CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using

CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
Annals of Nuclear Energy 131 (2019) 9–22



                                                                 Contents lists available at ScienceDirect


                                                              Annals of Nuclear Energy
                                                journal homepage: www.elsevier.com/locate/anucene




Uncertainty quantification of LWR-PROTEUS Phase II experiments using
CASMO-5
Jinsu Park a, Wonkyeong Kim a, Mathieu Hursin b,⇑, Gregory Perret b, Alexander Vasiliev b,
Dimitri Rochman b, Andreas Pautz b, Hakim Ferroukhi b, Deokjung Lee a
a
    Department of Nuclear Engineering, Ulsan National Institute of Science and Technology, UNIST-gil 50, Ulsan 44919, Republic of Korea
b
    Paul Scherrer Institut, Nukleare Energie und Sicherheit, PSI Villigen, 5232, Switzerland



a r t i c l e           i n f o                           a b s t r a c t

Article history:                                          This paper presents an improved uncertainty quantification technique for the validation of CASMO-5 on
Received 4 December 2018                                  the spent fuel reactivity worth experiments of the LWR-PROTEUS Phase II program. In the program,
Received in revised form 14 March 2019                    eleven spent fuel samples manufactured from rods irradiated in a Swiss PWR (discharge burnups of
Accepted 16 March 2019
                                                          20–120 MWd/kg) were measured in the PROTEUS research reactor. In this work, both irradiation and
                                                          reactivity worth measurement steps were modeled with CASMO-5 and the uncertainty on the code pre-
                                                          diction were calculated using SHARK-X. For the first time in this work, we propagated the nuclear data
Keywords:
                                                          uncertainties coming from cross-sections and fission yields for all samples in both fuel irradiations and
LWR-PROTEUS Phase II experiment
Spent fuel
                                                          the reactivity worth experiment models. We found that the fission yield and cross section uncertainties
Uncertainty quantification                                have similar contributions to the uncertainty of the reactivity worth prediction and that the reactivity
SHARK-X                                                   worth probability distribution is non-normal because of the non-normal distribution of the perturbed
CASMO-5                                                   fission yield data produced by the GEF code. Propagating input uncertainties through fuel irradiation
Fission yield perturbation                                as well as considering fission yield uncertainty led to larger computational uncertainty than previously
                                                          reported. However, the observed trends with respect to exposure, fuel type, and moderating conditions
                                                          are similar. In particular, a linear regression analysis showed that the predictions of reactivity with
                                                          exposure by CASMO-5 are very accurate for various moderating conditions and for very high burnup.
                                                             Finally, we estimated the effects of the irradiation history specifications on the relative reactivity worth
                                                          bias and its uncertainty using two independent irradiation histories. While the bias uncertainty coming
                                                          from uncertain cross sections was similar when considering either irradiation histories, we observed dif-
                                                          ferences in the bias value. The irradiation history specification is a significant source of modeling bias and
                                                          should be further investigated.
                                                                                                                               Ó 2019 Elsevier Ltd. All rights reserved.




1. Introduction                                                                              plants. By cutting samples from highly burnt fuel rods irradiated
                                                                                             in Swiss LWRs and oscillating them in the zero-power research
   With increasing discharge burnups of commercial Light Water                               reactor of PSI, PROTEUS, their reactivity worths were measured
Reactor (LWR) fuels, reactivity loss during irradiation must be val-                         to determine accurately the fuel reactivity loss during irradiation.
idated to support core design calculations as well as the use of bur-                        Such reactivity measurements were complemented by the deter-
nup credit for the storage and transportation of spent fuel. As                              mination of the burnt fuel sample isotopic compositions by chem-
comprehensive experimental data related to spent fuel is scarce,                             ical analysis.
a high-quality set of measured isotopic compositions and reactivity                              Those reactivity worth and isotopic concentration data have
effects measurements involving spent fuel has been carried out at                            been used extensively at PSI for validation activities related to
the Paul Scherrer Institut (PSI) in the framework of the LWR-                                the depletion capabilities of the CASMO-4/5 code (Grimm et al.,
PROTEUS Phase II program (Gunther-Leopold, 2007; Grimm et al.,                               2014, 2017, 2008). However, only a limited amount of work
2001, 2006). This experimental program aims at investigating the                             (Grimm et al., 2017; Leray et al., 2016; Rochman et al., 2018) has
physics of highly burnt fuel samples from Swiss nuclear power                                been dedicated to the quantification of computational
                                                                                             uncertainties required for meaningful validation. For these activi-
                                                                                             ties, the SHARK-X platform (Hursin et al., 2015; Wieselquist,
    ⇑ Corresponding author.                                                                  2012; Leray, 2015) has been used. SHARK-X is a set of Perl based
      E-mail address: Mathieu.hursin@psi.ch (M. Hursin).

https://doi.org/10.1016/j.anucene.2019.03.023
0306-4549/Ó 2019 Elsevier Ltd. All rights reserved.
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
10                                              J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22


