Variable Step Normalized LMS Adaptive Filter for Leak Localization in Water-Filled Plastic Pipes

• Published in: IEEE transactions on instrumentation and measurement Vol. 71; pp. 1 - 11
• Main Authors: Cui, Xiwang, Gao, Yan, Ma, Yifan, Liu, Yuyou, Jin, Boao
• Format: Journal Article
• Language: English
• Published: New York IEEE 2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Acoustic waves; Acoustics; Adaptive filters; Algorithms; Convergence; Cross correlation; Filled plastics; Iterative methods; Leak detection; leak localization; least mean square (LMS) adaptive filter; Localization; Localization method; Location awareness; Pipes; Piping systems; plastic pipe; Plastics; Sensors; time delay estimation (TDE); Water purification; Wave propagation
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2022.3169526

Acoustic waves induced by leakage attenuate as they propagate in the piping system. Difficulties are generally encountered in leak localization in plastic pipes since these waves are heavily attenuated than in metallic pipes. This study presents a leak localization method based on the variable step normalized least mean square (VS-LMS) adaptive filter. In the method, the optimal weight vector of the filter is determined for time delay estimation (TDE) without requirement of the prior knowledge of statistical characteristics of leak and noise signals. It has a desirable feature of sharpening the main peak of TDE curve through adaptive iterative learning process, thus improving the accuracy of TDE for leak localization. By adjusting the update mode of iteration step, the convergence speed of the VS-LMS algorithm can be greatly improved. A simulation model of leak wave propagation is established and is subsequently analyzed for leak localization with respects to the TDE accuracy. Experimental work is further carried out on a plastic pipe to verify the effectiveness of the proposed method. The simulation sand experimental results demonstrate that the proposed method outperforms the original LMS and basic cross-correlation methods for the leak localization in a water-filled plastic pipe.