基于高斯过程回归的连续式风洞马赫数控制

高赫; 刘学军; 郭晋; 吕宏强

doi:10.7638/kqdlxxb-2019.0018

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

基于高斯过程回归的连续式风洞马赫数控制

专栏——智能空气动力学研究探索 | 更新时间：2026-05-28

- 基于高斯过程回归的连续式风洞马赫数控制
- Mach number control of continuous wind tunnel based on Gaussian process regression
- 空气动力学学报 2019年37卷第3期页码：480-487
- 作者机构：
  
  1. 南京航空航天大学计算机科学与技术学院, 模式分析与机器智能工业和信息化部重点实验室,南京,211106
  2. 空气动力学国家重点实验室,绵阳,621000
  3. 软件新技术与产业化协同创新中心,南京,210023
  4. 中国航空工业空气动力研究院,沈阳,110034
  5. 南京航空航天大学航空学院,南京,210016
- 作者简介：
  
  高赫(1993-), 男, 硕士研究生, 研究方向:机器学习, 风洞控制.E-mail:gao_he@nuaa.edu.cn
  刘学军(1976-), 女, 教授, 主要研究方向:机器学习.E-mail:xuejun.liu@nuaa.edu.cn
- 基金信息：
  
  空气动力学国家重点实验室基金（SKLA20180102）;航天一院联合基金
- DOI：10.7638/kqdlxxb-2019.0018
  中图分类号： TP181;V211.71
- 纸质出版：2019
- 稿件说明：
移动端阅览
高赫, 刘学军, 郭晋, 吕宏强. 基于高斯过程回归的连续式风洞马赫数控制[J]. 空气动力学学报, 2019, 37(3): 480-487.

GAO He, LIU Xuejun, GUO Jin, et al. Mach number control of continuous wind tunnel based on Gaussian process regression[J]. 2019, 37(3): 480-487.
高赫, 刘学军, 郭晋, 吕宏强. 基于高斯过程回归的连续式风洞马赫数控制[J]. 空气动力学学报, 2019, 37(3): 480-487. DOI： 10.7638/kqdlxxb-2019.0018.

GAO He, LIU Xuejun, GUO Jin, et al. Mach number control of continuous wind tunnel based on Gaussian process regression[J]. 2019, 37(3): 480-487. DOI： 10.7638/kqdlxxb-2019.0018.

摘要

在风洞实验中保持实验段马赫数的稳定对实验的成功具有重要意义。传统的PID控制算法具有一定时滞性，不能满足连续变迎角实验模式下马赫数的控制精度要求。针对这一缺陷，提出了一种基于高斯过程回归的前馈控制策略，结合PID控制器共同完成马赫数控制任务。首先，对原始数据执行了预处理操作，将数据集中的异常数据进行清洗并且对清洗后的数据进行标准化；其次，选取迎角、实时马赫数、实验段截面积作为高斯过程回归模型的输入，压缩机转速作为输出，采用随机划分数据集与分组划分数据集两种策略进行建模，并将高斯过程回归与常用回归模型的预测精度进行了比较；最后，给出了利用高斯过程回归预测结果及预测置信度进行PID反馈控制的方法。实验结果表明高斯过程回归对风洞实验数据具有很好的建模能力，基于高斯过程回归的前馈控制与PID结合的控制策略能够提高连续变迎角模式下的马赫数控制精度。

Abstract

Maintaining the stability of Mach number plays an important role in the success of wind tunnel experiments. The traditional PID feedback control algorithm has a certain time lag and can not obtain the required control precision of Mach number in the experiments with continuously changing angle of attack. To overcome this defect

a feed forward strategy based on Gaussian process regression is proposed. The strategy combines the PID controller to carry out the control of Mach number. Firstly

the cleaned data is normalized in addition to cleaning the raw data to reduce the effects of noise. Secondly

the Gaussian process regression is employed to model the relationship between the inputs (angle of attack

real-time Mach number

and cross-sectional area) and the output (rotational speed of compressor) by two different strategies

which randomly divide data set and group data set by Mach number. The trained model is compared with other commonly-used approaches in terms of predictive accuracy. Finally

an improved PID control strategy that leverages the prediction and confidence results of Gaussian process regression is proposed. The experimental results confirm that Gaussian process regression has satisfactory ability of modelling wind tunnel data

and the control strategy that combines the PID with the feed forward control based on Gaussian process regression can improve the control precision of Mach number with continuous variable angles of attack in wind tunnel.

