École Nationale Supérieure de
Techniques Avancées

Abstract

My end-of-study internship for ENSTA Paris and INSTN GA took place between April 2023 and September 2023 at the Japan Atomic Energy Agency (JAEA) in Tokai-Mura, Japan. My tutor was Dr. TADA Kenichi. Neutronics simulation codes rely on evaluated nuclear data libraries to represent reality. The JAEA is developing the Japanese Evaluated Nuclear Data Library (JENDL). The purpose of this internship was to validate its latest version : the JENDL-5, which was released in December 2021. To achieve this purpose, it was compared with the previous version : the JENDL-4, using the idealized criticality experiments of the ICSBEP (International Criticality Safety Benchmark Evaluation Project) which are called “benchmarks”. A set of 123 benchmarks, which is used for the validation of the Joint Evaluated Fission and Fusion (JEFF) library developed in OECD/NEA, was selected to carry out this study. The benchmarks were simulated with the Japanese Monte Carlo simulation code : MVP. The first part of this work was to compare JENDL-4 and JENDL-5 simulations with the “experimental” benchmark. MVP input files for each of the 123 benchmarks were created for both JENDL-4 and JENDL-5. For better comparison but keeping a reasonable calculation time, their precision was leveled. The comparison using the χ^2 indicator showed that the new library : JENDL-5 is more accurate than the previous one : JENDL-4. Other investigations were conducted to estimate the impact of the Oxygen-in-water and the different Graphite porosities, which were introduced by the JENDL-5, on the neutronics calculations. For the Oxygen-in-water reference, it showed that it significantly affects the calculation results and that its use should be mandatory. Once JENDL-5 was validated as more accurate, in the second part, we wanted to know which and how nuclides are responsible for this result improvement between JENDL-4 and JENDL-5. Each nuclide contribution was assessed by modifying one at a time the nuclide references from JENDL-5 to JENDL-4. If calculation results are changed, it meant that the nuclear data of the processed nuclide had indeed changed between JENDL-4 and JENDL-5. The difference of k-effective in each benchmark showed if a nuclide is important. The individual nuclide changes induced by one nuclide were summed and compared with the result of changing all nuclides together. It allowed us to make sure that no major nuclides were left out of the study. The choice of nuclide was crucial. Eventually the result analyzes led to a list of key nuclide reference responsive of most of the result improvement. Some special studies on the impurities and statistical verification of Monte Carlo calculation results were also carried out. The estimation of each nuclide impact was verified using an independent approach : a sensitivity analysis in the third and the last part of the internship. A new estimation of the importance of each nuclide was based on their cross-section and sensitivity. The sensitivity estimates the impact on the k-effective of a cross section modification on the benchmark. The sensitivity is defined for each benchmark, for each nuclide, for each reaction, and in each energy group. This independent approach validated the results of the previous estimation. In other words, this internship is about validating nuclear data by representing a set of experiments with a neutronic simulation tool. This internship is also comparing them with experimental results (part I) similar simulations with different nuclear data (parts I & II), and another type of simulations (part III). This internship was rather exploratory, it led to several possible extensions : methods to assess the spectrum effect, methods to use the internship data to call for cross-section modifications or study on missing thermal scattering reference.