Water leak detection system uses machine learning
April 04, 2018
By Suzel Tunes | FAPESP Research for Innovation – Almost 40 of every 100 liters of water captured by Brazil’s supply networks is lost in distribution, according to statistics for 2016 from the National Sanitation Information System (SNIS). Stattus4, a startup located in the city of Sorocaba, São Paulo State, Brazil, proposes to reduce this waste by means of a leak detection tool.
“Eliminating leakage is most important from the standpoint of water resource management. Reducing waste also means reducing the need to capture and treat water, which results in lower operating costs for companies and enhanced protection of water sources,” says Antônio Carlos Oliveira Júnior, the principal investigator for the project.
Christened Fluid, the technology uses artificial intelligence to detect areas with potential leaks. A prototype has been developed with the support of FAPESP’s Innovative Research in Small Business Program (PIPE). It comprises a mobile sensor that records vibrations in the water flowing through the pipes and analyzes them with reference to a database stored in the cloud.
“Our approach is based on musical computing,” Oliveira explains. “The way Fluid works is comparable to the functioning of smartphone music recognition apps, which can identify a song from a small sample by searching a database.”
In the case of the leak sensor, different sounds captured from the water distribution network may point to water loss, a problem with a water meter, or even a clandestine connection. The software can make these distinctions. The more it analyzes sounds, the more robust its database becomes, and the more accurately it can diagnose the problem. “It’s what we call reinforcement learning,” Oliveira says.
A pilot of the Fluid system is in use by some ten companies. “Eight hundred new samples are entering the system per day. We already have over 40,000 real sound datasets,” he says. The first parameters fed into the algorithm were obtained from a school that trains operators of geophones, the instruments traditionally used to detect leaks.
Geophone operators are highly specialized in identifying leaks, usually in two stages. After scanning the city’s streets, they embark on a search for the exact location of a leak. The geophone is used to capture sounds, but the operator’s trained ear must analyze them.
Oliveira points out that he has no intention of replacing the work done by the geophone operator. The idea is to speed it up by enabling the street scanning stage to be skipped. “Fluid identifies areas with potential leaks and leaves the fine-tuning to the geophone,” he says.
In Votorantim, a city near Sorocaba, the private company that holds the basic sanitation concession, Águas de Votorantim, is already using this methodology. “They had just two geophone operators, who used to take 24 months to cover the entire concession area,” Oliveira says. “With three data collectors that perform a prior survey of potential water loss points, they shortened the time to four months. Now, the geophone operator visits only places where a problem has been detected. Fluid is easy to operate and doesn’t require a specialized professional, so cost reduction is guaranteed.”
The challenge his firm faces now is to advance toward pricing the system correctly. “Fluid isn’t just a device – it’s a management tool,” he says. In addition to detecting leaks, it enables management to monitor street surveys, often done by contractors, in real time. “Managers can check the data collector operator’s route and access sample audios to audit the efficacy of the system. Plus, they receive reports on the field data.”
Oliveira has a degree in electrical engineering and is working for a second in applied and computational mathematics, both from the University of São Paulo (USP). He is using his knowledge of both areas to develop the project.
The project’s initial goal, he explains, was to couple a sensor to every water meter and to be capable of locating every leak accurately, but the cost proved unviable. The decision to produce a mobile collector slashed the hardware investment required and was a key factor in enabling the firm to enter the market. “We were quick to make the decision. We used the lean startup concept, validating each move with potential customers and adapting with agility to the market,” says Marília Lara, co-founding partner and managing director.
Good acceptance of the product and recognition by the scientific community show that the firm is on the right track. Having begun at the Center for Innovation, Entrepreneurship & Technology (CIETEC), USP’s incubator, Stattus4 now has its own offices in Sorocaba with support from the city’s Technology Park and the Campinas-based startup accelerator Baita Aceleradora. It has undergone three rounds of angel investment and currently has 12 employees.
The firm won the Innovative Startup award at the 47th Conference of the National Association of Municipal Sanitation Service Providers (ASSEMAE) in June 2017 and has been booked to participate in many academic and corporate events.
On March 12, 2018, Marília Lara presented the Fluid project during the ILP-FAPESP Science & Innovation Cycle, an event organized by the São Paulo State Legislative Assembly Institute (ILP) in partnership with FAPESP.
On March 15, Stattus4 took part in an event hosted by the International Finance Corporation (IFC), the World Bank’s private-sector lending arm, and the World Economic Forum. It was one of 50 Latin American startups selected to participate in the Forum’s UpLink initiative, a digital community established to help startups interact with global peers, multinationals, governments, universities and investors. In addition, on March 22-19, Stattus4 participated in the 2018 World Water Forum in Brasília.
According to Marília Lara, winning contracts for use of the system and thinking about internationalization are the next steps. “Our target market consists of water supply concessionaires and their contractors. It’s a limited market. We see internationalization as a sine qua non for our existence,” she says.
Agência FAPESP licenses news reports under Creative Commons license CC-BY-NC-ND so that they can be republished free of charge and in a straightforward manner by other digital media or by print media. The name of the author or reporter (when applied) must be cited, as must the source (Agência FAPESP). Using the button HTML below ensures compliance with the rules described in Agência FAPESP’s Digital Content Republication Policy.