scripts developed by PSI for nuclear data uncertainty propagation              were irradiated in a Swiss PWR. Table 1 represents the description
in CASMO-5.                                                                    of each fuel sample. The exposure of the fuel samples varies from
    So far, only a reduced set of nuclear data was considered as               20 to 120 MWd/kg. To reach large burnup, samples went through
uncertain input and the propagation of input uncertainty was lim-              multiple reactor cycles and were relocated in several fuel assem-
ited to the modeling of the oscillation measurements (reactivity               blies, as recorded in Table 1. The burnup of a given fuel sample is
worth measurements) in PROTEUS (Grimm et al., 2017). The main                  determined using the concentration of 148Nd measured through
objective of the present paper is to summarize recent improve-                 chemical assay on neighboring pieces of the same fuel rod.
ments made to SHARK-X to carry out the Uncertainty Quantifica-                     The PROTEUS reactor is a zero-power experimental reactor
tion (UQ) of the simulation of the LWR-PROTEUS Phase II                        operated at PSI between 1968 and 2011. The reactivity worth of
program with CASMO-5. More specifically, the updated UQ analy-                 a given burnt fuel sample is measured by inserting and withdraw-
sis includes cross section and fission yield data perturbations and            ing the sample into a guide tube in the center of the PROTEUS core.
concerns both CAMSO-5 calculations required for the fuel sample                For each measurement, the absolute reactivity worth of burnt fuel
irradiation in the Swiss power plant as well as its oscillations in            samples is measured against the one of a fresh 3.5 wt% enriched
PROTEUS. Cross sections, averaged number of neutrons produced                  UO2 sample. This allows measurement of the reactivity loss of fuel
per fission (t), and fission spectrum (v) are perturbed for the full           caused by exposure in a nuclear power plant. In addition, the abso-
set of isotopes considered in CASMO-5 as well as the fission yields            lute reactivity worth of naturally enriched sample was measured.
of four major fissile isotopes following recent method develop-                Both measurements were performed for various moderating condi-
ments in this area at PSI (Leray et al., 2017). Their effects on the           tions. The ratio of the two reactivity worths is used for comparison
UQ results are investigated. Finally, as nuclear data uncertainty              with CASMO-5 instead of absolute reactivity worth for two rea-
may not be the leading source of uncertainty for such simulations,             sons. The first one is to cancel out possible errors in the experimen-
the effect of the irradiation history description on the computa-              tal design and measurement techniques, and the second is to
tional uncertainty is estimated by comparing two independent                   correct for the 2D, single assembly nature of the CASMO-5 model
irradiation histories (one determined internally at PSI and one pro-           used for comparison with measurements in the multi-zone driven
vided by the fuel vendor). A previous publication (Rochman et al.,             reactor PROTEUS. Strictly speaking, the relative reactivity worth
2018) summarized a detailed estimation of the effect of the irradi-            ratio is not a quantity of interest for burnup credit applications,
ation history on the computational uncertainty for the determina-              but it is the quantity computed by CASMO-5 that can be compared
tion of burnt fuel isotopic composition. As the quantity of interest           to the measurements. As a result, it is used for validation purposes
considered in the present work is different (reactivity worth as               in the present paper.
opposed to isotopic composition), using this information imposes
establishing a link between isotopic composition and reactivity                2.2. Modeling of the LWR-PROTEUS Phase II experiment
worth uncertainties. Such link would require determining sensitiv-
ity coefficients of the reactivity worth to each isotope of the burnt              The LWR-PROTEUS Phase II experiments are modeled using
fuel sample. Due to the number of isotopes considered, it is a fairly          CASMO-5. CASMO-5 (Rhodes et al., 2012) is a two-dimensional lat-
expensive calculation, which has not been envisioned in this work.             tice physics code with depletion capabilities developed by the
As such the conclusions of the previous study is not used. Instead             Studsvik Scandpower, Inc. In this work, CASMO-5 uses a 586
of performing a similar analysis in this paper, we chose a simpler             energy groups neutron cross section library based on ENDF-B/
approach as we are only interested in the magnitude of the compu-              VII.1 (Chadwick, 2011). Two separated procedures are required
tational uncertainty associated with the irradiation history. In               for the modeling of the LWR-PROTEUS Phase II experimental cam-
other words, two specific irradiation histories are considered in              paign. First, the irradiation of the burnt fuel samples in the nuclear
the depletion calculation without random perturbations of the                  power plant is modeled through a depletion calculation. Next, the
irradiation history related parameters such as power density, fuel             oscillations of the burnt fuel sample in the PROTEUS reactor are
and moderator temperatures, depletion step, etc.                               modeled. Both calculations are described in detail in the following
    This paper is organized as follows. Section 2 describes the LWR-           sections.
PROTEUS Phase II experiments, including details on the CASMO-5
calculations required for the modeling of the fuel sample irradia-             2.2.1. Determination of the fuel samples composition
tion and their oscillations in PROTEUS. Section 3 presents the                     Fig. 1 shows the geometrical configuration of the UO2 and MOX
SHARK-X methodology used for the propagation of nuclear data                   fuel assemblies for the CASMO-5 depletion calculation. The assem-
uncertainty in CASMO-5. The list of input parameters perturbed,                blies containing the UO2 samples are depleted considering a single
including fission yields is reviewed. Section 4 describes the numer-           fuel assembly and reflective boundary conditions. In the case of
ical results for the reactivity worth UQ for all fuel samples and              MOX fuel samples, the depletion calculations are performed using
moderating conditions as well as discusses the effects of the irradi-          a 2-by-2-assemblies model, a MOX assembly surrounded by UO2
ation history on the UQ results. Special consideration is given to             assemblies as shown in Fig. 1 on the right. The locations of each
the propagation of fission yields uncertainty. In the light of the             fuel sample in the fuel assembly are represented in Fig. 1.
uncertainty propagation results, the performance of the CASMO-5                    In this work, two irradiated histories are considered for the
for the modeling of the LWR-PROTEUS Phase II experiments is dis-               depletion calculation of fuel samples. The first irradiation history
cussed focusing on the prediction of reactivity loss of the fuel with          as well as the burnup of the irradiated fuel samples is provided
exposure. Finally, conclusions are presented in Section 5.                     by the fuel vendor based on the pre-cycle core analysis using a
                                                                               nodal core simulator with pin power reconstruction (not CMSYS).
                                                                               It is referred to as VENDOR history. The second irradiation history
2. LWR-Proteus Phase II experiments                                            was generated using the BOHR methodology of PSI (Herrero, 2017).
                                                                               It is referred to as BOHR history. The BOHR methodology allows re-
2.1. Measurements of the LWR-PROTEUS Phase II                                  running CASMO-5 with operating conditions extracted from refer-
                                                                               ence CASMO5-SIMULATE3 core models developed and validated
   A total of eleven 40-cm long PWR fuel samples were investi-                 for all Swiss reactors and operated cycles within the CMSYS plat-
gated during the LWR-PROTEUS Phase II experiments. Seven sam-                  form (Ferroukhi et al., 2008). The BOHR history contains more
ples are made of UO2 fuel and four of MOX fuel. Those samples                  detailed information about the irradiation conditions at each
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                          11


Table 1
Description of the fuel samples.

  Sample                  Fuel type        Initial fissile               No. of                 No. of             Burnup                DBurnup
  ID                                       contents                      cycles                 relocation         (MWd/kg)              VENDOR-BOHR
                                                                                                                                         (MWd/kg)
  U1                      UO2              4.1%                          2                      0                  40                   2.4
  U2                                       3.5%                          3                      0                  50                   1.9
  U3                                       3.5%                          5                      1                  70                   0.7
  U4                                       3.5%                          5                      1                  80                   0.4
  U5                                       3.5%                          7                      3                  90                   0.4
  U6                                       3.5%                          7                      3                  90                   0.1
  U7                                       3.5%                          10                     4                  120                  0.6
  M1                      MOX              5.5%                          1                      0                  20                   0.9
  M2                                       5.5%                          2                      0                  40                   0.4
  M3                                       5.5%                          3                      0                  60                   0.4
  M4                                       5.5%                          4                      0                  70                   0.6




                                                      Fig. 1. CASMO-5 models for depletion calculation.



burnup step. Fig. 2 illustrates the input parameters such as power                 power density for all burnup steps during the irradiation. Such
density, boron concentration, fuel and moderator temperatures as                   adjustment resulted in differences of calculated exposure when
a function of time for both VENDOR and BOHR irradiation histories                  considering the VENDOR or the BOHR irradiation history as
of the U1 sample. As shown in Fig. 2, the VENDOR irradiation his-                  reported in Table 1.
tory of U1 sample uses a finer burnup grid (30 and 28 burnup                          The effects of the different irradiation histories in terms of
points for first and second cycle, respectively) with only four oper-              uncertainty propagation are investigated in Section 4.3.
ating conditions per cycle for depletion calculation. In contrast, the
BOHR irradiation history uses a smaller number of burnup points                    2.2.2. Calculation of relative reactivity worth
(22 and 21 burnup points for first and second cycle, respectively)                     For the modeling of the LWR-PROTEUS Phase II experiments,
with different operating conditions for each burnup point. The                     the PROTEUS test zone, where burnt fuel samples were oscillated,
BOHR irradiation history always uses different operating condition                 is modelled with CASMO-5 through an 11-by-11 array of PWR UO2
for each burnup point.                                                             fuel rods. Fig. 3 shows the CASMO-5 model of the LWR-PROTUES
    A careful reader would have noticed in the BOHR history of                     Phase II experiments. The simulation of the Phase II experiment
Fig. 2, that while the power density increases slowly during the                   is performed in three different moderating conditions: pure light
course of the first cycle, the fuel temperature decreases by about                 water (H2O), a mixture of light and heavy water (DHO), and
100 K. Such phenomenon is due to the change of heat conduction                     borated light water (B2023).
properties of the fuel under irradiation (both the heat conductivity                   In the CASMO-5 model, the reference case, which uses the fresh
and gap conductance).                                                              fuel sample at the center of test zone, is made critical (keff = 1) by
    The burnt fuel sample exposure value obtained with the BOHR                    automatically adjusting the axial buckling. The reactivity of the
history is adjusted to match the measured 148Nd content within its                 naturally enriched and the burnt fuel samples are calculated by
experimental uncertainty: a constant coefficient is applied to the                 keeping this axial buckling value constant; the underlying assump-
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
12                                                           J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22




                                 Fig. 2. Input parameters as a function of time for the VENDOR and BOHR irradiation histories of the U1 sample.



