基于改进变分模态分解及循环相关熵谱的轴承故障诊断

针对强背景噪声下非高斯脉冲噪声和高斯噪声对滚动轴承故障诊断产生严重干扰的问题，提出了一种基于改进变分模态分解(variational mode decomposition, VMD)并与循环相关熵谱(cyclic correntropy spectrum, CCES)相结合的故障诊断方法。首先，针对VMD传统重构指标易受噪声影响的问题，引入相关熵峭度(correlation entropy kurtosis index, CEK)指标对VMD分解后的模态分量进行选择与重构，去除高斯噪声；然后针对重构后信号仍存在的脉冲噪声影响问题，对重构信号进行CCES投影融合去除非高斯脉冲噪声干扰并增强特征；最后对融合结果进行分析与故障诊断。经仿真测试与实验表明，所提出的方法可以在高斯噪声和非高斯脉冲噪声背景下有效提取滚动轴承故障特征频率并实现故障诊断。

Rolling bearing fault diagnosis based on CEK-VMD and CCES

To solve the problem that non-Gaussian impulse noise and Gaussian noise seriously interfered with rolling bearing fault diagnosis under strong background noise, a fault diagnosis method was proposed based on improved variational mode decomposition (VMD) combined with cyclic correlation entropy spectrum (CCES). Firstly, aiming at the problem that the traditional VMD reconstruction indexes are susceptible to noise, the correlation entropy kurtosis (CEK) index is introduced to select and reconstruct the modal components after VMD decomposition to remove gaussian noise. Then, aiming at the influence of pulse noise on the reconstructed signal, CCES projection fusion was performed to remove the interference of unless Gaussian pulse noise and enhance the features of the reconstructed signal. Finally, the fusion results are analyzed and the fault diagnosis is made. Simulation and experiment show that the proposed method can effectively extract the characteristic frequency of rolling bearing fault and realize fault diagnosis under gaussian noise and non-Gaussian pulse noise background.