GENPROM Development for Detection of Diversion Scenarios in Fuel Processing Plants

Under the sponsorship of US DOE/NA-22, Argonne National Laboratory (ANL) is currently investigating the concept of focusing the on-line monitoring and diagnostic systems developed specifically for nuclear power plant operations and predictive maintenance, on the subject of detection of material diversion in nuclear fuel reprocessing facilities as a means for proliferation resistance. ANL has been modifying the IGENPRO nuclear plant operator advisory system technology based on the MSET (Multivariate State Estimation Technique) and PRODIAG (Process Diagnostics) modules developed for monitoring and diagnosing thermal-hydraulics (T-H) systems. Once combined with chemical process knowledge databases it would become GENPROM (Generic Process Monitor) for fuel reprocessing plants. This paper presents recent accomplishments in the conversion of MSET from monitoring the thermal-hydraulic process variables of nuclear power plants to monitoring alarm conditions traceable to suspect reconfigurations of reprocessing plants for proliferation diversion of strategic nuclear material. Real-time data from a representative solvent extraction flow-sheet of the UREX+ process is being used to test the ability of the algorithms to demarcate the bounds of normal operation of the selected process. Experiments, conducted in the ANL solvent extraction facility, were performed to provide real-time process data from a representative solvent extraction unit operation. This experimental data was used in the initial evaluation of the MSET monitoring algorithms. The initial feasibility results show the successful application of the MSET package toward detection of anomalies in process variables of a representative solvent extraction unit operation of a reprocessing plant, for nominal and off-nominal operation conditions. But, there is a need to enhance the noisy signal robustness of the off-nominal hypothesis algorithms to produce stable alarms. This future concept development will be carried out by focusing on the DECIDE sub-module algorithms in MSET.