                                                                                                     1 X  M X N
                                                                                                                 ðC n;m =En;m  1Þ2
                                                                                            v2 ¼                                    :                         ð2Þ
                                                                                                   M  N m¼1 n¼1         r2n;m
                                                                                               The averaged chi-squared value for the VENDOR and BOHR irra-
                                                                                            diation histories are 0.86 and 1.39, respectively. Consequently, the
                                                                                            BOHR irradiation history does not produce more accurate relative
                                                                                            reactivity worth results, even though more detailed irradiation
                                                                                            conditions are provided during the depletion calculation. As a
                                                                                            result, the irradiation history will be considered as an uncertain
                                                                                            input parameter for the uncertainty propagation. As it is difficult
                                                                                            to formalize such source of uncertainty rigorously, the present
                                                                                            work only provides an attempt at estimating such uncertainty in
                                                                                            Section 4.3. For the reminder of the paper, the C/E produced with
          Fig. 3. CASMO-5 model of LWR-PROTEUS Phase II experiments.                        the VENDOR irradiation history are used.
                                                                                               The uncertainty of the calculated results needs to be assessed to
                                                                                            draw any conclusion from the C/E distribution. It is the subject of
                                                                                            the next sections of the present paper.
tion is that the neutron leakage out of the model is not influenced
significantly by the presence of the burnt fuel and natural uranium
sample. The keff values obtained for each of the burnt, fresh and                           3. Uncertainty quantification with SHARK-X
natural uranium samples are used to calculate the relative reactiv-
ity worth following Eq. (1).                                                                   SHARK-X (Hursin et al., 2015; Wieselquist, 2012; Leray, 2015) is
                                                                                            a set of Perl based scripts developed at PSI for nuclear data uncer-
              fresh           sample
          1=keff  1=keff                                                                   tainty propagation with CASMO-5. SHARK-X provides two major
Dqrel ¼      fresh           natural
                                       :                                          ð1Þ       approaches for UQ calculations. The first one relies on the calcula-
          1=keff         1=keff
                                                                                            tion of sensitivity coefficients using a Direct Perturbation (DP)
                      fresh   sample
   In Eq. (1), keff , keff , and keff
                                           natural
                                        are the values for the fresh,                       approach and subsequent use of the Sandwich rule. The second
                                                                                            method relies on Stochastic Sampling (SS). Because of the large
burnt, and natural uranium fuel samples, respectively.
                                                                                            number of uncertain inputs involved in the depletion calculations
   Fig. 4 shows the calculated over experimental (C/E) relative
                                                                                            (16,283 uncertain cross sections with a 19 energy group structure
reactivity worth obtained with both irradiation histories. The VEN-
                                                                                            and 1304 fission yields for 384 isotopes), the DP approach is not
DOR results are the ones reported in a previous publication
                                                                                            convenient due to its high computational cost. Consequently, the
(Grimm et al., 2017). For a detailed discussion about the C/E distri-
                                                                                            SS method is used.
bution, the reader is referred to this publication. The C/E values
range from 0.88 to 1.07 for both irradiation histories. The averaged
chi-square value for relative reactivity worths is calculated using                         3.1. Stochastic sampling method
Eq. (2) for each irradiation history. In Eq. (2), M is the number of
moderating conditions and N is the number of fuel samples. And,                                The SS method requires the perturbation of input data consid-
Cn,m and En,m are the calculated and experimental relative reactivity                       ered as random variables with presumed probability distribution
worth values of fuel sample n in moderating condition m. The bias                           functions. SS calculations consist in the process of assigning distri-
relative uncertainty of relative reactivity worth ratio is expressed                        butions to inputs, sampling the inputs (accounting for possible cor-
as rn;m , which is equal here to the experimental relative                                  relations between the inputs), performing independent
uncertainty.                                                                                calculations with each sample, and statistically analyzing the dis-
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                       13




                                    Fig. 4. Calculated over experimental value of relative reactivity worth using nuclear data nominal values.



tribution of outputs. The SS approach can produce meaningful and                             2017). The SCALE-6.0 VCM library was collaboratively developed
reliable UQ results with much fewer runs than the DP methods. For                            from Brookhaven National Laboratory (BNL), Oak Ridge National
relative reactivity worth UQ, each input sample is used in the three                         Laboratory (ORNL), and Los Alamos National Laboratory (LANL),
transport calculations needed to compute the relative reactivity                             and it contains 402 nuclides with various reaction channels. Table 2
worth as expressed in Eq. (1).                                                               represents the list of perturbed nuclides with corresponding
    The SS method produces an estimate for the variance VR of a                              nuclear reaction channels considered in this work. For the given
given response R, which converges to its true value (r2) as the                              isotope, only the covariance data between reactions and energy
number of samples increases. Because it is not possible to run an                            groups are available. The covariances between isotopes are
infinite number of samples, a confidence interval is a useful quan-                          neglected. The MT numbers listed in Table 2 are represented as
tity for evaluating the reliability of the estimate VR. Assuming the                         follows.
considered response has a normal distribution characterized by
its mean and variance (VR), the confidence interval shown in Eq.                              1. MT = 2: Elastic scattering, rs,el
(3) would contain the true variance (1  a)100% of the time. In                              2. MT = 4: Inelastic scattering, rs,in
practice, different values can be attributed to a to determine lower                          3. MT = 16: (n,2n) neutron production, rn,2n
and upper bounds. In the present work, both values for a are set to                           4. MT = 18: fission, rf
5% and the interval is referred to as the 95% confidence interval.                            5. MT = 101: capture, rc (including MT = 102–107, rn,c, rn,proton,
                          "                      #                                               rn,deuteron, rn,triton, rn,hydrogen, rn,a)
h                 i        ðN  1ÞV R ðN  1ÞV R                                              6. MT = 452: average number of neutron per fission, m
    r2lower ; r2upper ¼              ;            :                                ð3Þ
                           v1a=2;N1
                             2
                                        va=2;N1
                                        2                                                     7. MT = 1018: average fission spectrum, v

where v2N1 is the v2 distribution with N  1 degree of freedom. For
a desired probability level (1  a), the critical values v2a=2;N1 and                       3.3. Fission yields uncertainties

v   2
          of the v distribution are tabulated and can be looked
    1a=2;N1
                           2
                                                                                                 Fission yields are perturbed during the depletion calculations
up for a given number of degree of freedom N. In this paper, Eq.                             involved in the modeling of the nuclear power plant irradiation.
(3) is used to calculate the confidence intervals even for non-                              As the decay constants were shown to have a negligible contribu-
normally distributed responses.                                                              tion to the overall computational uncertainty (Ferroukhi et al.,
                                                                                             2014), they are neglected here. 500 ENDF-formatted files contain-
3.2. Cross sections, t and v uncertainty                                                     ing perturbed fission yields are produced by the PSI-GEF method as
                                                                                             reported in Leray et al. (2017) using a modified version of the GEF
   The SCALE-6.0 Variance Covariance Matrix (VCM) library                                    code (Schmidt et al., 2016; Schmidt and Jurado, 2014): the GEF
(Williams et al., 2009; Rearden, 2009) is used in this study to                              code is modified so that 21 model parameters are sampled inde-
describe the behavior of uncertain cross sections, averaged number                           pendently according to normal distributions (for the magnitude
of neutrons produced per fission (t), and fission spectrum (v).                              of their uncertainties, please refer to Table 1 in Leray et al.
Even if the more recent sources of nuclear data uncertainties can                            (2017). Then, for each set of sampled parameters, a corresponding
be also utilized in SHARK-X, the SCALE-6.0 VCM library is used in                            set of perturbed fission yields is calculated with GEF. This process
this study to compare consistently the behavior of uncertainty                               is repeated 500 times. This process gives the advantage of produc-
from this study with that from previous work (Grimm et al.,                                  ing random fission yields based on a theoretical model, as in the
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
14                                                                     J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22


Table 2
SHARK-X perturbed nuclides and reaction channels.

