Title: The Advanced Fuel Cycle Initiative Status of Neutronics Modeling
1The Advanced Fuel Cycle InitiativeStatus of
Neutronics Modeling
- Won Sik Yang
- Argonne National Laboratory
- NEAMS Reactor Simulation Workshop
- May 19, 2009
2Status of Neutronics Analyses
- Within the current knowledge of physics, theory
and governing equations are well known - Boltzmann equation for neutron transport
- Bateman equation for fuel composition evolution
- The coefficients of these equations are
determined by nuclear data, geometry, and
composition - Nuclear data are for the most part relatively
well known for the most commonly used nuclides - But still improved data are required to reduce
design uncertainties - Geometry and composition have stochastic
uncertainties and are affected by thermal,
mechanical, irradiation, and chemical phenomena - These coupled phenomena are not as well
described, and they can dominate the analysis
errors - The challenge in neutronics analysis is to
determine the solution efficiently by taking into
account geometric complexity and complicated
energy dependence of nuclear data
3Reaction Rate Traverse Example
- Monte Carlo simulation with MCNP5 (INL)
- Reaction rate tally uncertainties lt 1
- C/E values for U-235 fission rate distribution in
CIRANO-2A (Blanket) and CIRANO-2B (Reflector)
experiments
4Negative Reactivity Transients of PHENIX
- Four unexpected scrams occurred in 1989 - 1990
due to short negative reactivity transients (200
ms) with the same signal shape - Several potential explanations were given, but
not satisfactory - Experiments are planned for PHENIX end-of-life
tests for further investigation
5Generation IV Target Uncertainties
- Current and Target Uncertainties for sodium
cooled fast reactors
Parameter Current Uncertainty (SFR) Current Uncertainty (SFR) Targeted Uncertainty
Parameter Input data origin Modeling origin Targeted Uncertainty
Multiplication factor, Keff (?k/k) 1 0.5 0.3
Power peak 1 3 2
Power distribution 1 6 3
Conversion ratio (absolute value in ) 5 2 2
Reactivity coefficients (component) 20 20 10
Control rod worth (total) 5 4 2
Burnup reactivity swing (?k/k) 0.7 0.5 0.3
6Objectives and Requirements
- The final objective is to produce an integrated,
advanced neutronics code that allows the high
fidelity description of a nuclear reactor and
simplifies the multi-step design process - Integration with thermal-hydraulics and
structural mechanics analyses to account for
reactivity feedbacks due to geometry deformation
accurately - Required modeling capabilities
- Reactivity and power distribution (coupled
neutron and gamma heating) - Non-equilibrium and equilibrium fuel cycle
analyses - Refueling, fuel shuffling, and ex-core models
- Perturbation and sensitivity analyses
- Uncertainty analysis and optimization
- Transient analysis (coupled with T/H and T/M
analyses) - Reactivity coefficients and kinetics parameters
- Shielding, decay heat, coolant activation and
dose rate calculations, etc.
