Enhancing Nuclear Safety with a framework for Reproducable Nuclear Data Evaluation

Project title: Enhancing Nuclear Safety with a framework for Reproducible Nuclear Data Evaluation
Main applicant: Henrik Sjöstrand, Division of Applied Nuclear Physics
Co-applicants: Erik Andersson Sundén and Alf Göök, Division of Applied Nuclear Physics
Grant amount: SEK 3 000 000 for the period 2023-2026

The basis of all nuclear technology is nuclear data (ND), which is utilised as input in application codes to establish safety cases in the nuclear industry. To ensure full traceability of ND in the inspection responsibilities of the Swedish Radiation Safety Authority (SSM), comprehensive knowledge of all aspects of ND generation and use is required, including experimental analysis, ND modelling and evaluation, ND processing, and application calculations. Furthermore, well-founded uncertainty quantification (UQ) is necessary for all stages to ensure that applications meet safety standards. The project is a continuation of the work previously supported by SSM (Model calibration and model uncertainties-improved ND for improved nuclear security).

The project aims at

• further the use of ND uncertainties in application cases, such as criticality and material damage calculations,
• advancing the development in the domain of ND evaluation, and
• strengthening the general national competencies in model calibration and UQ.

The primary focus of this project is the continued development of an ND evaluation pipeline that enables transparent and reproducible ND evaluation, a requirement that is currently not widely met. This development will incorporate novel machine learning techniques, such as heteroscedastic Gaussian processes, to address complex experimental data structures, inconsistent data, and model inadequacies. Additionally, Bayesian networks will be employed to capture the fine structure in the mid-energy range.

In addition, the project will develop new techniques to integrate integral experiments into the calibration framework. Specifically, the project focuses on calibrating model parameters in the nuclear reaction code TALYS using both integral and differential experimental data. Integral data will be taken from, e.g. the international handbook of criticality safety. Information on differential data will be taken from the EXFOR database, from the Nuclear Energy Agency (NEA) Sub Group 50, and experiments conducted at the division. The use of machine learning methods will be explored to establish well-founded uncertainties, including addressing inconsistent data and model defects. The project's ultimate goal is to establish a pipeline that can be used for all isotopes of the periodic table, with a particular emphasis on important structural materials, such as iron, nickel, zirconium, copper, and chromium, that are critical to nuclear engineering applications.

The project aims to propagate ND uncertainties across a range of applications, including reactor safety [Hernandez2017], fuel storage and criticality safety [Sjostrand2013a, Östangård2013], back-end issues [TR 10 13], radiation protection [Sjostrand2013b], and material ageing [Sublet2019]. The project will involve developing an advanced uncertainty propagation framework. This framework will use both classical techniques based on random sampling and new techniques based on Halfway-Monte Carlo. Furthermore, the framework will be applied to various applications, with input from the reference group and funding agency influencing the final selection. The project offers flexibility in application selection. However, the current focus is to address fuel storage criticality and material damage under irradiation.

The research project will be led by principal investigator Henrik Sjöstrand, alongside co-applicants Alf Göök and Erik Andersson Sundén. The team will be supported by a reference group comprising national and Enhancing Nuclear Safety with a framework for Reproducible Nuclear Data Evaluation – SSM2023-1045 international experts. Project results will be disseminated through participation in evaluation projects at NEA and IAEA, as well as through contributions to relevant journals and conferences. The project aims to strengthen collaboration between Swedish ND groups and the international community, and its final products will include new open-source tools for ND evaluation, evaluated ND files for application use, and a well-founded quantification of ND uncertainties with an impact assessment on nuclear engineering applications in areas of SSM inspection responsibility.