Data-Driven Thermal Performance Prediction of ZnO/W:EG Nanofluid Radiators for Fuel Cell cooling

Nanofluids have demonstrated significant potential for enhancing heat transfer in various thermal engineering applications, owing to their superior thermophysical properties compared to conventional coolants. To accurately predict the thermohydraulic performance of radiators using nanofluids, this study proposes an advanced supervised machine learning (ML) approach. In particular, the influence of adding ZnO nanoparticles to a base fluid is investigated using predictive modeling to evaluate its impact on heat transfer characteristics in radiator. Four ML algorithms including Lasso Regressor (LR), Support Vector Regression (SVR), Gradient Boosting Regressor (GBR), and Random Forest Regressor (RFR) were applied to forecast the thermohydraulic efficiency of radiator; instead of the need for expensive laboratory testing. The input parameters were the coolant input temperature, coolant flowrate, and concentration, while the outputs were the outlet temperatures of nanofluids and air to calculate the heat transfer rate and efficiency of radiator. The inlet temperature and flow rate of air are considered as constant. The results demonstrated that LR model is the best for predicting the nanofluids outlet temperatures, achieving 0.9996 R2 with the lowest errors (RMSE=0.2037, MAE=0.1792) and minimal bias COV=0.00315, CRM =0.00116), outperforming the other models. Regarding the air outlet temperature, SVR performed slightly better (R2 0.9939) with LR close behind (R2=0.9934) and similarly low bias. Accordingly, LR is a strong candidate for modeling and prediction of the thermal performance of radiators using the ZnO/W:EG nanofluids.