基于全寿命周期退化信息的高速列车轮对轴承剩余寿命预测方法研究

Degradation process for wheelset bearings of high-speed trains exhibits many complex characteristics, e.g. strong noise interference and strong coupling modulation, which bring a big challenge in establishing a link between condition monitoring information and remaining useful life. This project focuses on the objective of wheelset bearings. By validating with accelerated degradation tests and using advances in dynamic signal processing, optimization theory and artificial intelligence, this project aims to study the following key issues. (1) Study an integrated optimization framework for local mean decomposition algorithm; study dynamic signal demodulation methods with the optimized local mean decomposition at the core; and study accurate location methods for sensitive bands in frequency domain and recognition methods for sensitive production functions. (2) Study nonlinear evaluation metrics and their adaptive analytical optimization methods for degradation characterization indicators; study multi-metric fusion strategy and selection methods for those indicators; and achieve an accurate evaluation and selection for the life-cycle degradation indicators. (3) Study remaining useful life prediction strategies and modeling methods based on health state assessment; build a complex mapping between the degradation indicators and remaining useful life; and achieve remaining useful life prediction of wheelset bearings with high precision. The project adheres to well accepted scientific research guidelines in the field of remaining useful life prediction of sophisticated engineering systems and complex infrastructures. The expected results from this project could provide theory support to safety monitoring and health management for high-speed trains, and thus the project has important scientific significance and engineering application value.

高速列车轮对轴承振动信号在退化过程中呈现出强噪声干扰与强耦合调制等复杂特点，建立其运行状态信息与剩余寿命之间的关联存在巨大挑战。本课题以轮对轴承为研究对象，采用动态信号处理、优化理论和人工智能等分析手段，结合加速退化试验验证，研究以下内容：（1）研究基于一体优化框架的局部均值分解理论及以其为核心的动态信号解调分析方法，研究退化过程不同阶段中敏感频带和敏感乘积函数分量的准确定位与识别方法；（2）研究退化状态表征的非线性评价指标及其自适应解析优化方法，研究多种评价准则的融合策略及表征优选方法；（3）研究基于健康状态评估的剩余寿命预测理论与建模方法，建立退化状态表征与剩余寿命之间的复杂映射关系，实现高精度的轮对轴承剩余寿命估计。本课题源于重大技术装备与重大设施寿命预测共性基础研究中的重要科学问题，将为高速列车的安全监测与健康管理提供理论支撑，具有重要的研究价值和工程应用前景。

轮对轴承是高速列车系统中的关键部件，利用振动监测进行剩余寿命预测和故障诊断是保障列车安全可靠运行和降低全寿命周期运维成本的有效手段。然而，强噪声干扰与强耦合调制导致了轮对轴承在故障机理、失效规律以及振动响应方面具有复杂性，这对建立振动响应与故障信息、剩余寿命之间的可靠关联关系带来了挑战。本项目以轮对轴承的内圈、外圈、滚动体和保持架等典型零部件为研究对象，采用了动力学、信号处理、最优化和人工智能等理论和方法，结合了仿真分析、故障试验和加速退化试验，主要开展了以下研究工作：1）在故障机理方面，基于势能法，以时变啮合刚度为切入点，建立了点蚀故障与振动响应之间的动力学模型，为旋转机械故障诊断提供了正向的逻辑关系依据；2）在振动信号处理方面，围绕解调分析和频带选择，发展了局部均值分解和希尔伯特黄变换等解调分析方法，提出了准确度谱图等频带选择方法，这些研究为解决强噪声干扰与强耦合调制提供了的技术手段；3）在故障诊断模型方面，提出了安全域、增量学习和一维深度残差卷积神经网络等异常检测和故障诊断模型，建立了振动数据驱动的动态响应与典型零部件故障模式之间的关联关系模型，并研究了其中的参数优化问题；4）在剩余寿命预测方面，提出了基于健康状态评估的剩余寿命预测模型，建立了振动响应与健康状态、振动响应与剩余寿命之间的关联关系; 5）在轴承试验方面，完成了滚动轴承加速退化试验和轮对轴承故障试验，积累了丰富的故障试验数据。综上所述，本项目研究了高速列车寿命预测共性基础研究中的重要科学问题，为解决其健康监测和评估提供了新思路和新方法，具有重要的科学意义和工程价值。