基于随机配置网络的机载电子吊舱多工况热模型

Avionics pod is a main carrier for multifunctional airborne electronic equipment, which effectively improves the performance of fighter. Increasing power of electronic equipment and low-pressure flight environment can exacerbate thermal environment in pod and can further affects the reliability of electronic equipment, so it is important to predict the thermal response of equipment under different flight conditions. In this paper, a thermal modeling method using thermal network analysis and stochastic configuration network is proposed and is further verified by experimental data of avionics pod with a ram air cooling system. The data of five conditions (high temperature storage, high temperature working, low temperature storage, low temperature accident and low temperature working) is divided into three groups according to heat transfer mechanism and is used to establish the storage thermal models, the working thermal models and the comprehensive thermal models, respectively. Thermal network analysis is used to obtain the input of network. Four-fold cross-validation and gray-scale analysis are used to determine hyper-parameters. The results show that the range sequence can be unified to [1-40] and the maximum number of hidden nodes of three thermal models can be set to 6, 9, and 11, respectively. The modeling results are positive, and the prediction error of electronic equipment temperature in the whole process of multi-conditions is within 3.512℃. Thus, the thermal modeling method that describes the thermal relationship of electronic equipment by data mining can be used to predict the avionics pod temperature in expected flight environment and evaluate the performance of thermal management system.