Utilização de redes neurais artificiais na indústria química para determinação de propriedades em leitos fluidizados: uma revisão da literatura

Abstract

This study aimed to identify the main machine learning methods, with an emphasis on artificial neural networks, applied to determining different properties in fluidized beds through a systematic review. Additionally, it analyzed the challenges associated with the application of these algorithms and explored future perspectives in this research area, highlighting the potential of artificial neural networks in modeling hydrodynamic phenomena. With the increasing industrial and scientific demands, understanding complex systems, such as fluidized beds, is essential to optimize processes and reduce costs. In this context, artificial neural networks emerge as promising tools, capable of identifying complex patterns, making accurate predictions, and reducing the need for a detailed understanding of physical processes. As a research protocol, the ProKnow-C method was used, with the Science Direct database, resulting in the selection of seventy-eight relevant articles, of which twenty-two met the study's objectives. From these, eleven were chosen for the review. A particularly promising approach is the integration of these networks with computational fluid simulators, enabling detailed prediction of properties such as flow, heat transfer, and particle distribution, thus optimizing the design and operation of industrial processes. Despite their potential, AI-based methods face significant challenges related to the quality and quantity of available data. The systematic review demonstrated that the use of artificial neural networks has yielded promising results, especially in industrial applications. Combined with ongoing advancements in machine learning and high-performance computing, these integrated solutions are expected to be widely adopted, contributing to significant improvements in efficiency and innovation. Furthermore, they highlight new research opportunities, such as the development of more robust hybrid approaches to optimize different types of properties in fluidized beds.