Qualifying Software for Nuclear-Safety-Related Calculations
- Friday, 04 September 2009
Following quality assurance procedures ensures that software used to perform analysis is producing the intended results.
The Department of Energy’s (DOE) Oak Ridge National Laboratory (ORNL) High Flux Isotope Reactor (HFIR) is the highest- flux reactor-based source of neutrons for condensed-matter research in the US. Thermal and cold neutrons produced by HFIR are used to study physics, chemistry, materials science, engineering, and biology, as well as produce unique radioactive isotopes for industry and research. As mandated by the DOE, ORNL must perform software quality assurance (SQA) procedures with special attention to nuclear-safety-related software applicable to the HFIR. The SQA process is followed to assure that the software used to perform an analysis is producing the intended results. The DOE requires that all software in their facilities satisfy a graded approach to SQA. For software used to perform nuclear-safety-related analysis, these requirements are more extensive.
The SQA procedure used at HFIR, called SBP-1300, was approved for use on June 6, 2001. Since that time, there have been approximately 58 separate computer codes approved through this procedure on about 33 separate computers. Recently, SBP-1300 was applied to COMSOL Multiphysics software so that nuclear-safety-related calculations may be performed using the software. In order for the calculations to be approved by the Research Reactors Division (RRD), the SQA must also be in place.
Examples of the nuclear-safety-related equipment at the HFIR facility that COMSOL might be applied to include safety plates that are inserted when the reactor needs to be shut down, pipes or valves that need to operate to keep the facility safe, or fuel plates that contain the nuclear fuel (see Figure 1). The nuclear-safety-related equipment being analyzed with a computer code might also require that the calculations be nuclear-safety-related. What makes a safety-related calculation different is that, not only does it go through a formal check and review process, but it also requires another independent review. The independent review process strongly encourages an alternative calculation be performed, or better yet, a test or experiment be performed to demonstrate the validity of the calculation.
The verification that a specific software code is installed and producing the results expected by the developer is the main focus and what is performed by the SQA procedure. Representative problems are taken from the software manuals that are typical of what is applied to HFIR, and these are executed on the qualified computer to verify that the results obtained are as intended by the code developers.
In addition to the representative problems extracted from the COMSOL manual, two additional problems were created to perform the SQA. These two problems were developed to verify both the finite-element convergence rate and the parallel-processing performance of the software. These problems are also typical of what might be analyzed at HFIR.
Finite-element convergence gives the user a desired numerical accuracy with the minimal level of resources required (number of nodes, amount of memory, CPU time, etc.). From these convergence curves, the user has some guidance for what level of noding to stop at to achieve a given level of accuracy (see Figure 2).
This work was performed by Dr. James D. Freels of the Research Reactors Division at Oak Ridge National Laboratory using COMSOL Multiphysics software. For more information, visit http://info.hotims.com/22926-127.