NUDUNA: “Nuclear Data Uncertainty Analysis”
Oliver Buss, Axel Hoefer, Jens-Christian NeuberAREVA GmbH, PEPA-G (Offenbach, Germany)
Meeting on uncertainty propagations in the nuclear fuel cycle Uppsala University, April 24-25, 2013
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.3
ContentMotivation Nuclear data uncertainty impact and criticality safety
Methods for estimating the impact of Nuclear Data Uncertainty
Monte-Carlo approach to the propagation of Nuclear Data uncertainty
NUDUNA framework Present issues with available uncertainty information Results for keff and depletion
Summary & Outlook
Motivation
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.5
MotivationCriticality safety evaluations for (re-) processing, storage, transport and final disposal of nuclear fuel Small Margins additional to the Regulatory Safety Margin Regulatory Safety Margin: keff<0.95, 0.98 or even keff<1.00
Analyses rely on transport codesDifferent types of analytical uncertainties:1. Geometrical and material data uncertainties 2. Burn-up uncertainties 3. Systematic uncertainties in the transport algorithms (Calc. Bias) 4. Nuclear data uncertainties
Status of nuclear data uncertainty estimation: USA, France: Demand for estimates of nuclear data uncertainties Otherwise nuclear data uncertainties considered to be covered by the safety margin
(which includes errors not accounted for)
We aim to provide a proper estimate for nuclear data uncertainties Ensure bounding estimates in safety margins Reduce conservatism in safety margins
►
Methods for estimating Nuclear Data Uncertainty
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.7
Using Evaluations in Transport Codes
Data Evaluations
Transport Code Library
1. Doppler Broadening (Temperature Effect)
2. Resonance+BG summation
3. Resonance Self-Shielding
4. Data compression / Discretization (Group-wise cross sections)
5. Formatting
Experimental DataModelling, Fitting, Expert Judgement
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.8
Approaches to the estimation ofNuclear Data Uncertainty impact
Generate random nuclear
data library
Perform analysis with random input
data
Store transport
result
1. Perturbation theory
Approximation, Fast!!
2. Monte Carlo
► Retrieve statistics for observable based on distribution of input parameters
All orders included! Flexible Tool!
ji
jiji
CovAAACov,
),()(
SensitivitiesNuclear DataCovariances
draw randomwith
111
)(2
11
2
i
iAN
iANN
NACovN
i
N
i
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.9
Available Codes for Nuclear Data Uncertainty Estimation
Monte Carlo Approach(All Orders) 1. Order Perturbation
Measured Data TMC (NRG)
Evaluations NUDUNA (AREVA NP)KIWI (LLNL)
Grouped DataXSUSA (GRS) TSUNAMI (ORNL)
TSURFER (ORNL)RIB (CEA)
App
roxi
mat
ions
Very Flexible, CPU intense
FAST!!
