College of Mechanical Engineering, North University of China, Taiyuan 030051, China.
Sensors (Basel). 2018 Nov 6;18(11):3804. doi: 10.3390/s18113804.
Aiming at the problem that the composite fault signal of the gearbox is weak and the fault characteristics are difficult to extract under strong noise environment, an improved singular spectrum decomposition (ISSD) method is proposed to extract the composite fault characteristics of the gearbox. Singular spectrum decomposition (SSD) has been proved to have higher decomposition accuracy and can better suppress modal mixing and pseudo component. However, noise has a great influence on it, and it is difficult to extract weak impact components. In order to improve the limitations of SSD, we chose the minimum entropy deconvolution adjustment (MEDA) as the pre-filter of the SSD to preprocess the signal. The main function of the minimum entropy deconvolution adjustment is to reduce noise and enhance the impact component, which can make up for the limitations of SSD. However, the ability of MEDA to reduce noise and enhance the impact signal is greatly affected by its parameter, the filter length. Therefore, to improve the shortcomings of MEDA, a parameter adaptive method based on Cuckoo Search (CS) is proposed. First, construct the objective function as the adaptive function of CS to optimize the MEDA algorithm. Then, the pre-processed signal is decomposed into singular spectral components (SSC) by SSD, and the meaningful components are selected by Correlation coefficient. For the existing modal mixing phenomenon, the SSC component is reconstructed to eliminate the misjudgment of the result. Then, the frequency spectrum analysis is performed to obtain the frequency information for fault diagnosis. Finally, the effectiveness and superiority of ISSD are validated by simulation signals and applying to compound faults of a Gear box test rig.
针对齿轮箱复合故障信号在强噪声环境下较弱且故障特征难以提取的问题，提出了一种改进的奇异谱分解（ISSD）方法来提取齿轮箱的复合故障特征。奇异谱分解（SSD）已被证明具有更高的分解精度，可以更好地抑制模态混合和伪分量。然而，噪声对其影响很大，难以提取弱冲击分量。为了提高 SSD 的局限性，我们选择最小熵反卷积调整（MEDA）作为 SSD 的预滤波器来预处理信号。最小熵反卷积调整的主要功能是降低噪声并增强冲击分量，从而弥补 SSD 的局限性。然而，MEDA 降低噪声和增强冲击信号的能力受到其参数（滤波器长度）的极大影响。因此，为了改进 MEDA 的缺点，提出了一种基于布谷鸟搜索（CS）的参数自适应方法。首先，将目标函数构建为 CS 的自适应函数，以优化 MEDA 算法。然后，通过 SSD 将预处理后的信号分解为奇异谱分量（SSC），并通过相关系数选择有意义的分量。对于现有的模态混合现象，对 SSC 分量进行重构以消除结果的误判。然后进行频谱分析，以获取用于故障诊断的频率信息。最后，通过仿真信号验证了 ISSD 的有效性和优越性，并将其应用于齿轮箱试验台的复合故障。
