Improving Acoustic Methods of Pipeline Leak Location with Distributed Sensing

Abstract

Abstract Leaks in water distribution mains are a big problem, with around 20% of supplied potable water lost to leaks during transport. Correlation-based acoustic techniques have provided an accurate and non-invasive way of detecting and locating these leaks for a few decades. These methods have almost exclusively been using two sensors, and so this paper presents work aiming to explore leak detection and location with multiple sensors distributed along a pipe. Beamforming is a well-established method for using arrays of sensors to locate sources, among other purposes. With this premise, the present work adopts an array processing algorithm (MUSIC) in the context of water leak detection, intending to develop a framework for detecting multiple leaks using a sensor array. The concept, processing and implementation details are first supported with numerical simulations using existing acoustic models of water pipes. Then, experiments are presented on a short section of water-filled pipe with leak-like disturbances. These are captured with an array of accelerometers and processed using an implementation of the algorithm, testing the impact of real-world effects studied in simulations. The study considers several aspects of practical interest: (i) the effect of noise, both correlated and uncorrelated; (ii) the effect of reflections from discontinuities, such as pipe fittings and connections; (iii) the number and the distribution of sensors; as well as (iv) the presence of multiple leaks. The results pave the way for implementing this algorithm on practical installation designs, including the rod method developed during a wider research project associated with this study.