Numerical analysis and artificial neural network-based prediction of two-phase flow pressure drop of refrigerants in T-junction

Abstract

The two-phase flow behaviors in T-junction are quite complex in energy transport systems. In this paper, the two-phase flow pressure drop of refrigerants in a horizontal branching T-junction was analyzed numerically and predicted using artificial neural network. Firstly, the distribution of static and total pressure was obtained based on Eulerian method, and the parametric studies on the local pressure drop were conducted. It is observed that the vortexes in the entrance of branch pipe lead to the pressure fluctuation and irreversible pressure losses, and the “descend-ascend” of static and total pressure happens under high mass flow split ratio in run pipe. Then, the ANN predicting model of local pressure drop coefficients was established. It shows that GA-BPNN and PSO-BPNN has the best predicting ability for K12J and K13J respectively, and the relative errors are within 10% for most cases. Finally, the sensitivity analysis was conducted, indicating that the effect of mass flow split ratio (F) and inlet quality (x1) is the most significant for K12J and K13J respectively.