Pipeline leak detection using the multiple signal classification-like method

Abstract

Leakages in pipelines can cause severe hazards to industries, the environment and people. For the purpose of an accurate identification of the leakage location, a transient-based leakage detection method using multiple signal classification (MUSIC)-like is applied to this paper. The localization is achieved by a one-dimensional search of leak location along the pipe, which means it involves low computational cost. The performance of the MUSIC-Like method in the cases of a single leak and multiple leaks is discussed by comparison with three spectral-based methods. In the single-leak case, the MUSIC-like algorithm provides precise localization estimation even for a high level of noise. For the multiple-leak case, the MUSIC-like method is superior to the other three methods. It is capable of identifying all leaks where the leak-to-leak distance is less than half the shortest probing wavelength. Therefore, the MUSIC-like method has an excellent performance in leak detection and location.