Previsão de propriedades reológicas em alta temperatura de nanossuspensões de sílica via redes neurais profundas

Abstract

Machine learning enables rheological predictions even beyond the limits of conventional hightemperature instrumentation. This work investigates the impacts of adding silica nanoparticles to a Newtonian base fluid (mineral oil), with an emphasis on analyzing the rheological properties under varying temperature conditions. The incorporation of these nanoparticles leads to pronounced modifications in the fluid’s rheological behavior, which become more prominent as temperature and nanoparticle concentration increase. The study focuses on extrapolated regions where conventional instruments, such as rheometers, face measurement limitations.To this end, machine learning techniques are employed, with particular emphasis on deep learning approaches, applied to flow curve data obtained experimentally for silica nanoparticle volumetric concentrations of 0.25%, 0.5%, 1.0%, 1.5%, and 2.0%, under various thermal conditions: 4◦C, 15◦C, 50◦C, 75◦C, 100◦C, and 125◦C. The results demonstrate that the developed model exhibits high predictive capability and robustness against overfitting, indicating its effectiveness in generalizing the rheological behavior of nanofluids. Furthermore, the model can provide consistent estimates in regions not experimentally accessible, thus enhancing the utilization of laboratory data and supporting a deeper understanding of the effects of temperature and concentration on the system’s rheological response.