基于SWT和NMF的管道泄漏声发射信号特征提取与判别

Extraction and Identification of Characteristics of Pipeline Leakage Acoustic Emission Signal Based on SWT and NMF

针对管道泄漏声发射信号存在的非平稳性以及易受噪声干扰、特征不易提取的问题,提出基于SWT和NMF的管道泄漏声发射信号特征提取与判别方法,该方法首先利用同步压缩小波变换的方法提取管道流体泄漏声发射信号特征空间;其次,采用非负矩阵分解(NMF)对信号特征空间进行精简和优化,提炼出用于管道泄漏诊断和模式识别的特征向量;最后,采用支持向量机(SVM)对提取的特征向量进行模式识别和分类。实测结果表明,所提出的方法具有较高的分类准确率。

According to the characteristics of pipeline leakage acoustic emission signal are unsteady,easy to be disturbed by noise and difficult to extract,a method based on SWT and NMF for feature extraction and discrimination of pipeline leakage acoustic emission signals is proposed.First of all,the feature space of the acoustic emission signal of pipeline fluid leakage is extracted by the method of Synchrosqueezed wavelet transform;secondly,non-negative matrix factorization(NMF)is used to simplify and optimize the signal feature space and the characteristic parameters for pipeline leak diagnosis and pattern recognition are extracted.Finally,support vector machine(SVM)is used to classify the extracted feature vectors.The results show that the proposed method has high classification accuracy.