欢迎访问《兵工学报》官方网站,今天是

兵工学报 ›› 2021, Vol. 42 ›› Issue (6): 1230-1237.doi: 10.3969/j.issn.1000-1093.2021.06.013

• 论文 • 上一篇    下一篇

高超声速飞行器烧蚀后退量时序提取及基于神经网络的预测

于哲峰1, 胥建宇1,2, 罗跃1, 杨鹰1, 刘进博1, 兰京川2   

  1. (1.中国空气动力研究与发展中心 超高速空气动力研究所, 四川 绵阳 621000;2.电子科技大学 自动化工程学院, 四川 成都 611731)
  • 上线日期:2021-07-19
  • 作者简介:胥建宇(1993—),男,硕士研究生。E-mail:xujianyuabc@163.com
  • 基金资助:
    国家部委科研项目(2020年)

Time Series Extraction of Ablation Recession of Hypersonic Vehicle and Its Prediction Based on LSTM Neural Network

YU Zhefeng1, XU Jianyu1,2, LUO Yue1, YANG Ying1, LIU Jinbo1, LAN Jingchuan2   

  1. (1.Hypervelocity Aerodynamics Institute, China Aerodynamics Research and Development Center, Mianyang 621000, Sichuan, China;2.School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731,Sichuan,China)
  • Online:2021-07-19

摘要: 为了观测高超声速飞行器头部模型在高温流场下的烧蚀现象,在电弧风洞上开展了热考核试验,基于电荷耦合器件相机图像比色测温方法,获得电弧风洞烧蚀模型表面温度随时间演化数据。通过追踪模型边缘温度场变化,给出模型烧蚀过程中形变过程,提取驻点及其他边缘点后退量随时间变化数据。利用最小二乘法和长短期记忆(LSTM)网络方法对烧蚀后退量数据进行拟合与预测,其中LSTM方法主要包括网络结构设计、模型训练、目标函数设定和预测过程的算法实现等。高超声速飞行器烧蚀模型不同位置后退量时间序列预测与试验对比发现:最小二乘法主要适合烧蚀后退量为线性区域的拟合与预测;LSTM方法不但适用于烧蚀后退量线性区域,还适合于端头边缘烧蚀后退量非线性变化区域数据的拟合与预测。

关键词: 高超声速飞行器, 烧蚀, 电荷耦合器件相机, 比色测温, 后退量, 长短期记忆网络模型

Abstract: The ablation phenomenon of hypersonic vehicle head model in high temperature flow field is studied. The thermal test is made in an arc wind tunnel. The temperature evolution data of the arc wind tunnel ablation model with time is obtained using the colorimetric temperature measurement method of CCD camera image. By tracking the change of the temperature field at the edge of the model, the deformation process of the model during ablation is given, and the time series of stagnation point and other edge points are extracted. Least squares method and long-term short-term memory(LSTM) network method are used to fit and predict the ablation recession data. The LSTM method mainly includes network structure design, network training, objective function setting and algorithm implementation of prediction process. Through comparison between the time series prediction and experiment of ablation recession at different positions of the ablation model, it is found that the least square method is mainly suitable for the fitting and prediction of ablation recession in the linear region; the LSTM method is not only suitable for the linear ablation recession, but also suitable for the fitting and prediction of the nonlinear change of ablation recession.

Key words: hypersonicvehicleablation, chargecoupleddevicecamera, colorimetrictemperaturemeasurement, recession, long-termshort-termmemorynetworkmodel

中图分类号: