摘要
在用声发射(AE)技术对低流量泄漏管道进行在线检测时,针对典型声发射特征参数无法有效区分低流量泄漏管道不同检测位置所采集的气动噪声信号的问题,以梅尔频率倒谱系数(MFCC)矩阵作为特征值,对泄漏信号提取特征,利用灰狼优化算法(GWO)将传统支持向量机(SVM)优化后对信号进行分类识别,并与传统支持向量机(SVM)的识别结果进行对比。结果表明,提取特征MFCC步骤中滤波器组数M的选取也对分类识别有一定影响,当M=12时,利用特征矩阵MFCC作为低流量管道泄漏气动噪声信号的特征识别准确率达到了90.67%,且随着传统支持向量机的优化识别准确率会进一步提高到94.67%。
When acoustic emission(AE)technology is used for online detection of low flow leakage pipeline,typical AE characteristic parameters are insufficient to effectively differentiate the aerodynamic noise signals collected from different inspection positions of low-flow leakage pipelines.Therefore,a method is proposed to extract features from the leakage signal with Mel frequency cepstrum coefficient(MFCC)matrix as the eigenvalue.The gray wolf optimization algorithm(GWO)is used to optimize the traditional support vector machine(SVM)and classify and recognize the signals,and the recognition results are compared with those of the traditional support vector machine(SVM).The results show that the selection of the number of filter banks M in the feature extraction MFCC step also has a certain impact on the classification and recognition.When M=12,the recognition accuracy of using the feature matrix MFCC as the signal feature of the low flow pipeline leakage aerodynamic noise signal reaches 90.67%,and with the optimization of the traditional support vector machine,the recognition accuracy will further improve to 94.67%.
作者
陈展
张颖
周发戚
束倩倩
CHEN Zhan;ZHANG Ying;ZHOU Faqi;SHU Qianqian(School of Environmental&Safety Engineering,Changzhou University,Changzhou Jiangsu 213164,China;不详)
出处
《工业安全与环保》
2023年第10期9-13,共5页
Industrial Safety and Environmental Protection
基金
中国石油天然气集团有限公司常州大学创新联合体科技合作项目(KC20210301)。
