Real-Time Cubesat Thermal Simulation using Artificial Neural Networks

Abstract

In space systems engineering, the Operational Simulator (OS) is a computational tool that can be used to test and validate the ground control system, to train the flight control operators, and to support the operation of spacecrafts. In order to accomplish these tasks, the OS must produce data of all the spacecraft subsystems in real-time. Among these subsystems, the thermal control subsystem is one of the most demanding in terms of computational cost. In this work we use Artificial Neural Networks (ANN) to learn the thermal behavior of a simple CubeSat model, generated by a thermal analysis software, and then apply it to reproduce that behavior and to generalize for scenarios not presented during training. The results show that the ANNs can simulate the temperatures of the CubeSat with good fidelity and very low computational cost.