Toward Uncertainty Quantification in Satellite Thermal Design

Abstract

In this study, an uncertainty quantification technique is applied to the thermal design of a pseudo-small satellite. An emulator that uses Gaussian process regression and the least absolute shrinkage and selection operator is employed to efficiently estimate the thermal margin based on the given uncertainties, and sensitivity analysis based on multiple regression analysis is employed to effectively reduce risk in the thermal design of the satellite. The uncertainty quantification results show that the emulator can reduce the uncertainty quantification cost compared to an ordinary simulator, and that the sensitivity analysis can clarify that only two factors are dominant in pseudo-small satellite thermal design uncertainty.