Back to overview


Type of publication Peer-reviewed
Publikationsform Proceedings (peer-reviewed)
Author Hursin Mathieu, Siefman Daniel, Perret Gregory, Rochman D, Vasiliev Alexander, Ferroukhi H,
Project Development of a Methodology for Nuclear Data Assimilation in Reactor Physics employing the PROTEUS Experimental Data Base
Show all

Proceedings (peer-reviewed)

Title of proceedings PHYSOR 2018: Reactors Physics paving the way towards more efficient systems
Place Cancun, Mexico

Open Access


SHARK-X is a set of Perl-based tools built around the lattice code CASMO-5 and developed at the Paul Scherrer Institut. SHARK-X is used to perform sensitivity analysis (SA), uncertainty quantification (UQ) and data assimilation (DA). The implementation of a new SA approach based on the calculation of Sobol Indices (SI) is presented, in which the correlations between input parameters are explicitly taken into account. The implementation in SHARK-X is verified using simple problems for which an analytical solution is available. An estimator of the probable error of the SIs is also provided. The numerical estimates of SHARK-X appear to converge after around 1000 samples. The large correlations existing in the input distributions of uncertain cross sections can be handled and SIs correctly calculated. Moreover large differences between SIs and sensitivity indices based on sensitivity coefficients are illustrated. Finally first order SIs were computed for a pincell depletion case involving linear and nonlinear neutronic responses. The analysis of the SHARK-X numerical estimates for the SIs showed that the current implementation can capture the non-linear behavior of a given model with respect to its input parameters. The computation of SIs is very expensive compared to other existing SA approaches when considering a large number of inputs. SIs based SA is better suited for models where non-linearity and correlations between a small number of inputs are involved.