Adaptive Independent Component Analysis–Based Cross-Correlation Techniques along with Empirical Mode Decomposition for Water Pipeline Leakage Localization Utilizing Acousto-Optic Sensors

Abstract

The leak localization in a water pipeline system plays a vital role in pipeline system modeling. The leak location is identified by estimating the time difference between two sensor signals placed on either side of the leak. The time difference estimation is achieved using the cross-correlation technique. The conventional cross-correlation technique fails to detect the leak when the vibrated sensor signal is corrupted by noise as well as convoluted with the impulse response of the pipe. In this paper, adaptive independent component analysis-based cross-correlation techniques along with empirical mode decomposition are implemented for processing the noisy and convolved impulse response output from acousto-optic sensors for locating the leaks in the water pipeline. The preprocessing of empirical mode decomposition is used to decompose the leak signal into several intrinsic mode functions (IMFs), and the external noise in the leak signal can be eliminated by removing the uncorrelated IMFs. The adaptive nature of independent component analysis in the proposed method makes the algorithm more robust in a real-time scenario. The probable benefit of using adaptive independent component analysis is to achieve a strong decorrelation and separation of the real-time vibrated sensor signal from the convoluted impulse response of the pipe along with noise. The proposed algorithm has been experimentally tested from a specially erected leak-detection facility located at Universiti Tunku Abdul Rahman Research site in Malaysia. The test result shows that the proposed method offers more than 95% accuracy of leak localization compared with conventional cross-correlation methods.