-
International Conference on Mathematics and Computational
Methods Applied to Nuclear Science and Engineering (M&C
2011)
Rio de Janeiro, RJ, Brazil, May 8-12, 2011, on CD-ROM, Latin
American Section (LAS) / American Nuclear Society (ANS) ISBN
978-85-63688-00-2
MULTIGROUP CROSS SECTION COLLAPSING OPTIMIZATION
OF A HE-3 DETECTOR ASSEMBLY MODEL USING
DETERMINISTIC TRANSPORT TECHNIQUES
Mi Huang and Ce Yi
Department of Nuclear and Radiological Engineering
University of Florida
202 Nuclear Sciences Bldg., Gainesville, FL 32611
[email protected]; [email protected]
Kevin L. Manalo and Glenn E. Sjoden
Nuclear and Radiological Engineering and Medical Physics
Program
Woodruff School of Mechanical Engineering, Georgia Institute of
Technology
770 State St, Atlanta, GA 30332
[email protected]; [email protected]
ABSTRACT
Multigroup optimization is performed on a neutron detector
assembly to examine the validity of
transport response in forward and adjoint modes. For SN
transport simulations, we discuss the
multigroup collapse of an 80 group library to 40, 30, and 16
groups, constructed from using the
3-D parallel PENTRAN and macroscopic cross section collapsing
with YGROUP contributon
weighting. The difference in using P1 and P3 Legendre order in
scattering cross sections is
investigated; also, associated forward and adjoint transport
responses are calculated. We conclude
that for the block analyzed, a 30 group cross section optimizes
both computation time and
accuracy relative to the 80 group transport calculations.
Key Words: SN transport, detector response, multigroup,
collapse, adjoint
1. INTRODUCTION
Transport simulation offers tremendous opportunities for highly
detailed information for neutron
spectroscopy applications. Concurrently, computer architectures
continue to rapidly evolve to
multicore CPUs and GPUs, which aids all engineering software
simulation. However, there still
remains a need to remain practical if one wishes to reduce
computation time from several hours
to minutes. Multigroup macroscopic cross section collapsing is
one obvious treatment. In
particular, we examine the validation of a multigroup collapse
sequence for a fixed source
detector problem using the PENTRAN 3-D SN Transport code and the
YGROUP Multigroup
cross section collapsing code.
We are interested in the construction of neutron detector
assembly models for SNM detection, for
which counts are optimized to specific energy ranges. The
evaluation and selection of moderator
materials was previously performed to optimize four distinct
energy bands [1]. We choose to
examine 1 of 4 blocks (corresponding to each energy band) from
the prior study, and present a
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
2/19
cross section view of the basic detector model in Figure 1 [1] .
The dimensions along a
prescribed x-axis are given as 16 cm concrete, 1 cm Hafnium, 1.5
cm High-Density Polyethylene
(HDPE), 1 cm He-3, 7.5 cm HDPE. The specific energy range
targeted by this block design is
approximately neutrons from 1.0 MeV to 3.68 MeV.
We began our study with an 80 group library (upper energy values
listed in Appendix A) and
worked to collapse the library further. An 80 group structure
may be computationally
burdensome, particularly with upscatter, as wall clock times for
transport (in parallel, with 80
CPUs) require on order of 7 to 8 wall clock hours (S30
Legendre-Chebyshev quadrature [960
directions per mesh cell] with a P5 Legendre Scattering order)
to appropriately converge the
adjoint block models. In order to improve computational
efficiency yet preserve group accuracy,
an optimal strategy would entail reduction of 80 energy groups
to an even broader group
structure, perhaps somewhere between 10 to 20 energy groups,
without (hopefully) any loss of
fidelity of the calculation. Originally, we performed adjoint
transport calculations using the
detector as an adjoint source. In order to complement these
runs, we perform standard forward
transport calculations, with the application of a half-isotropic
(directed towards detector),
uniform surface source located just outside the front detector
assembly face. For an individual
detector assembly, with paired calculations of forward and
adjoint transport, we can apply the
results and input them to the YGROUP code, which optimizes the
energy group collapse using
several options.
