基于低秩-稀疏分解的滚动轴承故障特征提取

Feature extraction for fault signals from a rolling bearing using low rank and sparse decomposition

针对旋转机械早期故障特征微弱且易受背景噪声影响而难以提取的问题,提出一种基于低秩-稀疏分解的轴承信号瞬态特征提取方法。研究了周期性瞬态信号的稀疏时频表示,建立了低秩-稀疏模型并从背景噪声中提取瞬态冲击信号。首先,通过高阶同步压缩变换(high-order synchrosqueezing transform, FSSTH)将测量信号变换到一个新的稀疏子空间;然后,使用鲁棒主成分分析算法(robust principal component analysis, RPCA)将稀疏时频矩阵分解为低秩部分和稀疏部分;最后,对低秩矩阵施加逆高阶同步压缩变换恢复得到瞬态冲击信号,并通过包络谱分析实现故障诊断。该方法由数据驱动实现,不需要任何先验信息。仿真信号和实际信号分析结果表明,所提方法可有效增强振动信号中故障引起的周期性瞬态冲击特征,能够实现强噪声背景下滚动轴承微弱故障特征提取。

The fault features of rotating machinery are weak in the early stages and easily affected by background noise, which makes them difficult to extract. A transient fault feature extraction method for use with rolling bearings based on low-rank and sparse decomposition is proposed in the paper. The sparse time-frequency representation of a periodic transient signal is exploited, a low-rank and sparse model is established, and the transient signal is extracted from the background noise. A measured signal is first transformed into a new sparse subspace through a high-order synchrosqueezing transform(FSSTH), and then the robust principal component analysis(RPCA) algorithm is used to decompose the sparse time-frequency matrix into a low-rank component and a sparse one. Finally the inverse FSSTH is applied to the low-rank matrix in order to recover the transient signal. The proposed method is data-driven and does not require any prior information. Simulated signal and experimental analysis results show that the proposed method can effectively enhance the periodic transient characteristics caused by faults in the vibration signal, and facilitate the feature extraction of weak faults for rolling bearings in the presence of strong background noise.