Page 74 - 《真空与低温》2025年第4期

第 31 卷 第 4 期 真空与低温
2025 年 7 月 Vacuum and Cryogenics 489

非 均 匀 温 度 空 间 中 光 束 内 温 度 预 测 方 法 及 优 化

冯泽域 ，佘少波 ，李春煜 ，蔡爱峰 ，杨 光 ，吴静怡 ，田 义 2
2
1*
1
1
1
1
（1. 上海交通大学，上海 200240；2. 上海机电工程研究所，上海 201109）
摘要：在红外制导半实物仿真系统中，飞行器舱体空间内温度场并非均匀，为保证飞行器舱体空间中红外光
束内温度均匀度良好，需对无法布置温度传感器的光路温度进行预测。建立了舱体三维模型，利用 ANSYS Fluent
进行仿真计算，获得 80 组不同入口速度及入口温度下舱内各监测点的温度值。以 80 组不同入口条件下的温度数
据为训练数据集，利用 Matlab 构建神经网络预测模型，对实际试验中的数据，以已知监测点温度对光束内未知温
度进行预测。研究了预测点数量及布置方式与预测精度的关系。结果表明，神经网络预测的总体均方误差不超
过 0.8%，基于神经网络模型的预测值与实际值误差不超过 1.1%，预测点的数量及布置方式对模型预测精度存在
影响。提供了有效可行的光束内部温度的监测方法，并基于预测点数量及空间位置与预测精度关系的研究对测
点布置进行优化。
关键词：温度监测；数值模拟；参数化研究；神经网络
中图分类号：TB66 文献标志码：A 文章编号：1006−7086（2025）04−0489−09
DOI：10.12446/j.issn.1006-7086.2025.04.010

Beam Temperature Prediction Method and Optimization in Non-uniform Temperature Space

2
1
1
1
1
1*
FENG Zeyu ，SHE Shaobo ，LI Chunyu ，CAI Aifeng ，YANG Guang ，WU Jingyi ，TIAN Yi 2
（1. Shanghai Jiao Tong University，Shanghai 200240，China；
2. Shanghai Institute of Mechanical and Electrical Engineering，Shanghai 201109，China）
Abstract： In infrared-guided semi-physical simulation systems， the temperature field inside the aircraft cabin is often
non-uniform，which can have a significant impact on the accuracy and performance of infrared guidance systems. To ensure
optimal temperature uniformity within the infrared beam，it is crucial to predict temperatures along the optical path，particu-
larly in regions where temperature sensors cannot be placed. This study focuses on developing a three-dimensional cabin
model，and utilizes ANSYS Fluent to simulate and calculate the temperature distributions at various monitoring points under
80 different inlet velocity and temperature conditions. These conditions cover a wide range of operational scenarios，thus pro-
viding a diverse and comprehensive dataset for further analysis. The temperature data obtained from these 80 inlet conditions
are then used as a training dataset to build a neural network prediction model in Matlab. The model aims to predict unknown
temperatures along the infrared beam based on the known temperatures at the monitoring points，which serve as the model's
input. In addition to developing the model，the study investigates the relationship between the number and spatial arrangement
of prediction points and the overall accuracy of temperature predictions. The results show that the neural network model
achieves an overall mean square error （MSE） of less than 0.8%，with the error between the predicted and actual temperatures
not exceeding 1.1%. This indicates that the model performs with a high degree of accuracy. The research highlights that both
the number of prediction points and their spatial arrangement significantly influence the model's accuracy. The arrangement
of prediction points is crucial for ensuring reliable temperature predictions，especially in regions where direct measurements
are not feasible. This study provides an effective and practical method for monitoring temperatures within the infrared beam.
It also offers valuable insights into optimizing the placement of measurement points to enhance prediction accuracy，ultimately

收稿日期：2024−11−14
基金项目：上海航天技术研究院产学研合作基金（USCAST2022-05）
作者简介：冯泽域，硕士研究生。E-mail：zeeyuuf123@sjtu.edu.cn
通信作者：李春煜，博士，助理研究员。E-mail：lcy621860@sjtu.edu.cn