一种基于SST-MOM的改进形态成分分析算法及其应用  

An Improved Algorithm of Morphological Component Analysis based on SST-MOM and Its Application

作　　者：杨杰[1] 郑海起[1] 关贞珍[1] 王彦刚[1] 田昊[1] 

机构地区：[1]石家庄军械工程学院,河北石家庄050003

出　　处：《机械传动》2012年第6期93-98,共6页Journal of Mechanical Transmission

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.

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 

分 类 号：TH132.41[机械工程—机械制造及自动化] TH165.3