Calibration Data Prediction of Wind Tunnel Balance Based on Regression Tree

The wind tunnel balance is mainly used to measure the aerodynamic load. However, due to the problem of the device itself, it needs to be calibrated before using. Traditional calibration methods mostly use the least squares method for polynomial fitting, which require manual computation and result in significant human and resource consumption. To break this situation and achieve efficient balance calibration, in this paper, the method of using machine learning models to predict balance calibration data is designed, including the generation model based on a Gaussian mixture model and five regression models. Furthermore, an evaluation metric dedicated to measuring the forecast results of balance data is proposed. Finally, extensive experiments on 12 real balance datasets are conducted with 5 prediction models, and experimental results demonstrate the high accuracy of using prediction models to predict balance calibration data.