Speaker
Description
Satellite thermal models are traditionally developed during the design phase and correlated using ground-based thermal vacuum tests. Their primary purpose is to predict temperature extremes under representative worst-case conditions and to demonstrate compliance with the spacecraft thermal requirements. While this approach is essential for spacecraft design and qualification, it provides only limited information on the dynamic thermal behaviour of the spacecraft during nominal operations. In particular, the actual in-flight temperature response results from the combined effects of the orbital environment, spacecraft operating modes, equipment dissipations and thermal control system behaviour, which are difficult to reproduce comprehensively during ground testing.
The use of in-flight telemetry provides an opportunity to further characterize this behaviour and to improve the representativeness of the thermal model under operational conditions. However, correlating a thermal model against flight data introduces additional challenges, as several physical phenomena and model parameters can produce similar temperature signatures. The identification of equipment dissipations is therefore an inverse problem involving a potentially very large number of parameters and requiring a large number of thermal simulations.
This work is in the frame of the ESA contract ESA/CNES study Rationalization of Thermal Simulators for Operations, led by DOREA with help of Thales Alenia Space.. This study focuses on the correlation of the Sentinel-3B thermal model developed by Thales Alenia Space. The starting model had previously been validated against thermal vacuum test measurements, providing a physically consistent basis for the in-flight correlation. The objective is to determine whether the model can reproduce the observed thermal behaviour in orbit while identifying parameters representative of the actual spacecraft operating conditions.
The proposed approach combines several levels of analysis. The orbital environment is reconstructed from the spacecraft position and orbital configuration, including eclipse entry and exit, variations in the Earth albedo resulting from the changing Earth projection, and other relevant environmental conditions. The scope was first to reach the convergence of the thermal model. The behaviour of the thermal control system is also analysed, including PI-controlled and thermostatically controlled equipment. Spacecraft operational modes are identified from telemetry to distinguish changes in thermal behaviour associated with different configurations and operating states.
The main identification task concerns the estimation of equipment thermal dissipations. Depending on the level of detail considered in the thermal model, several hundred dissipation parameters may potentially need to be adjusted within the ranges defined during the design phase. A systematic exploration of all possible combinations is therefore computationally impractical. Several optimization strategies are investigated, including gradient-based approaches and evolutionary algorithms, with the objective of efficiently minimizing the discrepancy between simulated and measured temperatures while maintaining physically meaningful parameter values.
The computational cost of the optimization process is addressed through the use of high-performance computing. minimizing the residuals (cost function) between simulated outputs and measured telemetry temperatures over the last orbit. A new convergence is needed by input dissipations. These steps resources provided by the CRIANN cluster. CRIANN is a French Normandy’s regional high-performance computing and digital research center. Its provides advanced computing infrastructure and expertise to support scientific research and innovation. The thermal simulations corresponding to successive optimization iterations are parallelized, allowing a large number of candidate solutions to be evaluated simultaneously and significantly reducing the overall computation time. Particular attention is given to the balance between optimization performance, computational resources and the physical interpretability of the resulting parameters.
The study also investigates the respective contributions of expert knowledge and automated optimization. While environmental conditions and thermal control behaviour can often be identified through expert analysis of telemetry profiles, the simultaneous estimation of a large number of equipment dissipations is better suited to automated optimization techniques. This combination aims to reduce the complexity of the identification process while preserving the physical understanding required to interpret the results.
The presentation will describe the methodology developed for the in-flight correlation, the main technological and computational challenges encountered, and the solutions implemented to overcome them. The resulting correlation performance will be presented together with an analysis of the identified thermal parameters and their consistency with the expected spacecraft behaviour. The respective advantages and limitations of the investigated optimization strategies will also be discussed.
Beyond the Sentinel-3B case study, the objective is to assess the applicability of this methodology to other spacecraft thermal models and operational scenarios. Improved knowledge of the actual in-flight thermal behaviour could provide a more accurate assessment of thermal margins and reduce uncertainties associated with thermal design assumptions. In the longer term, such an approach could contribute to more efficient thermal design processes and to the development of thermal simulators better suited to spacecraft operations.