An inverse-distance-based fitting term for 3D-Var data assimilation in nuclear core simulation

Abstract The evaluation of neutronic state is an important issue in the management and security monitoring of nuclear core. In particular, data assimilation is a powerful tool for such evaluation which combines the physical model and the observations in a coherent framework. In this article, an inverse-distance-based fitting term is proposed for the 3D-Var data assimilation, inspired from a recently-developed forecast-gradient formula. This term is constructed by a geometrical descriptor related to the Hessian matrix of the physical field, and doesn’t depend on the physical model nor the computation of the inverse of background error covariance matrix. A series of numerical tests are taken for various scenarios in nuclear core simulation, and it is observed that the proposed method performs effectively compared to other existing data assimilation methods. In addition, the proposed method has been integrated into the CORCA-3D code package to assimilate the observation data of nuclear reactors.