Computational Fluid Dynamics and Artificial Neural Network Integrated Framework for Rapid Prediction of Curved Fin Designs in Phase Change Material Systems

In phase change material (PCM) based thermal energy management systems, heat transfer and the melting‐solidification process determine the system performance. Recent studies have made it clear that the use of curved fins improves the PCM flow dynamics and reduces the total charging‐discharging time. However, a detailed study of the geometrical parameters of these fins through CFD requires significant time and computational cost. In the present work, a CFD and Artificial Neural Network (ANN) integrated framework is proposed for rapid prediction of the Complete Solidification Time (CST) in a PCM based system equipped with wave‐shaped curved fins. A total of 100 high‐fidelity CFD simulations were performed by systematically varying two key geometric parameters, fin amplitude ( = 1–10 mm) and number of peaks ( = 1–10). The generated dataset was used to train a feed‐forward ANN using the Levenberg–Marquardt algorithm. The trained model demonstrated excellent predictive accuracy, with strong agreement between ANN predicted and CFD derived CST values for both training and unseen testing cases. The results confirm that fin amplitude is the dominant parameter governing solidification performance, while the effect of peak number shows diminishing returns beyond a critical limit. Slightly higher prediction deviations were observed for high‐amplitude cases ( 8 mm) due to increased nonlinearity in PCM flow behavior, however, the maximum relative error remained within ~2%, which is negligible at the system level. Overall, the proposed ANN surrogate enables near instantaneous CST prediction, reducing computational cost by orders of magnitude compared to CFD.