Pipeline signal feature extraction with improved VMD and multi-feature fusion
Yina Zhou, Yong Zhang, Dandi Yang, Jingyi Lu, Hongli Dong, Gongfa Li
Abstract
This paper is concerned with the pipeline leakage detection problem. A pipeline signal feature extraction method based on improved variational mode decomposition (VMD) and multi-feature fusion are proposed. First of all, the mode number K-value of VMD decomposition is determined by combining the empirical mode decomposition (EMD) method and the centre frequency method. Next, according to the variance contribution rates, the effective mode components are selected from the mode components obtained by VMD, subsequently, the effective mode components are reconstructed to get the de-noised signal. Then, the characteristic parameters that might distinguish the different pipeline signals are selected from different aspects. Moreover, the selected characteristic parameters are formed feature vector to put into support vector machine (SVM) to recognize the different pipeline events. Finally, the laboratory pipeline samples are employed to verify the effectiveness and superiority of the proposed method.