26.02.2013 Views

Program - Brookhaven National Laboratory

Program - Brookhaven National Laboratory

Program - Brookhaven National Laboratory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

In the resolved resonance region nuclear resonance parameters are mostly deduced from experimental measurements<br />

using a least squares adjustment. Derived parameters can be mutually correlated through the<br />

adjustment procedure as well as through common experimental or model uncertainties. In this contribution<br />

we investigate four different methods to propagate the additional covariance caused by experimental or<br />

model uncertainties into the evaluation of the covariance matrix of the estimated parameters: (1) including<br />

the additional covariance into the experimental covariance matrix based on calculated or theoretical<br />

estimates of the data; (2) include the common uncertainty affected parameter as additional experimental<br />

data input and model parameter; (3) evaluation of the full covariance matrix by Monte Carlo sampling<br />

of the common parameter; and (4) retroactively including the additional covariance by marginalization<br />

proposed by Habert et al. [1]. These different methods are investigated based on simulated data and<br />

based on experimental data resulting from transmission and capture measurements on 197 Au covering the<br />

resolved and unresolved resonance region. Using the simulated data the impact of different components<br />

are investigated: counting statistics, the self-shielding term in the yield expression, and the inclusion of<br />

a thermal cross section on the deduced covariance matrix is demonstrated. The impact of the different<br />

procedures on the covariance of the final cross section calculated is shown.<br />

Corresponding author: Peter Schillebeeckx<br />

[1] Habert et al., Nucl. Sci. Eng. 166, 276 – 287 (2010)<br />

PR 134<br />

Comparison of Two Approaches for Nuclear Data Uncertainty Propagation In MCNP(X)<br />

for Selected Fast Spectrum Critical Benchmarks<br />

Ting Zhu, Alexander Vasiliev, Hakim Ferroukhi, Paul Scherrer Institut. Dimitri Rochman, Nuclear<br />

Research and Consultancy Group. William Wieselquist, Oak Ridge <strong>National</strong> <strong>Laboratory</strong>.<br />

Nuclear data uncertainty propagation based on stochastic sampling (SS) is becoming more attractive while<br />

leveraging the huge increase in modern computer power. Two variants of the SS approach are compared<br />

in this study. The Total Monte Carlo (TMC) method [1] developed at the Nuclear Research Consultancy<br />

Group (NRG) prescribes probability density functions to theoretic parameters of nuclear reaction models<br />

and uses them to create random ENDF nuclear data files, which are translated subsequently into the<br />

ACE-formatted nuclear data files by NJOY. At the Paul Scherrer Institut (PSI), the Nuclear data Uncertainty<br />

Stochastic Sampling (NUSS) system is under development and based on the concept of generating<br />

the randomly perturbed ACE-formatted nuclear data files directly from applying groupwise nuclear data<br />

covariances onto the original pointwise ACE-formatted nuclear data [2]. Both the TMC and PSI-NUSS<br />

methods are applied with the Monte-Carlo code MCNP(X) for the well-studied Jezebel and Godiva fast<br />

spectrum critical benchmarks in this paper. At first, using TENDL-2012 libraries and associated covariance<br />

matrices, uncertainty assessments for keff and spectral parameters are carried out and compared.<br />

The propagated uncertainties due to various reactions of 239 Pu, 235 U and 238 U are crosschecked between<br />

the two methods. Furthermore, with the PSI-NUSS method, an examination of applying different groupwise<br />

perturbations onto continuous nuclear data and the effect of using ENDF/B-VII.1 with its covariance<br />

library versus the SCALE6 44-group covariance library are assessed. With the latter, the results obtained<br />

by PSI-NUSS are compared to those calculated by the Sensitivity and Uncertainty (S/U) methodology of<br />

the SCALE6 package. Generated under different principles, the uncertainty results from the deterministic<br />

S/U method serve as a separate verification for the two SS-based methods. The observed differences in the<br />

propagated uncertainties of nuclear data among the considered methods and covariance files are discussed.<br />

[1] D. Rochman, A.J. Koning, S.C. van der Marck, A. Hogenbirk and C.M. Sciolla, ”Nuclear data uncertainty<br />

propagation: Perturbation vs. Monte Carlo” Ann. Nuc. En. 38, 942 (2011). [2] T. Zhu, A.<br />

331

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!