基于HHT和SVM的纤维拉伸断裂声发射信号的特征提取及分类研究

为了研究纤维拉伸断裂声发射信号的特征提取及分类方法,采用声发射技术采集了芳纶1313和阻燃黏胶2种纤维的拉伸断裂的声发射信号。通过小波变换,对采集的2种纤维的声发射信号进行消噪预处理以去除部分噪声,应用希尔伯特-黄变换对2种纤维去噪后的信号进行特征频率的提取,运用最小二乘支持向量机(LSSVM)对2种纤维的特征频率进行分类识别。结果表明:小波去噪方法可以去除信号的部分噪声;希尔伯特时频谱可以一定程度上反映2种纤维材料在时间维度上的断裂情况,边际谱上可以提取2种纤维材料声发射信号的特征频率;LSSVM能够对2种纤维材料拉伸断裂的特征频率分类识别,芳纶1313的识别率为40%,阻燃黏胶的识别率为80%,总的识别率为60%。

Research on feature extraction and classification of AE signals of fibers' tensile failure based on HHT and SVM

In order to study the feature extraction and recognition method of fibers'' tensile failure, AE technology is used to collect AE signals of fiber bundle''s tensile fracture of two kinds of fibers of Aramid 1313 and viscose. A transform called wavelet is used to deal with the signals to reduce noise. A method called Hilbert-Huang transform (HHT) is used to extract characteristic frequencies of the signals after the noise is reduced. And a classification method called Least Squares support vector machines (LSSVM) is used for the classification and recognition of characteristic frequencies of the two kinds of fibers. The results show that wavelet de-noise method can reduce some noise of the signals. Hilbert spectrum can reflect fracture circumstances of the two kinds of fibers in the time dimension to some extent. Characteristic frequencies'' extraction can be done from marginal spectrum. The LSSVM can be used for the classification and recognition of characteristic frequencies. The recognition rates of Aramid 1313 and viscose reach 40%, 80% respectively, and the total recognition rate reaches 60%.