摘要
机械或电气设备工作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,可采用独立分量分析(ICA)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独立分量分析,再做小波逆变换得到分离信号。与直接的ICA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 are important measures. Commonly, signal of sensor results from some aliasing sound sources. The signal of tested equipment can be separated by ICA method in order to analyse the fault signature. A new method is put forward to analyse the wavelet coefficient sequence of mixed signal for acquiring independent component, and to acquire separate signal by wavelet inverse transform. Since the kurtosis of wavelet coefficient is higher than that of original signal, this method can acquire better separating effect and faster convergence, compared with direct ICA.
出处
《计算机仿真》
CSCD
2008年第4期321-325,共5页
Computer Simulation
