摘要
针对强噪声环境中有用信号提取的难题,提出了基于广义形态分量分析的降噪方法。通过引入虚拟观测信号,将一维观测信号扩展为多维虚拟观测信号,再通过广义形态分量分析,实现观测信号的盲源分离,从而达到降噪的目的。通过仿真信号和齿轮磨损故障振动实验信号的研究结果表明:广义形态分量分析技术能有效分离强背景噪声中的微弱信号,有效提取故障特征,其降噪性能优于传统的独立分量分析。
Morphological component analysis (MCA) is a novel signal or image processing technique based on signal morphological diversity and sparse representation. MCA takes advantage of the sparse representation of analyzed data in over-complete dictionaries to separate features in the data based on their morphology. Aiming at the problem of extracting a useful signal from strong background noise, a novel de-noising approach based on generalized morphological component analysis (GMCA) was presented. By introducing the virtual observation signal into the original signal, the one dimensional observation signal vector was converted into multi-dimensional virtual observation signals. The GMCA was then applied to the virtual observation signals, the blind source separation was realized and the noise was eliminated. The simulation and test results showed that not only a weak signal is separated, but also the signal noise ratio of the separated signal is improved; the fault of gear wear can be effectively detected and diagnosed; the denoising performance is better than the traditional independent component analysis method.
出处
《振动与冲击》
EI
CSCD
北大核心
2013年第1期145-149,173,共6页
Journal of Vibration and Shock
基金
国家自然科学基金资助项目(50975185
50775219)
关键词
广义形态分量分析
稀疏分量分析
故障诊断
降噪
独立分量分析
generalized morphological component analysis
sparse component analysis
fault diagnosis
denoising
independent component analysis