基于分量加权重构和稀疏NMF的轮毂电机轴承复合故障特征提取方法

Feature Extraction Method Based on Component Weighted Reconstruction and Sparse NMF for Bearing Compound Faults of In-wheel Motor

为解决复合故障特征分离提取效果不佳的问题,针对轮毂电机轴承故障,提出一种基于分量加权重构(Component weighted reconstruction,CWR)与稀疏非负矩阵分解(Sparse nonnegative matrix factorization,SNMF)相结合的故障特征提取方法。首先,提出了一种融合指标CIH,从多角度评价振动信号并自适应选取局部均值分解后的乘积函数(Product function,PF)分量,进行CWR实现故障特征的增强表达;其次,分析重构信号时频能量矩阵的奇异值与隐含子空间的关系,基于奇异值方差比定义重构信号的平滑系数,对SNMF算法最优分解维数进行预估计;最后,引入板仓-斋藤(Itakura-Saito,IS)距离和稀疏约束建立SNMF算法,对时频能量矩阵进行分解降维,通过短时傅里叶逆变换获得子空间时域分离信号并进行频谱包络分析,提取故障特征。由仿真及试验结果表明,所提方法可以有效实现复合故障特征的分离提取,具有一定的工程应用价值。

To solve the problem of poor separation and extraction for compound fault features,a fault feature extraction method based on component weighted reconstruction(CWR)and sparse non-negative matrix factorization(SNMF)is proposed for the fault of in-wheel motor bearings.Firstly,a fusion index CIH is proposed to evaluate the information of the vibration signal from multiple perspectives and adaptively select the product function(PF)component by the local mean decomposition,then perform CWR to enhance the expression of fault characteristics.Secondly,the relationship between singular values and implicit subspace of the time-frequency energy matrix of the constructed signal is analysed,and the smoothing coefficient of the constructed signal is defined based on the variance ratio of the singular values,which is used to estimate the optimal decomposition dimension of the SNMF algorithm.Finally,Itakura-Saito(IS)distance and sparse constraint are employed to establish a SNMF algorithm,which is used to decompose the time-frequency energy matrix for dimension reduction.The subspace time-domain separation signals are obtained by the inverse short-time Fourier transform,and the fault features are extracted by spectral envelope analysis.Simulation and experiment results prove that the proposed method has realized effectively the separation and extraction of compound fault features,and has certain value in engineering application.