High-Fidelity Light Water Reactor Analysis with the Numerical Nuclear Reactor

Weber, David P.; Sofu, Tanju; Yang, Won Sik; Downar, Thomas J.; Thomas, J. W.; Zhang, Zhong; Cho, Jin Young; Kim, Kang Seog; Chun, Tae Hyun; Joo, Han Gyu; Kim, Chang Hyo

doi:10.13182/nse07-a2672

Cited by 40 publications

(18 citation statements)

References 15 publications

(16 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Discussion here is focused on updates and verification and validation activities of the SHARP neutronics code, DeCART [2], for application to thermal reactor analysis. As part of the development of SHARP tools, the different versions of the DeCART code created for PWR, BWR, and VHTR analysis [4][5] were integrated. Verification and validation tests for the integrated version were started, and the generation of cross section libraries based on the subgroup method [3] was revisited for the targeted reactor types.…”

Section: Discussionmentioning

confidence: 99%

FY2012 summary of tasks completed on PROTEUS-thermal work.

Lee

Smith²

2012

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

FY2012 summary of tasks completed on PROTEUS-thermal work.

Lee

Smith²

2012

View full text Add to dashboard Cite

“…As an example, the global vertex connecting to the fifth vertex in the quadratic Lagrangian (lower middle) will have exactly 9 connections on the row corresponding to each vertex of that element. With respect to the global system of equations, assuming a structured mesh where each element is adjacent to at most 8 other elements, we can further state for the quadratic Lagrangian quadrilateral that each corner associated vertex (1,3,7,9) will have 25 connections per row and each side associated vertex (2,4,6,8) will have 15 connections. If we linearly tessellate this element as shown in Figure 2.2 and consider the same structured grid, we find that every vertex will have exactly 9 connections per row.…”

Section: Development Of a Multi-grid Preconditionermentioning

confidence: 99%

“…Early in fiscal year 2010 (FY2010), a similar project focused on high fidelity modeling of the VHTR and other thermal reactor concepts, called NNR [4], was merged in with the SHARP project. While these two projects are conceptually different, they both contain good ideas for reactor analysis modeling and we are motivated to merge the concepts together to produce a single deployable set of tools.…”

Section: Introductionmentioning

confidence: 99%

FY2010 status report on advanced neutronics modeling and validation.

Smith¹,

Mohamed²,

Marin-Lafleche³

et al. 2011

View full text Add to dashboard Cite

SUMMARYUNIC is the neutronics component of the massively parallel, multi-physics SHARP (Simulation for High-efficiency Advanced Reactor Prototyping) framework under development at Argonne National Laboratory. During this fiscal year, the SN2ND solver, MOCFE solver, and NODAL solver received significant development to meet the needs of the SHARP project. Additional follow-on analysis of the ZPR-6/6A from the previous year was performed in addition to new analysis of the ZPR-6/7 experiments where UNIC predictions of ZPR foil activation were made against experimentally measured values.The SN2ND solver was applied to a plate-by-plate model of ZPR-6/6A using 294,912 cores of BlueGene/P and 222,912 cores of XT5, the two largest open-science high performance computing machines. The calculations proved that the SN2ND solver could be applied to heterogeneous reactor modeling problems, but more solver development (e.g., multigrid preconditioner) and computing power are required before such calculations are routine. As a consequence, the SN2ND solver was revised to incorporate a new multigrid preconditioner concept in addition to removing the remaining inappropriate spherical harmonics related quantities left over from its beginning (i.e. PN2ND). At the time of this report, the new version of SN2ND has not been completed, and the follow on work for SN2ND will continue to focus on updating the preconditioner as outlined in this report.The MOCFE solver was rebuilt into UNIC in the previous year such that it obeyed the basic concepts of parallelism (scalable memory and communication). The MOCFE parallel algorithm was fully debugged this year and initial scalability tests on over 2048 processors were carried out such that the parallel algorithm could be assessed. That work indicated that a significant load imbalance in the coefficient matrix-vector application exists. A possible solution was formulated, but it has not been fully tested at the time of this report. Various setbacks caused by numerous ray tracing problems and a mistake in the implementation were unanticipated thus delaying progress on the MOCFE solver and the targeted development tasks for MOCFE were not completed this year.In addition to the high fidelity solvers SN2ND and MOCFE, some time was spent implementing the NODAL solver. NODAL is similar to an existing legacy tool, but employs parallelism for enhanced performance and a capability to map a heterogeneous geometry into the homogenized geometry. This solver would provide a path to improve upon the existing homogenization approaches used for fuel cycle analysis, transient analysis, and perturbation theory calculations. An appropriate preconditioner was identified for NODAL this year and the solver algorithm was partially completed in UNIC. To facilitate the validation tests of UNIC using the ZPR-6 critical experiments, the BuildZPRmodel tool was also updated and a mesh merging algorithm was created. In addition to the newly implemented "solution along a line" analysis capability, these tools proved crucial t...