     Isotope (ZA)                                                                                                                                                             Perturbed MT number
     1     3    186-187   197     253
      H, He,            Re,     Au,      Es                                                                                                                                   2,   101
     2
      H, 9Be, 59Co, 59Ni                                                                                                                                                      2,   16, 101
     3
      H                                                                                                                                                                       2,   16
     4
      He, 7Be                                                                                                                                                                 2
     6-7
         Li, 10B                                                                                                                                                              2,   4, 101
     11
       B, 14-15N, 16-17O, 19F, 23Na, 24-26Mg, 27Al, 28-31Si, 32-34S, 36S, 35Cl, 37Cl, 36Ar, 38Ar, 40Ar, 39-41K, 40Ca, 42-44Ca, 46Ca, 48Ca, 46-47Ti, 49-50Ti, 50Cr, 52-54Cr,   2,   4, 16, 101
         55
            Mn, 54Fe, 56-58Fe, 58Co, 60mCo, 58Ni, 60-62Ni, 64Ni, 63Cu, 65Cu, 69Ga, 71Ga, 70Ge, 72-74Ge, 76Ge, 74-75As, 74Se, 76-80Se, 82Se, 79Br, 81Br, 78Kr, 80Kr,
         82-86
               Kr, 85-87Rb, 84Sr, 86-90Sr, 89-91Y, 90-96Zr, 93-95Nb, 92Mo, 94-100Mo, 99Tc, 96Ru, 98-106Ru, 103Rh, 105Rh, 102Pd, 104-108Pd, 110Pd, 107Ag, 109Ag,
         111
             Ag, 111mAg, 106Cd, 108Cd, 110-114Cd, 116Cd, 117mCd, 113In, 115In, 112-120Sn, 122-126Sn, 121Sb, 123-126Sb, 120Te, 122-126Te, 128Te, 130Te, 132Te, 133mTe,
         134m
               Te, 127I, 129-131I, 135I, 123-124Xe, 126Xe, 128-136Xe, 133-137Cs, 130Ba, 132-138Ba, 140Ba, 138-140La, 136Ce, 138-144Ce, 141-143Pr, 142-148Nd, 150Nd,
         147-149
                 Pm, 151Pm, 152mPm, 144Sm, 147-154Sm, 151-157Eu, 152-158Gd, 160Gd, 159-160Tb, 156Dy, 158Dy, 160-164Dy, 165Ho, 166mHo, 162Er, 164Er, 166-168Er,
         170
             Er, 175-176Lu, 174Hf, 176-180Hf, 181-182Ta, 182-184W, 186W, 191Ir, 193Ir, 196Hg, 198-202Hg, 204Hg, 204Pb, 206-208Pb, 209Bi, 225-226Ac
     45
       Sc                                                                                                                                                                     101
     48
       Ti                                                                                                                                                                     4, 16, 101
     227
         Ac, 230Th, 239Np, 237Pu, 243-244Pu, 241Am, 241Cm, 248Cm, 250Cf                                                                                                       2, 4, 16, 18, 101, 452
     227-229
              Th, 233-234Th, 232Pa, 232U, 234U, 236-237U, 239-241U, 235-238Np, 236Pu, 238Pu, 246Pu, 242-244Am, 242mAm, 244mAm, 242-247Cm, 249-250Cm, 249Bk,                   2, 4, 16, 18, 101
         249-251
                 Cf, 254Cf, 254-255Es
     232
         Th                                                                                                                                                                   2, 18, 101, 452, 1018
     231
         Pa, 233Pa                                                                                                                                                            2, 16, 18, 101
     233
         U, 235U, 238U, 239-241Pu, 252Cf                                                                                                                                      2, 4, 16, 18, 101, 452,
                                                                                                                                                                              1018
     242
           Pu                                                                                                                                                                 2, 4, 16, 18, 101, 1018
     253
           Cf                                                                                                                                                                 2, 18, 101, 452




Total Monte Carlo approach (Rochman and Koning, 2011). This                                               the perturbed CASMO-5 library that was used during the depletion
method is different from another attempt at propagating fission                                           calculation. The relative reactivity worth computational uncer-
yield uncertainties reported in Leray et al. (2016) where the per-                                        tainty is then determined from the set of 500 LWR-PROTEUS Phase
turbed fission yields were sampled from a VCM which was deter-                                            II model outcomes.
mined from correlations obtained from the GEF code and
variance from evaluated data files. As reported in Leray et al.
(2017), the chosen approach overestimates the fission yield uncer-                                        4. Numerical results
tainties and as such, the approach reported in Leray et al. (2016) is
probably better. However, it is chosen to allow for future activities                                        In this section, the results of the uncertainty propagation for the
where the C/E of the fission products concentrations will be used to                                      modeling of the LWR-PROTEUS Phase II experiments are presented
adjust directly the GEF model parameters.                                                                 and discussed. First, the effect of the cross section and fission yield
    The ENDF-formatted file contains perturbed fission yield data                                         perturbations on the relative reactivity worth uncertainty is ana-
for around 1300 isotopes. Because of the simplified decay chains                                          lyzed. Then, the UQ results are presented for all samples and mod-
of CASMO-5, the fission yield data contained in the ENDF file needs
to be reduced to the set of isotopes considered in CASMO-5. Conse-
quently, an additional treatment is required to convert the per-
turbed fission yield data available in the ENDF files into a format
usable by CASMO-5. Such process is done at PSI using a set of
scripts called ‘‘Efficace”. In the first step of Efficace, a mapping
between the independent and cumulated yields of the various iso-
topes of the respective chains of the ENDF file and CASMO-5 library
is determined. Next, the information in each perturbed ENDF file is
processed using the previously determined mapping to generate
the perturbed fission yields corresponding to the CASMO-5 decay
chains. To preserve the accuracy of the fission yields available in
the CASMO-5 library, the perturbed fission yield data are expressed
as relative perturbations. Fig. 5 illustrates the algorithm used in
Efficace. Only the fission yields of the 235U, 238U, 239Pu, and 241Pu
isotopes are currently perturbed in SHARK-X.

3.4. Uncertainty quantification of LWR-PROTEUS Phase II

   The uncertainty propagation for the simulation of the LWR-
PROTEUS Phase II experiment is conducted based on the computa-
tional scheme described in Section 2.2. The depletion calculations
are repeated 500 times based on the same input model but with
500 perturbed cross sections and fission yields libraries.
   With respect to the reactivity worth measurements, the isotopic
composition of the burnt fuel sample is extracted from each per-
turbed depletion output. The calculations required for the determi-
nation of the relative reactivity worth are then carried out using                                            Fig. 5. Algorithm of efficace to generate the fission yield data for SHARK-X.
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                                      15


erating conditions. Next, the effect of the choice of irradiation his-
tory option is investigated. The implications in terms of validation
of reactivity loss predictions are discussed at the end of the section.


4.1. Effect of fission yield perturbation

    The fission yield perturbation impacts only the depletion calcu-
lation, i.e. the determination of the burnt fuel sample composition.
Consequently, its effect is analyzed for the kinf uncertainty during
the depletion calculation. The cross sections are not perturbed in
this section. The depletion calculation is performed with the U1
sample and the VENDOR irradiation history. The dotted line of
Fig. 6 shows the evolution of kinf with burnup using the nominal
fission yields, whereas the solid line shows kinf absolute standard
deviation when propagating the uncertainty of the fission yields.
At the beginning of the shutdown cooling period, the reactivity
increases by about 9000 pcm. Because the fuel and moderator tem-
perature are set to room temperature at the beginning of shutdown
cooling period, the reactivity increases by about 4000 and
2000 pcm due to the moderator and fuel temperature, respectively.                       Fig. 7. Histogram of kinf at EOC for the U1 fuel sample with the VENDOR irradiation
After one cooling step, many of fission products, especially xenon,                     history due to fission yield variation.
have decayed out, so that the reactivity also increases by about
3000 pcm.
    The absolute kinf uncertainty increases with exposure. It is equal                  that the distribution of kinf values for a given exposure is not a nor-
to 470 pcm at end of cycle (EOC) before shutdown and cooling. As                        mal distribution. Fig. 7 shows the histogram of kinf samples at EOC
reported in previous publications (Leray et al., 2016, 2017), the                       before shutdown cooling for the U1 fuel sample. The kinf distribu-
PSI-GEF method utilized in this paper overestimates the kinf uncer-                     tion is clearly non-normal as revealed by a Shapiro-Wilk test
tainty. Other work in the field (Williams, 2013; Zwermann, 2014;                        (Yap and Sim, 2011) with 5% significance level.
Díez, 2015; Rochman et al., 2018; Gauld, 2011) have produced much                           The non-normal kinf distribution hints at an input parameter
lower estimates for kinf uncertainty (Williams, 2013) or showed that                    with a non-normal input probability distribution function (PDF)
the contribution of fission yield uncertainty to decay heat is small                    or at non-linearities introduced by CASMO-5 during the depletion
(Rochman et al., 2018; Gauld, 2011). It is also visible that the kinf                   calculation. The bottom panels of Fig. 8 show the distribution and
uncertainty increases during the cooling periods. Such behavior is                      convergence of the 134I cumulative fission yield data due to ther-
explained by the buildup of isotopes with large absorption cross sec-                   mal fissions in 239Pu considering the 500 perturbed ENDF files.
tion coming from the decay of daughter nuclides with short half-life                    The top panels show the resulting perturbation factors for the
(149Pm and 143Pr for 149Sm and 143Nd respectively) (Leray, 2017).                       CASMO-5 library obtained with Efficace. It should be noted that
    Moreover, the difference of kinf between the mean value of the                      the 134I fission yield is the largest among the yields for fissions in
                                                                                        239
perturbed samples and the nominal (unperturbed) value is mono-                             Pu. However other daughter isotopes have non-normal distribu-
tonically increasing up to around 200 pcm at 755 days suggesting                        tions. A detailed sensitivity analysis could be carried out to deter-




