基于分形维和独立分量分析的声发射特征提取

Feature Extraction of Acoustic Emission Signals Based on Fractal Dimension and Independent Component Analysis

针对噪声对声发射信号分形维的影响,提出了一种基于分形维和独立分量分析(ICA)的结构材料声发射信号特征提取方法.文中首先给出了分形维的概念,并从理论上分析了噪声对声发射信号分形维的影响.接着引入ICA进行信号预处理,以提取源独立的去噪信号进行分形维计算.最后进行了多组铅心模拟声发射实验.实验结果表明,不同的声发射源和传播介质下声发射信号的分形维表现出明显不同的特征,且与去噪前的分形维相比,能够更好地对应声发射事件数.分形维具有受研究者主观影响小、易于标准化的优点,可以作为一种新的结构材料声发射的特征识别方法.

According to the effect of noises on the fractal dimension of acoustic emission （AE） signals, this paper proposes a method based on fractal dimension and independent component analysis （ICA） to extract the AE signal feature of construction material. In the investigation, the concept of fractal dimension is first introduced, and the effect of noises on the fractal dimension is theoretically analyzed. Then, the ICA is introduced to preprocess AE signals, and the fractal dimension is calculated from the independent signal after ICA process, with the AE source lead simulation being finally performed. The results show that the fractal dimensions in different AE sources and transmission media display distinct features, and that the fractal dimension after de-noising corresponds with the emission event number more precisely than that without de-noising. It is thus concluded that the fractal dimension can be used to identify the AE signal feature of construction materials because it is almost independent of the subjects of researchers and is easy to standardize.