An Adaptive ICA-Based Cross-Correlation Techniques for Water Pipeline Leakage Localization Utilizing Acousto-Optic Sensors

Abstract

This paper describes the implementation of an adaptive ICA (Independent Component Analysis) based cross-correlation technique to process the noisy output from Acoustic optic sensor to locate the leaks in water pipes. In Digital Signal Processing methods, cross-correlation technique plays a prominent role in detecting the leak signals. In a real-time scenario, the leak signals get convolved with the impulse response of the pipe. In such cases, the conventional cross-correlation technique fails to detect the leak. This paper implements the maximization of non-gaussianity of ICA techniques along with wavelet decomposition as a preprocessing stage of frequency domain cross-correlation. The decomposition of the sensor data using wavelet reduces the computational complexity of the ICA. The adaptiveness of ICA in the proposed technique is more suitable for real-time water leak detection system. The probable benefit of using adaptive ICA is to enforce strong decorrelation and separation of the real-time leak signal from the convoluted impulse response of the pipe with noise. The above algorithm is validated using different experiments. The experimental result shows that the proposed method offers high accuracy of detecting the leak location compared to the conventional methods.