High-precision prediction model for supercritical co₂ flow and heat transfer in thermal control systems for mars exploration missions

Supercritical carbon dioxide (s-CO₂) plays a crucial role in the cooling of advanced aerospace engines. To this end, this study comparatively analyzed the performance of eight machine learning (ML) models in predicting the heat transfer characteristics of s-CO₂ flow within tubes. Additionally, the influence of different boundary conditions on flow heat transfer performance was investigated. Results indicate that operating parameters significantly influence the heat transfer coefficient( h ) and pressure drop( ΔP ). Variations in the mass flux affect both the h and ΔP by altering turbulence intensity and molecular friction. Higher heat flux leads to a thinner boundary layer. The boundary layer thickness affects thermal resistance, thereby influencing heat transfer performance. Notably, the trained ML models demonstrate high accuracy in predicting h and ΔP values, with linear correlation coefficients ( R² ) exceeding 0.98. More than 98% of the predicted h and ΔP values for the test set fall within a ± 2% error margin. The prediction time of the ML models is significantly shorter than that of computational fluid dynamics simulations, with a time difference of 41.23 times. The ML model successfully predicts the impact of inlet temperature, mass flux, and wall heat flux on flow and heat transfer. The optimal regression model is identified as Random Forest, followed by Ridge and Lasso regression. The proposed ML model offers an effective method for analyzing and predicting s-CO₂ flow and heat transfer.