Data-driven thermal state estimation for in-orbit systems via physics-informed machine learning.

Thermal analysis of spacecraft systems is a critical process for mission operations. However, knowing the temperature distribution of entire systems is not easy due to the uncertainty of the thermal mathematical model (TMM) and limited temperature sensors. This paper proposes a temperature estimation method using physics-informed machine learning (PIML). The PIML-based thermal analysis allows us to estimate the actual temperature distribution by seamlessly bridging the limited observations and the TMM. To evaluate the estimation accuracy of the proposed method, we conducted a numerical experiment using a pseudo small satellite model consisting of 100 nodes. The proposed method was applied to three different model error cases and was found to improve temperature estimation accuracy in all cases. In addition, the impact of the number of temperature sensors and their placement on estimation accuracy was investigated. • A data-driven thermal analysis that can bridge the uncertain model and limited measurements is proposed. • The physics-informed machine learning was applied to the thermal mathematical model. • A numerical experiment was conducted using a pseudo small satellite model. • The effect of the number and placements of sensors on the estimation accuracy for three model error scenarios was discussed. [ABSTRACT FROM AUTHOR]