Approximations
Long History, Not flexibleVery good Documentation
SCALE Cov based, FAST
Monte-Carlo approach to Nuclear Data Uncertainty Estimation
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.11
NUDUNA:Program Flow
Job card creation
Random library data bench (XML based)
Idealized HEU Shell Exp from LA4208. 2 7 5.0306E-05 1 -2 imp:n=13 2 4.8154E-02 2 -3 imp:n=14 7 5.0306E-05 3 -4 imp:n=15 3 4.8154E-02 4 -5 imp:n=16 7 5.0306E-05 5 -6 imp:n=17 4 4.8152E-02 6 -7 imp:n=18 7 5.0306E-05 7 -8 imp:n=19 5 4.8154E-02 8 -9 imp:n=110 6 4.7780E-02 9 -10 imp:n=111 0 10 imp:n=0
Idealized HEU Shell Exp from LA4208. 2 7 5.0306E-05 1 -2 imp:n=13 2 4.8154E-02 2 -3 imp:n=14 7 5.0306E-05 3 -4 imp:n=15 3 4.8154E-02 4 -5 imp:n=16 7 5.0306E-05 5 -6 imp:n=17 4 4.8152E-02 6 -7 imp:n=18 7 5.0306E-05 7 -8 imp:n=19 5 4.8154E-02 8 -9 imp:n=110 6 4.7780E-02 9 -10 imp:n=111 0 10 imp:n=0
Idealized HEU Shell Exp from LA4208. 2 7 5.0306E-05 1 -2 imp:n=13 2 4.8154E-02 2 -3 imp:n=14 7 5.0306E-05 3 -4 imp:n=15 3 4.8154E-02 4 -5 imp:n=16 7 5.0306E-05 5 -6 imp:n=17 4 4.8152E-02 6 -7 imp:n=18 7 5.0306E-05 7 -8 imp:n=19 5 4.8154E-02 8 -9 imp:n=110 6 4.7780E-02 9 -10 imp:n=111 0 10 imp:n=0
Idealized HEU Shell Exp from LA4208. 2 7 5.0306E-05 1 -2 imp:n=13 2 4.8154E-02 2 -3 imp:n=14 7 5.0306E-05 3 -4 imp:n=15 3 4.8154E-02 4 -5 imp:n=16 7 5.0306E-05 5 -6 imp:n=17 4 4.8152E-02 6 -7 imp:n=18 7 5.0306E-05 7 -8 imp:n=19 5 4.8154E-02 8 -9 imp:n=110 6 4.7780E-02 9 -10 imp:n=111 0 10 imp:n=0
Automated job card multiplication
Automated Output Analysis: Upper/Lower 95/95, Histograms
Generation of random libraries
Parallel execution of jobs: Linux or Windows Cluster
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.12
NUDUNA: Random input library creation
NUDUNA Scope: Provide libraries for SCALE
and MCNP transport suites Need to generate AMPX and
ACE formatted input files Present Project: Support
ARCADIANUDUNA makes use of: randomLib by AREVA PEPA „NJOY 99“ tool by Los Alamos
National Lab (LANL) SMILER/PUFF IV tool by
Oakridge National Lab (ORNL)
randomLib►Reads information encoded in ENDF6 tapes►Varies information randomly according to covariance matrices►Writes random information to ENDF6 tape
NJOY 99►Resonance + background treatment►Doppler broadening►S(α,β) treatment►Resonance self-shielding►ACE tape generation or multi group treatment
SMILER (PUFF IV)►Converts multigroup format to AMPX
MCNP SCALE
ENDF6 tape
AMPX tapeACE tape
GENDF tape
ARCADIA
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.13
NUDUNA: Creating random ENDF6 tapes
delayedprompttot mmm
File 1: Multiplicities►Restoration of sum rules
►Normal or log-normal model
File 2: Resonances►Respect positivity bounds for widths►Assume constant phases of amplitudes►Dimension ≤ 10.000 ►Normal or log-normal model
File 4: Angular Distributions►Check for angular distribution > 0►Normal or log-normal model
File 3: Cross sections► Define common energy grid acc. to
file 33 information► Average data on grid► Random draws for average data → scaling factors for energy ranges
► Rescale original data ► Restore ENDF6 sum rules
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.14
Problems with current ENDF6 encodings of Data Uncertainties
Missing LTY=0 Entries for Files 1&3 Arbitraryness in restoring sum rules
File 1 phasesCovariances of total and prompt neutron multiplicities must be consistent!Arbitraryness of Distribution ModelsWhy not distribution model in ENDF6? Especially problematic for forward-peaked angular
