Three artificial neural network (ANN) models were developed to predict the fluid properties of China RP3 kerosene under supercritical pressure in replacement of the time-consuming property calculations by the principle of Extended Corresponding State (ECS). The analysis shows that the properties predicted by the trained ANN models agree well with the calculations by the ECS method. The correlation coefficients (R) between the ANN predictions and the ECS calculations are higher than 0.99, and most of the relative errors are lower than 0.1%. The prediction by the ANN models is of several orders (104) faster than that by the ECS method, especially near the critical points. The trained ANN model was further coupled with the CFD modeling of a realistic kerosene jet, where high efficiency and satisfactory accuracy were shown compared with the direct ECS calculations.