     Fig. 6. Evolution of the nominal kinf and its absolute standard deviation with the VENDOR irradiation history when propagating the fission yield uncertainties.
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
16                                                 J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22




                                                                                                                                         Nominal




                                                                                                                                         Nominal




                                                                      134                                                    239
                               Fig. 8. Convergence and histogram of     I cumulative fission yield from thermal fission in         Pu.



mine the largest contributors to the kinf uncertainty. It will be done            moderating conditions. The correlations are shown in Fig. 10. All
in a later communication. As shown in the convergence behavior of                 Pearson coefficients are larger than 0.7 which suggests that even
fission yield value, the mean of the perturbed fission yields does                though the exposure of the fuel samples and the spectral condi-
not converge to the nominal (unperturbed) value. Moreover, the                    tions considered vary widely, the behavior of the relative reactivity
histograms of the ENDF and CASMO fission yields do not have nor-                  worth with respect to perturbations to cross sections and fission
mal distributions. Therefore, this suggests that the non-normal kinf              yields is similar.
distribution shown in Fig. 7 is due to the non-normal distribution                   The correlations are generally of lower magnitude when consid-
of fission yields obtained from the ENDF files. Of course, non-                   ering different fuel type (UO2 and MOX) as between the samples of
linearities due to the depletion chain itself cannot be excluded.                 the same fuel type but with different exposure. The moderating
However, normal kinf distributions cannot be excluded when per-                   conditions affect the correlation coefficients the most. Even though
turbing individually the fission yields of the other fissile isotopes,            the correlation coefficients are fairly large between the UO2 sam-
which suggests that the non-linearity due to the depletion chain                  ples in the light water and the mixture of light and heavy water
are small.                                                                        conditions, their magnitude is reduced in the borated water condi-
                                                                                  tions. The MOX samples are affected less by the type of the
                                                                                  moderator.
4.2. Relative reactivity worth UQ

    The uncertainty propagation results for the relative reactivity               4.3. Effect of irradiation history
worths are obtained assuming perturbations to the following set
of input parameters: cross sections, averaged number of neutrons                      In this section, the effect of the irradiation history description on
produced per fission (t), fission spectrum (v) and fission yields                 the relative reactivity worth UQ is analyzed by considering the
for all isotopes considered in CASMO-5. Fig. 9 shows the conver-                  VENDOR and BOHR irradiation histories. First, the depletion calcu-
gence behavior as well as the output distributions for the relative               lations and relative reactivity worth calculations are carried out for
reactivity worth of the U1 sample in the H2O moderating condition                 the U1 sample only considering the uncertainties on the cross sec-
with the VENDOR irradiation history. The 95% confidence intervals                 tions. Fig. 11 shows the evolution of the nominal kinf of the assembly
of the mean and the standard deviation of the population of results               containing the U1 sample together with its absolute standard devi-
are illustrated. Due to the proprietary nature of the data, the rela-             ation during irradiation in the nuclear power plant. Although the
tive reactivity worth is normalized to one and its distribution mod-              nominal kinf values are different with difference up to 1300 pcm
ified accordingly. The relative standard deviation is equal to 3.2%. It           at EOC for the two irradiation histories, the uncertainties of kinf
corresponds to the quadratic sum of the cross section and fission                 due to cross sections perturbations show a similar trend. (Note that
yield uncertainties, respectively 2.2% and 2.1% suggesting that                   the difference of discharge burnup between the two histories stems
cross sections and fission yields can be considered as independent                from the adjustment of the burnup in the BOHR methodology to
parameters and that no large interactions exist in the model.                     match the measured 148Nd content (see Section 2.2.1).)
    A normality test for the relative reactivity worth distribution of                Table 3 shows the C/E uncertainty values for the relative reac-
each fuel sample in each moderating condition is performed consid-                tivity worths obtained assuming the VENDOR and BOHR histories.
ering fission yields perturbations separately from the cross sections             The magnitudes of the C/E uncertainties are similar for both irradi-
perturbations. Considering only the cross sections perturbations, the             ation histories. With 500 samples, the observed differences are not
null hypothesis that the distribution is normal is not rejected. How-             statistically significant. It suggests that uncertainty estimates for
ever, the hypothesis is rejected when fission yields are perturbed.               the relative reactivity worth are not affected by the modeling of
Similarly to the kinf distribution at EOC investigated in Section 4.1,            the irradiation history.
the relative reactivity worths have non-normal distributions due                      Second, the comparison of the relative reactivity worths pre-
to the non-normal distributions of the fission yields.                            dicted with CASMO-5 for all burnt fuel samples, all moderating
    Using the perturbed samples for the relative reactivity worth, it             conditions and for both irradiation histories is detailed below. It
is possible to compute the correlations (Pearson correlation coeffi-              allows estimating the variations of the computational results due
cients) between the various fuel samples and for the three                        to the specification of the irradiation history. Because we are com-
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                                17




Fig. 9. Histogram and convergence behavior for the normalized relative reactivity worth of the U1 sample in the H2O moderating condition with the VENDOR irradiation
history when considering uncertainties on cross section (left), fission yield (middle) and cross section plus fission yield (right).


paring only two irradiation histories (VENDOR and BOHR), we can-                       ference is in the percent range (0.93–2.20%) and is similar irrespec-
not strictly speaking extract an uncertainty on the sample reactiv-                    tive of the moderating condition. At the same time, the difference
ity worth due to this parameter. We can, however, calculate a                          for the UO2 fuel samples is larger than that for the MOX fuel sam-
difference between the results obtained with either irradiation                        ples. As shown in Table 3, the relative uncertainty of the C/E value
modeling and compare its magnitude with that of the uncertainty                        from cross sections perturbation is around 2.3% regardless of the
due to nuclear data.                                                                   irradiation history. Therefore, the uncertainty propagated from
    The difference of C/E values due to the irradiation history is cal-                cross section is only slightly larger than the difference in reactivity
culated by Eq. (4) which represents the mean of the C/E bias                           worth coming from the irradiation history. This estimation is only
obtained for a given group of C/E values. In order to identify poten-                  preliminary. A more complete analysis would require investigating
tial trends, the C/E values are grouped by fuel or moderator type.                     many more plausible irradiation histories (e.g. different discharge
The full set of groups considered here is shown in Table 4.                            burnup, admissible power, etc.) for a stochastic treatment of the
                                                                                     uncertainty such as in Rochman et al. (2018) or defining ‘‘en-
                        VENDOR       BOHR 
                 1X N b
                         c;g      bc;g                                               velope” irradiation histories to obtain a conservative estimate of
dirradiation   ¼                                                           ð4Þ
 g
                 N c¼1        BOHR                                                   the impact of this parameter.
                             b     
                                  c;g

               VENDOR      BOHR
where bc;g     and bc;g are the respective C/E value for the VENDOR                    4.4. Validation of CASMO-5 calculations for the reactivity worth
and BOHR histories in a given group g. N is the total number of C/E                    measurements of LWR-Proteus Phase II
values in the group g. The goal of this section is to estimate how
dependent are the simulation results to the irradiation history with                      The relative reactivity worth bias uncertainties obtained with
respect to the other source of computational uncertainty considered.                   SHARK-X are listed in Table 5. Only the uncertainties due to cross
   Table 4 shows the differences in reactivity worth due to the                        sections and fission yields are considered and the VENDOR irradi-
specifications of the irradiation history. The amplitude of the dif-                   ation history is considered. The latter choice is made as impact of
CASMO-5 - Uncertainty quantification of LWR-PROTEUS Phase II experiments using
18                                                        J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22




        Fig. 10. Pearson coefficients for the relative reactivity worth of the eleven burnt fuel samples and three moderating conditions based on the 500 samples.




