System developed by researchers increases the productivity of industrial processes

The company I.Systems is the first to be funded by Natura and Banco Itaú’s investment fund.

By Elton Alisson

Agência FAPESP – A company founded by four graduates of Universidade Estadual de Campinas (UNICAMP) with degrees in Computer Engineering and Mathematics—has been chosen to receive the first capital injection from the Pitanga Venture Capital fund. The fund was created by the founders of Natura and Itaú Unibanco and is managed by biologist Fernando Reinach, who is the former director of Votorantim Novos Negócios.

After analyzing nearly 700 projects over a period of a year and a half, the fund managers decided to make their first investment in the company I.Systems, which develops software for industrial automation.

“They developed a technology that seems very innovative and promising,” says Reinach. The technology in question is advanced control systems software for industrial processes, which is based on fuzzy logic, to increase the efficiency of production lines in large industries. The company’s development efforts were funded by FAPESP’s Innovative Research in Small Business (PIPE).

Expansion of the mathematical theory of binary sets and fuzzy logic enables the control of a greater number of variables in a production line—for example, on an intermediary scale between 0 and 1—than does classic binary logic, which is restricted to exact values of the variables 0 and 1. As a result, fuzzy logic is utilized in image recognition, air conditioners, ABS brakes and photographic cameras—to automatically adjust the focus according to the distance of an object or the environmental light conditions.

“This expansion of mathematical theory has generated major benefits in the area of mathematical modeling,” comments Igor Bittencourt Santiago, the president of I.Systems.

According to Santiago, one of the limitations of applying fuzzy logic in control systems is the mathematical modeling of the system’s functional rules and the ability to manually synchronize such rules—which is laborious and expensive—or training an artificial neural network (a computer technique that presents a mathematical model inspired by the neural structure of intelligent organisms) to absorb the characteristics of the process.

Mathematical modeling of a system, however, is very complex because the behavior of all of the variables, and the relationships among them, must be established. Once simplified, the mathematical model does not utilize its full potential and requires the presence of an operator to intervene in the process.

One of the obstacles encountered in training an artificial neural network is the uncertainty in what the system is learning despite a complete understanding of the characteristics of a process. “If there is some point outside the training zone of an artificial neural network, it can present very unstable behaviors in the process,” explains Santiago.

To overcome these technical obstacles and to enable a broader use of fuzzy logic in control systems, Santiago and his partners at I.Systems—Leonardo Freitas, Ronaldo Silva and Danilo Halla—began developing algorithms to automatically generate the functional rules of a process.

The four partners produced software that runs on a Windows operating system, collects very simple information from an industrial process—classified at three levels (minimum, average and maximum)—and instantly generates all of the correlations among these variables.

The software, known as Leaf, increases stability, reduces costs and improves the productivity of automated processes. “We made systems control with fuzzy logic much cheaper and usable by people who have no knowledge about it or even knowledge of the industrial process that it intends to automate,” says Santiago.

In 2009, the research company PBM Industrial Automation initiated a Phase 1 PIPE project that was funded by FAPESP to conduct a software feasibility test.

To conduct the test, Santiago sought the assistance of the industrial manager of the Femsa Coca-Cola factory in Jundiaí—the largest bottler of Coca-Cola in Latin America in terms of production volume. The manager described one of his technological challenges in the bottling lines. Small variations in the pressure used to fill bottles with soda can affect the filling speed and cause beverage loss, because of bubbling and variations in the levels of injected liquid.

The system, which was implemented by PBM at the factory, enabled simultaneous control of the pressure valves of the bottling line and generated a saving of 500,000 liters of Coca-Cola and 100,000 PET bottles per year.

In addition to the Coca-Cola plant, the system is currently employed by large companies, such as Rhodia, Ajinomoto and Lanxess, and sugar and bioenergy mills, Tonon, São Martinho and Pederneiras.

“In order for our solution to be economically feasible, there must be a certain production scale. In smaller production units, the return on investment becomes much longer,” says Santiago. However, he stresses that the company has not ruled out the possibility of developing a solution for small-scale companies.

In sugar and bioenergy mills, the system is employed to establish optimal boiler vapor generation. By adding this function to energy generation, the mill can generate more energy with the same quantity of biomass that was utilized previously.

“Because this software executes multivariable control of the process, it manages to have more global information on the boiler state, for example, and to control the generation of energy,” explains Santiago.

The next steps

In 2010, the company initiated a Phase 2 PIPE Project, in which the researchers improved the performance of the Leaf system and developed additional software for industrial equipment.

“Now we are working toward installing the software in equipment and initiating the sales process,” explained Santiago. According to Santiago, the investment made by the Pitanga fund will help to improve the sales team, as well as the development of new products.

In the meantime, I.Systems does not have competitors in the Brazilian market. In international markets, where it seeks partners to distribute products, it will encounter competition from large companies, such as Siemens and General Electric.

“We are evaluating whether to patent our technology in Brazil and abroad and if we will work with industrial secrets on the U.S., Asian and European market,” says Santiago.

In the short term, the company intends to secure Pitanga as its sole partner. According to the researcher, the possibility of attracting other investors in the long-term has not been rejected.

The Pitanga fund intends to invest in twelve additional innovative companies in the future. Future investment volume may reach R$ 100 million.

“We are a classic venture capital fund. For this reason, we are seeking companies that have a very different business model or that employ innovative technology and have the potential to grow a lot,” says Reinach.

Reinach was one of the idealizers of the FAPESP Genome Project, which completed sequencing of the Xylella fastidiosa bacterium—the cause of the citrus variegated chlorosis (CVC) disease—in 2000. In 2000, Reinach was a full professor at the Chemistry Institute of the Universidade de São Paulo.