Performance evaluation of various training functions using ANN to predict the thermal conductivity of EG/water-based GNP/CNC hybrid nanofluid for heat transfer application

Thermal management efficiency is still a significant problem in many industries and techniques due to the ultimate limitations in the performance of conventional heat transfer fluids. The present research focuses on predicting the thermophysical properties of hybrid graphene nanoplatelet (GNP) and cellulose nanocrystal (CNC) nanoparticles to improve the thermal performance of heat transfer systems. Resolving the thermal management issues can be critical for saving energy, enhancing the effectiveness of the systems, and advancing the existing and emerging technologies needed to handle high temperatures. GNP-CNC/ethylene glycol–water hybrid nanofluids were prepared in volume concentrations from 0.01 to 0.2%. Thermal conductivity was measured from 30 to 80 °C, providing comprehensive data for analysis. The most important resolution was formulated at 0.1% volume concentration within a 60:40 volume ratio of ethylene glycol and water, with UV–Vis analysis showing absorption peaks in the highest order at 0.10% and 0.2% concentrations. Thermogravimetric analysis has shown an increase towards thermal resilience, with the mass decline beginning at 130 °C and full degradation at 500 °C. An interesting observation was invested for 0.20% GNP: CNC, where the onset of degradation occurred at 150 °C, providing an increased variety of potential high temperatures. An artificial neural network (ANN) model was implemented to predict thermal conductivity, and 15 training functions were examined for the ANN structure. The model's best prediction results were obtained by utilizing tansig and Purlin transfer functions in a single hidden layer with ten neurons, which employed the Bayesian regularization function. It reached R² = 99.99%, MSE = 4.8352 × 10⁻⁷, and RMSE = 1.2083 × 10⁻³, which is superior to other functions, e.g. trainlm. The novelty is successfully synthesizing a stable GNP-CNC hybrid nanofluid with excellent thermophysical properties and establishing a highly accurate predictive model. The impact could be widespread in various industries, from better cooling to more efficient energy systems, and even the applicability of this effect in improving industrial processes.