Load prediction model for wind tunnel balance based on combinatorial optimization methods
Physica Scripta · 19.06.2026 · 101 (24) · 245929
DOI: 10.1088/1402-4896/ae6bbe ↗Kurz gefasst
Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 18.06.2026. Verfasst von XiaoGang Li und 5 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben.
Abstract (Originalsprache)
To address the issue of significant prediction errors in traditional machine learning models for wind tunnel balance load prediction, this paper thoroughly explores the enhancement effects of various machine learning algorithms on the predictive performance of wind tunnel balance models. Experimental results demonstrate that support vector machines (SVM) achieve strong predictive performance for small sample data by balancing generalization and fitting through regularization parameters and efficiently handling nonlinear relationships via kernel functions, yielding lower absolute errors, relative errors, and extreme errors in load prediction. To further improve model prediction accuracy, different machine learning models based on distinct principles are integrated, resulting in a two-stage prediction model. For various components of the wind tunnel balance, multiple error-compensated regression models are proposed for load prediction. These include: Support Vector Regression with Random Forest Error Compensation (SVR-RF); Bayesian Regression with Support Vector Error Compensation (BY-SVR); Bayesian Regression with Neural Network Error Compensation (BY-BP); Support Vector Regression with Neural Network Error Compensation (SVR-BP); and Backpropagation Neural Network with Random Forest Error Compensation (BP-RF). First, the regression model predicts the data and calculates the prediction error. Second, an error compensation model is trained to effectively predict the prediction error. Experimental results show that, compared to traditional methods such as least squares, SVMs, Bayesian regression, and multilayer perceptrons, the proposed error-compensated regression models significantly improve prediction accuracy. This provides an effective solution for high-precision, highly robust sensor calibration.
Zitieren
APA
Li, X., Wang, B., Shen, L., Zheng, Y., Liu, B., & Tan, S. (2026). Load prediction model for wind tunnel balance based on combinatorial optimization methods. Physica Scripta, 101(24), 245929. https://doi.org/10.1088/1402-4896/ae6bbe
Vancouver
Li X, Wang B, Shen L, Zheng Y, Liu B, Tan S. Load prediction model for wind tunnel balance based on combinatorial optimization methods. Physica Scripta. 2026;101(24):245929. doi:10.1088/1402-4896/ae6bbe
BibTeX
@article{Li2026,
author = {Li, XiaoGang and Wang, Biling and Shen, Lihua and Zheng, Yayuan and Liu, Boyu and Tan, Sen},
title = {Load prediction model for wind tunnel balance based on combinatorial optimization methods},
journal = {Physica Scripta},
year = {2026},
volume = {101},
number = {24},
pages = {245929},
doi = {10.1088/1402-4896/ae6bbe}
}
EUVIMED