Analysis of the effect of roughness and concentration of Fe3O4/water nanofluid on the boiling heat transfer using the artificial neural network: An experimental and numerical study

Since experimental studies in the field of nanofluid pool boiling requires costly and time-consuming tests, numerical methods such as artificial neural networks with higher predictability and nonlinear features are suitable for prediction and modeling of problem parameters. In this paper, 180 pool boiling laboratory data of Fe3O4/water nanofluid are employed as datasets used for network training to determine the effect of different parameters of nanofluid pool boiling on Boiling Heat Transfer Coefficient (BHTC) and wall superheat. The concerned input parameters for the neural network include concentration, roughness, and heat flux, while the network outputs are the BHTC and wall superheat. Finally, it becomes clear that the trainbr training algorithm with the optimal quantity of 41 neurons within the hidden layer shows the best performance. In addition, the present model can accurately predict the BHTC and wall superheat with correlation coefficients (R) of 0.99936 and 0.9986 and the mean square error (mse) of 0.103 and 0.013, respectively. Also, given the optimization objectives considered in this research, including maximizing the heat transfer coefficient and minimizing the wall superheat in the nanofluid pool boiling process, the multi-objective genetic algorithm has been used to optimize the two objective functions concerned.

Automated summary

The study utilizes artificial neural networks to predict and model parameters in nanofluid pool boiling, achieving high accuracy and performance through training with experimental data.