A Highly-Precise Digital Wind Tunnel

A digital wind tunnel (WT) can learn from historical aerodynamic coefficients derived from experiments and simulations to build a prediction model. This model enables the prediction of aerodynamic coefficients without requiring additional experiments or simulations, significantly reducing their overall frequency and cost. In 2020, Huang et al. adopts multi-output back-propagation neural network (BPNN) to simultaneously predict lift coefficient and drag coefficient. However, Huang et al. does not investigate about the following issues, (1) abnormal data detection, (2) other aerodynamic coefficients, (3) overfitting problem. In order to solve these problems, this paper proposed a data-validate strategy and re-designed the single-output BPNN. Experimental results showed that both the mean square error and maximum absolute error of the proposed method is smaller than those of Huang et al. 's method.