Modelling the effects of mixing ratio and temperature on the thermal conductivity of GNP-Alumina hybrid nanofluids: A comparison of ANN, RSM, and linear regression methods

This research aimed to evaluate and compare the efficacy of three distinct methods for forecasting the thermal conductivity of GNP-Alumina hybrid nanofluids. The methods under consideration were artificial neural network (ANN), response surface methodology (RSM), and linear regression (LR). The predictive performance of the ANN model was investigated in relation to the number of neurons in the hidden layer. The findings revealed that the optimal number of neurons was 7, which produced the best performance with an overall mean square error (MSE) of 1.08E-06. The correlation coefficient was also high at 0.99799. The RSM approach involved testing linear, quadratic, cubic, and quartic models, with the quadratic model showing the highest predicted R² (0.9721) values, indicating that it provided the best fit to the data. Finally, the LR model was developed using a backward elimination approach, with temperature and volume fraction being the significant variables in the final model. Overall, the ANN model produced the most accurate predictions, followed by the RSM and LR models. These findings suggest that the ANN and RSM techniques can be effective tools for forecasting the thermal conductivity of nanofluids, and highlight the importance of selecting appropriate model parameters for optimal performance.