Research On Acoustic Leak Detection Technology For Gas Pipeline

During the long distance pipeline operation,leakage occurs occasionally due to the natural corrosion,aging or vandalism by people.Because nature gas is highly flammable and explosive,if the leakage could not be found in time it would bring great loss to the society and the environmental,even cause people's damage or death.So it's very important to study the technology of leak detection for gas pipeline.In this paper,using the high-pressure gas pipeline device,the technology of leak detection is investigated based on the acoustic method.With the purpose of de-noise the collected the acoustic signals,wavelet transform is firstly adopted.The optimal wavelet basis,scales,threshold selection rules,threshold functions and other parameters are studied in the paper.The characteristics of acoustic signal are extracted in the time,frequency and wavelet domain.Then,some of these characteristics are input to BP neural network,the leakage is diagnosed by the output of the neural network.Finally,a leak-detection system is designed based on LabVIEW.Through the theoretical development and experimental analysis,some conclusions can be drawn as follows:(1)The collected acoustic signals are non-stationary stochastic signals,which can be de-noised by the wavelet transform.(2)The parameters of wavelet transform are determined as followed: ‘sym2' is chosen as the optimal wavelet basis,11 scales,threshold selection rule is 'heursure',threshold function is ‘soft'.(3)Continue wavelet transform is used to implement the time-frequency analysis.The optimal wavelet basis is chosen as‘mexh'.(4)The average value,variance,self-correlation,gradient,frequency spectrum,power spectrum,wavelet packet power spectrum can used as the characteristics of theleakage.(5)A leakage detection system with multiple variables is established.Five most significantly characteristics are chosen as the inputs,including the average value,variance,self-correlation,density of the frequency spectrum and power spectrum of the wavelet package.The leakage detection system with multiple variables is established by using BP neural network,which can correctly implement the diagnosis function for the gas pipeline.(6)A leakage detection system is designed based on LabVIEW.The wavelet de-noising,characteristics extraction,NNT detection algorithm,interference exclusion can be realized in LabVIEW language.The work can provide the foundation practical application of the algorithm in the engineering.