Water Pipe Leak Detection using the k-Nearest Neighbor Method

Abstract

Water loss due to pipe leakage is a common phenomenon that occurs and is often overlooked, pipeline leaks that carry water are not too dangerous as in gas or oil pipes, but the losses will be even greater if the leak is neglected, infrastructure around the pipeline leak may suffer damage, impacting water companies and most importantly will affect the availability of clean water. These losses can be minimized if the pipeline leak is immediately known. This paper presents the detection of water pipe leakage by detecting changes in pressure at several points on the water pipe network. A Variable Speed Drive (VSD) is used to control the performance of the water pump, controlled water pump performance by VSD aims to regulate the pressure on the water pipe network, the performance of the 20% strength water pump produces weak water pressure. The water pressure on a weak pipeline network with leakage at 2 points will be analyzed and processed using the k-Nearest Neighbor (k -NN). Data processing was conducted in several experiments, initially pump performance of 40%, 60%, 80% and 100%. Pipeline leak detection did not experience difficulties with 100% accuracy if the pump strength is known and held constant at those specified pump strengths. Then, in the case of combined data from 40%, 60%, 80% and 100% pump strength, with and without leaks, the k-NN method was able to detect leakage with an accuracy of 95.62%. The leak detection accuracy of the method drops to 88.46% when the pump performance was 20 % which produces a weak water pressure in the pipeline.