Detection And Location Of Pipeline Leak Based On Acoustic Emission Method

Pipeline transportation is the most important oil and gas transmission in the petroleumindustry. Aging, breakage, corrosion, environmental corrosion, wear, fatigue and defects mayoccur during operation which will lead to pipeline leak. Finally resulting in the oil economiclosses, environmental pollution and other serious consequences. Acoustic emission (AE)detection technology can achieve early detection and warning of the leak, than find the leaksource quickly and easily, which is conducive to the timely repair of the leak, thus reduceleakage losses caused. The study has important significance and application value.Experimental studies such as the simulation of leak conditions, AE signal acquisition,AE signal propagation characteristics were carried out based on the gas and liquid leak anddiffusion large-scale experimental system of Safety Engineering in China University ofPetroleum (East China). At first, pipeline leak AE signal (non-buried pipeline and buriedpipeline) characteristics, the signal amplitude, frequency as the variation rule of pressure andflow, and non-buried pipeline leak AE signal amplitude, frequency as the variation rule ofleak aperture were been analysis. Then, the influence of pressure, flow, leakage aperture andpipeline surrounding medium (air and soil) on pipeline leak AE signal amplitude attenuationwhen the signal propagation in a long straight pipeline was been studied. At last, theattenuation rule (amplitude and frequency) of pipeline leak AE signal propagated through theflanges, valves and waveguide was been discussed.Take non-buried pipeline liquid, gas leak AE signals and buried pipeline gas leak AEsignals as the object, the feasibility of four feature extraction methods were been studied.They were wavelet packet decomposition (WPD), empirical mode decomposition (EMD), thecombination of EMD and FFT, the combination of EMD and WPD. The results show that themethods of WPD, the combination of EMD and WPD could achieve good results in thepipeline leak state feature extraction. Support vector machine (SVM) was been applied forleak pattern recognition of AE signal. The leak state of pattern recognition included the longstraight pipeline leak state, the complex pipeline (with pipeline fittings) leak state and no-leakstate.Finally, a method called multi-level leak source localization which combined of regional localization and precise localization was presented. Certain improvements was been made ofpositioning method based on signal amplitude attenuation, then the regional localizationwhich was more convenient and can be achieved by the program was been proposed. Theprecise positioning method based on WPD was been raised from the computational efficiencypositioning accuracy, the range of WPD components selection was been reduced and theWPD level was been determined. Through the source location experiments of pipelineleakage detecting by AE under different environmental (air and soil), the study confirmed thatwith the change in distance, pressure, the positioning error fluctuated between0-5%, theaverage location error controlled in about2.5%.