Hybrid AI framework for the predictions of film cooling effectiveness distribution with various surface curvatures and compound angles

Film cooling is essential for protecting high-temperature components in gas turbines. For efficient cooling design, it is vital to predict film cooling effectiveness based on flow and geometrical parameters accurately and rapidly. Artificial intelligence (AI) models have recently been popular for this task due to their strong fitting nature. However, standard image-based AI models require significant training data to cover different surface curvatures and compound angles. Moreover, incorporating curvature complicates the input format for these models. Therefore, the current study proposes a hybrid AI framework for the rapid prediction of the film cooling effectiveness distribution under the influence of surface curvature and compound angle, which requires a relatively small amount of training data by integrating film cooling knowledge and different neural networks (Back-propagation neural network and U-Net). The proposed model achieves high accuracy across diverse flow and geometrical parameters, with a mean absolute error of approximately 0.006 or 0.012 for the film holes without or with compound angles on curved surfaces, respectively. The hybrid AI framework offers an accurate and rapid prediction method, offering a valuable tool for turbine cooling design.