Turbulent convective heat transfer enhancement modeling of water-Al2O3 nanofluid using CFD mixture model and adaptive neural fuzzy inference system

• Main Authors: Al Mahmud, Suaib, Rahman Khan, Mazbahur, Ibne Noor, Wazed, Ismail, Ahmad Faris, Momin, Md. Abdul, Bappy, Jamirul Habib
• Format: Article
• Language: English
• Published: Taylor & Francis 2022
• Subjects: TJ Mechanical engineering and machinery; TP155 Chemical engineering
• Online Access: http://irep.iium.edu.my/101875/1/101875_Turbulent%20convective%20heat%20transfer%20enhancement.pdf; http://irep.iium.edu.my/101875/; https://www.tandfonline.com/doi/full/10.1080/10407790.2022.2149197; https://doi.org/10.1080/10407790.2022.2149197

ANFIS and multiphase Mixture model, two different methods from two very distinct engineering domains: Machine Intelligence and Computational Mechanics, have been put into a good amount of use over the past few years for modeling of nanofluids heat transfer enhancement. However, not only the previous investigations using both of these approaches suffer from the use of narrow range of nanofluid and flow properties, but also never have these two particular approaches been juxtaposed to point at a superior approach for specific flow regimes for predicting nanofluids heat transfer enhancement. In this study, water-Al2O3 nanofluid has been simulated using CFD multiphase Mixture model and ANFIS in order to assess the precision of both approaches to predict heat transfer enhancement of water-Al2O3 nanofluid for a very wide range of nanofluid configurations and flow properties, and to suggest the better approach for prediction of heat transfer enhancement for each specific flow regime. The results suggest that almost in every single case ANFIS is able to predict the heat transfer enhancement of nanofluids very efficiently with a maximum error of 0.35%, but the Mixture models’ predictions deviate significantly from the experimental correlation in some cases, though for intermediate nanofluid configurations, yielded results by Mixture model could be reliable with error around 1%.