Performance assessment of adjusted nuclear data along with their covariances on the basis of fast reactor experiments

Highlights: • Covariances and basic data need to be consistent when analyzing fast systems. • Efficient adjustments can be obtained even on a limited assimilation database. • Improvement of targeted ratios of computed to experimental values is then achieved. • Correspondingly, reduced uncertainties due to data uncertainties appear justified. - Abstract: In view of fast reactor analyses, it is shown that efficient nuclear data adjustments can be obtained on a limited assimilation database consisting of just six well documented integral parameters, i.e. the central spectral indices measured in Godiva and ZPPR-9. This study uses a Generalized Linear Least-Squares (GLLS) based data assimilation method by means of Asymptotic Progressing Incremental nuclear data Adjustment (APIA) simulations with two incremental steps, one involving Godiva; the other one ZPPR-9. Consistent JEFF-3.3 and TENDL based prior data including their covariances are used; correspondingly, the assimilation leads to posterior JEFF-3.3 and TENDL data. 34 target experiments are then investigated by means of both prior and posterior data. These experiments consist of spectral indices as well as multiplication factors which pertain to 11 fast spectrum configurations including the six integral parameters which are part of the assimilation. It is found that (1) after adjustment the mean χ² is strongly reduced to values smaller than 2, in each case. (2) The performance of the adjustment is comparable between JEFF-3.3 and TENDL also in terms of the Gaussian Coverage Factor (GCF), which is the common surface spanned below two normal probability density functions associated with data means and variances. Correspondingly it is found by comparing JEFF-3.3 and TENDL data among each other in a similar way by computing GCFs of cross-sections, that (3) posterior data overall appears less deviating than prior data. It seems worthwhile investigating whether similar promising results and trends assessed based upon a deterministic code, namely ERANOS, are reproducible with a stochastic method which is deemed to be a reference tool.