Abstract
The importance of maintenance activities for improving the quality of water sources and guaranteeing a steady supply of water has increased significantly because of current social concerns. Water supply pipe corrosion is an issue that can cause leaks and lower water quality. The identification of hydraulic anomalies in water pumping systems is the subject of this project. A database was created of data acquired from a water supply network with pipes of various lengths and sizes. In hydraulic systems, sensor meters are mounted at various sites with distinct physical features, pipe sizes, and vital supply points. The input parameters used for a model are the sensor parameters, and the model analyzes the correlation between the input parameters (sensors) and determines which parameters are the most important, deciding on the output of the model, and thereby building the simplest model, which requires the least input parameters and gives the most accurate prediction results. In this project, using on the input signal from the sensors, the k-nearest neighbors machine learning algorithm was used to correlate/predict whether the pump was shut down (broken) for a certain period of time.