               Fig. 11. Evolution of the nominal kinf and its absolute standard deviation using the VENDOR and BOHR irradiation histories for the U1 sample.




the irradiation history modeling on the uncertainty of the relative                          The bias uncertainties vary from 1.2% to 8.9% as a function of
reactivity worth is limited (see previous section). With the esti-                       the moderating conditions as well as of the fuel type and burnup
mates for the calculated and experimental uncertainties of relative                      of the LWR-PROTEUS Phase II sample. The uncertainty of the rela-
reactivity worth, the z-score can be calculated using Eq. (5).                           tive reactivity worth tends to decrease with burnup. As reported in
                                                                                         Grimm et al. (2017), the major contributors to the relative reactiv-
       jbc  1j
ez ¼              :                                                            ð5Þ       ity worth computational uncertainty for all burnt fuel samples is
         rbc                                                                             the average number of neutrons per fission for 239Pu because of
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                                               19


Table 3                                                                                        smaller, but increases with exposure in absolute term (the relative
Uncertainty of C/E for the U1 relative reactivity worth considering the VENDOR and             uncertainty decreases). Also, the bias uncertainty of the UO2 fuel
BOHR irradiation histories.
                                                                                               samples in the B2023 moderating condition is smaller than that
      Moderator          History      C/E Rel. std.        Confidence Interval (95%)           in the H2O and DHO moderating conditions. In contrast with the
      H2O                VENDOR       2.26%                [2.06%, 2.46%]                      UO2 fuel sample, the bias uncertainty of the MOX fuel samples in
                         BOHR         2.30%                [2.10%, 2.51%]                      B2023 moderating condition is larger than that in the H2O and
      DHO                VENDOR       2.25%                [2.05%, 2.44%]                      DHO moderating conditions.
                         BOHR         2.29%                [2.08%, 2.48%]                          In any case, the z-scores as obtained from these UQ estimates
      B2023              VENDOR       2.25%                [2.05%, 2.45%]                      are lower than 2 for most of the burnt fuel samples when consid-
                         BOHR         2.30%                [2.10%, 2.51%]                      ering the uncertainties on the cross sections and fission yields.
                                                                                               CASMO-5 performs very well at predicting the relative reactivity
                                                                                               worth of the LWR-PROTEUS Phase II experiment.
                                                                                                   For the sake of completeness, Table 5 also shows the bias uncer-
the large input uncertainties in the VCM library used. As a result,                            tainty and z-score for each burnt fuel sample and each moderating
the magnitude of the bias uncertainty of the UO2 fuel samples is                               condition from a previous work (Grimm et al., 2017). In this work,
smaller than that of the MOX fuel samples due to larger Pu content                             cross section data were only perturbed in the simulations of the
of the MOX fuel as shown at the bottom of Table 5. The MOX sam-                                reactivity worth measurements at PROTEUS. The isotopic composi-
ples have a larger uncertainty (both relative and absolute), which                             tion of each burnt fuel sample was assumed to be a normally dis-
decreases with exposure, while the UO2 samples uncertainty is                                  tributed random variable with its mean equals to the nominal



Table 4
Differences in reactivity worth due to the irradiation history.

      Case                                                                    Number of responses                           dirradiation (%)                Mean standard error
                                                                                                                             g
                                                                                                                                                            (68% confidence)
      All sample                                                              32                                            1.73%                           [1.67%,   1.78%]
      Fuel type                                    UO2                        20                                            2.20%                           [2.09%,   2.31%]
                                                   MOX                        12                                            0.93%                           [0.86%,   1.01%]
      Moderating condition                         H2O                        11                                            1.55%                           [1.41%,   1.69%]
                                                   DHO                        11                                            1.52%                           [1.38%,   1.66%]
                                                   B2023                      10                                            2.15%                           [1.93%,   2.36%]




Table 5
Bias uncertainty and z-score for each burnt fuel sample and each moderating condition with VENDOR irradiation history. (The approximate Pu content of each sample is also
provided.)

      H2O                                                         U1        U2         U3        U4         U5        U6             U7        M1      M2       M3             M4
      Previous work (Grimm et al., 2017)      Nominal C/E         0.921     0.972      0.937     0.931      0.948     0.950          0.983     0.954   0.940    0.953          0.996
                                              Mean C/E            0.921     0.972      0.937     0.931      0.948     0.950          0.983     0.954   0.940    0.953          0.996
                                              C/E unc. (%)        2.2%      1.7%       1.6%      1.7%       1.6%      1.6%           1.7%      4.9%    3.3%     2.6%           2.4%
                                              ez
                                              *
                                                                  3.6       1.6        3.8       4.1        3.2       3.2            1.0       0.9     1.8      1.8            0.2
      Present work                            Nominal C/E         0.921     0.971      0.933     0.927      0.942     0.944          0.971     0.959   0.941    0.949          0.992
                                              Mean C/E            0.930     0.981      0.943     0.936      0.952     0.954          0.981     0.972   0.953    0.961          1.004
                                              C/E unc. (%)        3.4%      2.7%       2.6%      2.6%       2.5%      2.5%           2.5%      5.2%    3.9%     3.3%           3.1%
                                              ez                  2.1       0.7        2.2       2.4        1.9       1.8            0.8       0.5     1.2      1.2            0.1

      DHO                                                         U1        U2         U3        U4         U5        U6             U7        M1      M2       M3             M4
      Previous work (Grimm et al., 2017)      Nominal C/E         0.972     0.999      0.974     0.983      0.987     0.972          0.980     0.983   0.985    0.980          0.995
                                              Mean C/E            0.972     0.999      0.974     0.983      0.987     0.972          0.980     0.983   0.985    0.980          0.995
                                              C/E unc. (%)        2.2%      1.9%       1.8%      1.8%       1.8%      1.8%           1.9%      4.3%    3.1%     2.5%           2.4%
                                              ez                  1.3       0.0        1.4       1.0        0.7       1.6            1.1       0.4     0.5      0.8            0.2
      Present work                            Nominal C/E         0.972     1.000      0.973     0.979      0.982     0.967          0.970     0.989   0.988    0.977          0.993
                                              Mean C/E            0.980     1.009      0.982     0.989      0.992     0.977          0.979     1.000   0.999    0.988          1.003
                                              C/E unc. (%)        3.2%      2.6%       2.5%      2.5%       2.5%      2.5%           2.4%      4.5%    3.5%     3.1%           2.9%
                                              ez                  0.6       0.3        0.7       0.4        0.3       1.0            0.9       0.0     0.0      0.4            0.1

      B2023                                                       U1        U2         U3        U4         U5        U6             U7        M1      M2       M3             M4
      Previous work (Grimm et al., 2017)      Nominal C/E         0.950     1.012      0.975     –          0.999     0.986          0.992     1.036   1.051    1.014          1.029
                                              Mean C/E            0.950     1.012      0.975     –          0.999     0.986          0.991     1.035   1.052    1.014          1.029
                                              C/E unc. (%)        2.0%      1.4%       1.3%      –          1.2%      1.2%           1.3%      8.8%    4.3%     2.8%           2.4%
                                              ez                  2.5       0.9        2.0       –          0.1       1.2            0.7       0.4     1.2      0.5            1.2
      Present work                            Nominal C/E         0.951     1.011      0.973     –          0.996     0.983          0.984     1.050   1.056    1.013          1.027
                                              Mean C/E            0.959     1.019      0.981     –          1.004     0.991          0.992     1.075   1.072    1.026          1.039
                                                                                                                                               **
                                              C/E unc. (%)        3.1%      2.2%       2.0%      –          1.9%      1.9%           1.9%       8.9%   4.9%     3.6%           3.2%
                                              ez                  1.3       0.9        1.0       –          0.2       0.5            0.4       0.8     1.5      0.7            1.2
      239
            Pu content (mg/gU)                                    6.5       5.6        5.5       5.8        5.5       5.5            6.0       40      26       18             15
  *
       ez: z-score.
 **
       Extreme value of bias uncertainty.
20                                                         J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22


Table 6
Component of relative reactivity worth uncertainty of the U1 sample in each moderating condition with the VENDOR irradiation history. (XS stands for cross sections and FY for
fission yields.)

