脊线优化与多尺度稀疏融合的转速大波动工况轴承故障诊断研究

The roller element bearing is one of the most failure-prone mechanical elements in the rotating machinery, particularly when operating in the speed condition with large oscillating. The resulting catastrophic consequence is even more severe than that in stationary operating condition. Therefore, it is of great significance to monitor the health condition of bearing under the speed situation with large fluctuation. Although numerous approaches have been developed for bearing fault detection, its health condition monitoring under time-variant speed operation still encounters the challenges summarized as follows: (1) the shaft rotating speed is either dependent on tachometers or has poor robustness when extracted from mechanical signals, and (2) the faulted signature extraction is confronted with the difficulties of the localization of fault-related frequency band and the decoupling of interfering components. To address such problems, we propose a novel method via the joint application of ridge optimization and multi-scale sparse fusion for bearing fault diagnosis under the large fluctuation speed condition. More specifically, the ridge path optimization framework is, firstly, constructed based on the time-frequency enhancement technique to accurately predict the rotational speed. Subsequently, the energy operator which is featured by its multi-resolution capability is used to generate the multi-scale signals to enrich the fault-induced information. Then, the sparse filtering algorithm is explored to eliminate the random components and coupling components buried in the resampled multi-scale signals. The multi-scale sparse fusion (MSSF) method is finally established by incorporating the fusion technique to extract the weak fault-related impulse, from which the bearing health condition under nonstationary speed with large fluctuations can then be monitored.

转速大波动工况下，轴承等机械设备核心部件因恶劣工况极易出现故障，且故障对设备的危害相比定速情形更加严重。因此，开展转速大波动工况下轴承故障诊断具有重要意义。然而，现有变转速轴承故障诊断方法普遍存在如下两类问题：①转速信息的获取依赖于转速传感器或基于设备动态信号的转速识别方法鲁棒性差；②为提取故障特征需要准确定位故障信息频带和难以去除耦合干扰分量。为解决上述问题，提出脊线优化与多尺度稀疏融合的变转速轴承故障诊断方法。首先，基于增强时频表示建立复杂形态脊线优化识别框架，准确地提取设备转速；其次，利用能量算子的多分辨特性构建多尺度信号，丰富、扩展信号中的故障信息；然后，研究稀疏滤波算法对设备动态信号的适应性，利用该理论去除基于转速重采样多尺度信号中的随机、耦合干扰成分；最后，引入融合技术建立多尺度稀疏融合方法，提取多尺度信号中的故障冲击特征，实现转速大波动工况下轴承等复杂设备核心部件的健康监测。

轴承广泛应用于轨道车辆走行系统、航空发动机等高端机械装备。转速大波动工况下轴承等设备核心部件因承受恶劣工况，而容易出现故障，高效准确地监测轴承健康状态对于保障高端机械装备服役安全十分关键。本项目研究主要是为克服转速大波动工况下轴承故障诊断的难题：1）在无转速计下实现转速信息及故障相关特征脊线准确追踪识别；2）强背景噪声干扰下轴承微弱故障信息增强与提取，以去除耦合干扰分量。.为了实现脊线准确追踪识别，从时频表示增强与脊线融合以及时频分解角度开展了相关研究。建立了基于时频表示增强与融合的脊线识别方法；基于脊线优化趋势时频分解的脊线识别方法；基于脊线同步特性的时频分解的脊线快速识别方法。在轴承故障脊线特征信息识别方面，所提方法在抗噪性以及对先验知识依赖性相比于现有脊线识别方法都得到了很大程度的改善。特别是利用脊线同步特性显著提升了时频分解算法的效率，在实时性要求高的场景具有潜在的应用价值。.针对强背景噪声干扰下轴承微弱故障信息增强与提取问题，分别提出了多尺度能量算子融合轴承故障信息增强方法、广义稀疏解卷积轴承故障信息增强方法和调制基约束下微弱故障信息增强与提取方法等，实现了不同转速工况下轴承微弱故障信息的准确提取与噪声抑制。特别是发现了调制基约束下变分模式信号分解方法具有单调收敛性，在一定假设条件下初步给出了证明，对于揭示频谱结构具有潜在的应用价值，例如复杂多源激扰下轮对轴承状态监测问题。.基于本项目相关研究成果，标注基金号的学术论文共发表24篇（一作/通讯论文16篇），其中SCI期刊论文16篇，授权发明专利5项，获得省部级科技进步二等奖2项（分别排名第3和7），ESI高被引论文1篇。第一作者发表于MSSP和JSV论文都被引用超过60次，相关研究成果得到了清华大学、杰青褚福磊教授，北京航空航天大学、杰青/长江学者林京教授，英国哈德斯菲尔德大学谷丰收教授，香港城市大学Kwok Leung Tsui教授等著名专家学者引用评价。