摘要
传统管道漏磁检测信号处理出现混叠、过包络发散、低频异变等问题,导致缺陷信号特征量提取与识别效果不理想。针对上述问题,基于变分模态分析-支持向量机(Variational Mode Decomposition-Support Vector Machines, VMD-SVM)算法完成管道漏磁信号特征辨识。采用四阶VMD处理管道漏磁信号,解决了经验模态分解(Empirical Mode Decomposition, EMD)的过包络引发的信号发散问题,也解决了小波分解(Wavelet Transform, WT)的低频信号异变问题。同时,以峭度最大原则选择最佳的模态分量(IMFm),提取模态分量的特征量,建立样本集。最后,采用SVM算法对信号特征量进行辨识分类,优选核函数,提高辨识精度。利用现场采集信号进行验证,结果表明:VMD-SVM算法抗干扰性强、识别精度高。
The traditional pipeline leakage detection signal processing has problems such as overlapping, over-envelope divergence and low-frequency variation, which leads to the unsatisfactory extraction and recognition of defective signal feature quantity. Aiming at the above problems, based on the Variational Mode Decomposition-Support Vector Machines(VMD-SVM) algorithm was proposed to identify the characteristics of pipe leakage signal. The fourth-order VMD algorithm was used to process the pipeline signal, which solved the signal divergence problem caused by the over-envelope of the empirical mode decomposition(EMD), and also solved the low-frequency signal divergence problem of the wavelet decomposition(MT). Meanwhile, the best modal component(IMFm) was selected with the principle of maximum kurtosis, and the eigenvalue of the modal component was extracted to establish the sample set. Finally, the SVM algorithm was used to identify and classify the signal eigenvalue, and the kernel function was optimally selected to improve the discrimination accuracy. Therefore, the results of the validation using field-acquired signals show that the VMD-SVM algorithm has strong anti-interference and high recognition accuracy.
作者
张敏
王德国
郭岩宝
张咪
张政
ZHANG Min;WANG Deguo;GUO Yanbao;ZHANG Mi;ZHANG Zheng(College of Mechanical and Transportation Engineering,China University of Petroleum-Beijing,Beijing102249,China)
出处
《石油矿场机械》
CAS
2022年第6期9-17,共9页
Oil Field Equipment
基金
国家重点研发计划项目“基于多源数据分析的承压类特种设备典型缺陷预测技术研究”(2018YFC0809004)。
关键词
VMD-SVM
管道缺陷
漏磁检测
特征识别
VMD-SVM
pipeline defects
magnetic flux leakage testing
characteristics identification