关键词

Keywords

references

LI Z F. Handbook of wind tunnel test[M]. Beijing:Aviation Industry Press, 2015.(in Chinese)

WANG K W. Application of neural network PID control in measured sub-flow field calibration of FD12 wind tunnel[D]. Harbin Engineering University, 2011.(in Chinese)

LUO Y, XING S, SONG W, et al. Fuzzy and PID compound control of mach number for 0.6 meters supersonic wind tunnel[C]//Artificial Intelligence, Management Science and Electronic Commerce (AIMSEC) 2011. IEEE, 2011: 3992-3995.

CAMACHO D E F, BORDONS D C. Model predictive control[J]. International Journal of Robust Nonlinear Control, 2007, 18(8):799-799.

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, 2002, 11(1):63-72.

YUAN P, YI F, XIAO Y H, et al. Wind tunnel flow field model predictive controller for angle of attack variation[J]. Journal of Control and Decision, 2018, 33(6):1026-1032.(in Chinese)

YUGENG X I, DEWEI L I, LIN S, et al. Model predictive control-Status and challenges[J]. Acta Automatica Sinica, 2013, 39(3):222-236.

RASMUSSEN C E, WILLIAMS C K I. Gaussian process for machine learning[M]. MIT Press, 2006:106-107.

MURPHY K P. Machine learning:A probabilistic perspective[M]. MIT Press, 2012.

GHAHRAMANI Z. Probabilistic machine learning and artificial intelligence[J]. Nature, 2015, 521(7553):452.

RICHARDSON R R, OSBORNE M A, HOWEY D A. Gaussian process regression for forecasting battery state of health[J]. Journal of Power Sources, 2017, 357:209-219.

PILON B H A, MURILLOFUENTES J J, COSTA J P C L D, et al. Predictive analytics in business intelligence systems via Gaussian processes for regression[C]//International Joint Conference on Knowledge Discovery, 2015.

DEISENROTH M P, FOX D, RASMUSSEN C E. Gaussian processes for data-efficient learning in robotics and control[J]. IEEE Transactions on Pattern Analysis Machine Intelligence, 2015, 37(2):408-423.

DUVENAUD D. Automatic model construction with Gaussian processes[D]. University of Cambridge, 2014.

KUMARSINGH B, VERMA K, THOKE A S. Investigations on impact of feature normalization techniques on classifier's performance in breast tumor classification[J]. International Journal of Computer Applications, 2015, 116(19):11-15.

CHEN Z, BO W. How priors of initial hyperparameters affect Gaussian process regression models[J]. Neurocomputing, 2017, 275:1702-1701.

MICCHELLI C A, XU Y, ZHANG H. Universal kernels[J]. Journal of Machine Learning Research, 2006, 7:2651-2667.

DUVENAUD D, NICKISCH H, RASMUSSEN C E. Additive Gaussian processes[J]. Neural Information Processing Systems, 2012:226-234.

HASTIE T J, TIBSHIRANI R J. Generalized additive models[J]. Statistical Science, 1986, 1(3):297-318.

ZHOU Z H. Machine learning[M]. Beijing:Tsinghua University Press, 2016.(in Chinese)

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于多元正交函数的风洞天平校准建模方法

基于GPR模型的多保真气动力建模方法

基于条件扩散模型的火箭气动外形快速设计方法

基于机器学习的极端风场短时预测研究进展与思考

大型结构风效应流固耦合机器学习研究进展