Thermal models for avionics pod cabin based on Stochastic Configuration Network (SCN)

Abstract

Abstract The thermal failure of airborne avionics equipment is not optimistic. It is very necessary to establish relatively accurate thermal models for predicting thermal response of avionics equipment under different flight conditions. Traditional thermal modeling methods are often difficult to obtain accurate temperature response in complex conditions. This has severely restricted the application of these models. However, the Stochastic Configuration Network (SCN) model based on random algorithm can weaken the heat transfer mechanism and pay attention to the mining of experimental data, so that a more accurate thermal relationship might be obtained. In this paper, the SCN was used to analyze the experimental data of the avionics pod with a Ram Air Turbine (RAT) cooling system. The thermal models based on the SCN were finally built for avionics pod. Compared with the commonly used Random Vector Functional Link Network (RVFLN) thermal models, the SCN thermal models not only inherit the advantages of simple network structure and low computational complexity, but also have some merits, such as the better learning performance and the less human intervention. The presented SCN models provide a way to predict the thermal response of avionics pod cabin under the full flight envelope for a fighter.