机械噪声信号的基于小波变换的快速盲源分离法  

Blind Source Separate of Mechanical Noise based on Wavelet Transform

作　　者：王春[1] 彭东林[2] 

机构地区：[1]重庆科技学院机械工程学院,重庆市博士研究生400042 [2]重庆工学院

出　　处：《工具技术》2008年第3期41-45,共5页Tool Engineering

基　　金：国家自然科学基金资助项目(项目编号:50375163)

摘　　要：机械噪声的测试分析是实时故障检测诊断的重要手段,通常情况下,传感器检测得到的信号是多个噪声源叠加的结果,可采用盲源分离分析方法分离出待检测对象信号。提出先对混合信号小波系数序列进行盲源分离分析,再做小波逆变换得到分离信号。与直接的盲分离比较,小波系数比原始信号的超高斯性更强,因此分离处理的收敛速度更快,分离效果更好。由于在小波变换的过程中可以引入阈值去噪,因此基于小波变换得到的分离结果较之常规方法有更强的抗噪能力。The analysis and test of mechanical noise is important measure. Commonly, signal of sensor result from some aliasing sound source, The signal of tested equipment can be separated by BSP method in order to analysis fault signature, Put forward a new method that analysis the wavelet coefficient sequence of mixed signal to acquire independent component, and acquire separate signal by wavelet inverse transform, Since the kurtosis of wavelet coefficient is large than original signal, this method can acquire more distinctly separate effect and faster Convergence, Compare with direct BSP.

关 键 词：盲分离 超高斯信号 小波 机械噪声 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TH17[自动化与计算机技术—计算机科学与技术]