Finite element model reduction for space thermal analysis

Abstract

To alleviate the computational burden of the finite element method for thermal analyses involving conduction and radiation, this paper proposes an automatic conductive-radiative reduction process based on the clustering of a detailed mesh coming from a structural model for instance. The proposed method leads to a significant reduction of the number of radiative exchange factors (REFs) to compute and size of the corresponding matrix. It further keeps accurate conduction information by introducing the concept of physically meaningful super nodes. The REFs between the super nodes are computed through Monte Carlo ray-tracing on the partitioned mesh, preserving the versatility of the method. The resulting conductive-radiative reduced model is solved using standard iterative techniques and the detailed mesh temperatures can be recovered from the super nodes temperatures for further thermo-mechanical analysis. The proposed method is applied to a structural component of the Meteosat Third Generation mission and is benchmarked against ESATAN-TMS, the standard thermal analysis software used in the European aerospace industry.