Determination of thermal parameters in thermal mathematical models for spacecraft thermal design

Abstract

The spacecraft operating in Earth orbit must maintain the appropriate temperature of onboard equipment amidst extreme conditions such as the 5700 K from the Sun and the frigid temperatures of outer space, to safely perform its mission throughout its designed lifetime. Moon, natural satellites of Earth, experiences average temperatures in the space environment ranging from approximately -53.15°C (equator) to -123.15°C (at 85° north latitude), depending on the surface reflectivity and thermal conductivity of the lunar soil. These temperatures can exceed the operational limits of typical onboard devices. Therefore, spacecraft employ thermal design to maintain suitable temperatures, requiring precise temperature prediction through thermal mathematical models. However, thermal parameters inputted into these mathematical models carry uncertainties, necessitating reduction through verification against data from thermal vacuum tests and comparative analysis. Such test data are generally limited, posing an undetermined problem when facing complex missions or spacecraft designed for new space environments where thermal design/environmental variables outnumber equations. As designs become more intricate, resolving multivariate undetermined problems via iterative variable determination by engineers becomes increasingly challenging. This thesis proposes a technique to swiftly determine various thermal parameters of spacecraft, acknowledging their inherent uncertainties. Thermal mathematical models are structured using the widely adopted lumped-parameter method to configure thermal nodes of spacecraft, with component properties inputted based on well-established, measured, or theoretical values. However, the values with lesser-known or high uncertainties are defined as thermal parameters. To enhance accuracy for thermal parameters, quasi-steady test data from thermal vacuum tests are utilized, correlating and verifying the thermal mathematical models to ensure their temperature prediction align within criteria specified by the European Cooperation for Space Standardization (ECSS). This approach incorporates a data-driven surrogate modeling approach using radiation basis function networks and deep neural networks to predict test data, setting the objective function to minimize temperature deviations between quasi-steady state of the thermal vacuum test and predictions by surrogate models. By imposing temperature constraints to maintain the physical validity of thermal parameters, adjustments are made to achieve solutions consistent with test results, thus verifying the potential of surrogate modeling based on data for determining thermal parameters.