Application of integrated relap5-transuranus approach to a RIA benchmark exercise including uncertainty and sensitivity analysis

A Reactivity Initiated Accident (RIA) Fuel Codes Benchmark was organized by the OECD/NEA Working Group on Fuel Safety. While the phase-I of the exercise focused on the analysis of power pulse tests performed on research reactors, the phase-II dealt with the benchmarking and comparison of participants’ modelling approaches: (task 1) to the analysis of several idealized RIA cases, and (task 2) with sensitivity analysis and uncertainty quantification (through the analysis of one particular RIA case). This paper describes the work performed by NINE as one of the participants to phase-II, and particularly on task 2. Such work involved the integrated use of two different codes, namely RELAP5 and TRANSURANUS for the thermal hydraulic and the fuel performance parts of the analysis respectively, the integration consisting in an off-line one-way sequential coupling. According to the benchmark specifications, the uncertainty analysis was performed by adopting a Monte Carlo statistical approach, requiring 200 runs with a number of uncertain input parameters randomly sampled from a specified Gaussian distribution, and by applying a 95/95 statistics to the selected output variables. Moreover, the sensitivity analysis was performed by calculating the Spearman’s rank correlation coefficients of specified output variables versus specified input parameters. The whole analysis process was handled through a Python-based “Statistical RELAP5-TRANSURANUS” (StaRT) coupled analysis programme developed on purpose. StaRT takes care of input parameter sampling, update of RELAP5 and TRANSURANUS input decks, post-processing of both codes’ results, uncertainty and sensitivity analyses. All the results produced at key steps of the process are stored in a single HDF5 file. StaRT turned out to be an effective tool for the purposes of the benchmark as well as for future applications to similar studies.