2. CALCULATION OF FORWARD AND ADJOINT TRANSPORT RESPONSE
In this section we briefly discuss the evaluation of forward and
transport response (in the He-3
detector), since this is computed in this work, and is used to
evaluate model consistency. Similar
analyses have been performed [2], but we now consider
calculation of response in terms of the
multigroup collapse; we proceed to outline the basic theory and
calculation steps.
Figure 1. Detector Assembly Block
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
3/19
2.1. Multigroup Cross Section Collapsing
The YGROUP code performs the cross section collapsing [3], and
was developed as part of a
code (funded through computational NNSA research with Sandia
National Laboratory) to speed
up deterministic particle transport simulations by reducing the
number of discrete energy groups
while maintaining computational transport accuracy [4]. The
YGROUP code helps to achieve
this through application of the contributon approach originally
developed by Alpan and
Haghighat to automate collapsing/group selection [5]. This
contributon collapsing option aims
to divide the broad group bin structure uniformly in the sense
of relative contributons from
each fine group. Contributons are defined by
3 *, , , ,V
C E d r d r E r E
(1)
Where and * are the forward flux and adjoint functions,
respectively. Equation (1) specifies
the energy-dependent contributon for a particular calculation
objective (as defined by the adjoint
function). For example, if the detector response is the
objective, the detector response cross
section can be used as the adjoint source. The calculated
adjoint function is a measurement of a
particles importance to the detector.
The group-dependent discretized contributon (only considering
zeroth order moments) is given
by:
*g gg ii
C V (2)
where iV
is the discretized fine mesh volume with the same material in
the model, g and *g
are the average forward scalar flux and inverse adjoint
function. Therefore, Cg is a measurement
of a Group g particles importance to the objective. By using Cg
as the weighting function during
cross section data collapsing, the broad group cross section
could conserve the detector response
better than using forward flux as weighting function.
YGROUP operates in the following manner. First, forward and
adjoint deterministic transport
calculations are performed on a smaller problem model, or on one
section of a large problem
model representative of problem physics using a fine group
structure. Then, the calculated
forward flux and adjoint function moments are used by YGROUP to
collapse the fine group
cross section library and generate a problem-dependent broad
group cross section library.
Finally, the broad group library is used for new transport
calculations on the full scale/ refined
problem model, and results are compared to evaluate the
effectiveness of the group collapse.
YGROUP provides several weighting options to both determine
group structure, and also
collapse the cross sections. Users can also specify fine groups
in specific energy ranges of
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
4/19
interest to be reserved after collapsing. YGROUP also can be
used to evaluate the Feynman-Y
asymptote characterizing neutron multiplicity, although this
feature is not used here.
2.2. Transport Response
Since the user must select a material as the basis for flux
weighting, in all detector blocks, we
considered two materials/regions in ths study: He-3 + HDPE
(evenly weighted 50%, 50%) in the
detector assembly, as well as for He-3 alone (100%). There are
other considerations one could
make, such as using all materials defining the assembly,
although we are primarily concerned
with the neutron response of the He-3 tube. The detector
assembly, with forward transport (with
an isotropic uniform energy source, as mentioned before) and
adjoint transport models (aliased to
the absorption responses in He-3) were validated with regard to
convergence by computing
detector response based on both the forward flux and
independently using the adjoint response,
calculated using the expression for detector response R :
*det,g g g gR q (3)
From equation (3) , it is clear that detector response can be
obtained by complete integration of
the source distribution with the adjoint functionfor any
arbitrary source distribution.
Therefore, R can be computed directly from the results of either
of several forward transport
computations for each neutron source distributions, or one
single adjoint transport computation
with coupling to each source density distribution. The true
calculated ratio between these two
methods of computing responses should be unity, but deviations
from unity result typically as a
result of numerical truncation error; in theory, forward
response and adjoint response
independently computed should be equal to each other. In the
final results, we will identify
common trends by examining response ratios.
2.2.1. Response Calculation in Discretized Form
First, we discuss the calculation of transport detector response
in discretized form. We can
simply sum over the groups, from 1 to G, using index g. The
units are shown in square brackets.
3det, det, det 21
1 #
sec
G
g g
g
V cmcm cm
(4)
We assume that the detector is usually comprised of a volume and
not a surface. The units show
clearly that we expect a response rate.