     U1                                                         C unc. (%)                             H2O                           DHO                            B2023
     Previous work (Grimm et al., 2017)                         XS                                     2.07%                         2.11%                          1.88%
                                                                Tech. Param.                           0.57%                         0.58%                          0.55%
     Present work                                               XS                                     2.20%                         2.19%                          2.19%
                                                                FY                                     2.27%                         2.07%                          1.87%




depletion calculation results and its uncertainty taken as the                            1.87% and 2.27%, which is similar to the uncertainties resulting
experimental uncertainty of the isotopic composition measured                             from cross section perturbations. As discussed above, in the previ-
through chemical assay. The effects of both source of uncertainties                       ous work, the isotopic composition of burnt fuel sample is assumed
on the relative reactivity worth (combination of three keff calcula-                      to be a normally distributed random vector. The resulting uncer-
tions – see Eq. (1)) were summed quadratically to determine the                           tainties on the reactivity worth bias (labelled Technological Param-
computational uncertainty of relative reactivity worth.                                   eter in Table 5) ranged from 0.55% to 0.58% and impact
    In the present work, the cross section and fission yield perturba-                    significantly less the bias uncertainty than the fission yields uncer-
tions are considered throughout the modeling of the irradiation of                        tainty. As the fission yields uncertainties drive the uncertainty of
the fuel samples in the nuclear power plant as well as that of the                        the fission products composition (Aures et al., 2017), their effects
oscillation measurements in PROTEUS. The burnt fuel sample com-                           can be compared to the technological parameter uncertainties
position uncertainties are a byproduct of the uncertainty propaga-                        reported in the previous work, as both represent the effect of the
tion during the depletion calculation. Propagating input                                  fuel composition uncertainty on the relative reactivity worth of
uncertainty through the irradiation of the fuel samples with                              the burnt fuel sample. It should be noted however, that the uncer-
CASMO-5 as well as considering fission yield uncertainties among                          tainty on the relative reactivity worth due to the fission yield per-
the input uncertainties lead to larger computational uncertainty on                       turbation should be lower than the technological parameter
the relative reactivity worth than previously reported.                                   uncertainty of (Grimm et al., 2017) because the effect of the fission
    In order to better understand the larger uncertainty estimates                        yield uncertainty does not affect the actinide concentration uncer-
obtained in this work, the breakdown of the contributions of the                          tainty. As reported in Leray et al. (2017), Rochman (2017), fission
various type of input parameters to the computational uncertainty                         yield input uncertainties vary widely depending on the library
of the U1 sample is reported in Table 6. The computational uncer-                         and method used to determine them. The method used in this
tainty coming from cross section perturbations is slightly larger                         work (PSI-GEF method) appears to overestimate the uncertainty
than that reported in the previous work since the cross section                           (Leray et al., 2017). Finally, the behavior of bias uncertainty with
uncertainties have been propagated through the fuel irradiation                           respect to exposure, fuel type, and moderating condition does
and the fuel sample oscillations calculations (while they were                            not change between the previous and present works.
propagated only through the oscillation calculations in the previ-                           Combining the C/E and its uncertainty, the performance of
ous work). Moreover, the uncertainties on the calculated relative                         CASMO-5 to predict the LWR-PROTEUS Phase II reactivity worth
reactivity worth due to fission yield uncertainties range between                         measurements is assessed. Fig. 12 shows the mean values of the




                                 Fig. 12. Mean of the C/E distributions for the relative reactivity worth of all samples with 2-r error bars.
J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22                                            21


Table 7
Fitting results of linear regression analysis on the C/E values.

  Fuel type                 Moderating condition                   C/E                  Confidence interval               Slope            Confidence interval
                                                                                        of C/E (95%)                                       of slope (95%)
  UO2                       H2O                                    0.954                [0.939, 0.969]                    3.53E04         [4.16E04, 1.12E03]
                            DHO                                    0.987                [0.978, 0.995]                    1.86E04        [6.19E04, 2.46E04]
                            B2023                                  0.991                [0.974, 1.007]                    8.07E05         [7.84E04, 9.45E04]
  MOX                       H2O                                    0.972                [0.950, 0.994]                    6.72E04         [2.49E03, 3.83E03]
                            DHO                                    0.997                [0.991, 1.004]                    3.29E05        [1.11E03, 1.05E03]
                            B2023                                  1.053                [1.029, 1.077]                    9.82E04        [3.38E03, 1.42E03]



C/E distributions for the relative reactivity worth with its uncer-                          tions. Indeed, at first the authors did not consider any uncertainty
tainty for all samples. As mentioned in Section 4.2, nominal and                             on the code values and in a later stage considered only an approx-
mean values are different due to the non-normal distribution of                              imate uncertainty estimation in which they propagated the cross
relative reactivity worth. The differences are small even though                             section uncertainties in the reactivity measurements at PROTEUS
statistically significant (see Table 5).                                                     but not in the model for the fuel irradiation. Instead, they assumed,
    The distributions of the C/E ratios grouped by moderator and                             as a proxy, that the uncertainty on the burnt fuel sample isotopic
fuel type are normal according to a Shapiro-Wilk test (Yap and                               composition (labelled technological parameters) was that of the
Sim, 2011) with a 5% significance level. They can be subjected to                            burnt fuel isotopic composition measurements.
a paired t-test to see if their means are similar or not. Considering                            In this paper, we proposed an improved treatment of the
the UO2 samples, the t-tests show significant differences at the 95%                         sources of uncertainties for the reactivity worth code predictions,
confidence level between the mean values for H2O and B2023 con-                              which was tested previously on a single spent fuel sample (Leray
ditions as well as between the H2O and DHO conditions, but no sig-                           et al., 2016; Rochman et al., 2018). We propagated the cross section
nificant differences between the DHO and B2023 conditions.                                   and fission yield uncertainties in the simulations of the NPP and
Considering the MOX samples, the paired t-tests show significant                             PROTEUS irradiations and quantified the impact of refining the
differences at the 95% confidence level between the values for                               irradiation history. The resulting total relative uncertainties on
H2O and B2023 as well as between the DHO and B2023 conditions,                               the C/E values for the relative reactivity worth varied from 1.2%
but no significant difference between the H2O and DHO conditions.                            to 8.9% for all moderating conditions and fuel types. The uncertain-
For the H2O moderating condition and the UO2 samples, all                                    ties on cross sections and fission yields contributed at the same
CASMO-5 simulations under-predict the experimental values for                                level to this uncertainty (2.0%). The cumulative fission yields of
                                                                                             239
relative reactivity worth by an average of 0.046, with a 95% confi-                             Pu produced by the GEF code had non-normal probability distri-
dence interval of ±0.020 around that value. Consequently, on aver-                           butions and led to non-normal distributions for the relative reac-
age simulations and experiments show a significant bias for this                             tivity worths. The non-normal distributions were coming from
moderation condition. The conclusions are the same for the                                   the data produced by the GEF code itself and not by the treatment
B2023 moderating conditions, as all CASMO-5 simulations over-                                to adapt them to the simplified decay chains of CASMO-5.
predict the experimental value by an average of 0.053 with a 95%                                 As we observed no linear trend with burnup in the bias between
confidence interval for the mean of ±0.024 around that value.                                the measured relative reactivity worth and that predicted with
    Finally, a linear regression analysis is performed on the C/E val-                       CASMO-5, this work confirms that CASMO-5 can predict accurately
ues for each population defined by the previous statistical analysis                         the changes of reactivity with fuel exposure up to large burnup val-
with respect to burnup, considering the updated bias uncertainties                           ues (120 MWd/kg).
determined in the present work. The fitting results as well as their                             Finally, we estimated the effect of the irradiation history
uncertainties are shown in Table 7. The zero slope is within the                             description on the relative reactivity worths by comparing two
confidence interval for each fit which suggests that there is no lin-                        independent irradiation histories (a detailed one determined inter-
ear trend with burnup in the accuracy of the prediction of relative                          nally at PSI and a coarser one provided by the fuel vendor). The dif-
reactivity worth. That is, the change of reactivity with increased                           ference in the C/E values for the relative reactivity worth obtained
fuel exposure is captured accurately by CASMO-5, even for the                                with the two irradiation histories is around 2%, i.e. on par with the
cases for which CASMO-5 may have a computational bias (the                                   uncertainty coming from nuclear data. This suggests that we
H2O moderating condition). It should be noted here that the fitting                          should more carefully estimate the role of the irradiation history
is done assuming normal distribution for the relative reactivity                             modelling in the future.
worth. As shown in Section 4.2, it is not the case due to the fission
yields uncertainties. However, as the difference between the mean
and nominal values is small as reported in Table 5, it is expected                           Acknowledgements
that such non-normal distribution may not affect significantly
the validation outcomes of this study.                                                          The work was partially funded by the Swiss Nuclear Power
                                                                                             Plants (swissnuclear) in the framework of the QUASAR project.
                                                                                             The LWR-PROTEUS program was conducted jointly by PSI and
5. Conclusion                                                                                swissnuclear, with specific contributions from Kernkraftwerk Goes-
                                                                                             gen (KKG) for Phase II. Many thanks are also due to the PROTEUS
   This paper summarizes additional work performed towards the                               experimental team for their precise work and to the operational
validation of the CASMO-5 lattice and depletion code against the                             team for reliable operation and maintenance of the facility.
reactivity worth measurements of the 11 PWR spent fuel samples
of the LWR-PROTEUS program. In a previous publication (Grimm
et al., 2017), the authors determined biases between measured                                Appendix A. Supplementary data
and calculated values for all these burnt fuel samples and dis-
cussed trends of the bias with burn-ups, fuel type and moderation                               Supplementary data to this article can be found online at
conditions. However, the bias analysis suffered from several limita-                         https://doi.org/10.1016/j.anucene.2019.03.023.
22                                                             J. Park et al. / Annals of Nuclear Energy 131 (2019) 9–22


