8

Neural networks evaluate pipeline properties
Award-winning research contributes to the understanding of reinforced plastics production processes

MANUEL ALVES FILHO

Professor Liliane Lona (on the left) and the author of the dissertation, Sheila Contant: results similar to those obtained by the industry

Research developed for Sheila Contant's master's thesis, defended at the Faculty of Chemical Engineering (FEQ) at Unicamp, brought new contributions to the understanding of one of the main production processes of polymer composites (reinforced plastics), filament winding (in Portuguese , filament winding or continuous filament). The work, which was ranked third in the fifth edition of the “Petrobrás Pipeline Technologies Award”, used neural networks, a computational technique inspired by the behavior of the human brain, to evaluate the mechanical and thermal properties of the final product (pipelines ), as well as the temperature profile inside the parts during manufacturing, a stage called “curing”. According to the author of the study, the method proved to be efficient, presenting results similar to those obtained experimentally by the industry.

According to Professor Liliane Lona, who guided the research, a better understanding of the reinforced plastics production process directly contributes to reducing the time and cost of its development. Just to get an idea of ​​what this could represent, it is enough to know that the use of polymeric compounds has grown significantly in recent decades, due to the series of advantages they present over other materials, especially metals. In addition to being lighter and more resistant to corrosion than conventional products, they are also durable and have excellent mechanical properties.

“The resistance of these materials to corrosion, for example, is of fundamental importance for the oil industry, which often faces leakage problems due to wear and tear on metal pipelines”, says the professor. The trend, according to her, is for parts produced from polymer composites to gradually replace current pipes, as they provide greater safety. The author of the dissertation recalls that the literature mentions that reinforced plastics can have around 40 thousand different applications, ranging from the manufacture of water tanks to chairs, obviously including pipes.

Sheila also draws attention to the fact that South America only represents around 3% of the global polymer composites market. However, the annual growth of this sector in the region has been the largest in the world in recent years, which demonstrates that there is an immense market for these materials, as well as space for the development of new studies around them. “As Brazil is the market leader in South America, there is a clear need for the country to increase its competitiveness, especially through research and improvement of human resources”, analyzes the author of the dissertation.

Neural networks – Sheila explains that she chose the filament winding method to investigate, as it is one of the main composite production processes. Highly resistant plastic is nothing more than a polymer matrix, reinforced by a fiber. The option for neural networks was due to the advantage it presents over other computational models, such as shorter processing time and the ability to apply it to highly complex processes. The author of the work analyzed two aspects of the parts produced by the filament winding or continuous filament system.

First, she checked the properties of the final product to verify its uniformity. Afterwards, the thermal behavior of the parts was monitored. To this end, Sheila developed computational programs for training neural networks. The computational tool was supplied by data provided by a company in the sector. When comparing the experimental results with those obtained by the simulation, the researcher found that they were very close to each other. “The results showed the efficiency of the proposed methodology in all cases studied”, she assures.

This is not the first time that Sheila has been awarded the “Petrobrás Pipeline Technology Award”. In 1999, she was ranked second in the undergraduate category. “She was the only person to receive two awards in the five editions of the event”, highlights professor Liliane, who supervised both works. “I consider this award an important incentive for the development of new national technologies”, says Sheila. In her doctorate, which is already underway, the postgraduate student will take on another challenge, this time working in the area of ​​polymerization engineering.

Top