Thermal conductivity modeling of nanofluids with ZnO particles by using approaches based on artificial neural network and MARS

Nanofluids are attractive alternatives for the current heat transfer fluids due to their remarkably higher thermal conductivity which leads to the improved thermal performance. Nanofluids are applicable in porous media for improving their heat transfer. Proposing accurate models for forecasting this feature of nanofluids can facilitate and accelerate the design and modeling of nanofluids' thermal mediums with porous structure. In the present study, three methods including MARS, artificial neural network (ANN) with Levenberg-Marquardt for training and GMDH are employed for thermal conductivity of the nanofluids containing ZnO particles. The confidence of the models is compared according to various criteria. It is observed that the most accurate model is obtained by using ANN with Levenberg-Marquardt followed by GMDH and MARS. R-2 of the mentioned models are 0.9987, 0.9980 and 0.9879, respectively. Finally, sensitivity analysis is performed to find the importance of the input variables and it is concluded that the thermal conductivity of the base fluids has the highest importance followed by volume fraction of solid phase, size of particles and temperature.

Automated summary

Nanofluids with high thermal conductivity are used to improve heat transfer in porous media. In this study, different methods were used to model the thermal conductivity of nanofluids containing ZnO particles, with artificial neural network (ANN) proving to be the most effective.