Machine learning regression analysis of impinging slot jet with various aspect ratios to heated rough surface

Abstract

This study aims to investigate the parametric relationships in the convective cooling process when a slot jet with various aspect ratios (ARs) impinges on a heated surface. To achieve this objective, an experimental study and a machine learning (ML)-based regression analysis applied to the experimental results were used as the main methodology. The effects of independent variables -particularly AR (w/h = 4, 6 and 8), jet-to-plate distance (z/ w = 1, 2, 6 8, and 10), Reynolds number (Re = 10,000, 20,000, and 40,000), and surface roughness (smooth, inline-dimpled, and staggered-dimpled) - on heat transfer, namely Nusselt number, along with their interactions, were analyzed using ML techniques. Applying ML regression to such studies, especially when the variation in AR is within a narrow range, is a key innovation of this work. Five ML regression models, Multiple Linear Regression (MLR), Decision Tree Regression (DTR), Support Vector Regression (SVR), Gaussian Process Regression (GPR), and Multi-Layer Perceptron (MLP) were used to evaluate the relationship between the dependent and independent variables. Among the tested regression models, GPR provides the most accurate overall predictions, yielding the lowest RMSE (2.2016) and MAE (1.6183), while DTR performs the worst. The results demonstrate that ML-based regression models, particularly GPR, provide a reliable and efficient framework for estimating heat transfer behavior in complex impingement-cooling applications.