Evolutionary optimization of thermo-physical properties of MWCNT-Fe₃O₄/water hybrid nanofluid using least-squares support vector regression-based models[Formula presented]

Decisions on optimizing design and operating parameters are challenging when using hybrid nanofluids (HNFs). A procedure is proposed and implemented for predicting and optimizing the thermal conductivity and dynamic viscosity of MWCNT-Fe₃O₄/water HNF. The procedure involves using precise least-squares support vector regression (LSSVR) models, multi-objective genetic optimization of thermal properties, and automated selection of optimal design conditions. Tuned parameters are the volume fractions of nanoparticles and the operating temperature. The cross-validated and carefully optimized LSSVR models for thermal conductivity and dynamic viscosity showed excellent performances, with testing mean percentage errors of −0.246 and −0.103%, and relative root mean square errors of 1.325 and 2.165%, respectively. By assigning equal importance to the two response parameters, an HNF with volume fractions of 0.302 (Fe₃O₄) and 0.183% (MWCNT), operating at 55 °C, is highlighted as the optimal design within the considered range of tuned parameters. This corresponds to a particle mixing proportion (PMP) of 0.605, and the corresponding values of thermal conductivity and dynamic viscosity are 0.803 W/m K and 0.625 mPa s, respectively.