Improving Thermal Uniformity in Refrigerators Through AI-Supported Analysis of Wall-Outlet Arrangement

This study proposes an integrated approach to enhance temperature uniformity in household refrigerators by optimizing the arrangement of wall outlets using computational fluid dynamics (CFD) and artificial intelligence (AI) techniques. 3D CFD model reflecting the actual geometry and control logic of the refrigerator was developed, and both numerical and experimental analyses were conducted to evaluate its accuracy. The CFD model was validated using thermocouple measurements, showing good agreement. To efficiently investigate design configurations, a feed-forward neural network (FFNN) combined with a Bagging ensemble method was employed to predict temperature deviation and identify optimal outlet arrangements. The optimized design achieved a 14% reduction in temperature deviation in CFD simulations and a 4% reduction in experiments compared with the reference model. Flow field analysis revealed that the optimized configuration strengthened circulation in the lower compartment and mitigated jet concentration in the upper region, leading to a more balanced airflow distribution. These results demonstrate the effectiveness of combining CFD, AI, and experimental methods for achieving uniform temperature distribution and suggest the potential applicability of this approach to other thermal-fluid systems requiring improved flow uniformity.