7Selected Approaches
- Utilize modern computing power and computational
techniques - Meshing, domain decomposition strategies,
parallel linear solvers, new visualization
techniques, etc - Allow uninterrupted applicability to core design
work - Phased approach for multi-group cross section
generation - Simplified multi-step schemes
- Online cross section generation
- Adaptive flux solution options from homogenized
assembly geometries to fully explicit
heterogeneous geometries in serial and parallel
environments - Allow the user to smoothly transition from the
existing homogenization approaches to the
explicit geometry approach - Rapid turn-around time for scoping design
calculations - Detailed models for design refinement and
benchmarking calculations
8Adaptive Flux Solution Options
- Unified geometrical framework
- Unstructured finite element analysis for coupling
with structural mechanics and thermal-hydraulics
codes
9Flux Solvers Available in UNIC
- PN2ND
- Second-order, even-parity transport equation (CG
solve) - 1-D, 2-D, 3-D Cartesian with general reflected
and vacuum b.c.s - Spherical harmonics combined with Serendipity and
Lagrangian FE - SN2ND
- Second-order even-parity transport equation (CG
solve) - 2-D 3-D Cartesian with general reflected and
vacuum b.c.s - Discrete ordinates combined with Serendipity and
Lagrangian FE - MOCFE
- First-order transport equation (long
characteristics) - 3-D Cartesian with general reflected and vacuum
b.c.s - Discrete ordinates combined with Serendipity and
Lagrangian FE - NODAL hybrid finite element method for
structured geometries - Will replace nodal diffusion and VARIANT options
in DIF3D - Use as an multi-grid preconditioner for other
solvers
10Takeda Benchmark 4
Control Rod In Control Rod Half Control Rod Out
Reference 0.88001 0.00038 0.98340 0.00039 1.09515 0.00040
PN2ND 0.87960 0.98365 1.09599
SN2ND 0.87877 0.98275 1.09494
MOCFE 0.87796 0.98164 1.09353
11ABTR Whole-Core Calculations
Angular Directions Spatial Mesh Approximation Spatial Mesh Approximation Spatial Mesh Approximation Spatial Mesh Approximation Spatial Mesh Approximation
Angular Directions 78243 113873 461219 671219 785801
32 -241 -233 -69 -64 -59
50 -220 -210 -47 -40 -37
72 -225 -217 -51
98 -216 -207 -43
288 -216
12ZPPR-15 Critical Experiments
Flux expansion order Scattering order Eigenvalue
P1 P1 0.99258
P3 P3 0.99640
P5 P3 0.99651
Monte Carlo (VIM) Monte Carlo (VIM) 0.996160.00010
Computational Mesh and Example Flux Solutions of
ZPPR-15 Critical Experiment
132D OECD/NEA C5G7 Benchmark
Reference 1.18655 0.00010
MOCFE 1.18649
14Parallel Implementation
- The scalability to peta-scale computing resources
has been demonstrated - 163,840 cores of BlueGene/P (Argonne)
- 131,072 cores of XT5 (ORNL)
- Over 75 weak scalability
Weak Scaling Study by Angle on BlueGene/P (PHENIX
EOL test)
Cores 4p Angles keff Fission Iters. / Time Total Time (sec) Source Update (sec) Weak Scaling
32,768 32 0.96006 23 / 152 3493 2934 100
49,152 48 0.96004 23 / 152 3510 2933 100
65,536 64 0.96007 23 / 153 3526 2934 99
73,728 72 0.96015 23 / 156 3593 2934 97
131,072 128 0.96019 27 / 156 4209 3437 83
163,840 160 0.96019 27 / 173 4676 3436 75
15Parallel Implementation
Weak Scaling Study by Angle on XT5 (PHENIX EOL
test)
Cores 4p Angles keff Fission Iters. / Time Total Time (sec) Source Update (sec) Effective Weak Scaling
32,768 32 0.96017 25 / 63 1574 851 100
49,152 48 0.96014 22 / 64 1399 746 99
65,536 64 0.96017 22 / 64 1402 745 99
98,304 96 0.96017 25 / 65 1623 847 97
114,688 112 0.96017 26 / 65 1687 882 97
131,072 128 0.96029 28 / 68 1902 948 93
16PHENIX End-of-Life Experiments
- Participating in the PHENIX end-of-life
experiments - Whole-core geometry is required (no symmetry)
using homogenized fuel and explicit control rods - Space/angle convergence study completed using
over 4 billion DOF on up to 163,840 cores of Blue
Gene/P - Energy discretization study is ongoing
0.4 MeV Max/Min1.78 900 eV Max/Min16.9 2 eV Max/Min84.2
