College of Engineering Unit:
The Dragon assembly was constructed during the Manhattan Project to test the feasibility of a nuclear chain reaction sustained by prompt neutrons alone. The Dragon assembly core consisted of a fuel region of highly enriched uranium hydride combined with a plastic binder. The core was surrounded by a thin layer of cadmium and a beryllium oxide tamper at the outermost layer. A descending fuel slug driven by gravity caused the system to briefly go prompt supercritical as it traveled through a cavity within the core.
Previously, the Dragon assembly was modeled using a Point Reactor Kinetics Equations (PRKEs) approach within the article “Critical Assemblies: Dragon Burst Assembly and Solution Assemblies” by Dr. Robert Kimpland et al. The PRKEs in the article were modified to include multiphysics and moving-geometry parameters. The work done in the article presented an opportunity for future work involving the expansion of the spatial dimensions and implementation of multi-physics feedback regarding the fuel temperature.
It is necessary to develop computational models of rapid transients such as those from the Dragon experiment to accurately predict the behavior of the neutron population throughout the transient for safety and experimental design. Historically, rapid reactivity transients have been primarily modeled using deterministic transport methods. Monte Carlo methods allow for dynamic phase space sampling, eliminating discretization errors. The error incurred by Monte Carlo simulations is statistical and the solution follows a convergence rate of 1/N, where N is the number of particles simulated. Although this convergence rate is significantly slower than deterministic methods, the high-fidelity geometry and non-homogenized continuous energy cross-section data allow for accurate modeling of complex physical systems.
A moving geometry model of the Dragon assembly has been developed in the VTT Technical Research Centre of Finland Monte Carlo code, Serpent. The initial Dragon model was constructed to capture the general material and geometric parameters from the literature. Transient simulations with this model produced peak assembly powers far greater due to incorrect maximum and minimum reactivity conditions. However, the Serpent transient results compared favorably to a PRKE model using the same reactivity curve and mean generation time. This indicates that Serpent accurately captured the neutron population's overall behavior versus time. In order to generate a Serpent model more representative of the neutron population in the physical experiment, a scheme was devised and is currently being carried out. This scheme modifies the assembly geometry to generate reactivity values that are equal to the experiment at set points. Neutron dose was calculated through the Monte Carlo N-Particle (MCNP®) stochastic neutron transport code; this was achieved by using the outgoing current data from Serpent as a point source to find energy deposited per mass in water one meter away from the experiment. Under a modern lens, the materials used in the experiment pose a significant health hazard. Considering this, alternative fuel and tamper materials, typically used in contemporary criticality experiments, were explored with the goal of a modern redesign. The best candidates are uranium zirconium hydride fuel with a polyethylene moderator and a copper tamper.
