结构健康监测中常用声发射信号进行声发射源的定位及特征描述。多个冲击事件发生时,声发射信号是多个信号的混叠,而且混合方式未知,这使利用声发射信号对冲击源进行定位变得非常困难。而近年来兴起的基于独立分量分析的盲源分离技术为解决这一难题提供了可能。本文采用基于信息极大化原理的反馈网络结构对同时作用在铝梁上的两个冲击事件产生的声发射混合信号进行分离,估计出各个源信号到达传感器的时延后,运用两点直线定位公式对两个冲击源进行定位。混合仿真实验验证了基于信息极大化原理的独立分量分析方法估计时延的有效性,铝梁上的两源冲击实验,进一步表明运用独立分量分析方法能较好的解决多冲击源定位问题。
Acoustic emission (AE) signal is used for localization and characterization of acoustic emission source. When there are multiple impact sources, it is a difficult problem that locating impact sources because AE signals are mixtures of multiple statistically independent signals and the mode of mixing is unknown. Fortunately, blind source separation (BSS) by independent component analysis (ICA) can solve the problem in recent years. In the paper, the feed-back neural network based on information maximization (informax) principle is used to separate the signal mixed by two independent continuous AE signals induced by two simultaneously impact events on an aluminum beam. Meanwhile, time delay between two PZT sensors for each source is obtained. Then two sources position are determined based on two-point localization formula. Time delay estimation by ICA based on informax principle is verified by simulation experiment. Furthermore, experiment on aluminum beam indicates that multiple impact sources locating can be solved by ICA.