基于平行因子分析的振动信号盲分离方法研究  

Study on Blind Source Separation Method of Vibration Signals Based on Parallel Factor Analysis

作　　者：张英堂[1] 孙宜权[1,2] 范红波[1] 李志宁[1] 许瑞刚[2] 

机构地区：[1]军械工程学院车辆与电气工程系 [2]66267部队

出　　处：《内燃机工程》2014年第3期118-124,共7页Chinese Internal Combustion Engine Engineering

基　　金：国家自然科学基金资助项目(50775219)

摘　　要：发动机缸盖振动信号为卷积混合信号,为消除延时滤波的干扰,提出了平行因子分析和快速独立成分分析结合的盲源分离方法,并对振动仿真信号与缸盖振动信号进行分析,结果表明:消除干扰后,分离得到的信号有更高的信噪比。利用分离信号的奇异值特征向量动态聚类对发动机进行故障诊断,应用实例表明:消除延时滤波影响后,发动机故障诊断精度提高47.17%。Signals of the engine cylinder head vibration are convolutive mixed signals. In order to eliminate delay filtering interference,a blind source separation method based on parallel factor analysis and fast independent component analysis was presented and used to analyze the simulation and measurement signals of cylinder head vibration. The results indicate that eliminated the delay filtering interference, the obtained separated signals are of much higher signal-to-noise ratio. The singular value characteristic vector dynamic clustering of separation signals can be used to diagnose engine failure. Practical application demonstrates that the engine fault diagnosis precision increases 47.17% by eliminating the delay filtering interference.

关 键 词：内燃机 盲源分离 卷积混合信号 发动机缸盖 平行因子分析 快速独立成分分析 

分 类 号：TK413[动力工程及工程热物理—动力机械及工程]