600 eV Flux and Radial Mesh
17ZPR-6 Critical Experiments
- Two ZPR-6 critical experiments are targeted for
VV in 2009 (Assemblies 6A and 7) - Explicit fuel plate representation allows direct
comparison to legacy homogenization methods - Spatial mesh requirements are large U-235 plates
are 1/16th in thick - Preliminary studies performed on BG/P and Jaguar
up to 130,000 processors indicate that over 10
billion DOF will be required to resolve the
space-angle-energy mesh
18ZPR-6 Critical Experiments
14 MeV Flux / Mesh
U-235 Plate Power
19Advanced Multi-group Cross Section Generation
Code MC2-3
- A modular version has been integrated into UNIC
for on-line generation of multi-group cross
sections of each spatial region with given
material and temperature distribution - Standalone code to generate ISOTXS datasets for
legacy tools - Ultrafine group (2082 groups) transport
calculations - Homogeneous mixture, and 1-D slab and cylindrical
geometries - Resolved resonance self-shielding with numerical
integration of point-wise cross sections using
the narrow resonance (NR) approximation - Unresolved resonance self-shielding with the
generalized resonance integral method - Elastic scattering transfer matrices obtained
with numerical integration of isotopic scattering
kernel in ENDF/B data
20Advanced Multi-group Cross Section Generation
Code MC2-3
- 1-D hyperfine group (100,000) transport
capability - Consistent P1 transport calculation for entire
resolved resonance energy range (lt 1 MeV) with
anisotropic scattering sources - Optionally used for accurate resolved resonance
self-shielding and scattering transfer matrix
generation - Efficient strategy to generate accurate
multi-group cross sections for heterogeneous
assembly or full-core calculations is being
developed by combining various solution options - 1-D hyperfine group cell calculation
- 1-D ultrafine group whole-core calculation (with
homogenized regions) - 2-D MOCFE calculation in several hundred groups
21MC2-3.0 and Coupling with UNIC
22Reconstructed Pointwise Cross Sections
(ENDF/B-VII.0)
23Hyper-Fine-Group Spectrum Calculation
- Inner core composition of ZPR-6/6A
24Hyper-Fine-Group vs. Ultra-Fine-Group Spectra
25LANL Criticality Assembly Benchmarks (UFG
Calculation)
- Multiplication factors are in an excellent
agreement within 0.15 ?? by taking into account
the anisotropy of inelastic scattering
26MC2-3 vs. VIM for ZPR-6/7 (Standalone UFG
Calculation)
Region VIM MC2-3 (Dk pcm)
Inner Core 1.22945 0.00038 10
Outer Core 1.22482 0.00048 -36
Radial Blanket 0.33513 0.00043 485
Axial Blanket 0.33215 0.00048 440
27ZPPR-15 Critical Experiments
28A Realistic View of ZPPR-15 Double Fuel Column
Drawer
29ZPPR-15 Critical Experiments
- Three loading configurations of ZPPR-15 Phase A
were analyzed - Loading 15 initial criticality
- Loading 16 reference configuration for sodium
void worth measurement - Loading 20 configuration with an 18 sodium void
in part of inner core
VIM - Exp DIF3D - Exp
Data Configuration Experiment VIM ?k, pcm DIF3D Sn ?k, pcm
ENDF/B-V.2 L15 1.00046 0.99647 -399 0.99525 -521
ENDF/B-V.2 L16 0.99627 0.99200 -427 0.99104 -523
ENDF/B-V.2 L20 0.99853 0.99529 -324 0.99428 -425
ENDF/B-V.2 Void Worth (pcm) 226 329 324
ENDF/B-VII.0 L15 1.00046 0.99985 -61 0.99905 -141
ENDF/B-VII.0 L16 0.99627 0.99571 -56 0.99489 -138
ENDF/B-VII.0 L20 0.99853 0.99742 -111 0.99741 -112
ENDF/B-VII.0 Void Worth (pcm) 226 171 252
Standard deviations of Experiment and VIM 0.00021 Standard deviations of Experiment and VIM 0.00021 Standard deviations of Experiment and VIM 0.00021 Standard deviations of Experiment and VIM 0.00021
30Summary
- An initial version of new multi-group cross
section generation code MC2-3 has been developed - Preliminary tests showed significantly improved
performance relative to MC2-2 - Integrated with UNIC for online cross section
generation - Consistent thermal feedbacks
- Account for spectral transition effects
- Second order solvers PN2ND and SN2ND have been
