Physics-assisted Recursive Method for Sample Selection from Wall-Bounded Turbulence Data

The purpose of this work is to improve the efficiency of sample selection with the recursive method for wall-bounded turbulent flows. In the proposed physics-assisted recursive method, the flow field is divided into several sub-regions along the wall distance. Since the distributions of the flow variables have certain similarity along the wall normal direction, fast clustering of similar data can be achieved, which paves the way for the rapid elimination of redundant data. Then, the recursive method is used for sample selection in each sub-region. The effectiveness of the proposed method is investigated through several cases. The results show that the proposed method has good convergence and grid independence and improves the computational efficiency of the recursive method for sample selection. Since the amount of training data is reduced, the time consumption of model training is decreased. In addition, it is demonstrated that sample selection can also be helpful to achieve more balanced model performance by changing the distribution of training data.