Research On Leakage Detection Technology Of Liquid-Filled Pipeline Based On Vibration Signal

As one of the five major modes of transportation,pipeline transportation has the characteristics of low cost,large transportation volume,small footprint,safety and reliability,and is dominant in fluid transportation.However,due to corrosion,artificial damage,natural disasters and other reasons,pipeline leakage occurs from time to time,resulting in several difficult problems in pipeline operation and maintenance.How to use new technologies in the fields of signal processing and basic algorithms to accurately determine pipeline leakage status is to overcome this problem.key to the puzzle.In this paper,the liquid-filled pipeline is taken as the research object,and the leakage problem of the liquid-filled pipeline is studied through the pipeline leakage simulation experiment platform.Starting from the leakage noise mechanism of the liquid-filled pipeline and the acoustic wave conduction model,the vibration signal is used to detect the leakage state and extract the leakage information.Combined with the interpretable machine learning model,an optimization method for pipeline leakage detection is proposed to realize the ability to identify leakage modes of different apertures.And improve the accuracy of leak state detection.The main research contents of this paper are as follows:(1)The research background and significance of this paper are clarified,the research status of domestic and foreign pipeline leak detection hotspot technologies is reviewed,and the applicable scope of different detection technologies and the development trend of pipeline leak detection technology are summarized.(2)Using the knowledge of fluid mechanics,elasticity and acoustics,the pipeline acoustic wave conduction model was established,and the correlation between the sensor signal and the leakage noise power spectral density was described;the prediction model of the pipeline leakage noise power spectrum was expounded;Pipeline leakage noise mechanism generation mode,pipeline leakage noise mode is divided into turbulent noise mode and bubble cavitation mode;the response equation of pipe wall vibration is expounded,and the corresponding frequency band information of the two noise mechanism modes in the sensor signal is given.(3)According to the physical pipeline leakage simulation test bench,the vibration signals under the two sets of data acquisition devices are compared,and the images that can clearly distinguish the pipeline leakage state are obtained through Fourier transform and continuous wavelet transform.The sensor selection scheme and the comparative experimental design scheme under this study are clarified.By analyzing the experimental data,the applicability of the pipeline leakage noise generation mechanism in this experimental bench and the feasibility of the slope characteristics under the pipeline leakage noise spectrum prediction formula to identify the pipeline leakage state are verified.(4)Based on the pipeline leakage acoustic wave conduction model and the pipeline leakage noise generation mechanism,a method of using the piecewise power spectrum entropy to describe the pipeline leakage vibration signal is proposed,and it is added to the feature space to compare with other time domain and frequency domain features.For comparison,the optimized feature space under the experimental conditions was established.Then combined with the Boruta algorithm and the interpretable ensemble learning tree model,a set of pipeline leakage state identification methods under the typical feature space set is formed,and the corresponding relationship between the leakage mode and the typical features is given.The robustness of the method is verified by comparing the performance of the method under different detection distances.By comparing the optimization method proposed in this paper with the traditional method of clustering modal decomposition and random forest algorithm,the good performance of this method is verified.