Water leak detection system using machine learning

• Grant number: 15/01100-4
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: July 01, 2016
• End date: March 31, 2017
• Field of knowledge: Engineering - Sanitary Engineering - Basic Sanitation
• Principal Investigator: Antonio Carlos Oliveira Júnior
• Grantee: Antonio Carlos Oliveira Júnior
• Company: Stattus4 Cidades Inteligentes e Sustentabilidade Ltda
• Associated researchers: Anderson Fraiha Machado
• Associated research grant(s): 17/00798-3 - Automatic leak detection system in water pipes network, AP.PIPE
• Associated scholarship(s): 16/12288-7 - Water leak detection system using machine learning, BP.PIPE

Abstract

With the growth and development of humanity, mankind has to deal with important challenges to their survival. One of them is how to supply the cities with drinking water in a sustainable manner. The demand for drinking water has grown worldwide. However, periods of drought, pollution of water sources and water use for non-consumptive purposes (eg. Electricity generation), end up impacting negatively in the volume of available drinking water of the population. Thus, control of water losses in the distribution network becomes crucial. However, what we see in Brazil is currently a very considerable loss level. According to ABES (Brazilian Association of Sanitary and Environmental Engineering) on average 40% of the captured volume of water is lost in the system, some sanitation companies that loss may reach 60%. Keeping this in view, this research aims to develop a leak detection system in water extensions. From the vibration data of building extensions, more precisely collected in racks measuring water consumption, will be extracted the best transcribers to characterize the acoustic vector and subjected to different methods of clustering and choose the more robust a posteriori. Seeking greater efficiency in results when compared to the techniques used by recent research, homeomorphic processing techniques will be incorporated in addition to the expanded exploration classifiers - clusters. As a result of this research is expected to have an efficient algorithm for identifying leaks from the vibration monitoring. With this result in hand, we intend to plead the PIPE Phase 2 so that it can develop a second stage of research (types of sensors and geolocation of leaks), to then get to the final commercial product. (AU)