improved - SN2ND demonstrated good scalability to gt100,000
processors - Working on enhancing the anisotropic scattering
iteration - Fixing the load imbalance for reflected boundary
conditions - Starting next phase of pre-conditioner
development - p-refinement multi-grid and
- Algebraic multi-grid beyond that or possibly
h-refinement
31Summary
- First order solver MOCFE
- Improving parallel performance with Krylov Method
- Added more elements to ray tracing capabilities
- Adding back projection for parallel
- Started NODAL
- Implement Krylov solution technique to fix some
convergence problems - Eliminate memory problems and 1970s architecture
- Will investigate energy parallelization on
multi-core machines (8-32 cores)
32Backup Slides
33Perturbation Evaluation with MCNP (LANL)
- The MCNP perturbation option was used to
determine the difference in net neutron
production in every fuel assembly as a resulting
of reducing - Fuel density by 2, cladding density by 5, and
coolant density by 50 - While the fuel density reduction showed
reasonable results, the clad and coolant density
effects still showed significant statistical
variations - Observed statistical errors are less than 2 for
the fuel density perturbation - However, as large as 41 for the cladding density
perturbation and 100 for the coolant density
perturbation - Direct perturbation calculations showed even
worse results - Relative statistical uncertainties of the
re-converged production rates are often above
50, and in some cases reach 100 - The re-converged calculation ran 50,000 histories
per cycle for 160 active cycles, each of which
took 1000 minutes on a 2.7-GHz Opteron processor
34Convergence of Assembly Power Distribution
- NGNP with 60-degree periodic symmetry
- Core multiplication factor converges relatively
quickly - Power distribution converges very slowly
- Asymmetric assembly power distribution is
observed - Extremely large number of histories would be
required for converged pin power distribution
Number of neutron histories Number of neutron histories 100M 20M 5M
Eigenvalue Eigenvalue 1.45598 ? 0.0001 1.45599 ? 0.0002 1.45607 ? 0.0003
CPU time, hr CPU time, hr 1765 360 145
Variation, RMS 0.3 1.6 2.4
Variation, Max 0.6 3.2 5.4
34
35Depletion with Monte Carlo Method
- DB-MHR benchmark
- Cycle length 540 EFPD
- Total 7 cycles
- 6 burn steps per cycle (90 days interval)
- 50K and 100K neutron histories per burn step
- Note that there are 3 billion fuel particles
- Comparison of whole core depletions performed by
GA, BNL, and ANL - MONTEBURNS (MCNP5ORIGEN2)
- Simple cubic lattice model
- CPU time 40 hours for 50K and 100 hours for
100K histories - Much larger number of histories are required for
converged flux solutions
35
36CEA NEPHTIS Verification Results
Control Rod Position Control Rod Worth Control Rod Worth Control Rod Worth
Control Rod Position TRIPOLI4 38 pcm NEPHTIS, Diff. NEPHTIS, Diff.
Control Rod Position TRIPOLI4 38 pcm Homogeneous Heterogeneous
ARI 18,341 0.90 -1.05
ORI 7,083 0.98 0.64
SRI 5,676 -3.87 -3.35
- APPLO2172-group CP and 28-group MOC calculation
- CRONOS2 8-group diffusion calculation (finite
element method)
difference in fission rate distributions from
MCNP4C (3D core)
Heterogeneous Element
36
37Power Distribution of Fuel Block (CR Inserted)
37
38Effective Multiplication Factors for 2D and 3D
VHTRs with Heterogeneous Fuel Compact
Geometry Control Rod Position MCNP5 20 pcm DeCART , ?? pcm DeCART , ?? pcm
Geometry Control Rod Position MCNP5 20 pcm 190 Group 47 Groups
2D ARO- Standard block 1.46245 187 573
2D ARI 1.09752 14 788
3D ARO- Standard block 1.46379 439
3D ARO 1.45791 123
All Rods In (ARI)
All Rods Out (ARO)
Operating Rods In (ORI)
38
392D Power Distributions
ORI
ARO
ARI
39
402D Block Power Comparison with MCNP5
ORI
ARO
ARI
40
413D Flux Distribution for All Rods Out (ARO) Case
7 eV
1 MeV
0.13 eV
1 eV
41
423D Flux Distribution for Operating Control Rods
In (ORI)
7 eV
1 MeV
0.13 eV
1 eV
42