Analysis of leakage in high pressure pipe using acoustic emission method

摘要

Leak detection is one of the most important issues in the oil and gas pipelines, as it can lead to financial losses, as well as severe human and environmental impacts. Acoustic emission test is a new technique for leak detection. Leakage in high pressure pipes creates stress waves caused by localized loss of energy. Stress waves are transmitted through the pipe wall which can be recorded by using acoustic sensor or accelerometer installed on the pipe wall. Knowledge of how the pipe wall is vibrated by acoustic emission resulting from leakage is a key parameter for leak detection and localization. This paper aims to model acoustic emission generated by pipe vibration due to leakage. Donnell's non-linear theory for cylindrical shell was used to derive motion equation under simply supported boundary condition. Then, the motion equation was solved by using Galerkin method that resulted in a system of non-linear equations with 6 degrees of freedom. To solve these non-linear equations, ODE tool of MATLAB software and Runge-Kutta numerical method was employed to obtain pipe wall radial displacement. For verifying this method, acoustic emission by a continuous leak source was measured. Experiments were carried out with a linear array of sensors on steel pipe (ASTM A 106/99) of nominal length 6 m, 7.35 mm wall thickness and external diameter of 169 mm. The pressurized air was flown inside the pipe through the compressor. Two simulated continues leak sources with 0.6-mm and 1 -mm diameter holes were used under 5 bar air pressure. This source propagated waves in a large of frequencies about 0-400 kHz. In this study the vibration behavior of the pipe is investigated per resonance frequencies of the used AE sensors which are near 150 and 300 kHz. Signals generated by the pipe wall vibration were recorded by using acoustic emission sensors. In the next step, Fast Fourier Transform (FFT) was used in the signal analysis. Comparison of the obtained results, indicate the good agreement between the experimental and modeled frequencies ranges. The mean error between analytical modeling and experimental results is less than 6%.