For adjoint transport, the equivalent response is given by
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
5/19
33
*
,
1'
, #sec
1)( cm
cmVq srcgsrc
g
g
gsrc
(5)
By unit analysis, the forward source term provides a group
dependent rate (and can be per unit
volume, the importance function is unitless, and the volume will
cancel the preceding term. So
again, we arrive with adjoint response having the same
units.
But if we define a surface source (in our case a J surface
source), then the respective adjoint
equation changes to the form of
22,
*
1
, #sec
1))(( cm
cmAJq srcgsrc
g
g
gsrc
(6)
We use a convention where we normalize the response by dividing
by the integral source term.
The normalization step is important for cross-comparison to
Monte Carlo and other transport
codes.
2.3. A Brief Description of Codes Used
PENTRAN is a parallel discrete ordinates SN solver in Cartesian
Coordinates [6]. There are
many recent advances with the code, with both high performance
computing and adaptive
differencing recently investigated [7]. In a parallel
computation environment, files are also
created based on the decomposition strategy used.
Post-processing tools are quite useful, such as
the PENDATA code, which extracts parallel data from PENTRAN. For
the analysis presented in
this paper, we have also incorporated appropriate scripting to
auto-calculate transport response
from PENTRAN using the PENGRAB code.
As previously discussed, material fluxes associated with full
multigroup transport are used to
serve as weighting functions for the calculations performed in
YGROUP. To be clear, both the
selection of groups collapsed and subsequent reweighting of
collapsed cross sections are handled
by YGROUP. As an example, if a user requests 35 groups, YGROUP
may enforce collapsing to
38 groups. Users can also supply a fixed list of fine groups to
collapse into a broad group. The
options provided by YGROUP should be considered carefully
especially when collapsing to less
than 10 groups. For this study, we use the contributon weighting
option available with the
collapsing code.
2.3. Model Parameters for Transport Codes
For the SN transport model, the forward model applies a
half-isotropic source in the directed
toward the detector assembly, on the detector assembly face.
Vacuum boundary conditions are
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
6/19
used in the transport calculations. All models were run using
S30P1 and S30P3, using a Legendre-
Chebyshev (Pn-Tn) quadrature. All or some models employed both
angular and group
decomposition, with a minimum of group decomposition performed.
Instead of a standard
multigroup iteration, we applied the Hiromoto-Weinke multigroup
iteration method as applied by
the PENTRAN code, which performs a single iteration in each
group while stepping toward
convergence. Also, we applied the default DZ-DTW-EDI adaptive
spatial differencing strategy.
The inner flux tolerance (infinity norm) was set to 1.0E-03.
3. RESULTS
A series of 16 (2 x 4 x 2) transport calculations were performed
on the detector assembly block
to investigate model performance. They are broken down into the
combinations of : forward v.
adjoint, 16 vs. 30 vs. 40 vs. 80 groups, P1 vs. P3 Legendre
order in scattering cross section.
The first two figures give a detailed view of an 80-group flux
in the detector in Figure 2, and also
of the adjoint current importance (exiting thru the detector
face) in Figure 3.
Figure 4 is a more general examination of the spatial
distribution of the adjoint importance
function for the mean-weighted and max-weighted energy groups of
the P1-80 group (red block
is the detector assembly, outside of the block is air). We can
see that the peaks in spatial adjoint
importance occur in the polyethylene, outside of the He-3.
A combination of all the case results for the adjoint are given
by Figure 6. Similarly,
combinations of case results are given for forward transport in
Figure 7. Immediately, we can
see calculation fidelity is not well maintained with the 16
group structure in red in the epithermal
range, and is more apparent in the adjoint transport.
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
7/19
Figure 2. Detector Flux (in He-3 region) in
Forward Transport for 30 Groups and 80
Groups.
Figure 3. Partial Adjoint Leakage (at the Detector
Assemblys Front Surface) for 30 Groups and 80
Groups.
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
8/19
The average energy as calculated by Figure 4 is calculated
(along a prescribed x-axis, 1D) is given
in equation (7), where gE is the average energy for group g, is
the polar angle, and ,adj
g x
is the group adjoint function.
1 0
1 0
,
,
Gadj
g g
g
Gadj
g
g
E d x
E
d x
(7)
Defining the partial adjoint leakage function as
0
,adj adjg gJ x x d
(8)
Figure 4. Adjoint Importance along x-axis at
Average Forward Group 11 (of 80 groups) and
Max Forward Group 13 (of 80 groups).
