometries were quantified with 3D laser scanning technology so that it provides intuitive references for evaluating effectiveness of signal processing approaches with respective to bearing damage status. Instead of calculating how much RUL left by physics model or neural network model, the proposed approach determines if the useful life can be extended from one grade level to another or to next overhaul under continuous monitoring. The proposed model establishes an initial database and knowledgebase for HSTWSB monitoring. This model can be dynamically enhanced with involvement of AI technology and accumulation of tested bearing database in the future.

中文翻译：

---

高速列车轮对轴承故障诊断与预测：一种基于可延长使用寿命的新预测模型

摘要 滚动轴承的诊断和预测近年来得到了广泛的研究，但很少涉及高速列车轮对轴承（HSTWSB）的研究。大多数预测和健康管理 (PHM) 模型通常基于获得相关轴承的剩余使用寿命 (RUL)。由于很难从振动信号中量化和监控轴承状态，并且没有明确定义轴承使用寿命的结束时间，因此确定 RUL 在工业实践中是不现实的。为了实现HSTWSB可靠的故障诊断和预测，在现实或接近现实的运行条件下进行深入研究是非常重要和必要的。因此，在本文中，两种技术，即振动和声发射，专门研究过。与以往许多PHM研究通过建立物理模型或人工神经网络模型来寻求轴承RUL不同，提出了一种基于连续监测和轴承状态分类下的可延长使用寿命（EUL）的新混合模型。研究了从振动和声发射中提取的典型时域特征的统计特性。审查这些参数与轴承状态的相关性，并评估轴承状态量化的可行参数。通过驱动电动联动车组（EMU）时速达到350km/h，介绍了一种接近真实运行环境的测试装置。特别选择了一批具有不同程度的自然缺陷而不是伪影的轴承作为本文的数据库样本。设计测试程序以验证低、中、高速下的故障诊断，并建立相应的轴承状态评估数据库和知识库。缺陷几何形状通过 3D 激光扫描技术进行量化，从而为评估信号处理方法的有效性与轴承损坏状态提供了直观的参考。所提出的方法不是计算物理模型或神经网络模型剩余多少 RUL，而是确定使用寿命是否可以从一个等级延长到另一个等级或在持续监控下延长到下一次大修。所提出的模型为 HSTWSB 监测建立了一个初始数据库和知识库。