distributions (File 4)Incomplete Information in File 35 Need for uncertainty of underlying parameters and not of the
sum itself
Results
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.16
Results IGodiva (HEU-MET-FAST-001)
Evaluation Transport Suite
Average keff
Standard Deviaton due to
Nuclear Data Uncertainty
[pcm]
JENDL 4.0SCALE 6238 Groups
0.99891 955
JENDL 4.0 MCNP 5 0.99937 953
ENDF 7.1βSCALE 6238 Groups
0.99900 1016
ENDF 7.1β MCNP 5 0.99920 1018
►Tsunami (SCALE 6): 930 pcm (ENDF/B VII)
Evaluation Transport Suite
Average keff
Standard Deviaton due to
Nuclear Data Uncertainty
[pcm]
JENDL 4.0SCALE 6238 Groups
0.99891 955
JENDL 4.0 MCNP 5 0.99937 953
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.17
Results II: Adding the benchmark loop
Generate random nuclear data library
Perform Transport calculation for
application case
Apply order statistics to the database of result+bias
→ Upper confidence limit
Evaluate representative Experiments
Store transport result
Extract calc. bias
for application case
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.18
Results II:Adding the benchmark loop
►Simple Pin-Cell Model: UO2 with 5% enrichment water reflected►Bias estimation based on LCT51 experiments►ck≈0.9 for all experiments
Upper 95%/95%
19.51,...,2.51
)()(181
LCTLCTi
calceffbenchb ikikk
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.19
Results III:Depletion calculations for BUC
►OECD Phase 1b Benchmark, Sample A (3.2 wt.-% enrichment, thermal pin-cell)
►Impact of U238 XS uncertainty based on TALYS
►TRITON / SCALE 6.0
Generate random nuclear
data library
Perform Depletion analysis
Store Isotopic Composition
Sample A
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
1.2%
1.4%
1.6%
Mo-
95
Tc-9
9
Ru-
101
Rh-
103
Ag-
109
Cs-
133
Nd-
143
Nd-
145
Sm-1
47
Sm-1
49
Sm-1
50
Sm-1
51
Sm-1
52
Eu-
153
Gd-
155
O-1
6
U-2
34
U-2
35
U-2
36
U-2
38
Pu-
238
Pu-
239
Pu-
240
Pu-
241
Pu-
242
Pu-
243
Am-2
41
Am-2
43
Isotope
Rel
. Err
or o
f Num
ber D
ensi
ty
NUDUNA by AREVA
TMC by NRG (Rochman et al.)
Succesful Benchmarks with TMC (NRG)
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.20
Burnup-Credit with Gd-Fuel
Problem: No experiments for validating Burn-Up of Gd fuel at low burn-up!!!How to estimate uncertainty of number densities? NUDUNA !!!
Variation of 155Gd cross sections according toENDF 7.1β
Assumption: XS uncertainty dominates BU-Algorithm uncertainty
Results:
►155Gd XS uncertainties do not induce seizable relative errors on other number densities
►Expected relative uncertainty of 155Gd number density :
ca. 10%
13 GWd/MTU
0% void
All rights are reserved, see liability notice.NUDUNA - Meeting on uncertainty propagations in the nuclear fuel cycle, Uppsala, 24/04/2013 - Oliver Buss - AREVA NP GmbH Proprietary © AREVA - p.21
SummaryNUDUNA provides nuclear data uncertainty impact estimatesSupport for continuous and group-wise XS librariesNUDUNA improves on the perturbation theory methodologyFlexibility of MC ⇨ easy to embed in hierarchical schemes for uncertainty estimationSuccesful benchmarks for keff and depletion Consistent results for MCNP and SCALE NUDUNA compatible to TSUNAMI (ORNL) Succesful TMC (NRG) vs. NUDUNA benchmarks for OECD Phase-1b
depletion problem
Problems with ENDF6 standard and present evaluations Strong support by Nuclear Data Community needed
End of presentationNUDUNA –“Nuclear Data Uncertainty Analysis in Criticality Safety”Oliver Buss, Axel Hoefer, Jens-Christian NeuberAREVA NP GmbH, PEPA-G (Offenbach, Germany)
Workshop FDN & PEPA08/03/2012