Intern

Background

Today’s core design, reload analysis and safety analysis are performed with several codes often developed in-house and licensed by regulators. The calculation chain is being executed serially and if optimization is performed it is for a specific software not for the overall process, i.e. only sub-optimization is performed.

If instead a parameter study for all possible solutions, this is solved and optimized defining objective functions, the overall most optimal solution can be found. A methodology to optimize a calculation chain is physically agnostic and could be applied to any analysis executing multiple software codes. It will be possible to incorporate high-fidelity simulations in workflows using Artificial Intelligence/Reduced Order Models. Implementation of all used computer software’s to one platform will enable model exploration and decision making through customized analysis workflows. A suite of core neutronics computer code, thermal hydraulics computer codes (system level, subchannel or CFD), and fuel performance computer code can be integrated to the platform.

Proposed decomposition of work

1. Review available platforms to be used for optimization of a workflow consisting of several computer codes. The platform should be able to create surrogate models of existing computer codes, define and optimize a workflow. List pros and cons for the reviewed platforms as input for decision.
2. Review already performed work within Fuel Engineering on the use of AI and constructing Reduced Order Models.
3. Run a selected in-house software code (POLCA/STAV/BISON/POLCA-T) for a range of input parameters in order to create a Reduced Order Model
4. Create a Reduced Order Model using data obtained from simulations
5. Define an objective function on an optimization platform and create a workflow with at least two different Reduced Order Models