Speaker
Description
To prevent thermal induced breakdown, thermal modelling is a crucial task during each design phase of a new space mission. Monte Carlo ray tracing, while computationally demanding, is usually the go-to method to simulate thermal energy transfer. State of the art thermal modelling suites, such as ESATAN-TMS, usually rely on a CPU-based implementation of Monte Carlo ray tracing.
Due to the relatively long simulation times, only the mission points with the highest thermal loads, which have to be determined in advance, can be evaluated.
In 2018 NVIDIA announced their RTX GPUs, which feature dedicated ray tracing cores. This, together with their powerful OptiX ray tracing API allows for the
development of performant and highly optimized programs, where the developer
can solely focus on the ray behaviour.
The potential of these technologies for thermal modelling was investigated as part of an ESA project.
In the scope of this project, Fraunhofer EMI developed RayNer, a command-line-based technology demonstrator capable of determining black body view factors. The algorithm was verified against simple analytically solvable scenarios. Complex geometries can be built using the open source modelling software blender and
imported to RayNer in the GLTF format. Using realistic test geometries and similar sampling routines, RayNer determines the view factors up to 100 times faster than the ESATAN-TMS reference. The obtained view factors deviate only by ${\sim}10^{-3}$.
The complete pipeline and technology of RayNer will be discussed. Obtained results will be presented and compared to ESATAN-TMS.
As our results suggests, existing solutions, such as ESATAN-TMS, might experience a drastic speed increase when switching from CPU- to GPU-based ray tracing. RayNer could be developed further into a thermal ray tracing API, or into a slimmed down stand-alone tool for quick first thermal assessments.
Currently the implementation of environmental fluxes and a physically accurate
reflection model is being investigated.
