Application Research Of The Improved LMD Algorithm In Pipeline Leakage

In view of the problem that it's difficult to extract leakage feature information and the low accuracy of leak positioning in the natural gas pipeline leakage detection process,the local mean decomposition algorithm(LMD)is applied to the pipeline leakage detection in this paper for realizing pipeline leakage signal decomposition,feature extraction and leakage location.Firstly,this paper introduces the local mean decomposition algorithm in theory and its application in signal decomposition.The LMD as an effective means of dealing with non-stationary random signal,it has the characteristics of adaptive and completeness for signal decomposition.But due to the influence of the algorithm itself,it's easy to generate modal aliasing.So the ensemble local mean decomposition algorithm that assisted by noise is used to suppress the modal-aliasing problem during the process of LMD decomposition.Secondly,In the process of transmission,the signal is often affected by various noises and interference,so that the useful signal in the original signal source is cut or confused.In order to enhance the useful signal,restrain the noise interference and ensure the extracted eigenvalue can represent the signal characteristic.The de-noising pretreatment of the acquired original signal is needed.At the same time,the ELMD and spectral kurtosis joint de-noising algorithm,which based on wavelet packet,is proposed to avoid the technical gaps during the process of the wavelet decomposition in this paper.Based on the effective PF component by ELMD decomposition,this algorithm use the spectral kurtosis optimal parameters and the wavelet packet energy distribution to determine the signal reconstruction node,completed the signal de-noising of all effective PF component.The PF component with noise reduction can characterize the original signal features in different scales.Then,through the analysis of the characteristics of pipeline signal and the time-frequency theory,this paper proposes a spectral entropy parameter to describe the signal frequency characteristics quantitatively with adaptive optimal kernel time-frequency distribution(AOK).The AOK spectral entropy that extracted from all effective PF components is used to determine preliminarily whether the pipeline have leaked and the kinds of pipeline situation.It have a good discrimination and recognition accuracy for three pipeline situation which including normal operation of pipeline,pipeline leakage,pipeline tapping.Finally,this paper introduces the leak detection algorithm based on the ELMD multi-scale and correlation analysis.Get the different characteristic scale time delay by combining the correlation analysis with the de-noising PF component which dealt with the ELMD decomposition and realizes the pipeline leak location.Compared with Correlation calculation using the original signal without ELMD decomposition,the algorithm can get more accurate results,which is helpful to improve the accuracy of pipeline leakage location.