基于降噪及独立分量分析的轴承故障声信号特征提取

针对传统降噪算法的缺点，提出了将局部投影用于故障声信号的降噪。该算法具有较高的计算效率及广泛的应用前景，不仅可用于线性系统，而且还可用于非线性系统。而独立分量分析可用于分解相互独立的信号，它解决了多传感器信号的信息融合与特征提取问题。综合局部投影算法及独立分量分析算法两者的优点，提出了一种轴承弱故障特征识别算法。试验表明，该方法能有效地分离背景信号及特征信号。

Feature extraction of the acoustic signal of faulty bearing based on noise reduction and independent component analysis

In light of the shortcoming of traditional noise reduction algorithms, local projection noise reduction method based on high embedding dimension of phase space reconstruction is proposed for noise reduction of acoustic signal. The method, with high calculation efficiency, has a great prospect of application, and can be used for both linear and nonlinear systems. Independent component analysis can be used for separating mutually independent signals, which solves the problems of information fusion and feature extraction of multi-sensor signals. An algorithm for the feature identification of bearing acoustic signal that combines the merits of local projection and independent component analysis is proposed in the paper. Analysis indicates that the algorithm can effectively separate background signal from characteristic signal.