轴承复合故障严重影响设备的可靠性和使用寿命。为研究滚动轴承复合故障特性变化规律, 本文基于ANSYS workbench, 建立了角接触球轴承复合故障动力学仿真模型, 研究了不同转速下复合故障轴承的动态响应规律; 提出一种基于优化变分模态分解(VMD)的复合轴承故障特征识别方法, 能有效消除噪声对轴承故障特征提取产生的干扰。采用改进的北方苍鹰寻优算法对VMD的参数进行自适应寻参, 在确定最佳参数后对振动信号进行变分模态分解, 基于最大峭度原则对本征模态函数(IMF)进行信号重构后做频谱分析。研究结果表明：复合缺陷对滚动体和外圈的接触应力分布影响显著, 等效应力主要集中在凹坑缺陷的边缘; 滚动体通过外圈缺陷时会导致轴承内部局部的应力集中和释放, 应力会下降。改进的北方苍鹰寻优算法优化过的VMD可以有效提取轴承中的复合故障特征频率成分, 频率峰值清晰地对应了理论计算的故障频率, 准确率高于其他广泛使用的特征提取方法。

Objective: Rolling bearings are a key component in the operation of numerical control machine tool, directly affecting the reliability and safety of the entire equipment. Under large load impacts, defects such as pits, scratches, and spalling can form on the bearing raceway and rolling element surfaces. Compared with a single defect, a composite defect results in a larger impact amplitude in the time-domain waveform and more complex main frequency components in the spectrum. Therefore, it is essential to study the compound failure characteristics of rolling bearings to further clarify their failure mechanisms and dynamic behaviors. Methods: The present study uses ANSYS Workbench to establish a dynamic simulation model of an angular contact ball bearing with composite faults. The simulation results of the bearing composite fault characteristics were compared with the theoretical calculations. The comparison shows an error of less than 3%, thereby verifying the accuracy of the model. The dynamic response behavior of the composite fault bearing at different rotational speeds was then analyzed. Additionally, a composite bearing fault feature identification method based on optimized variational modal decomposition was proposed. The proposed method can be employed to effectively eliminate noise interference in the extraction of bearing fault features. The original search method of the northern eagle optimization (NGO) algorithm exhibits a linear decreasing trend, which hinders the effective balance between global search and local development capabilities. To address this issue, the positive cosine strategy and position optimization search algorithm are employed to replace the position update method in the NGO algorithm during the search phase, thereby improving the original search method. The improved NGO algorithm is then applied to the adaptive parameter search for VMD. Once the optimal parameters are determined, the simulated vibration signals undergo variational mode decomposition, followed by signal reconstruction of the eigenmode function based on the maximum kurtosis principle, enabling spectral analysis. Furthermore, the rolling bearing accelerated life test dataset from Xi'an Jiaotong University was used for bearing fault feature extraction to verify the effectiveness of the SPNGO-VMD composite fault feature extraction method proposed in this study. Finally, a quantitative comparison is presented to assess the differences in feature extraction performance between SPNGO-VMD and the other methods. The analysis is based on search efficiency, global convergence, and fault feature extraction accuracy, and different methods are evaluated using quantitative indexes. Results: The research findings indicate that before and after the defective rolling element passes through the outer ring, the equivalent stress is mainly concentrated at the edge of the outer ring defect. The passage of a rolling element through a defect causes localized stress concentration and release within the bearing, decreasing stress. The VMD optimized by the improved NGO algorithm effectively extracts the characteristic frequency components of the composite faults in the bearings, with the frequency peaks corresponding to the theoretically calculated fault frequencies. The improved algorithm demonstrated better search efficiency, global convergence, and fault feature extraction accuracy than other feature extraction models. The algorithm can consistently find the global optimal solution in fewer iterations, avoiding local optima and extracting fault frequency features with greater accuracy. Conclusions: The simulation model provides a foundation for the experimental design of defective composite bearings. The study results not only aid in identifying and diagnosing potential faults but also offer a theoretical basis for designing more reliable bearing systems.