Author(s): Xun Wang, Mohamed S. Ghidaoui
Linked Author(s): Xun Wang
Keywords: Defect detection, leak localization, fluid transient, matched-field processing.
Abstract: This paper proposes a transient leak detection method in a pipeline system which uses head measurements in the frequency domain. A matched-field processing (MFP) approach is used to separately estimate location and size of a leak. The leak location is determined by matching the model with the measurement, more exactly by maximizing the inner product between a unit weighting vector and the measurement. This maximum occurs in the case that the weighting vector and the measurement have the same direction implying that the parameters behind them are identical. Then, the leak size is estimated via least square. This MFP approach is able to use not only the resonant frequencies but also all the frequencies being invoked. Besides, the proposed method is a maximum likelihood estimation and equivalent to a matched-filter approach which maximizes the signal-to-noise ratio (SNR). Experiments on simulated data for a single pipe system illustrate the estimation efficiency of the MFP method. The simulation results justify that the proposed leak detection method is able to localize not only a single leak but also multiple leaks
Year: 2017