|
|
|
| News |
| Nuclear Energy |
| Petascale Algorithms for Reactor Hydrodynamics |
| Computations on the IBM Blue Gene/P are yielding important results in the analysis of reactor flows, thanks to new techniques developed at Argonne National Laboratory. One of the principal bottlenecks in simulating unsteady incompressible and low-Mach number flows typical of proposed fast nuclear reactors has been the elliptic problem governing the pressure. To overcome this, Argonne researchers, led by Dr. Paul Fischer, have developed a state-of-the-art code featuring high-order numerical discretizations, a new algebraic multigrid solver characterized by robust convergence properties, and a new technique for restructuring the central communication kernel. With this code, the researchers have been able to attack large problems capable of using up to 64,000 processors on the DOE's leadership-class machines. |
 |
| Figure 3. Turbulence in a seven-pin wire-wrapped subassembly visualized by axial velocity distributions. |
|
Recently, researchers achieved remarkable success in simulations of turbulent coolant flow in multipin assemblies (figure 3). The pins are separated by helically wrapped spacer wires that also divert flow between channels and enhance mixing. The simulations provide new understanding of the fundamental thermal mixing phenomena in reactor cores. These insights will drive improvements in safety, economy, and design of next-generation nuclear reactors.
|
