基于DPFNN网络机载武器战备高温贮存环境预计  

High Temperature Storage Environment Prediction Technology of Airborne Equipment Combat Readiness Based on DPFNN Network

作　　者：郝宇君 蔡良续[1] 李贺 张建军[1] HAO Yu-jun;CAI Liang-xu;LI He;ZHANG Jian-jun(China Aero-Polytechnology Establishment,Beijing 100028)

机构地区：[1]中国航空综合技术研究所,北京100028

出　　处：《环境技术》2024年第8期28-34,共7页Environmental Technology

摘　　要：由于机载外挂武器战备贮存时存在复杂的传热关系,为克服传热传质方程建模过程中对物性参数要求较高、求解困难等缺陷,准确预测极限环境条件下某型导弹舱内高温极值,提出一种基于DPFNN神经网络的机载外挂高温预计方法。以某型导弹为研究对象,地表温度Te、大气温度Ta、大气相对湿度h、太阳辐射强度G为输入,建立DPFNN过程神经网络结构,为提加快收敛速度,提出参数独立自适应学习率的梯度下降算法PALA,建立某型导弹地面挂机时舱内温度预测模型,对比DPFNN与ANN、多元线性回归MLRM方法的泛化能力。基于DPFNN的温度预测模型能够根据连续三日环境条件的变化确预测舱内温度变化,最大绝对误差仅1.17℃,具有较好的泛化能力。基于DPFNN的过程神经网络温度预计模型具备准确预计机载外挂温度的能力,该方法可用于机载外挂贮存温度预计工作,为确定机载外挂贮存温度环境适应性要求与试验验证条件提供参考。To overcome the difficulty of using thermodynamic equations to describe the relationship between the cabin environment and the external environment,and to accurately predict the extreme high temperature was proposed due to the complex heat transfer relationship during the airborne external combat readiness storage.A process neural network structure was established and a direct method of solving orthogonal transformation coefficients for discrete inputs was proposed.In order to accelerate the convergence speed,a gradient descent algorithm PALA was proposed to establish the temperature prediction model of air-to-air missile when it is hung up on the ground.Compare the generalization ability of DPFNN,ANN and MLRM.The temperature prediction model based on DPFNN can predict the temperature change in the cabin according to the change of environmental conditions for three consecutive days,and the maximum absolute error is only 1.17 ℃,which has good generalization ability.The process neural network temperature prediction model based on DPFNN has the ability to accurately predict the airborne external temperature,and the method can be used to predict the airborne external storage temperature,and provide reference for determining the environmental adaptability requirements of airborne external storage temperature.

关 键 词：过程神经网络 机载外挂 贮存环境 温度预计 

分 类 号：V267[航空宇航科学与技术—航空宇航制造工程] V216.5