Speaker
Description
In the design, construction and operation of breeder blankets, there is large uncertainty in a range of properties including the geometry, material properties, physical parameters, isotope composition and nuclear cross-sections. Uncertainty quantification (UQ) methods aim to quantify how variability in each of the properties drives variability in the performance of the system. This allows design engineers to understand the key drivers and variability in their system, leading to more robust, feasible and reliable designs.
CFMS has been collaborating with UKAEA to perform UQ on the pin-cell breeder blanket design used in the LIBTRI programme. Two models were considered to simulate different processes in the system: an OpenMC neutronics model for tritium production, and a MOOSE finite element based model for tritium advection and extraction. Each model was fully parametrised, allowing for arbitrary variation of various parameters within given bounds. Automated pipelines were created for geometry generation, meshing, simulation execution and results extraction which allowed for a large number of models to be evaluated on HPC.
A fourier-based UQ method was applied to both models and the first order sensitivity indices were extracted. The key drivers for both models were identified and can now be targeted by UKAEA to reduce variability. Parameters with little contribution can have their tolerances loosened, saving cost. Overall, a better understanding of the system was achieved, which can help inform future design choices.
| Speaker affiliation | Centre for Modelling and Simulation |
|---|
