Using Novel Complex-Efficient FastICA Blind Deconvolution Method for Urban Water Pipe Leak Localization in the Presence of Branch Noise

Abstract

Leaks in water pipelines can create many problems, such as financial loss, environmental pollution, and public health hazards. Previous scholars and engineers have researched leak detection and localization methods for water-filled pipe systems, which have proven to be effective in pipe fault detection, even if under external noise. However, the performance of these techniques is unavoidably affected by the noise from the pipe, which includes branch noise, and has motivated researchers to explore more high-performance technologies further. For this purpose, this paper mainly focuses on the problem of locating leaks in a branch water pipe system. Then, a novel blind deconvolution algorithm, called the complex-efficient FastICA (C-EFastICA) algorithm, is proposed to extract the original leak signal from the mixed leak acoustic signal. Unlike the other complex field blind deconvolution methods, the proposed C-EFastICA can adaptively select the nonlinear function g to establish the cost function and iterative learning rules according to the different generalized Gaussian characteristics, which makes the algorithm more accurate in decomposing mixed leak signals in a complex field. The experiment results show that the C-EFastICA is faster than the classical complex FastICA (C-FastICA) algorithm. The relative precision of the leak localization reached approximately 88% under the interference of branch noise.