Supervised Learning Algorithm for Water Leakage Detection through the Pipelines

The water distribution system is an efficient, quick, cost-effective, and ecologically friendly method of providing water.Groundwater pipeline networks frequently lack planning as the population expands.Freshwater is often provided to end customers through a water supply network, which consists of subterranean and above-ground pipelines. The water distribution sector is increasingly concerned about leakage in water pipeline networks. This necessitates substantial development in leakage sensing technologies to both avoid and mitigate the potential of leakage. As a solution to this, in this study,a supervised learning algorithm-based water leakage detection through the pipeline idea is proposed. The dataset required for this study is collected in real-time using a hardware set. The hardware setup consists of a water storage tank that is connected with pipelines thatare directed toward the consumers. The pipeline is fitted with microphones that measure the sound of the water flow through the pipeline and the values are stored in a Data Acquisition (DAQ) module. The values in DAQ are further analyzed by the computer system that is integrated with the DAQ. The dataset is further pre-processed and dimensionally reduced to make them compatible with Machine Learning (ML) models. The ML models used in this study further evaluate and classify the data values and helps in detecting the water leakage in pipelines. The ML model Naïve Bayes used in this study shows an accuracy of 97.5% and has the highest accuracy among the three ML models used in this study and is concluded as the most efficient model among the other ML models.