Speaker
Description
The current trend in the space industry is to create less expensive satellites in shorter periods of time. In such context, space systems thermal analyses require dealing with models with different levels of fidelity. Detailed thermal mathematical model (TMM) can imply high computational cost due to the complexity of the orbital environment of the missions, but also due to the many iterations that are required to be run in thermal design processes. Thus, low-fidelity or reduced-models are generally developed and used to shorten analysis time. The main challenge with building a reduced model is to have the right level of accuracy while having the smallest possible computational time compared to a detailed model. Indeed, when simplifying a thermal model, one needs to make assumptions and to ensure that the model gives the right predictions at critical locations. Thus, space thermal engineers must correlate reduced and detailed models before using a reduced model for the multiple operating conditions involved in a mission analysis.
In this presentation, we will introduce an automated and efficient method of correlating and calibrating a reduced-thermal model with a detailed model using an adjoint-based solver implemented in the TMG Correlation tool. It takes advantage of the adjoint-based formulation of the energy equation to get the variation of an objective function, which represents the temperature difference between the reduced and detailed models at specific locations, with respect to the design variables of the correlation/calibration problem using a single solve of the thermal problem. By combining this adjoint-based solver with an optimization algorithm, the correlation/calibration procedure can be automated and performed much faster than using a manual procedure.
The presentation will first describe the adjoint-based approach used for model correlation. Then, the detailed and reduced models of a spacecraft sub-system will be described. Using the adjoint-based approach, we will correlate and calibrate the reduced model with the detailed model results. Finally, we will discuss the results and give some perspective of work for calibrating thermal reduced models.
