Robust data assimilation for LFR nuclear data improvement

Neutronics models are extremely well-understood, with their major source of uncertainty being mostly associated to nuclear data. When it comes to fast reactor technology - such as a lead-cooled fast reactor (LFR) - the quality of these nuclear data fails to comply with the target accuracy requirements defined by the reactor design and safety studies. This affects the reliability of the neutronics analysis and it increases the associated costs by introducing non-negligible safety and operational margins. In this regard, SCK CEN has produced in the last decade an invaluable collection of experimentally measured neutronics characteristics (integral experiments) of the zero-power reactor VENUS-F: historically employed as a LFR mock-up. The past and recent developments in the field of inverse uncertainty quantification suggest that these experiments, when combined with a solid data assimilation technique, can be used to adjust nuclear data to reduce their uncertainty. This process will increase the reliability of the neutronics model predictions and decrease the large uncertainty margins that are currently considered in the LFR design.

The objectives of this PhD topic are the following:

• to identify what nuclear data uncertainties can be reduced by defining priorities in the form of target accuracy requirements;
• to investigate and implement robust data assimilation techniques, e.g., Bayesian statistics or machine learning, to reconcile nuclear data and integral experiments;
• to provide recommendations for novel nuclear data evaluations;
• to verify and validate the data assimilation methodology in the framework of a LFR design.

This PhD student will combine the knowledge in modelling with the boundary conditions defined by the experimental techniques and nuclear data evaluation processes. The PhD candidate will integrate in a recognized team of experts involved in national and international high-impact neutronics R&D.

La durée estimée