UC Berkeley

Safeguarding spent fuel reprocessing facilities and other bulk material handling facilities is a challenge. These high throughput facilities typically operate continuously and produce thousands of significant quantities worth of special nuclear material annually. This, combined with the high measurement uncertainties and extremely high radiation fields present, makes the timely detection of the diversion of a single significant quantity difficult. It is in these facilities where accurate and precise techniques are vital for the detection and prevention of both inadvertent and deliberate hold-up.

Our goal was to develop novel near-real-time accountancy techniques that take advantage of advanced radiation detection and imaging technologies, in combination with non-radiation signals present in bulk handling facilities. Fault detection and isolation methods were used to investigate material holdup and diversion in a closed-loop hydraulic system. An initial experimental loop was constructed to model the movement of nuclear material between material balance areas in a commercial PUREX reprocessing plant, and accurately induce material holdup and diversion. This process was done using commercially available and diluted radiotracers such as Tc-99m and F-18, to simulate the various radiation signals present in spent fuel. Not all faults in this loop were completely isolable and more collimation was needed to reduce the cross talk between radiation detectors.

The Next Generation Loop (NGL) was a re-designed reprocessing loop which can operate in a variety of configurations and boasts control of individual process streams while employing both radiation and non-radiation sensors to observe and detect off-normal behavior. Off- normal behavior that can be simulated may include material holdup and diversion, which can correspond to plant inefficiencies, faults such as blockages or leaks, or unknown/unauthorized nuclear material streams. A model network was utilized to detect and quantity off-normal behavior. With this improved reprocessing loop, all faults were completely isolable and material diversion was detected with reasonable accuracy.

Using advanced hybrid methods to reduce the computational burden associated with running high precision simulations of complex facility models, the previous two experiments served as benchmarks for variance reduction tools such as ADVANTG. An MCNP model of a concrete shielded Input Accountability Tank (IAT) inside of a commercial PUREX re- processing plant was created. The source term was generated using the Separations and Safeguards Performance Model (SSPM) developed at Sandia National Laboratories. This data was used to inform the type and placement of radiation detectors as well as the collimation and shielding required to obtain and verify fuel characteristics such as burn up, initial enrichment, and cooling time. A high purity germanium detector and a 1 cm diameter cylin- drical collimator through the concrete hot cell wall produced count rates on the order of 105 cps. Pu-239, Eu-154, Ag-110, Bi-214, and Y-88 were detected in the input accountability tank. Cs-134 and Cs-137 were expected to be observed but were not.

This work serves as a basis for the safeguards by design process within bulk handling facilities including advanced reactor facilities such as molten salt reactors. Using fault diagnostics and all available signals bulk handling facilities can leverage the collective expertise of the various stakeholders in the system design process to ultimately improve plant performance, reduce maintenance concerns, and communicate information about the facility’s operation to better safeguard nuclear material.