Research On Acoustic Detection And Leak Localization Technology Of Pipelines

With the laying length of the pipeline exceeding one million kilometers,the pipeline plays a vital role in the construction and operation of the city.Energy transportation and water transportation are inseparable from the normal operation of the pipeline.The economic loss and social impact caused by pipeline leak are very bad,and the safety accidents caused by pipeline leak also have a negative impact on the long-term stability of society.Therefore,pipeline safety monitoring is of great significance for long-term dynamic management of pipelines.Pipeline safety monitoring includes pipeline leak detection,leak localization,and pipe damage assessment.Aiming at the above problems,this paper focuses on the principle of acoustic detection and leak localization of pipelines,proposes a pipeline leak localization method based on wavelet packet decomposition and higher-order cumulant methods.The particle swarm optimization least squares support vector machine method is used to identify the leak aperture of pipelines.The main research work of this paper is as follows:Firstly,the principles of acoustic detection and leak localization and the principle of higher-order cumulant method are analyzed.Firstly,the principles of acoustic detection and leak localization are analyzed.Based on the principle of higher-order cumulants,the feasibility of time delay estimation using higher-order cumulants is verified by Matlab simulation signals.Compared with the basic cross-correlation method,the higher-order cumulant method has higher time delay estimation accuracy.By analyzing the wavelet packet decomposition principle,an index evaluation system for simulation signals is established to select the optimal wavelet packet basis function.Secondly,based on the research of water pipeline leak,a water pipeline leak localization method based on wavelet packet decomposition and higher-order cumulant is proposed.The wavelet packet decomposition method is used to extract the characteristic frequency band from the collected acoustic signal,and the higher-order cumulant method is used to estimate the time delay of the characteristic frequency band signal.The localization of the pipeline leak point is determined based on the delay time.Compared with basic cross correlation and generalized cross correlation methods,higher order cumulant methods have higher accuracy in time delay estimation.By analyzing the delay estimation curves and positioning errors of the third order cumulant and fourth order cumulant methods,it can be seen that the fourth order cumulant method has smaller positioning errors.Finally,based on Particle Swarm Optimization(PSO),the Least Squares Support Vector Machine(LSSVM)method is used to identify the leak aperture of the pipeline.Feature extraction is performed on the experimental data of 0-4mm leak aperture,and then the Pearson correlation coefficient between the feature and leak aperture is calculated to select features with an absolute value of correlation coefficient greater than 0.7.Using the selected feature values as input,leak aperture identification is performed based on the PSO-LSSVM algorithm.Compared with support vector machines and least squares support vector machines,the PSO-LSSVM method has higher recognition accuracy.