Investigation Into The Model Of Leak Noise,Leak Detection And Location Methods In Urban Gas Pipelines

Gas pipelines are important lifeline systems of modern cities.However,due to factors such as construction quality,degradation,corrosion,third-party damage and so on,gas leaks occurred frequently during the entire life of gas pipelines,leading to serious consequences.Therefore,timely and accurate detection and location for gas leaks are of great importance to the operational safety of gas pipelines.This paper aims to develop the acoustic techniques for gas pipes,and focuses on the whole process of the generation,propagation and collection of leak noise to establish a systematic framework for leak detection and location.This paper first designed an indoor gas pipe rig to study the characteristics of leak noise under different physical parameters,and analyzed its general trend and behavior under different pipe pressure and leak sizes.On this basis,outdoor gas pipe rigs were established to further study the characteristics of leak noise in the presence of gas pipe complexities,such as bends and branches,and the backfill soil.The attenuation properties of leak noise at a rather long propagation distance were also examined.These experiments provided necessary data which will be used for the validation of the proposed models in the following study.Leak noise is the main research object of the acoustic techniques.In order to explore the generation mechanism of leak noise in gas pipes,the physical process of gas leaks was studied and the main acoustic source was then determined.A general framework for the modeling of leak noise was proposed based on the aeroacoustics and the Green function of gas pipeline systems.Based on the assumption of isotropic turbulence at the leak point,the Liepman spectrum was introduced to describe the behavior of leak velocity.Then the leak noise model was established and the generation mechanism of leak noise in gas pipes was revealed.The proposed model predicted the linear relationship between the overall sound pressure level and the fourth power of the leakage radius and the square of the leak velocity,which was further demonstrated by the indoor experimental data.For the accurate detection of gas leaks,a robust and effective feature was proposed based on the established leak noise model.The wavelet transform was used to denoise the measured data,and a feature selection algorithm was employed to select the most discriminative features to form the feature vector.The gas leak identification methodology was then established using artificial neural network,support vector machine and random forest with the selected feature vector.High performance was achieved on both experimental data,which is thus promising and encouraging for leak detection in practical applications.The propagation characteristics of leak noise in gas pipelines were further explored to determine the wavespeed and attenuation coefficients.A model of the cross correlation function of leak noise for gas pipelines was established by combining the leak noise model and the basic mathematical equations of the cross correlation function.The proposed model takes into account the whole physical process of the generation,propagation and collection of leak noise,which is capable of predicting the theoretically maximum distance between sensors.The theoretical model was further compared with the experimental data from the outdoor rig.The results showed the proposed model could not only portray the main characteristics of the cross correlation coefficients,but also reflect the general trend at different lower limit frequencies.The time delay estimated from the peak of the cross correlation was combined with the theoretical wavespeed to locate the gas leak,which provided a quite satisfactory accuracy.Finally,the research results and conclusions of this study are briefly summarized,and some research contents requiring further investigation are outlined.