show abstract

“…Table 3. 4 shows the details of the space-angle convergence for the k eff results. With the exception of the calculations using the finest mesh, all of the other calculations utilized the coarsest mesh.…”

Section: Partially Heterogeneous Modeling Of Zpr-6 Assembly 7 Experimentsmentioning

confidence: 99%

“…Given the widened scope of reactor types and the differences associated with cross section processing [4,5], a single neutronics code is not practical and thus development on UNIC as a single entity has ended. Instead, the individual solvers developed thus far: SN2ND, PN2ND, MOCFE, and NODAL have been separated such that they can be incorporated in the specific end use reactor projects they are being developed for.…”

Section: Introductionmentioning

confidence: 99%

FY2011 Status Report on SN2ND neutronics solver development.

Smith¹,

Mohamed²,

Marin-Lafleche³

et al. 2011

View full text Add to dashboard Cite

SUMMARYA change in focus of NEAMS to a broader set of reactor types necessitated significant effort being diverted away from SN2ND development in FY2011. Nevertheless, significant gains were made with regard to building the new preconditioner and meeting the objectives of 1) parallelism in energy, 2) a spatial multigrid concept, 3) using S N vector spaces to accommodate the adjoint and time dependent functionalities needed for NEAMS work, and 4) complete manuals and documentation. While the desired goal of an exportable SN2ND solver usable within the NEAMS community for neutronics analysis was not completed, we can be confident that this year's work puts the SN2ND solver well on its way to meeting that objective. It is important to note that SN2ND is just as capable as any other methodology for doing thermal reactor analysis, but the motivation to execute on smaller scale machines has initiated focused development on simplified treatments such as those proposed in the MOCARV tool.All of the cited coding tasks were completed but not all are fully working (i.e. debugged). The spatial multigrid technique was by far the most time consuming part of the SN2ND development process. It required multiple new data structure concepts along with identifying the necessary changes in the vectors and their associated functions. Combining the spatial multigrid as part of a GMRES preconditioner operating on the full vector space, necessary for parallelization in energy, increased the complexity of that development. Research into spatial multigrid techniques for unstructured meshes, void treatments, and code optimization and acceleration are key development needs for SN2ND.Similar to previous years, a considerable amount of effort was spent on validation work. Again we focused on fast reactor problems such as ZPR experiments and the MONJU reactor in Japan. Conventional drawer homogenized models and partially heterogeneous models were used on the ZPR calculations. The drawer homogenized models produced excellent results for both eigenvalue and foil measurements on four loadings of ZPR-6/7. The partially heterogeneous models of ZPR-6/7 also produced good results, but additional work is necessary to generate consistent effective cross section data. Calculations for MONJU were homogeneous and also gave excellent results.Finally, mock-up heterogeneous calculations were setup and attempted to scope out the performance limitations of SN2ND on whole core heterogeneous calculations. Specifically, heterogeneous models of a PWR, a VHTR, and the MONJU reactor were created using anywhere from 2 group to 33 group cross section data. Meshes with greater than 100,000,000 vertices are necessary for accuracy and a specific non-scaling memory issue in PETSc restricted the SN2ND modeling (VHTR and PWR could not be solved). All of the calculations performed this year indicate that significant computational resources beyond those available today are required to obtain a high fidelity solution possible with SN2ND.The future goals for SN2ND are to get th...

show abstract

High-Fidelity Light Water Reactor Analysis with the Numerical Nuclear Reactor

Abstract: Abstract

Cited by 40 publications

References 15 publications

FY2012 summary of tasks completed on PROTEUS-thermal work.

FY2012 summary of tasks completed on PROTEUS-thermal work.

FY2010 status report on advanced neutronics modeling and validation.

FY2011 Status Report on SN2ND neutronics solver development.

Contact Info

Product

Resources

About