基于随机森林的风洞马赫数预测模型
收稿日期: 2015-06-17
修回日期: 2015-08-19
网络出版日期: 2015-08-26
基金资助
国家自然科学基金(61473073,61333006)
Wind tunnel Mach number prediction model based on random forest
Received date: 2015-06-17
Revised date: 2015-08-19
Online published: 2015-08-26
Supported by
National Natural Science Foundation of China (61473073, 61333006)
在风洞试验中,马赫数的稳定性和快速性对风洞流场品质有着重要影响。为了实现马赫数的精确控制,必须对马赫数进行快速、准确的预测。风洞试验积累了大量数据,大数据集包含了更多的有益信息,为实现马赫数的精确预测提拱了可能性,但也增加了建模的复杂度。通常高度复杂的模型会加重其在实际使用时的计算负担。针对大数据集问题,本文将随机森林方法应用于风洞马赫数建模。随机森林是一种集成模型建模方法,它从3方面降低模型的复杂度:产生多个样本子集,减少了子模型的训练样本个数;具有并行集成结构,子模型可在不同的CPU上运行,提高了运行速度;以简单学习算法回归树作为基学习机,降低了子模型的复杂度。试验证明基于随机森林的马赫数预测模型能够有效利用试验积累的大数据,满足工程上预测速度及精度的要求。
王晓军 , 袁平 , 毛志忠 , 杜宁 . 基于随机森林的风洞马赫数预测模型[J]. 航空学报, 2016 , 37(5) : 1494 -1505 . DOI: 10.7527/S1000-6893.2015.0229
In the measurements of wind tunnel, the stability and the rapidity of the Mach number produce an important effect on quality of the flow field. To realize precisely controlling of the Mach number, it is required that the Mach number prediction should be speed forecasting and accurate. Large-scale data ste are accumulated from measurements. Although large-scale data set contain more useful information to improve the accuracy on the Mach number prediction, it increases the complexity for modeling. In general, high complexity models also increase the computational burden at the phase of active use. To deal with the large-scale set issue, the random forest method is applied to predicting the Mach number in the wind tunnel. Suitable for large-scale problem, random forest reduces the complexity in the following three aspects:generating training subset and decreasing the size of training samples; with the parallel ensemble structure, running sub-models on different CPUs and saving the running time; selecting a simple base learner, thus reducing the complexity of sub-models. The test demonstrate that the random forest-based Mach number prediction model can successful utilize the large-scale data accumulated from measurements and meet the requirements of the forecasting speed and the accuracy.
Key words： wind tunnel test; Mach number; large-scale data set; random forest; regression tree
[1] 孙智伟, 白俊强, 高正红, 等. 现代超临界翼型设计及其风洞试验[J]. 航空学报, 2015, 36(3):804-818. SUN Z W, BAI J Q, GAO Z H, et al. Design and wind tunnel test investigation of the modern supercritical arifoil[J]. Acta Aeronautica et Astronautica Sinica, 2015, 36(3):804-818(in Chinese).
[2] 路波, 吕彬彬, 罗建国, 等. 跨声速风洞全模颤振试验技术[J]. 航空学报, 2015, 36(4):1086-1092. LU B, LYU B B, LUO J G, et al. Wind tunnel technique for transonic full-model flutter test[J]. Acta Aeronautica et Astronautica Sinica, 2015, 36(4):1086-1092(in Chinese).
[3] SOETERBOEK R A M, PELS A F, VERBRUGGEN H B, et al. A predictive controller for the Mach number in a transonic wind tunnel[J]. IEEE Control Systems, 1991, 11(1):63-72.
[4] YAN Y, XI Z, ZHANG S. Numerical simulation and transonic wind-tunnel test for elastic thin-shell structure considering fluid-structure interaction[J]. Chinese Journal of Aeronautics, 2015, 28(1):141-151.
[5] LIU W, MA X, LI X, et al. High-precision pose measurement method in wind tunnels based on laser-aided vision technology[J]. Chinese Journal of Aeronautics, 2015,28(4):1121-1130.
[6] GREEN J, QUEST J. A short history of the European transonic wind tunnel ETW[J]. Progress in Aerospace Sciences, 2011, 47(5):319-368.
[7] 钱卫, 张桂江, 刘钟坤.飞机全动平尾颤振特性风洞试验[J]. 航空学报, 2015, 36(4):1093-1102. QIAN W, ZHANG G J, LIU Z K. Flutter characteristics for aircraft all-movable horizontal tail through wind tunnel test[J]. Acta Aeronautica et Astronautica Sinica, 2015, 36(4):1093-1102(in Chinese).
[8] 张贵军. 风洞过程建模与控制问题的研究[D]. 沈阳:东北大学, 1998. ZHANG G J. Research on model and control of wind tunnel process[D]. Shenyang:Northeastern University, 1998(in Chinese).
[9] LEONTARITIS J, BILLINGS S A. Input-output parametric models for nonlinear systems, Part I:Deterministic nonlinear systems; Part II:Stochastic nonlinear system[J]. International Journal of Control, 1985, 41(1):303-344.
[10] DANDOIS J, PAMART P Y. NARX modeling and extremum-seeking control of a separation[J]. Journal of Aerospace Lab, 2013(6):1-12.
[11] BOUBEZOUL A, PARIS S. Application of global optimization methods to model and feature selection[J]. Pattern Recognition, 2012, 45(10):3676-3686.
[12] CEPERIC V, GIELEN G, BARIC A. Recurrent sparse support vector regression machines trained by active learning in the time-domain[J]. Expert Systems with Applications, 2012, 39(12):10933-10942.
[13] SU Y, GAO X, LI X, et al. Multivariate multilinear regression[J]. IEEE Transactions on Systems, Man and Cybernetics-Part B:Cybernetics, 2012, 42(6):1560-1573.
[14] SPEYBRIECK N. Classification and regression trees[J], Hints & Kinks International Journal of Public Health, 2012, 57(1):243-246.
[15] BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1):5-32.
[16] BREIMAN L. Bagging predictors[J]. Machine Learning, 1996, 24(2):123-140.
[17] NORGAARD M. Neural network based system identification toolbox, Version 2, For use with MATLAB, Department of Automation, Department of Mathematical Modelling, Technical University of Denmark[EB/OL]. (2000-01-23)[2015-06-17]. http://www.iau.dtu.dk/research/control/nnsysid.html.
[18] 郭昌辉, 刘贵全, 张磊. 基于回归树与K-最近邻交互模型的存储设备性能预测[J]. 南京大学学报:自然科学版, 2012, 48(2):123-132. GUO C H, LIU G Q, ZHANG L. An interactive model based on regression tree and K-nearest neighbor for storage device performance prediction[J]. Journal of Nanjing University:Natural Sciences Edition, 2012, 48(2):123-132(in Chinese).
[19] 陈湘芳, 陈明, 冯国富, 等. 多变量时序回归树的黄瓜产量预测模型[J]. 计算机工程与设计, 2012, 33(1):407-411. CHEN X F, CHEN M, FENG G F, et al. Yield prediction model of cucumber based on multivariate time series regression tree[J]. Computer Engineering and Design 2012, 33(1):407-411(in Chinese).
[20] PERRONE M P, COOPER L N. When networks disagree:Ensemble methods for hybrid neural networks[M]. In:MAMMONE R J, editor. Artificial neural networks for speech and vision. London:Chapman & Hall, 2001:126-142.
/
| 〈 |
|
〉 |
CJA