Model-based neural network correlation for refrigerant mass flow rates through adiabatic capillary tubes

Abstract

A capillary tube is a common expansion device widely used in small-scale refrigeration and air-conditioning systems. A generalized correlation of refrigerant mass flow rate through adiabatic capillary tubes covering both subcooled and two-phase inlet conditions is expected for multiple purposes. Based on the homogeneous equilibrium flow model, a new group of dimensionless parameters has been proposed. To express the nonlinear relationship between the mass flow rate and the associated parameters, the multi-layer perceptron neural network is employed as a universal function approximator. Simulated data from a validated homogeneous equilibrium model are used for the neural network training and testing. A 5-6-1 network trained with the simulated data of R600a and R407C shows good generality in predicting the simulated data of R12, R22, R134a, R290, R410A, and R404A. Also, the deviations between the trained neural network and the experimental data from the open literature fall into ±10%.