Prediction of leak flow rate in plastic water distribution pipes using vibro-acoustic measurements

Leakage from water distribution systems is a worldwide issue with consequences including loss of revenue, health and environmental concerns. Leaks have typically been found through leak noise correlation by placing sensors either side of the leak and recording and analysing its vibro-acoustic emission. While this method is widely used to identify the location of the leak, the sensors also record data that could be related to the leak’s flow rate, yet no reliable method exists to predict leak flow rate in water distribution pipes using vibro-acoustic emission. The aim of this research is to predict leak flow rate in medium-density polyethylene pipe using vibro-acoustic emission signals. A novel experimental methodology is presented whereby circular holes of four sizes are tested at several leak flow rates. Following the derivation of a number of features, least squares support vector machines are used in order to predict leak flow rate. The results show a strong correlation highlighting the potential of this technique as a rapid and practical tool for water companies to assess and prioritise leak repair.