基于相空间重构与奇异值分解的刀具磨损信号降噪方法研究

Research on Tool Wear Signal Noise Reduction Based on Phase Space Reconstruction and Singular Value Decomposition

在机械故障诊断过程中实测信号通常是非平稳,非线性的含噪信号,针对这一特点提出了基于相空间重构的奇异值分解降噪方法。此方法根据时延嵌入理论对采集的时序信号进行相空间重构,由自相关法和Cao方法分别求时延和嵌入维,对重构的坐标矩阵进行奇异值分解,分析奇异谱特点,分离有用信息和噪声平台,对信号进行降噪。重构的相空间同时也构造了动力系统吸引子,吸引子的相空间轨迹可以直观反映去噪效果。通过对MATLAB环境下的bumps信号和实际采集的刀具磨损声发射信号进行验证。结果表明,该方法降噪效果比较明显。

The signals from mechanical fauh diagnosis are generally non-stationary, nonlinear and noise corrupted. So the method of singular value decomposition （SVD） is proposed based on phase space reconstruction. Phase space of timing signals is reconstructed based on delay-embedding theory. The embedding dimension m and delay time τ are computed respectively with Cao method and autocorrelation method. After SVD, the main information is separated from noise. The reconstructed phase space reflects attractor＇ s structure too. Its phase space trajectory is sensitive to noise, so it can reflect de-noising effect. The test of bumps from MATLAB and real tool wear AE（ Acoustic Emission） signals shows that the method is effective.