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
9/19
we can simplify equation (7) to
1
1
Gadj
g g
g
Gadj
g
g
E J x
E
J x
(9)
Figure 5. Adjoint Transport - Relative Jx-1/2 Adjoint Leakage
Function for All Groups on Log-
Log scale.
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
10/19
Another interesting examination to be made is comparing P3 vs.
P1 Legendre order.
Ratios by energy are given on a Log-Log scale in Figure 7 and 8
for adjoint and forward
transport, respectively. From Figure 7, the 80 group case
oscillates more due to a better
representation of cross section detail. An effective smoothing
occurs as a function of increasing
group collapse. It is also apparent that the 16 group case is
losing information in the epithermal
range, decreasing accuracy. Figure 8 showcases the ratio of P3
to P1 as a function of log energy;
we see that scatter effects are noted in the fast energy range
(as expected), likely due to
improvements in hydrogen scatter cross section in polyethylene
surrounding the He-3.
Figure 6. Forward Transport - Relative Detector
Flux for All Groups on Log-Log scale.
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
11/19
Figure 7. Adjoint Transport P3/P1 Partial
Adjoint Leakage Ratio for All Groups on
Log-Log scale.
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
12/19
Finally, we present the tabulated results of the response,
presenting forward response, adjoint
response, and their ratio (adjoint/forward) in Table I for P1
and Table II for P3. If we ratio the
results, we can affirm that P3 is higher in the forward and
adjoint cases than compared to P1.
As was previously discussed in the basic theory section, we note
that all forward/adjoint
response ratios are consistently close to 1 (worst percent
difference was 9.7 % in the P3-16 group
case). There is a monotonic trend decrease with increasing
groups, which may also be indicative
of Legendre term expansion impact with more scattering cross
section values.
Figure 8. Forward Transport - P3/P1 Detector
Flux Ratio for All Groups on Log-Log scale.
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
13/19
Table I. P1 Transport Results: Forward Response, Adjoint
Response, and Response Ratio
P1; # Groups Forward Adjoint RR
16 2.784 10-12
2.889 10-12
1.037
30 1.260 10-12
1.255 10-12
0.9961
40 1.325 10-12
1.276 10-12
0.9630
80 1.164 10-12
2.500 10-2
0.9317
Table II. P3 Transport Results: Forward Response, Adjoint
Response, and Response Ratio
P3; # Groups Forward Adjoint RR
16 3.143 10-12
3.448 10-12
1.097
30 1.416 10-12
1.496 10-12
1.056
40 1.490 10-12
1.517 10-12
1.018
80 1.304 10-12
1.274 10-2
0.9773
Table III. P3 to P1 Ratio: Ratios of Forward Response, Adjoint
Response, and Response
Ratio
# Groups Ratio of Forward
Ratio of Adjoint
Ratio of RR
16 1.129 1.194 1.057
30 1.124
1.192
1.061
40 1.125
1.189
1.057
80 1.120
1.175 1.049
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
14/19
3. CONCLUSIONS
From examination of a variety of transport models, computations
using the 30 group library
collapsed using the contributon approach in YGROUP yielded the
minimum group structure
while preserving problem accuracy. The P1-30 group calculation,
and associated multigroup
cross sections from YGROUP contributon weighting process,
yielded a reduced transport
computation time (< 10 minutes), and that response is
calculated to a reasonable accuracy with
response ratio calculated to 0.996. Computation times of all
cases are listed in Appendix A.
Future work will continue with similar analysis of the three
other detector blocks targeting the
distinct neutron energy bands of 31.8-369 keV, 0.369-1.0 MeV,
and 3.68-17.3 MeV [1].
ACKNOWLEDGMENTS
This research was made possible with support by NNSA.
REFERENCES
1. Ghita, G., G. Sjoden, and J. Baciak, "On Neutron Spectroscopy
Using Gas Proportional
Detectors Optimized by Transport Theory". Nuclear Technology,
(2009). 168(3): p. 620-
628.
2. Sjoden, G.E., "Deterministic adjoint transport applications
for He-3 neutron detector
design". Annals of Nuclear Energy, (2002). 29(9): p.
