Akademik Veri Yönetim Sistemi
4th International Conference on Modern and Advanced Research, Konya, Türkiye, 6 - 07 Kasım 2025, (Tam Metin Bildiri)
Microchannel heat sinks are widely utilized in compact thermal systems due to their superior heat transfer performance; however, optimizing their design remains challenging because of the nonlinear coupling between geometric parameters and two-phase flow behavior. This study aims to accurately predict the flow boiling heat transfer coefficient in microchannels and determine the optimal channel configuration that maximizes thermal performance while satisfying the dimensional limits defining microchannel structures. A hybrid Particle Swarm Optimization–Artificial Neural Network (PSO-ANN) model was developed to capture the complex thermo-hydraulic behavior using dimensionless numbers and geometric descriptors as input features. The model was trained using 70% of the dataset with 10-fold cross-validation, while the remaining 30% was used to evaluate generalization capability. Compared with a conventional ANN, the PSO-ANN demonstrated significant improvements, achieving R² values of 0.9991 in training and 0.9878 in testing, along with reductions in RMSE and MAE of up to 45%, confirming enhanced predictive accuracy and stability. The optimized model was then integrated with the Grey Wolf Optimization (GWO) algorithm to maximize the predicted heat transfer coefficient under a geometric constraint ensuring valid microchannel dimensions. The optimal design was obtained as 300 m channel width and 60 m channel height, resulting in a maximum heat transfer coefficient of 40.652 kW/m²K. Overall, the results demonstrate that coupling machine learning-based predictive modeling with metaheuristic optimization offers a powerful and efficient approach for improving the design of next-generation microchannel heat sinks.