高超声速风洞短时气动力智能辨识算法研究  

RESEARCH ON INTELLIGENT IDENTIFICATION ALGORITHMS FOR SHORT-TERM AERODYNAMICS OF HYPERSONIC WIND TUNNELS

作　　者：王钦超 李世超[1] 高宏力[1] 马贵林 伍广 段志琴 Wang Qinchao;Li Shichao;Gao Hongli;Ma Guilin;Wu Guang;Duan Zhiqin(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《力学学报》2022年第3期688-696,共9页Chinese Journal of Theoretical and Applied Mechanics

基　　金：国家自然科学基金资助项目(52105562)。

摘　　要：风洞测力试验是高超声速飞行器研发的重要环节,随着这项技术的发展,试验模型的大尺度化成为高超声速风洞试验的趋势.在几百毫秒的有效测试时间内,大尺度测力系统刚度减弱等问题会严重导致气动力辨识精度变差,试验模型大尺度化对短时脉冲燃烧风洞精确气动力辨识带来了挑战.对此本文提出了一种新的基于传统信号处理结合深度学习的智能气动力辨识算法,该框架分解两个主要阶段:(1)信号分解,(2)数据训练.其中信号分解阶段通过变分模态分解将原始数据分解为不同模态子信号,随后通过Pearson相关性分析筛除干扰子信号;在训练阶段通过深度学习模型提取训练数据集中含有有效特征的子信号,最终得到真实气动力信号.此外,为增强算法的鲁棒性,在算法框架不同阶段通过不同方法对算法中的超参数进行优化得出最优参数组合.此算法在气动力辨识精度以及抗干扰等方面都得到了比较理想的结果.通过悬挂测力实验台进行验证,结果表明该算法可以有效滤除由大尺度模型带来的传统方法难以消除的干扰分量.最后应用于脉冲燃烧风洞的大尺度模型测力系统,气动力辨识精度得到有效提高.Pulse combustion wind tunnel force measurement is an important step in the research and development process of hypersonic aircraft,and with the development of hypersonic aircraft technology,large-scale and heavy-load aircraft test models has become the trend of hypersonic pulse combustion wind force test.During the effective test time of several hundred milliseconds,large-scale force measurement system stiffness weakened and other issues will seriously lead to poor aerodynamic identification accuracy.The large-scale measurement model poses a challenge to the accurate aerodynamic identification of the short-term pulse combustion wind tunnel.To solve this problem,a new intelligent aerodynamic identification algorithm based on traditional signal processing combined with deep learning is presented in this paper.The algorithm framework is mainly divided into two stages for signal processing:(1)signal decomposition,(2)data training.In the signal decomposition stage,the original data is decomposed into different modal sub-signals through variational modal decomposition(VMD).In the training stage,the effective features in the remaining datasets containing characteristic sub-signals are extracted by deep learning model,and the real aerodynamic signals are obtained.In addition,in order to enhance the robustness and applicability of the algorithm,different optimization methods are used to optimize the hyperparameters in the algorithm at different stages of the algorithm framework to obtain the optimal parameter combination.This algorithm model has obtained relatively ideal results in terms of aerodynamic recognition accuracy and anti-interference.Finally,the algorithm is validated on a suspended force test bench,and the results show that the algorithm model can effectively identify and filter out the interference components that are difficult to eliminate by the traditional methods brought by the large-scale model.Finally,the algorithm is successfully applied to the large scale model force measurement system of pulse combus

关 键 词：脉冲风洞 深度学习 气动力辨识 应变天平 测力系统 

分 类 号：V211.751[航空宇航科学与技术—航空宇航推进理论与工程]