Experimental energy and exergy assessments of a mechanical vapor compression refrigeration system having low-GWP alternative gas of pure R1234yf and its nanoparticle mixtures using artificial intelligence models.

Concerns about the effects of current refrigerants on global warming and ozone depletion have accelerated the development of alternatives. In the current work, two different Artificial Neural Network (ANN) structures were established to estimate the parameters Coefficient of Performance (COPR), exergy efficiency (ηex), isentropic efficiency (ηisen) and total exergy destruction (Exdest_Total) as outputs based on the experimental data of the alternative refrigerant R1234yf containing aluminum oxide (Al2O3), graphene, and Carbon Nanotubes (CNTs) nanoparticles instead of the conventional refrigerant R134a in a mechanical Vapor Compression Refrigeration System (VCRS). In experimental cases, the use of pure R1234yf, R1234yf+Al2O3, R1234yf+graphene, and R1234yf+ CNTs instead of R134a was experimentally investigated with energy and exergy approaches at same system. In network topologies, 70 % of all data points were utilized for training, 15 % for validation, and 15 % for testing. Finally, the Levenberg-Marquardt learning algorithms with measured and calculated values as input parameters were assessed as the training ones in Multilayer Perceptron (MLP) models. The coefficients of determination were greater than 0.99. The average deviations were smaller than 0.01 %. The high-accuracy predictions enable rapid performance optimization of R1234yf-based nanorefrigerants, providing manufacturers with a validated tool to comply with impending refrigerant regulations while minimizing experimental costs and system redesign efforts. This AI framework bridges the gap between nanoparticle-enhanced thermodynamics and industrial deploy ability.