Abstract This investigation diverges from traditional studies concentrating on single-component nanofluids, instead examining the thermophysical benefits of hybrid nanofluids, like Al₂O₃/MWCNT, aimed at improving thermal conductivity and heat transfer efficiency in passive systems. Machine learning is a promising solution for designing efficient heat exchangers by understanding intricate relationships and utilizing suitable modelling techniques. Numerical simulations were conducted to validate the benchmark results; later, passive techniques were incorporated into the numerical model to predict the heat transfer characteristics. The dataset derived from numerical simulation results is employed to train contemporary machine learning methodologies, including support vector regression (SVR), decision tree (DT), and random forest (RF). Data from experiments and CFD analysis were gathered for preprocessing and machine learning (ML) analysis. The preprocessing phase involved the application of a standard scaler operation to enhance accuracy levels. The models underwent validation using ten experimental data samples to assess their performance against statistical tool metrics. A higher thermal performance factor (ThPF) is observed with the divergent nozzle insert in the plain tube at 0.028 vol% of MWCNT/Al2O3 (HNF3) at Reynolds number 3093. R2 values of SVR, DT and RF are predicted as 0.95, 0.98 and 0.99, respectively, for the case of HNF3 fluid flowing through the divergent nozzle insert. The investigation broadens its conclusions to include improvements in passive heat transfer, encompassing extended surfaces and geometric alterations, offering practical guidance for developing advanced heat exchanger designs.
Read full abstract