Neural Network Modeling for Predicting Local Heat Transfer in Jet Impingement: A Comparison With Traditional Correlations

Abstract Numerous authors have reported piecewise correlations in the literature to capture local heat transfer characteristics in jet impingement for various jet-to-target plate spacings and Reynolds numbers. However, developing a unified, comprehensive correlation using traditional data-fitting techniques poses significant challenges due to the complex and nonlinear nature of the phenomenon. To address this, a neural network model is trained using three experimental datasets (Katti and Prabhu, 2008; Lee & Lee, 1999; Lytle & Webb, 1994) to create a unified predictor for local heat transfer performance. A hyperparameter tuning study is conducted to determine the optimal neural network architecture. The input variables include the Reynolds number, jet-to-target plate spacing, and radial distance from the jet’s geometric center, while the output variable is the local Nusselt number. To evaluate the model’s performance, key metrics such as the regression coefficient (R2), mean absolute error (MAE), and root mean square error (RMSE) are computed. The neural network outperformed traditional correlations in predicting local heat transfer characteristics. By leveraging the neural network’s ability to capture complex, nonlinear relationships, the study demonstrates that robust and accurate predictors can be developed for a wide range of geometrical parameters and operating conditions. This approach holds promise for advancing heat transfer research and providing engineers with more reliable tools for designing efficient thermal systems.