Developing the machine learning models to estimate thermodynamic and transport properties of refrigerants

Abstract

Listen

Translate article

ABSTRACT The thermophysical properties of refrigerating systems should be accurately understood for designing low-temperature refrigeration cycles of economic acceptance. The thermal conductivity, density, velocity of sound, entropy, and enthalpy of various refrigerating systems from four differing classes, namely halocarbon, inorganic, hydrocarbon, and cryogenic fluids, were investigated here by the use of Multivariate Nonlinear Regression (MNR), Genetic Programming (GP), and Particle Swarm Optimization-Adaptive Neuro-Fuzzy Inference System (PSO-ANFIS). The development of a new and simple correlation was for the first time introduced to estimate saturated thermodynamic and transport properties of refrigerants without having in-depth knowledge on complicated parameters. Our research demonstrates that the PSO-ANFIS model is representative of an outstanding alternate for estimating the thermodynamic and transport properties of various refrigerating systems with a proper precision because Absolute Average Relative Errors (%AARD) for liqid and vapor thermal conductivity, liquid density, velocity of sound, entropy, and enthalpy were estimated as 3.0651, 7.9934, 1.0681, 1.2155, 1.8603, and 2.4835. This model is generally capable of estimating the thermodynamic and transport properties of various refrigerants with a superior accordance with data obtained experimentally.