A Novel Deep Learning Method for Water leakage Detection Using Acoustic Features

Conservation of water is one of the major objectives for any country around the world. Water management plays a very important role in a society, as it is one of the basic needs for the mankind. Water leak detection is an important task for ensuring the safety and efficiency of water systems. Deep learning techniques, combined with real-time sensor data, can greatly enhance the accuracy and efficiency of water leak detection. Water leak detection plays a crucial role in maintaining the safety, efficiency, and sustainability of water systems. This paper presents an investigation of the capacity of deep learning methods (DL) to localize leakage in water distribution systems (WDS). Progress in real-time monitoring of WDS and DL has inspired towards new opportunities to develop data-based methods for water leak localization. However, the handlers of WDS need recommendations for the selection of the optimal DL methods as well their practical use for leakage localization. This paper contributes to this issue through an investigation of the capacity of DL methods to localize leakage in WDS. Water leak detection is crucial in conversation of water, cost saving, infrastructure protection prevention of water damage and operation efficiency. This paper shows a proposal for a system based on a wireless sensor network designed to monitor water distribution systems, such as hotel industry, irrigation systems, which, with the help of an autonomous learning algorithm, allows for precise location of water leaks. Autoencoder neural network (AE), an unsupervised DL model, is further developed to detect leak with unbalanced data. The results show AE-DL model achieved high accuracy when leaks occur in pipes inside the sensor monitoring area, while the accuracy is compromised otherwise. This observation will provide guidelines to deploy monitoring sensors to cover the desired monitoring area. A novel strategy is proposed based on multiple independent detection attempts to further increase the reliability of leak detection by the AE-DL and is found to significantly reduce the probability of false alarm.