Inverse Problem Approach for the underwater localization of Fukushima Daiichi fuel debris with fission chambers

Fuel debris have a distinct neutron signature that can be detected to locate the said debris in a damaged nuclear power plant. Neutron measurement in a damaged PCV environment is however submitted to severe deployments constraints, including a high-dose-rate gamma background and limited available space. The study was therefore oriented towards small fission chambers (FC), with U-235-enriched active substrates. To investigate the expected performance of the FC in various irradiation conditions, a numerical model of the detector head was built. We describe the elaboration and experimental calibration of the numerical model and the Monte Carlo study of the fission rate inside U-235 coatings per generated neutron. The evaluation of a representative calibration coefficient then allowed us to carry out a multi-parameter performance study of a FC underwater, aiming at computing an explicit response function linking, on the one hand, the activity and spatial distribution of neutron emitters in a water container, with, one the other hand, the expected count rates measured by a fission chamber as a function of its radial and axial position inside the water volume. The FC underwater behavior was subsequently corroborated by a measurement campaign on a FC response, set at different positions inside a water drum, as a function of its axial and radial distance to a Cf-252 neutron source attached near the center of the container. We finally present an approach in which fuel debris localization is defined as an Inverse Problem, solvable with a Maximum-Likelihood Expectation Maximization (ML-EM) iterative algorithm. The projector matrix is built by capitalization on the results of the previously consolidated numerical studies. The ML-EM was tested on simulated data sets with a varying number of active voxels. Our first results indicate that, for a thermal neutron flux in the order of 10 n.cm(-2).s(-1) at the detector, originating voxels are identified with a spatial resolution in the radial plane in the order of 10 to 100 cm(2).