应用过程神经网络的航天器瞬态温度预测方法  

Transient Temperature Prediction Method for Spacecraft Using Process Neural Network

作　　者：陈冠宇 杨森 彭威 多乐乐 张健鹏 刘宪闯 CHEN Guanyu;YANG Sen;PENG Wei;DUO Lele;ZHANG Jianpeng;LIU Xianchuang(System Design Institute of Hubei Aerospace Technology Academy,Wuhan 430040,China)

机构地区：[1]湖北航天技术研究院总体设计所,武汉430040

出　　处：《航天器工程》2023年第6期25-30,共6页Spacecraft Engineering

摘　　要：航天器在轨温度受空间热环境影响变化较大,同时研制阶段的热分析与热试验往往也耗时较长,因此通过准确有效的预测方法为其提供在轨温度预警信息、提高热仿真与热试验效率至关重要。文章提出航天器瞬态温度预测方法,依据航天器在轨温度实测数据,采用相空间重构理论构建样本集完成训练,应用过程神经网络建立瞬态温度预测模型,并对温度进行外推预测。经验证,根据温度预测方法建立的温度预测模型绝对误差最大值为0.746 K,可在满足工程精度的情况下实现对航天器瞬态温度的快速预测。The temperature of spacecraft in orbit is greatly affected by the space thermal environment,and the thermal analysis and testing during the development stage are usually time-consuming.Therefore,it is crucial to provide accurate and effective prediction methods for warning temperature information in orbit and improving the efficiency of thermal simulation and testing.A transient temperature prediction method for spacecraft is proposed in this paper.Based on the measured temperature data of spacecraft in orbit,a sample set is constructed using phase space reconstruction theory to complete training.A transient temperature prediction model is established using process neural networks,and the temperature is extrapolated for prediction.After verification,the maximum absolute error of the temperature prediction model established based on the temperature prediction method is 0.746K,which can achieve rapid prediction of transient temperature of spacecraft while meeting engineering accuracy.

关 键 词：航天器瞬态温度预测 过程神经网络 相空间重构理论 

分 类 号：V474[航空宇航科学与技术—飞行器设计]