MPA-MMD方法在变转速齿轮箱振动信号特征提取中的应用

Application of MPA-MMD method for gearbox vibration signal feature extraction under variable rotating speed condition

变转速工况下齿轮箱振动信号的分量通常具有时频重叠和跨频带特征,分量直接分离非常困难。对此,引入一种新的多通道多分量分解(multichannel multipoint distribution, MMD)方法,并利用新型群体智能优化算法——海洋捕食者算法(marine predators algorithm, MPA)求解MMD方法中的关键优化问题,进而提出了基于MPA优化的MMD(MPA-MMD)方法。MPA-MMD方法将每一个分量表示为一组加权特征向量的线性组合,因为不依赖时间尺度特征,所以特别适合分解具有时频重叠或跨频带特征的复杂信号。通过设置具有分量重叠、跨频带和波动性特征的加噪仿真信号,将MPA-MMD与基于其他优化算法的MMD,以及多通道变分模态分解进行了对比分析,结果表明MPA-MMD在分解效果、收敛性和抑噪性方面的优势;在此基础上,针对变转速工况下齿轮箱振动信号具有分量重叠和跨频带的复杂特征,将MPA-MMD应用于变转速工况下齿轮箱振动信号的特征提取,具有针对性的试验信号分析结果表明,MPA-MMD可直接准确地获得受转速影响的故障分量。

The components of gearbox vibration signals under variable speed conditions usually have time-frequency overlap and cross-band features,which make it very difficult to separate the components directly.To this end,a new multichannel multicomponent decomposition(MMD)method was introduced,which utilizes a new group intelligent optimization algorithm-marine predators algorithm(MPA)to solve the key optimization problem in the MMD method.An MPA-MMD method was proposed based on the MPA optimization.The MPA-MMD method represents each component as a linear combination of a set of weighted feature vectors,and is particularly suitable for decomposing complex signals with time-frequency overlap or cross-band features because it does not rely on time-scale features.By setting up noisy simulation signals with component overlap,cross-band and fluctuation features,MPA-MMD is compared with MMD based on other optimization algorithms and multi-channel variational mode decomposition(MVMD).Results show the advantages of MPA-MMD in terms of decomposition effect,convergence,and noise suppression.On this basis,MPA-MMD is applied to the feature extraction of gearbox vibration signals under variable speed conditions,which have complex features of component overlap and cross-band.The targeted experimental signal analysis results show that MPA-MMD can directly and accurately obtain the fault components affected by speed.