1055-1071.
3. Yi, C. and G. Sjoden, "YGROUP User Manual: Multigroup Cross
Section Data
Collapsing Code Using Contributon Weighting Scheme". (2010),
University of Florida.
4. Yi, C., et al. "Computationally Optimized Multigroup Cross
Section Data Collapsing
Using the YGROUP Code". in PHYSOR 2010. (2010). Pittsburgh, PA:
American Nuclear
Society.
5. Alpan, A. and A. Haghighat, "Development of the CPXSD
methodology for generation of
fine-group libraries for shielding applications". (2005), La
Grange Park, IL: American
Nuclear Society. 14.
6. Sjoden, G. and A. Haghighat, "PENTRAN - Parallel Environment
Neutral-particle
TRANsport, Code Users Guide/Manual, Version 9.4X.5". (2008), HSW
Technologies
LLC.
7. Sjoden, G., et al. "Automatically Tuned Adaptive Differencing
Algorithm For 3-D Sn
Implemented in PENTRAN". in International Conference on
Mathematics,
Computational Methods & Reactor Physics (M & C 2009).
(2009). Saratoga Springs,
New York: American Nuclear Society.
-
International Conference on Mathematics and Computational
Methods Applied to Nuclear Science and Engineering (M&C
2011)
Rio de Janeiro, RJ, Brazil, May 8-12, 2011, on CD-ROM, Latin
American Section (LAS) / American Nuclear Society (ANS) ISBN
978-85-63688-00-2
APPENDIX A
The wall-clock times for PENTRAN calculations are provided in
Tables A.I and A.II for forward
and adjoint transport, respectively. Calculations were performed
on nodes of dual quad-core
E5405 Xeon processors, on up to 5 nodes. The 16 group
calculations were performed on 2 nodes,
and the rest were performed on 5 nodes. There is a minimum of 4
GB available per node.
Tables A.III A.V give the YGROUP collapse results with
associated broad group/contributon
weighting. Table A.VI is the 80 group energy structure used.
Table A.I. Forward Transport Wall Clock Times in Seconds
# Groups P1 P3 CPUs
16 172
(2:52 min) 622
(10.3 min) 16
30 323
(5.3 min)
1249 (21 min)
30
40 1252
(20 min)
1664 (28 min)
40
80 2268
(38 min) 6755 sec (112 min)
40
Table A.II. Adjoint Transport Wall Clock Times in Seconds
# Groups P1 P3 CPUs
16 376
(6.25 min) 1411
(23.5 min) 16
30 734
(12.25 min) 2647
(44 min) 30
40 1567
(26 min) 5573
(93 min) 40
80 6367
(106 min) 14751
(246 min) 40
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
16/19
Table A.III. 40 Group Collapse Results
Broad group
#
Fine group range
Broad group wt.
Broad group
#
Fine group range
Broad group wt.
1 1 to 1 8.49395E-01 21 40 to 41 1.94042E-03
2 2 to 3 6.34773E-05 22 42 to 43 4.75340E-03
3 4 to 5 6.74168E-05 23 44 to 45 2.18759E-03
4 6 to 7 8.92143E-05 24 46 to 47 2.89751E-03
5 8 to 9 7.55904E-05 25 48 to 49 2.62035E-03
6 10 to 11 8.42971E-05 26 50 to 51 1.79396E-03
7 12 to 13 1.10408E-04 27 52 to 53 9.96407E-04
8 14 to 15 2.70830E-05 28 54 to 55 3.09489E-03
9 16 to 17 1.73748E-05 29 56 to 57 3.49418E-03
10 18 to 19 3.81938E-05 30 58 to 59 3.43235E-03
11 20 to 21 3.71920E-05 31 60 to 61 5.98998E-03
12 22 to 23 4.35763E-05 32 62 to 63 4.79148E-03
13 24 to 25 4.68327E-05 33 64 to 65 1.04344E-02
14 26 to 27 6.11489E-05 34 66 to 67 1.08146E-02
15 28 to 29 5.52832E-05 35 68 to 69 2.31466E-02
16 30 to 31 5.12889E-05 36 70 to 71 2.29577E-02
17 32 to 33 7.76942E-05 37 72 to 73 1.43598E-02
18 34 to 35 1.21201E-04 38 74 to 74 2.68466E-02
19 36 to 37 3.31489E-04 39 75 to 76 7.17634E-04
20 38 to 39 1.01197E-03 40 77 to 80 9.24418E-04
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
17/19
Table A.IV. 30 Group Collapse Results
Broad group
#
Fine group range
Broad group wt.
