CCWi2017: F150 'APPLICATION OF LEAST SQUARES SUPPORT VECTOR MACHINES AND ACOUSTIC MEASUREMENTS FOR LEAK FLOW RATE PREDICTION'

Abstract

Leakage from water distribution pipes represent a huge issue worldwide with economic. Leaks are normally found by placing sensors either side of the leak recording the leaks acoustic emission as it discharges the leak hole. As the leak noise is intrinsic to the leak, it contains information which could provide information about the leak, including the leak flow rate. Any tool which can accurately determine the leak’s flow rate could be used by water industry practitioners in order to prioritise leakage repair activities by repairing the higher leak flow rates first. This will result in economic savings through reduced water lost and better allocation of company resources. This paper demonstrates a small element of research undertaken at the University of Sheffield in collaboration with several UK water companies. The aim of the research is to develop a tool in order to predict leak flow rate using acoustic emission sensors. The research uses Least Squares- Support Vector Machines in order to predict leak flow rate in MDPE pipe using high quality data from a unique experimental pipe rig. The results demonstrate that there is sufficient information within the leak’s acoustic emission signal in order to predict leak flow rate. Therefore the research represented in this paper presents a tool which can be used by water industry practioners to prioritise leak repair.