ABSTRACT
This chapter reviews methods for discovering the size and location of leaks in water distribution systems. Detecting and localizing leaks can be performed by observing the leak influences and using the information to estimate the location and size of the leaks. Machine learning for leak quantification can be categorized into two methods: optimization and classification algorithms. Optimization algorithms have similar principles to mathematical modelling techniques; but with lower computational cost. Classification algorithms use a training model that contains classes to identify which class an observation profile belongs to. The chapter reviews techniques to solve leak quantification using machine learning algorithms, sensor placement strategies, and sectorization of a large water distribution system (WDS). The optimization algorithms work by comparing all possible sensor places and choosing positions with the highest detection and localization abilities. However, conventional optimization techniques, such as brute force, have a high computational cost.
