SEMI-EMPIRICAL THERMAL MODELLING OF LAPAN-A3 SATELLITE USING MACHINE LEARNING METHOD

• Main Author: Sutardi, Ricky
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/67739
• Institution: Institut Teknologi Bandung

From the satellite temperature sensor readings, it is found that LAPANA3, an Indonesian microsatellite, requires special maneuver to keep its main payload, a multispectral imager, above 0 °C. It can then be concluded that the satellite passive thermal control system is inadequate for the satellite mission needs. To prevent this problem in future satellite thermal design, a new simple thermal model which can model LAPAN-A3 thermal characteristic accurately is needed. Conventional thermal analysis requires accurate calculations of many variable which results in increased complexity. Meanwhile, data-driven approaches in solving satellite thermal problem are becoming increasinly common to reduce the complexity and number of variable calculations needed in thermal modelling. Thus, a semi-empirical approach which uses empirical telemetry data and observations of the satellite to simplify the theoretical calculations of satellite thermal equation hopefully can result in a quicker and easier thermal modelling process. In this paper, a semi-empirical 7-node thermal model of LAPAN-A3 satellite is developed with machine learning method using linear regression approach. The model is trained and tested with satellite data for the observation period of 19 and 20 May 2018 to solve the unknown variables in LAPAN-A3 satellite thermal equation in order to predict the satellite nodes’ temperature change. Initial evaluation shows that the thermal model is generally able to predict LAPAN-A3 node temperature changes and has potential usage in reallife satellite development. Even though the model can still be improved upon, it may serve as a basis for future satellite thermal model.