School of Mechanical Engineering and Automation, Northeastern University, Shenyang 110819, China.
Key Laboratory of Vibration and Control of Aero-Propulsion System Ministry of Education, Northeastern University, Shenyang 110819, China.
Sensors (Basel). 2023 Apr 17;23(8):4044. doi: 10.3390/s23084044.
The quantitative diagnosis of rolling bearings is essential to automating maintenance decisions. Over recent years, Lempel-Ziv complexity (LZC) has been widely used for the quantitative assessment of mechanical failures as one of the most valuable indicators for detecting dynamic changes in nonlinear signals. However, LZC focuses on the binary conversion of 0-1 code, which can easily lose some effective information about the time series and cannot fully mine the fault characteristics. Additionally, the immunity of LZC to noise cannot be insured, and it is difficult to quantitatively characterize the fault signal under strong background noise. To overcome these limitations, a quantitative bearing fault diagnosis method based on the optimized Variational Modal Decomposition Lempel-Ziv complexity (VMD-LZC) was developed to fully extract the vibration characteristics and to quantitatively characterize the bearing faults under variable operating conditions. First, to compensate for the deficiency that the main parameters of the variational modal decomposition (VMD) have to be selected by human experience, a genetic algorithm (GA) is used to optimize the parameters of the VMD and adaptively determine the optimal parameters [, ] of the bearing fault signal. Furthermore, the IMF components that contain the maximum fault information are selected for signal reconstruction based on the Kurtosis theory. The Lempel-Ziv index of the reconstructed signal is calculated and then weighted and summed to obtain the Lempel-Ziv composite index. The experimental results show that the proposed method is of high application value for the quantitative assessment and classification of bearing faults in turbine rolling bearings under various operating conditions such as mild and severe crack faults and variable loads.
滚动轴承的定量诊断对于实现维护决策的自动化至关重要。近年来，Lempel-Ziv 复杂度（LZC）已被广泛用于机械故障的定量评估，作为检测非线性信号动态变化的最有价值指标之一。然而，LZC 侧重于 0-1 代码的二进制转换，这容易丢失时间序列的一些有效信息，并且不能充分挖掘故障特征。此外，LZC 对噪声的免疫力无法保证，并且很难在强背景噪声下定量表征故障信号。为了克服这些局限性，开发了一种基于优化变分模态分解 Lempel-Ziv 复杂度（VMD-LZC）的定量轴承故障诊断方法，以充分提取振动特征，并在变工况下定量表征轴承故障。首先，为了弥补变分模态分解（VMD）的主要参数必须由人为经验选择的缺陷，使用遗传算法（GA）优化 VMD 的参数，并自适应地确定轴承故障信号的最佳参数[，]。此外，基于峰度理论，选择包含最大故障信息的 IMF 分量进行信号重构。计算重构信号的 Lempel-Ziv 指数，然后对其进行加权求和，以获得 Lempel-Ziv 综合指数。实验结果表明，该方法对于在各种工况下（如轻度和重度裂纹故障以及变载荷）对涡轮滚动轴承的轴承故障进行定量评估和分类具有很高的应用价值。
