Abstract
<title>Abstract</title> <p> This study presents a comprehensive thermodynamic and artificial intelligence (AI)-assisted framework for evaluating the performance of nano-refrigerants in a vapor compression refrigeration system (VCRS). The framework integrates thermodynamic modelling using CoolProp and Python, heat-transfer analysis, first- and second-law thermodynamic analyses, sensitivity analysis, uncertainty quantification, and machine learning (ML) prediction to investigate the influence of nanoparticle type, density, and concentration on system performance. Four nano-refrigerants comprising R134a with Al₂O₃ (2200 and 3690 kg m⁻³), TiO₂, and CuO nanoparticles were investigated. The developed model was validated against published experimental and numerical studies, yielding excellent agreement with numerical results, with average deviations ranging from <bold>0.23% to 2.72%</bold> , while comparison with experimental data showed acceptable agreement for engineering applications. The results demonstrate that incorporating nanoparticles significantly enhances the thermophysical properties of the refrigerant, leading to increased cooling capacity, reduced compressor work, improved heat transfer, lower entropy generation, and higher exergy efficiency. Among the investigated nano-refrigerants, <bold>R134a/CuO</bold> exhibited the best overall performance, achieving the highest cooling capacity ( <bold>147.5 kJ kg⁻¹</bold> ), the lowest compressor work ( <bold>49.0 kJ kg⁻¹</bold> ), the highest coefficient of performance ( <bold>COP = 3.01</bold> ), the highest convective heat-transfer coefficient ( <bold>702 W m⁻² K⁻¹</bold> ), the lowest entropy generation ( <bold>0.97</bold> ), and the highest exergy efficiency ( <bold>70.4%</bold> ). Performance improved with increasing nanoparticle concentration up to an optimum value of approximately <bold>0.60 wt.%</bold> , beyond which viscosity-related effects are expected to limit further enhancement. The machine learning models accurately reproduced the thermodynamic simulation results, demonstrating the feasibility of hybrid physics-based and data-driven modelling for rapid performance prediction and optimization. The proposed framework provides a reliable multi-physics platform for the design, optimization, and intelligent operation of high-efficiency nano-refrigeration systems. </p>