Information processing in buried pipeline leak detection system

Abstract

Detecting leak noises of buried pipelines is the most effective method of pinpointing a leak or leaks in buried pipelines. However, the acquired leak signals are usually blurred with ambient noises and therefore they have low signal-to-noise ratio. It is essential to use appropriate information processing scheme in order to acquire the leak signatures. The sensors are positioned on either side of the suspected leak position. By estimating the time delay between the two sensor signals, the leak can be localized. However the features of leak noise vary with the materials, sizes and inbuilt conditions of pipes. It is difficult to predetermine the spectral knowledge of a leak noise. The spectral knowledge is a prerequisite for the commonly used generalized correlation method to get the time delay estimation between the sensor signals. Here, the LMS adaptive filter is used to determine the time delay. As leak noise is stationary, the mean response of the LMS adaptive filter can converge to the discrete Winner filter and be peaked at the delay. This way, the estimation is adaptive and does not depend on either the spectrum of leak noise or the spectra of ambient noises. In practice, bursting interferences are inevitable where detection operates. The bursting interferences produce nonstationary characteristics in a signal and in turn result in diverging of the adaptive filtering process. The Wavelet decomposition and thresholding is deployed to remove the bursting interferences and recover the stationary behaviors.