摘要
形态成分分析(MCA)是最新提出的一种基于稀疏表示的信号和图像分解(分离)方法,其扩展算法GMCA(Generalized MCA)可用于超定和欠定情形的盲源分离。为了降低GMCA算法中重构信号的均方差,提高分离信号的精度,将半软阈值函数和MOM阈值更新机制相结合,提出了SST-MOM(Semi-soft Thresholding MOM)阈值更新策略,仿真结果表明,新算法较原GMCA算法提高了分离信号的信噪比,将其应用于齿轮箱复合故障诊断中,有效地识别出了两路观测信号中的3种故障,表明了该算法的有效性。
Morphological Component Analysis(MCA) based on sparse representation is a novel decomposition method of signals and images,Generalized MCA can be used for over-determined and underdetermined Blind source separation(BSS).In order to decrease the mean square error and enhance the precision of separated signals,combining the semi-soft threshold function and MOM strategy,SST-MOM strategy is proposed.Simulation shows that new algorithm improves the SNR of separated signals.It can tell three kind of faults lying in the two observed vibration signals which demonstrates the effectiveness of the algorithm in the complex fault diagnosis of gearbox.
出处
《机械传动》
CSCD
北大核心
2012年第6期93-98,共6页
Journal of Mechanical Transmission
关键词
形态成分分析
盲源分离
齿轮箱
复合故障诊断
半软阈值
Morphological component analysis Blind source separation Gearbox Compound fault diagnosis Semi-soft threshold