Abstract

Today, 844 million humans around the world have no access to safe drinking water. Furthermore, every 90 seconds, one child dies from water-related illnesses. Major cities lose 15% - 50% of their water and, in some cases, losses may reach up to 70%, mostly due to leaks. Therefore, it is paramount to preserve water as an invaluable resource through water networks, particularly in large cities in which leak repair may cause disruption. Municipalities usually tackle leak problems using various detection systems and technologies, often long after leaks occur; however, such efforts are not enough to detect leaks at early stages. Therefore, the main objectives of the present research are to develop and validate a leak detection system and to optimize leak repair prioritization.
The development of the leak detection models goes through several phases: (1) technology and device selection, (2) experimental work, (3) signal analysis, (4) selection of parameters, (5) machine learning model development and (6) validation of developed models. To detect leaks, vibration signals are collected through a variety of controlled experiments on PVC and ductile iron pipelines using wireless accelerometers, i.e., micro-electronic mechanical sensors (MEMS). The signals are analyzed to pinpoint leaks in water pipelines. Similarly, acoustic signals are collected from a pilot project in the city of Montreal, using noise loggers as another detection technology. The collected signals are also analyzed to detect and pinpoint the leaks. The leak detection system has presented promising results using both technologies. The developed MEMS model is capable of accurately pinpointing leaks within 12 centimeters from the exact location. Comparatively, for noise loggers, the developed model can detect the exact leak location within a 25-cm radius for an actual leak.
The leak repair prioritization model uses two optimization techniques: (1) a well-known genetic algorithm and (2) a newly innovative Lazy Serpent Algorithm that is developed in the present research. The Lazy Serpent Algorithm has proved capable of surpassing the genetic algorithm in determining a more optimal schedule using much less computation time. The developed research proves that automated real-time leak detection is possible and can help governments save water resource and funds. The developed research proves the viability of accelerometers as a standalone leak detection technology and opens the door for further research and experimentations. The leak detection system model helps municipalities and water resource agencies rapidly detect leaks when they occur in real-time. The developed pinpointing models facilitate the leak repair process by precisely determine the leak location where the repair works should be conducted. The Lazy Serpent Algorithm helps municipalities better distribute their resources to maximize their desired benefits.