Multiphysics Simulations Transmuting Designs for Safer Nuclear Power

HFIR diagram showing co-centric cores.

Like the rest of the US’s nuclear research reactors, Oak Ridge National Lab’s (ORNL) high flux isotope reactor (HFIR) is moving from high-enriched uranium (HEU) fuel to low-enriched uranium (LEU). As such, the safety of the system must be assessed to incorporate the changes in fuel properties and the subsequently modified fuel plate.

Due to the recent growth in multiphysics, fluid-structure dynamics calculations can be coupled using a fluid-structure interaction (FSI) solver. The FSI solver is the key to the analysis of the HFIR system. Due to a built-in fully coupled FSI solver, and implicit solution capabilities, ORNL chose COMSOL to obtain their stable and precise solution.

The HFIR core features two co-centered fuel rings. The inner ring contains 171 fuel plates, while the outer ring contains 369 fuel plates. Each plate is 50 mils thick, and the shape of the reactor ensures constant cooling from the high-speed coolant. Due to changes in the design for the newer fuel, the structural integrity of the reactor must be studied once again.

Studying the theoretical model of this and similar setups has led to the Miller Critical Velocity, or M_c. Some experimentations are able to handle velocities twice the theoretical critical, however, which suggests that the Miller Critical value is conservative at best.

Model Computational domain. Y axis is into the paper.

A more robust method of analysis was needed. Until recently, however, simulations required too much CPU power to perform a coupled analysis. Furthermore, decoupling the fluid and structural analysis has led to model instability. ORNL’s simulation is designed to simulate the dynamic deflections of the LEU and HEU fuel plates. The light-water (coolant) is assumed to be incompressible and governed by Reynolds-averaged Navier-Stokes model. As for the structural mechanics, it is modeled as a linear elastic governed by Newton’s Second Law of motion, strain-displacement, and constitutive Hooke’s Law equations.

To improve stability, a fully-coupled steady state solver was used to determine the results of the system. To reduce the calculation times associated with coupled solvers, a course mesh was first used with a one-way coupled FSI solver. This was then used as an initial condition for a fully coupled solver, and the subsequent result was used as an initial condition with an increased mesh density. It now took only 3-4 hours to solve the system using 12 cores and 96 GB of memory.

Leading edge velocity field (6 m/s). Legend (m/s). Cut along the plate center (span-wise direction).
Simulations of ORNL’s HFIR have yet to be verified through experimentation. However, similar models have been verified with this simulation technique with the help of data from the University of Missouri (MU).

The results show the velocity of the channel increasing suddenly as the plate deflects the flow. Flow separation was present at the corners of the leading edge. The 23 mils deflection can be seen as represented by the black rectangle.

Comparison of simulation and experimental results of a 40 mil thick plate.

Currently, an experimental set up is under construction at Oregon State University. ORNL expects that the simulation will also be able to closely match these results; and with the system properly simulated, I’m sure that future design optimizations are just around the corner.

Source & images courtesy of Curtis, Ekici, & Freels, Comsol Conference 2013