References                                                                                    Leray, O., 2015. SHARKX v.2.1 – Fission Yield Perturbation Methodology:
                                                                                                  Implementation of Correlations and Application on the LWR-PROTEUS Phase
                                                                                                  II U1 Sample, TM-41-15-22 October. Paul Scherrer Institut, Villigen, CH.
Aures, A., Bostelmann, F., Hursin, M., Leray, O., 2017. Benchmarking and application
                                                                                              Leray, O. et al., 2017. Fission yield covariances for JEFF: a Bayesian Monte Carlo
    of the state-of-the-art uncertainty analysis methods XSUSA and SHARK-X. Ann.
                                                                                                  method. EPJ Web Conf. 146, 09023.
    Nucl. Energy 101, 262–269.
                                                                                              Leray, O., Rochman, D., Grimm, P., Ferroukhi, H., Vasiliev, A., Hursin, M., Perret, G.,
Chadwick, M.B. et al., 2011. ENDF/B-VII.1 nuclear data for science and technology:
                                                                                                  Pautz, A., 2016. Nuclear data uncertainty propagation on spent fuel nuclear
    cross sections, covariances, fission product yields and decay data. Nucl. Data
                                                                                                  compositions. Ann. Nucl. Energy 94, 603–611.
    Sheets 112, 2887–2996.
                                                                                              Leray, O., Fiorito, L., Rochman, D., Ferroukhi, H., Stankovskiy, A., Van den Eynde, G.,
Díez, C.J. et al., 2015. Comparison of nuclear data uncertainty propagation
                                                                                                  2017. Uncertainty propagation of fission product yields to nuclide composition
    methodologies for PWR burn-up simulations. Ann. Nucl. Energy 77, 101–114.
                                                                                                  and decay heat for a PWR UO2 fuel assembly. Progr. Nucl. Energy. https://doi.
Ferroukhi, H., Hofer, K., Hollard, J.-M., Vasiliev, A., Zimmermann, M.A., 2008. Core
                                                                                                  org/10.1016/j.pnucene.2017.05.033.
    Modelling and Analysis of the Swiss Nuclear Power Plants for Qualified R&D
                                                                                              Rearden, B.T., 2009. Overview of SCALE 6.2. Proceeding of ANS NCSD, Wilmington,
    Applications. Proceeding of Physor 2008, Interlaken, Switzerland, September
                                                                                                  NC, USA.
    14–19.
                                                                                              Rhodes, J., Gheorghiu, N., Ferrer, R.M., Hykes, J., Smed, T., 2012. CASMO-5 User’s
Ferroukhi, H., Leray, O., Hursin, M., Vasiliev, A., Perret, G., Pautz, A., 2014. Study of
                                                                                                  Manual Rev. 5. SSP-07/431 Rev. 5.
    nuclear decay data contribution to uncertainties in heat load estimations for
                                                                                              Rochman, D. et al., 2017. Nuclear data uncertainties for typical LWR fuel assemblies
    spent fuel pools. Nucl. Data Sheets 118, 498–501.
                                                                                                  and a simple reactor core. Nucl. Data Sheets 139, 1–76.
Gauld, I.C. et al., 2011. Isotopic depletion and decay methods and analysis
                                                                                              Rochman, D., Koning, A.J., 2011. How to randomly evaluate nuclear data: a new data
    capabilities in SCALE. Nucl. Technol. 174 (2), 169–195.
                                                                                                  adjustment method applied to 239Pu. Nucl. Sci. Eng. 169, 68–80.
Grimm, P., Meister, A., Jatuff, F., Benger, H.D., Chawla, R., 2001. The LWR-PROTEUS
                                                                                              Rochman, D., Vasiliev, A., Ferroukhi, H., Janin, D., Seidl, M., 2018. Best estimate plus
    Phase II experimental programme on high-burnup reactivity effects and
                                                                                                  uncertainty analysis for the 244Cm prediction in spent fuel characterization.
    isotopic compositions. 2001 Annual Meeting on Nuclear Technology,
                                                                                                  BEPU 2018, Lucca, Italy, May 13–19.
    Jahrestagung Kerntechnik, Dresden, Germany, May 15–17.
                                                                                              Rochman, D., Vasiliev, A., Dokhane, A., Ferroukhi, H., 2018. Uncertainties for Swiss
Grimm, P., Gunther-Leopold, I., Berger, H., 2006. Burnup calculations and chemical
                                                                                                  LWR spent nuclear fuels due to nuclear data. EPJ Nucl. Sci. Technol. 4, 6.
    analysis of irradiated fuel samples studied in LWR-PROTEUS Phase II.
                                                                                              Schmidt, K., Jurado, B., 2014. GEF A General Description of Fission Observables.
    Proceeding Physor 2006, Vancouver, Canada, September 10–14.
                                                                                                  OECD Nuclear Energy Agency. JEFF Report 24, NEA/DB/DOC.
Grimm, P., Murphy, M.F., Jatuff, F., Seiler, R., 2008. Analysis of reactivity worths of
                                                                                              Schmidt, K., Jurado, B., Amouroux, C., Schmitt, C., 2016. General description of
    highly-burnt PWR fuel samples measured in LWR-PROTEUS Phase II. Physor
                                                                                                  fission observables: GEF model code. Nucl. Data Sheets 131, 107–221.
    2008, Interlaken, Switzerland, September 14–19.
                                                                                              Wieselquist, W., 2012. CASMO-5MX: Tools for Sensitivity Analysis and Uncertainty
Grimm, P., Perret, G., Ferroukhi, H., 2014. CASMO-4E and CASMO-5 analysis of the
                                                                                                  Quantification with respect to Nuclear Data in CASMO-5M, TM-41-12-09 July.
    isotopic compositions of the LWR-PROTEUS Phase II burnt PWR UO2 fuel
                                                                                                  Paul Scherrer Institut, Villigen, CH.
    samples. Physor 2014, Kyoto, Japan, September 28–October 3.
                                                                                              Williams, M.L. et al., 2013. Statistical sampling method for uncertainty analysis
Grimm, P., Hursin, M., Perret, G., Siefman, D., Ferroukhi, H., 2017. Analysis of
                                                                                                  with SCALE and XSUSA. Nucl. Technol. 183 (3), 515–526.
    reactivity worths of burnt PWR fuel samples measured in LWR-PROTEUS Phase
                                                                                              Williams M., Wiarda D., Arbanas G., Broadhead B.L., 2009. SCALE Nuclear Data
    II using a CASMO-5 reflected-assembly model. Progr. Nucl. Energy 101, 280–
                                                                                                  Covariance Library, ORNL/TM-2005/39 Version 6.
    287.
                                                                                              Yap, B.W., Sim, C.H., 2011. Comparisons of various types of normality tests. J. Stat.
Gunther-Leopold, I. et al., 2007. LWR-PROTEUS Programme, Phase II, Final Report,
                                                                                                  Comput. Simul. 81, 2141–2155.
    TM-43-06-05. Paul Scherrer Institut, Villigen, CH.
                                                                                              Zwermann, W. et al., 2014. Nuclear data uncertainty and sensitivity analysis with
Herrero, J.J. et al., 2017. Impact of nuclear data uncertainty on safety calculations for
                                                                                                  XSUSA for fuel assembly depletion calculations. Nucl. Eng. Technol. 46, 343–
    spent nuclear fuel geological disposal. EPJ Web Conf. 146, 9028.
                                                                                                  352.
Hursin, M., Perret, G., Pautz, A., 2015. Verification of the new implementations in
    SHARKX against TSUNAMI to perform pinpower UQ and representativity
    analysis. Ann. Nucl. Energy 77, 300–309.
You can also read
Next part ... Cancel