Project Highlights
High-Resolution Fluid Dynamics, Heat Transfer Simulations Conducted on Blue Gene First Step towards Petascale Simulations
 |
The accompanying image shows the coolant velocity distribution for a fuel pin in an advanced burner reactor (ABR) that is being proposed as one of the central elements in the U.S. Department of Energy's Global Nuclear Energy Partnership program. A key objective of ABRs is to reduce radioactive wastes and effectively increase the capacity of geologic repositories by a factor of 100.
Proposed ABR designs call for liquid-metal-cooled cores comprising thousands of fuel pins that are helically wrapped with wire. In addition to maintaining the proper pin spacing, the wire wrap serves to promote coolant mixing between pins and thereby leads to a more uniform temperature distribution. High-resolution fluid dynamics and heat transfer simulations such as those shown here are a first step towards high-fidelity petascale simulations of the reactor thermal distribution that will ultimately lead to improved economy and safety of ABRs.
The image shows an isosurface of axial velocity near the fuel pin and wire wrap. Ripples in this surface are markers of low-speed streaks associated with turbulent boundary layers. The streak pattern indicates that the flow tends to remain axial until it is on the lee side of the wire, at which point it follows the helical wire path. This observation is in contrast to the commonly made modeling assumption that the wire deflects the flow into adjacent channels. Further insight into the flow physics obtained from these and more detailed computations will lead to improved low-dimensional models that can be applied to the reactor design process.
The simulations were made with the fluid simulation code Nek5000 on the IBM Blue Gene computer at Argonne National Laboratory. Nek5000 is based on the spectral element method and is designed specifically for DNS, LES, and transitional flow simulations in complex domains. It is highly scalable and was recognized with the 1999 Gordon Bell prize for algorithmic quality and sustained performance on 4096 processors of ASCI-Red.
The discretization for the current problem comprises 25920 spectral elements of order 7 (approximately 9 million gridpoints) and required 16 hours on 1,024 processors to compute a single flow-through time. The flow reaches a fully turbulent state after a single flow-through time.
This work is supported by the Argonne Laboratory Directed Research and Development (LDRD) program and is one of the projects under the Petascale Computing and Lab-wide Theory Strategic Initiative.
