Mean cross section prediction in PWR-MOX using neural network
Résumé
Fuel depletion calculation codes require one-group mean cross sections of many
nuclides to solve Bateman's equations. In such codes, the mean cross sections are assed by the
mean of iterative calls of Boltzmann equation solver thanks to neutron transport codes. This is a
time consuming task, especially with Monte Carlo codes such as MCNP. This paper presents a
methodology based on neural network for building a cross section predictor for a PWR reactor
loaded with any MOX fuel. This approach allows performing fuel depletion calculation in less
than one minute with an excellent accuracy. A maximum deviation of 3% on actinides is obtained
at the end of cycle between inventories calculated from neural networks and from the reference
coupled neutron transport / fuel depletion calculation.