摘要
为了克服压力波动下输气管道泄漏信号变化幅度大导致孔径识别准确率低的问题,提出了一种基于WPT-PCAGMHMM的泄漏源特征识别模型。开展了压力波动下管道泄漏的声发射检测实验,通过小波包变换(WPT)提取了不同工况下声发射信号的小波包能量谱,随后通过主成分分析(PCA)对频带能量进行去相关性与降维。最后将数据及标签分为训练集与测试集,采用高斯混合-隐马尔可夫模型(GMHMM)实现了对管道压力与泄漏孔径的分类识别。结果表明,所提出的模型整体准确率最高达到95.20%,泄漏孔径准确率达到99.95%,显著泄漏识别准确率达到100%,在充足样本及小样本的环境下相比BPNN、SVM均有优秀的表现。
In order to overcome the problem of low aperture recognition accuracy caused by large amplitude change of leakage signal of gas pipeline under pressure fluctuation,a leakage source feature recognition model based on WPT-PCA-GMMMM is proposed.The acoustic emission detection experiment of pipeline leakage under pressure fluctuation was carried out,and the wavelet packet energy spectrum of acoustic emission signal under different working conditions was extracted by wavelet packet transformation(WPT).Then the frequency band energy was decorrelated and dimensionally reduced by principal component analysis(PCA).Finally,the data and labels were divided into training set and test set,and the Gaussian mixed-hidden Markov model(GMHMM)was used to realize the classification and identification of pipeline pressure and leakage aperture.The results show that the overall accuracy of the proposed model reaches 95.20%,the accuracy of leakage aperture reaches 99.95%,and the accuracy of notable leakage identification reaches 100%,which has excellent performance compared with BPNN and SVM in the environment of both sufficient samples and small samples.
作者
喻可
张宏南
金建新
曾磊
林志明
金其文
吴迎春
吴学成
YU Ke;ZHANG Hongnan;JIN Jianxin;ZENG Lei;LIN Zhiming;JIN Qiwen;WU Yingchun;WU Xuecheng(Qingshanhu Energy Research Center,Zhejiang University,Hangzhou 311305,China;Ningbo Innovation Center,Zhejiang University,Ningbo 315100,China;Zhejiang Zheneng Jiahua Power Generation Co.,Ltd,Jiaxing 314201,China)
出处
《能源工程》
2024年第2期56-66,共11页
Energy Engineering
基金
宁波市“科技创新2025”重大专项项目(2018B10024)。
