Image
Dorea Technology is a small company founded in 2006, based in Sophia Antipolis, specialized in thermal and mechanical simulation software development for the space sector. The company works with prestigious partners (Thales Alenia Space, ESA, CNES) and is a member of the Network of Experts for Space Thermal Analysis (NESTA).
Technical/scientific Challenge:
In-flight thermal correlation of satellite models requires the optimization of numerous uncertain parameters, including optical properties, contact conductances, and environmental conditions. Evolutionary optimization algorithms are well-suited for these high-dimensional problems, but each iteration requires simulating at least one complete orbit, leading to prohibitive computation times on standard workstations. As part of an ESA project on a satellite thermal digital twin, Dorea aimed to leverage high-performance computing to accelerate these automated correlations and efficiently explore the parameter space.
Solution:
CRIANN supported Dorea through seven technical meetings, assisting with supercomputer familiarization, software environment setup, and code parallelization. The solution involved parallelizing evaluations required by evolutionary optimization algorithms across multiple CPU cores on CRIANN's Austral supercomputer. Python with the GEMSEO library (Generic Engine for Multi-disciplinary Scenarios, Exploration, and Optimization) was used to implement the parallel approach. This strategy allows simultaneous execution of orbit simulations for each algorithm iteration, drastically reducing convergence times while enabling comprehensive exploration of the parameter space.
Business impact:
Initial tests achieved a threefold performance gain using 48 CPU cores compared to 20 cores on a standard desktop machine. This acceleration enables efficient distribution of orbital simulations required by evolutionary optimization algorithms, significantly reducing convergence times and allowing quantitative sensitivity assessments of solutions. As a result, the project now enables correlation quality that would have been impossible under conventional computational constraints. The work has been submitted to the ICES 2026 conference (International Conference on Environmental Systems) organized by NASA, highlighting its scientific relevance and impact.
Benefits:
- Parallelized evolutionary optimization for high-dimensional satellite thermal problems
- Threefold performance improvement using 48 CPU cores on the Austral supercomputer
- Efficient exploration of large parameter spaces and reduced convergence times
- Enables quantitative sensitivity assessment and high-quality orbital correlation
- Supports ESA satellite thermal digital twin research with HPC-driven acceleration