Broad group
#
Fine group range
Broad group wt.
1 1 to 1 8.49395E-01 16 58 to 61 9.42234E-03
2 2 to 5 1.30894E-04 17 62 to 65 1.52259E-02
3 6 to 9 1.64805E-04 18 66 to 67 1.08146E-02
4 10 to 13 1.94705E-04 19 68 to 68 8.50400E-03
5 14 to 17 4.44578E-05 20 69 to 69 1.46426E-02
6 18 to 21 7.53859E-05 21 70 to 70 1.43243E-02
7 22 to 25 9.04090E-05 22 71 to 71 8.63349E-03
8 26 to 29 1.16432E-04 23 72 to 72 9.68713E-03
9 30 to 33 1.28983E-04 24 73 to 73 4.67268E-03
10 34 to 37 4.52691E-04 25 74 to 74 2.68466E-02
11 38 to 41 2.95239E-03 26 75 to 75 4.33843E-04
12 42 to 45 6.94100E-03 27 76 to 77 6.05606E-04
13 46 to 49 5.51786E-03 28 78 to 78 2.85618E-04
14 50 to 53 2.79037E-03 29 79 to 79 2.97477E-04
15 54 to 57 6.58907E-03 30 80 to 80 1.95068E-05
-
Mi Huang, et al.
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
18/19
Table A.V. 16 Group Collapse Results
Broad group
#
Fine group range Broad group
weight
1 1 to 1 8.49395E-01
2 2 to 10 3.41808E-04
3 11 to 19 2.31248E-04
4 20 to 28 2.09249E-04
5 29 to 37 6.16458E-04
6 38 to 46 1.12705E-02
7 47 to 55 1.00260E-02
8 56 to 64 2.19866E-02
9 65 to 67 1.69703E-02
10 68 to 68 8.50400E-03
11 69 to 69 1.46426E-02
12 70 to 70 1.43243E-02
13 71 to 71 8.63349E-03
14 72 to 73 1.43598E-02
15 74 to 74 2.68466E-02
16 75 to 80 1.64205E-03
-
Multigroup Cross Section Optimization of a Detector Assembly
Model
2011 International Conference on Mathematics and Computational
Methods Applied to
Nuclear Science and Engineering (M&C 2011), Rio de Janeiro,
RJ, Brazil, 2011
19/19
Table A.VI. 80 Group Upper Energies (MeV)
Group Upper MeV Group Upper MeV Group Upper MeV
1 20 31 0.03 61 4e-07
2 17.333 32 0.025 62 3.5e-07
3 15.683 33 0.017 63 3.25e-07
4 12.84 34 0.013 64 2.75e-07
5 10 35 0.0095 65 2.25e-07
6 8.1873 36 0.006 66 1.75e-07
7 6.434 37 0.00374 67 1.5e-07
8 4.8 38 0.00155 68 1.25e-07
9 4.304 39 0.00055 69 1e-07
10 3 40 0.00021 70 7e-08
11 2.479 41 0.000108 71 5e-08
12 2.354 42 3.7e-05 72 4e-08
13 1.85 43 1e-05 73 3e-08
14 1.5 44 5e-06 74 2.53e-08
15 1.4 45 4e-06 75 1.5e-09
16 1.356 46 3.05e-06 76 1.2e-09
17 1.317 47 2.38e-06 77 1e-09
18 1.25 48 1.86e-06 78 7.5e-10
19 1.2 49 1.45e-06 79 5e-10
20 1.01 50 1.3e-06 80 1e-10
21 0.82 51 1.12e-06
22 0.75 52 1.08e-06
23 0.6 53 1.04e-06
24 0.49952 54 1e-06
25 0.33 55 8.5e-07
26 0.27 56 8e-07
27 0.2 57 7e-07
28 0.1 58 6.25e-07
29 0.073 59 5.5e-07
30 0.045 60 5e-07